Search/filter projects:
Evie Stergiakoulis (lead), Dr Sarah Lewis, Prof Michael Owen Prof Marianne van den Bree
Cleft of the lip and/or palate is a common birth defect worldwide and occurs at a rate of one in 650 live births in the UK. Being born with cleft places a significant burden on children, their families and the health system as they require surgery (multiple times depending on cleft type), and other interventions to improve appearance, speech, hearing, dentition and other adverse outcomes. They are also at increased risk of psychological, psychiatric and cognitive problems [1]. The UK-based Cleft Collective is a unique resource comprising the world’s largest cohort study of individuals affected by cleft and their families [2]. Rich longitudinal information is collected on the children’s mental health, parental, prenatal and early life factors as well as genetic data, providing unique opportunities to study genetic as well as environmental influences on risk of development of mental health conditions over time.
The PhD project will provide the first detailed description of neurodevelopmental and mental health outcomes in children with cleft and examine the contributions of genetic and environmental factors. We will use two unique genetically informative clinical cohorts of children; the University of Bristol Cleft Collective and the Cardiff University longitudinal ExperiencCes of people witH cOpy number variants (ECHO) study. Control samples will consist of the Avon Longitudinal Study of Parents and Children (ALPSAC) and the Millennium cohort which are deeply-phenotyped cohorts of typically developing children.
The aims of the study are: 1) To improve understanding of risk of neurodevelopmental and mental health outcomes in children born with cleft. This will be achieved by comparing children born with cleft to those at high genetic risk of neurodevelopmental and mental health outcomes but without cleft (children from the ECHO study) and typically developing children.
2) To improve understanding of the causes of neurodevelopmental and mental health outcomes in children born with cleft. This will be achieved by determining in children born with cleft the contribution of: a) composite genetic (polygenic) risk scores for neurodevelopmental and psychiatric disorders and b) rare genetic mutations.
3) To improve understanding of non-genetic factors, the project will also examine contributions of early developmental problems, family socio-economic status, family relationship quality, and traumatic experiences to risk of childhood psychiatric disorders in children born with cleft.
This multidisciplinary project will integrate expertise across cleft, genetic epidemiology, psychiatry, and epidemiology. Advanced epidemiological methods will be applied to study repeated measures and compare outcomes across cohorts. Copy Number Variants (CNVs) and other rare variants will be identified in children from the Cleft Collective and compared with mutations identified in children with neurodevelopmental disorders and children from the general population. Genetic epidemiological methods, such as polygenic risk scores, will be used to study the contribution of common variants and causally informative designs will be employed to test the causal link between cleft and mental health problems. The candidate will have the opportunity to develop into one of few experts globally with in depth understanding across these fields. The project will include training in advanced epidemiology, genetic epidemiology, bioinformatics and a mini-MD in the MRC Centre for Neuropsychiatric Genetics and Genomics at Cardiff.
[1] doi: 10.1016/j.jaac.2018.06.024
[2] http://www.bristol.ac.uk/dental/cleft-collective/
[3] doi: 10.1038/s41380-020-0654-3
[4] doi: 10.1371/journal.pgen.1007501
[5] doi: 10.1016/S2215-0366(19)30123-3
Dr Luisa Zuccolo (lead), Dr Laura Johnson, Dr. Laura Howe Dr. Maria Carolina Borges
The World Health Organisation recommends exclusively breastfeeding for 6 months, but in the UK, only 1% of infants meet the guidelines (1). Supporting mothers to maintain breastfeeding is a key priority for promoting child health. Breastfeeding for longer is associated with a reduced childhood obesity risk in observational studies(2), but breastfeeding is socially patterned meaning associations may be confounded. Also, the most common reason mothers report for stopping breastfeeding early is perceived milk insufficiency (often not real insufficiency)(1), so it is possible that mothers are more likely to worry about insufficient milk supply, and therefore to stop breastfeeding, if their baby is growing fast (which is associated with a higher risk of later obesity(3)). Thus a shorter breastfeeding duration may cause rapid growth or a faster growth rate may cause breastfeeding to stop. Disentangling this will help provide women with informed decisions about whether to stop breastfeeding.
This project aims to disentangle the relationship between infant feeding and infant growth and childhood obesity. This evidence will contribute to developing decision support systems to help mothers with infant feeding choices. Specific aims are to:
1. Establish whether and to what extent rapid infant growth affects infant feeding patterns, in particular shorter breastfeeding duration
2. Establish whether and to what extent longer breastfeeding duration protects against excessive weight gain in infancy
3. Investigate the persistence of these relationships with respect to childhood obesity
This project will use several methods developed and/or expertly applied within the MRC IEU, including Mendelian randomization, longitudinal modelling of childhood growth, and triangulation of epidemiological evidence.
Mendelian randomization is a study design that mimics a randomised experiment, by comparing groups of individuals randomly allocated a genetic predisposition for a trait or behaviour. Randomisation minimises confounding and reverse causation, both of which could explain the observed associations between breastfeeding and infant growth.
This project will harmonise phenotypic data on infant feeding across multiple cohorts and combine with data on maternal/offspring genetics and infant growth outcomes. The Early Growth Genetics (EGG) consortium, a collaboration of birth cohorts with genotype data combined with phenotypes including infant feeding, growth and later childhood obesity (5), will contribute data to this project, and so will several large biobanks internationally including the Norwegian HUNT study, the China Kadoorie Biobank and the UK Biobank.
Analyses will include:
1) Comparing the infant feeding patterns of children with different genetic predispositions to growing faster or having overweight/obesity;
2) Comparing infant growth or obesity in children genetically predisposed to being breastfed for longer, to those whose genetic makeup predicts shorter breastfeeding duration;
3) Cross-cohort comparisons utilising cohorts in diverse settings e.g. where the confounding of infant feeding by social class differs.
4) Formal triangulation of the evidence produced through 1-3 with previous evidence from Randomised Controlled Trials of breastfeeding promotion interventions.
1. DoH. Infant Feeding Survey 2010; http://www.ic.nhs.uk/catalogue/PUB00648/infaseed-serv-2010-earl-resu-rep.pdf
2. Yan J et al. BMC Public Health. 2014;14(1):1267
3. Zheng M et al. Obesity Reviews. 2018;19(3):321-32
4. Johnson L et al. Int J Obesity (2005). 2014;38(7):980-7
5. Middeldorp CM et al. Eur J Epidemiol. 2019;34(3):279-300
Dr Evie Stergiakouli (lead), Prof Claire Haworth, Dr Oliver Davis, Dr Alexandra Havdahl
Children are now exposed to and use digital media from a very young age both for entertainment and educational purposes. Parents and clinicians are concerned about the potential harmful effects that digital media use may have on children’s mental and physical health [1,2,3]. However, most studies have been cross-sectional and have shown positive as well as negative effects (with the exception of excessive use and extreme online harmful behaviour) [4] and they have not considered the effect of digital media use on children from the general population. COVID-19 and lockdown increased digital media use for the majority of children and this has been linked to positive effects, such as increased connectedness, at the time that it was most needed [5].
The aim of this project is to examine if digital media use is having an impact (positive or negative) on children’s mental health and explore the mechanisms behind any associations. We also aim to determine if there are sensitive windows in development or situations (such as lockdown) when digital media are most/least harmful and if there are groups of children (e.g. with pre-existing neurodevelopmental problems) that are particularly vulnerable.
The PhD student will use genetic epidemiological methods to test if digital media use is causally associated with mental health in children. Methods will include performing GWAS for digital media use, polygenic risk score analysis [5] and Multivariable Mendelian randomization [6]. These methods will be applied across large cohorts of children from the general population with mental health data across time, such as the Avon Longitudinal Study of Parents and Children (ALSPAC), the Millennium Cohort and the Norwegian Mother, Father and Child Cohort Study (MoBa).
1. Canadian Paediatric Society, Paediatr Child Health. 2019; 24(6): 402–408.
2. Ra CK, et al. JAMA. 2018;320(3):255–263. doi:10.1001/jama.2018.8931
3. Hoge E et al. Pediatrics. 2017;140(Suppl 2):S76-S80. doi: 10.1542/peds.2016-1758G.
4. Orben, A. Soc Psychiatry Psychiatr Epidemiol 2020; 55, 407–414.
5. Widnall E et al. 24th August 2020, NIHR School for Public Health Research
6. Davey Smith G, Hemani G, Human Molecular Genetics 2014;23(R1):R89–R98, https://doi.org/10.1093/hmg/ddu328
7. Leppert B et al. JAMA Psychiatry. 2019;76(8):834–842. doi:10.1001/jamapsychiatry.2019.0774
Dr. Siddhartha Kar (lead), Prof. Paul Brennan (IARC), , Prof. Tom Gaunt
Cancer is a disease of the genome. Certain changes that are acquired over the course of life in the genomes of healthy cells in the human body (somatic genomic changes) dysregulate the fine balance between cell death and proliferation. These somatic genomic aberrations are the cornerstone of malignant cellular transformation. Targeting somatic genomic changes is fundamental to the practice of precision cancer medicine. We understand that common exposures and cancer risk factors such as ultraviolet light and smoking accelerate the acquisition of these changes. However, little is actually known about how everyday exogeneous and endogenous factors such as diet, obesity, and insulin resistance relate to, and likely drive, carcinogenic changes in the somatic genome. This is because it is difficult to measure lifelong trajectories of the factors retrospectively at cancer diagnosis and expensive to measure them prospectively in large numbers of individuals until some of them develop cancer. Such one-time "snapshot" measures, even where feasible, are prone to bias and confounding. Specific inherited or germline genetic variants have been found to be robustly associated with these exposures or factors. Since genetic variants are allocated at random at conception and fixed thereafter, they are less affected by bias and confounding. The factor-associated variants provide remarkable proxies for the lifetime levels of these factors even in patients in whom the factor itself has not been measured. These variants collected into polygenic scores can serve as instruments to evaluate association between the germline genetically inferred levels of the factor and somatic/tumour molecular features and mechanisms that operate within the cancer.
1. To identify tumour molecular features associated with common exposures or putative cancer risk factors
Genome-wide association studies involving hundreds of thousands of individuals have identified germline variants that are robustly associated with different factors, ranging from body mass index to blood-levels of protein markers. This variation will be leveraged to generate personalised life-course profiles of these factors in cancer patients using germline genotype data. The association of these profiles with tumour gene expression, methylation, copy number, and mutations will then be evaluated at the level of single genes and multi-gene biological pathways in >11,000 tumours that have been subjected to deep germline-somatic molecular and clinical phenotyping in The Cancer Genome Atlas (TCGA) project.
2. To investigate the association between common exposures or putative cancer risk factors and cancer drug sensitivity
Over 1,000 cancer cell lines from the Genomics of Drug Sensitivity in Cancer project have been screened for their response to >450 cancer drugs either approved for use in patients or in development. Germline genotypes from the cell lines will be used to index the factors and the association of each index with therapeutic response assessed.
A key aspect of this project is flexibility in terms of the scientific direction taken with these rich data sets that will be provided to the student. It is envisioned that such flexibility may manifest in various ways such as (but certainly not limited to) encouraging an investigation of cancer risk factors in the context of genetic ancestry and sex (for example, the student may wish to study whether the impact of body mass index on the tumour genome differs by sex or ancestry).
The student will receive exceptional training in the handling and statistical analysis of large-scale, high-dimensional cancer genetic, genomic, transcriptomic, and epigenomic data sets and in the interpretation of findings based on these data sets. The student will apply a range of computational techniques including state-of-the-art Mendelian randomisation methods implemented in MR-Base, polygenic scoring approaches such as LD-Pred, and expression quantitative trait locus analysis using the R package Matrix eQTL. The project seeks to encourage a high-degree of flexibility both in terms of the scientific questions being asked and in terms of the methods being applied and should the student choose to do so, there is ample scope for implementation of artificial intelligence/machine learning-based methods in these data sets, etc. Relevant training in these methods will be provided. It is envisioned that the work will lead to multiple high-profile and highly interdisciplinary publications that will be led by the student, providing an excellent foundation for future scientific leadership.
The Cancer Genome Atlas project: Ding, L. et al. Cell 173, 305-320.e10 (2018).
The Genomics of Drug Sensitivity in Cancer project: Iorio, F. et al. Cell 166, 740–754 (2016).
Prof Deborah Lawlor (lead), Dr Maria Carolina Borges, Dr Ge Zhang Dr Nicole Warrington
Disruptions in maternal metabolism during pregnancy (e.g. maternal hyperglycaemia) potentially affect several aspects of maternal health and fetal development. Understanding which molecular pathways are implicated in adverse child and maternal health is key to inform effective interventions to improve maternal-child health; however, producing reliable evidence on that is challenging due to issues of confounding and reverse causation in observational studies and due to the scarcity of evidence from high-quality randomised controlled trials. Mendelian randomisation is a method that uses genetic variants robustly associated with modifiable exposures to generate more reliable evidence regarding which risk factors to intervene to produce health benefits. Mendelian randomisation can be used to discriminate causal from non-causal molecular traits, which may facilitate targeted development of interventions and inform evidence-based recommendations for pregnant women.
To use Mendelian randomization to systematically assess the effect of the maternal molecular profile (methylome, proteome and metabolome) during pregnancy on a wide-range of perinatal health outcomes (e.g. gestational hypertension, gestational diabetes, perinatal depression, need of induction of labour, caesarean delivery, early-membrane rupture, preterm delivery, birth weight, miscarriage, stillbirth).
1. To develop genetic instruments for probing the effect of changes in maternal molecular traits during pregnancy
2. To use novel statistical methods for partitioning genetic effects at single loci into maternal and offspring genetic components
3. To use (one-sample and two-sample) Mendelian randomisation to probe the causal role of maternal molecular traits on perinatal health
4. To triangulate Mendelian randomisation findings with findings from other study designs (e.g. negative paternal control, randomised controlled trials)
https://wellcomeopenresearch.org/articles/2-11
https://www.ncbi.nlm.nih.gov/pubmed/26978208
https://www.ncbi.nlm.nih.gov/pubmed/30815700
https://www.ncbi.nlm.nih.gov/pubmed/31335958
https://www.biorxiv.org/content/10.1101/737106v2.full
Dr Rebecca Richmond (lead), Gemma Sharp, Prof Deborah Lawlor
A large body of literature from animal studies has highlighted the role of core clock genes in relation to reproductive function [1]. In contrast, only a few epidemiological studies have implicated the circadian clock with human reproduction. These have largely focused on investigating the modification of menstrual cycle patterns in relation to shift work [2,3]. Circadian dysregulation as captured by sleep traits have also been associated with menstrual-related disorders [4] and fertility [5]. In turn, menstrual cycle characteristics 6 and reproductive events such as menopause [7] have been previously linked to changes in sleep, which may imply a bi-directional relationship.
Large epidemiological resources such as ALSPAC and the UK Biobank include detailed data on both sleep characteristics, reproductive traits and menstrual disorders which may be used to investigate associations between these traits. Genetic variants for a range of sleep characteristics, reproductive traits and menstrual conditions have been identified in recent genome-wide association studies (GWAS) which can be used establish and orient causal relationships. In particular, bivariate LD score regression calculates the genetic correlation between two traits [8] and Mendelian randomization (MR) uses those genetic variants most strongly associated with one trait to establish a causal effect on another [9-11].
To use observational and genetic epidemiological approaches within UK Biobank, ALSPAC, and other genetic studies and consortia (ReproGen, Human Reproductive Behaviour consortium, FinnGen) to: i) investigate cross sectional and prospective associations between sleep characteristics and reproductive/menstrual traits , ii) estimate the genetic correlations between sleep characteristics and reproductive/menstrual traits and iii) evaluate causal effects of sleep on reproductive traits and vice-versa.
1) Derive sleep, reproductive and menstrual measures from ALSPAC and UK Biobank
2) Perform multivariable regression analyses to investigate the associations between these traits, both cross sectionally and prospectively
3) Identify genome-wide association studies related to a series of sleep characteristics (insomnia, chronotype, sleep duration, daytime sleepiness, circadian disruption), reproductive traits (age at menarche, age at menopause, age at first birth, number of births, infertility, sex hormones) and menstrual conditions (dysmenorrhea, PCOS, endometriosis, menstrual cycle length) from GWAS data repositories, specifically the GWAS Catalog (ebi.ac.uk/gwas) and IEU Open GWAS (gwas.mrcieu.ac.uk/).
4) Perform bivariate LD score regression analysis to investigate genetic correlations between sleep and reproductive traits.
5) Extract genome-wide significant genetic variants related to those traits where there is evidence for genetic correlation.
6) Perform both one- and two-sample MR analyses to establish causal relationships.
7) Conduct sensitivity analyses to evaluate the robustness of findings.
8) Compare MR estimates with those obtained from multivariable regression.
1. Kennaway DJ, Boden MJ, Varcoe TJ. Circadian rhythms and fertility. Mol Cell Endocrinol. 2012;349(1):56-61.
2. Stocker LJ, Macklon NS, Cheong YC, Bewley SJ. Influence of shift work on early reproductive outcomes: a systematic review and meta-analysis. Obstet Gynecol. 2014;124(1):99-110.
3. Gamble KL, Resuehr D, Johnson CH. Shift work and circadian dysregulation of reproduction. Front Endocrinol (Lausanne). 2013;4:92.
4. Baker FC, Driver HS. Circadian rhythms, sleep, and the menstrual cycle. Sleep Med. 2007;8(6):613-622.
5. Goldstein CA, Smith YR. Sleep, circadian rhythms and fertility. 2016. 2016;2(4):206-217.
6. Baker FC, Lee KA. Menstrual Cycle Effects on Sleep. Sleep Med Clin. 2018;13(3):283-294.
7. Eichling PS, Sahni J. Menopause related sleep disorders. J Clin Sleep Med. 2005;1(3):291-300.
8. Zheng J, Erzurumluoglu AM, Elsworth BL, et al. LD Hub: a centralized database and web interface to perform LD score regression that maximizes the potential of summary level GWAS data for SNP heritability and genetic correlation analysis. Bioinformatics. 2017;33(2):272-279.
9. Davey Smith G, Ebrahim S. 'Mendelian randomization': can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol. 2003;32(1):1-22.
10. Davey Smith G, Hemani G. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet. 2014;23(R1):R89-98.
11. Davies NM, Holmes MV, Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. BMJ. 2018;362:k601.
Dr Anya Skatova (lead), Dr Andy Skinner , Prof Tom Gaunt, Prof Richard Martin
Shopping history records, collected via purchases tracked on loyalty cards, can provide a new perspective on lifestyle choices and behaviours and how these relate to health outcomes such as cancer. Shopping histories can provide information, which is otherwise difficult to measure such as granular, population level, objective data on lifestyle behaviours and risk factors (e.g., smoking, alcohol consumption) that can be tracked longitudinally. However, shopping history data also have inherent biases. For example, despite providing details on purchasing habits and basic individual characteristics, patterns in the data could be explained by other factors (e.g., the gap between purchase and consumption). Reliability of health information that is derived from shopping history data can be assessed through integrating these data with detailed self-reports of behaviour collected through Ecological Momentary Assessment (EMA). This work will improve detection of cancer risk as well assess validity of integrated data sources in risk prediction.
The overall aim of this PhD is to integrate supermarket loyalty cards data with EMA data and conventional epidemiological measures (eg questionnaires, biomarkers, etc) in Avon Longitudinal Study of Parents and Children (ALSPAC) to predict risk factors for cancer. The innovative aspect is the use of transactional and EMA data, which provide higher-density time-series data with different biases from conventional questionnaire/interview data. The ability to predict risk factors using such data could produce novel insights of early cancer symptoms and associated consumption patterns.
Identify patterns in standalone shopping history data that can be reflective of consumption association with known risks of cancer (Years 1 – 2). Collect EMA data on behaviours related to known risks of cancer using wearable technology (e.g. smartwatches) (Year 2). Use statistical methods (e.g., linear and logistic regression) to validate shopping histories patterns through EMA and conventional self-report/biomedical data in ALSPAC (Years 2-3). Use statistical and machine learning methods to predict cancer risk factors in the ALSPAC dataset in a sample of thousands of ALSPAC participants as well as standalone supermarket loyalty cards data in population-wide sample of millions of supermarket customers (Years 2-4)
.
Dr Anya Skatova (lead), Prof Deborah Lawlor ,
Shopping history records collected by supermarkets contain population level health information which could be missing from traditional health research data such as medical records. For example, shopping transactions can provide granular and objective data on under/unreported behaviours and outcomes in reproductive health domain – related to pain and weight management, vitamins consumption, infant feeding, etc - that can be tracked longitudinally. Combining shopping history datasets with epidemiological methods has potential for health research and might improve diagnosis, disease prevention and planning of interventions.
The aim of the PhD is to explore the potential of shopping history data to identify key reproductive events and lifestyle choices around these in real time. The specific focus of the PhD will be developed by the student, with potential objectives including: (1) determining the accuracy of shopping history to determine one or more reproductive events, such as conception, pregnancy, breastfeeding or parenthood; (2) whether shopping histories can identify lifestyle changes around these events, such as pre-conception, pregnancy and breastfeeding related changes in diet; (3) the extent to which shopping histories enhance repeat data collected in cohort studies, for example, shopping histories with data in real time might be able to pinpoint the timing of events such as planning a pregnancy and conception, whereas cohort data collected from movement sensors over periods that coincide with the timing of these events might better identify changes in physical activity and sleep patterns. The PhD will work with standalone population level supermarket shopping histories data, as well as a subset of shopping histories data linked into Avon Longitudinal Study of Parents and Children (ALSPAC).
The student will mainly work with shopping histories data of a large UK health and beauty retailer, both standalone (>12.5m customers, >1.5 billion transactions) and linked into ALSPAC (for ~1,500 index ALPSAC participants). There is a scope for additional new quantitative data collection with ALSPAC participants where it is needed to meet research aims of the PhD project.
Shopping histories data will be used first to identify a reproductive life event of interest (e.g., pregnancy) and a time window associated with it. Products that are bought during this time window will be then explored. This will allow to identify other behaviours (e.g., pain management, fertility issues) and health outcomes (e.g., miscarriages) associated with this life event. Those behaviours and outcomes will be then validated through the contextual variables using the data available in ALSPAC (and new data collected through surveys) related to causes and consequences of the life event. The student will be expected to explore the structure of the repeat shopping data and identify appropriate methods for analysing those data. For example, repeat purchasing of sanitary products (indicative of menstruation) which change over time might be analysed by multilevel models or structural equations depending on the structure of the data and the specific research question.
This is a data intensive quantitative PhD. The successful candidate would be expected to have had experience of statistical analysis in their first degree, be competent in handling complex large-scale data and eager to learn new quantitative methods and/or about new topic areas in a multidisciplinary team. Depending on their previous experience the successful candidate will obtain training in epidemiology, survey design and data collection, advanced statistical methods, and data science, including the ethics and governance and management and use of data, through the completion of the research project and through postgraduate short courses.
Dr Philip Haycock (lead), Dr Gibran Hemani , Prof Tom Gaunt
Summary data from genome-wide association studies (GWAS) are a valuable resource for many post-GWAS analytical tools, including Mendelian randomisation (MR), fine-mapping and linkage disequilibrium score regression. Access to GWAS summary data is increasingly supported by a number of online repositories, such as Open GWAS (https://gwas.mrcieu.ac.uk/), PhenoScanner (http://www.phenoscanner.medschl.cam.ac.uk/), the GWAS catalog https://www.ebi.ac.uk/gwas/ and GWAS central (https://www.gwascentral.org/). Online repositories may, however, be susceptible to meta-data errors, such as mis-specification of the effect allele or allele frequency columns, which has the potential to introduce substantial bias into downstream analyses. These errors occur because conventions for the inclusion or naming of data fields that avoid ambiguity have not been widely adopted by the GWAS community. Anecdotal evidence suggests that around 2% of datasets in Open GWAS may be affected by such errors. In this project, the student will conduct a systematic quantification of meta-data errors in online repositories of GWAS summary data. The work will be supported by the CheckSumStats R package (https://github.com/MRCIEU/CheckSumStats).
To quantify the extent of meta-data errors in online post-GWAS summary data platforms
Comparison of meta-data fields between GWAS studies and reference studies using the CheckSumStats R package. https://github.com/MRCIEU/CheckSumStats
Dr Richard Parker (lead), Prof Kate Tilling, Prof Laura Howe Dr Jon Heron
Whilst the association of mean blood pressure (BP) with cardiovascular disease (CVD) has been the subject of considerable investigation, interest is also growing in the role of within-individual variability in blood pressure (blood pressure variability; BPV), with evidence suggesting it may be an independent CVD risk factor above and beyond mean BP.(1)
A number of genetic variants (single nuclear polymorphisms; SNPs) associated with mean BP have been identified, and further analyses have in turn explored the association of these SNPs – together with those for body mass index (BMI) and height – with rates of change in BP (in addition to mean BP at baseline) across childhood and adolescence.(2)
Further broadening this work to additionally consider BPV would allow us to gain a better understanding of the association of these genetic variants with broader aspects of phenotypic BP. This could help provide further insight into CVD pathogenesis, as well as suggesting potential interventions to reduce parameters of mean BP and BPV implicated in CVD.
To investigate whether genetic variants associated with mean BP – together with SNPs for body mass index (BMI) and height – help explain differences in BPV, as well as differences in mean BP at baseline and rates of change in BP.
This project will provide experience in using ALSPAC data, handing genetic data and deriving polygenic scores, and statistical methods for repeated measures. We will support you to write the work up for publication if you wish to do so.
Multilevel growth curve models (which estimate population average and individual-specific trajectories)(2) will be further extended to allow for within-individual variance – typically assumed to be constant – to instead depend on genetic variants associated with mean BP, BMI and height (in mixed-effects location scale (MELS) models).(3) As such, the extent to which these SNPs predict BPV – above-and-beyond their association with mean BP and rates of change in BP – can be investigated in a model controlling for other predictors (e.g. age, sex, etc.) as appropriate.
Using data from the Avon Longitudinal Study of Parents and Children (ALSPAC) this will build on work characterising BPV across childhood and adolescence in this cohort.(4)
MELS models can be fitted using a number of different software packages, including Bayesian estimation via MCMC methods (using the R libraries brms or rstan, for example) or by maximum likelihood (using MIXREGLS, for instance). MELS models for repeatedly-measured BP, which can be adapted to include genetic variants, in ALSPAC have already been fitted, and code to do so in brms and rstan is available. (4)
1. Rothwell PM, et al. Lancet. 2010; 375: 895-905.
2. Howe LD, et al. Circulation-Cardiovascular Genetics. 2013; 6: 608-14.
3. Hedeker D, et al. Biometrics. 2008; 64: 627-34.
4. Parker RMA, et al. American Journal of Epidemiology. 2021; 190: 652-62.
Dr Laurence Howe (lead), Dr Neil Davies, Professor Kate Tilling Professor George Davey Smith
Adulthood height is a key indicator of population health because it is strongly affected by early-life environmental factors such as nutrition and childhood disease. The average height of individuals in European countries substantially increased during the 20th century illustrating the effects of environmental factors on adulthood height 1 2. The exact environmental mechanisms underlying this increase are not well understood but are thought to relate to the reduction in childhood diseases, better nutrition, and improvements in medicine. Particularly surprising is that the observed increases in average height persisted through two world wars and the great depression, periods marked by economic hardships and food shortages 2.
In this project, you will use data from UK Biobank to further explore the relationship between year of birth and adulthood height in the 20th century. You will also consider the components of height; leg length and trunk length (waist to crown) which may have distinct aetiologies and consequences 3. First, you will investigate whether the increases in height and its components varied by geographic region, socio-economic position (SEP) or by germline genetic factors. Second, you will use a regression discontinuity approach to evaluate if events relating to World War 2 (e.g., the introduction of food rationing) had a discernible effect on adulthood height. For example, comparing individuals born the year before the introduction of rationing to children born in the year after may inform the effects of maternal nutrition on their offspring’s growth.
1. Evaluate if the increases in height and its components were uniform across the population by determining if the relationship between birth year and height varies by:
a. Birth coordinates (North/South & East/West)
b. Socio-economic position (e.g., using Townsend Deprivation Index or family-size as proxies).
c. Polygenic scores (height, education).
2. Use regression discontinuity methods to estimate the effects of policy changes on adulthood height and components such as the introduction of food rationing in World War 2 and the removal after World War 2.
Linear regression models, regression discontinuity
1. Perkins JM, Subramanian SV, Davey Smith G, Özaltin E. Adult height, nutrition, and population health. Nutrition reviews 2016; 74: 149-65.
2. Hatton TJ. How have Europeans grown so tall? Oxford Economic Papers 2013; 66: 349-72.
3. Regnault N, Kleinman KP, Rifas-Shiman SL, Langenberg C, Lipshultz SE, Gillman MW. Components of height and blood pressure in childhood. Int J Epidemiol 2014; 43: 149-59.
Apostolos Gkatzionis (lead), Kate Tilling, Evie Stergiakouli Jessica Tyrell
Mendelian randomization (MR) is an increasingly popular technique to adjust for confounding bias in epidemiological studies. However, MR analyses can still suffer from selection bias, when the study sample is not representative of the study population, or when risk factor or outcome values are unobserved for some individuals (Hughes et al. 2017, Gkatzionis et al. 2019). Traditionally, approaches like inverse probability weighting can be used to adjust for selection bias. However, such approaches assume a “data missing at random” framework, where selection into the study can be fully modelled using observed data. This assumption will not always hold in practice. When it is violated, an alternative method is to use instrumental variables for selection into the study in order to adjust for selection bias. This was originally proposed for observational studies (Heckman 1979, Tchetgen Tchetgen and Wirth, 2017) and has recently been extended to Mendelian randomization by the supervisors of this project (paper forthcoming).
The aim of the project will be to apply selection bias adjustment methods in real-data analyses using data from the optional participation components of the UK Biobank. Optional participation components include the food frequency questionnaire, physical activity monitoring and the mental health questionnaire. Tyrell et al. (2021) showed that individuals who chose to participate in these components were a non-random subset of the UK Biobank sample and were likely to be taller and have better education, lower levels of adiposity and dyslipidaemia and lower prevalence of schizophrenia and Alzheimer’s disease. Building on the work of Tyrell et al. (2021), we will implement inverse probability weighting and the “instruments for selection” approaches to adjust for bias due to selective participation in applied MR analyses. Our plan is to use depression as an outcome (Davis et al. 2020) and one or more of the following as exposures: height, body mass index, years of education, intelligence, alcohol consumption. The applied questions can be tailored to suit the PhD student’s interests.
Mendelian randomization analysis of the effects of height, body mass index, education, intelligence and alcohol consumption on depression. The analysis will include the implementation of selection bias adjustment methods. Programming will be done in the R language, where the “instruments for selection” approaches have been implemented.
1) R. A. Hughes, N. M. Davies, G. Davey Smith, and K. Tilling (2019). Selection bias when estimating average treatment effects using one-sample instrumental variable analysis. Epidemiology, 30(3), 350-357.
2) A. Gkatzionis, and S. Burgess (2019). Contextualizing selection bias in Mendelian randomization: how bad is it likely to be? International Journal of Epidemiology, 48(3), 691-701.
3) J. J. Heckman (1976). The common structure of statistical models of truncation, sample selection and limited dependent variables and a simple estimator for such models. Annals of Economic and Social Measurement, 5(4), 475-492.
4) E. J. Tchetgen Tchetgen, and K. E. Wirth (2017). A general instrumental variable framework for regression analysis with outcome missing not at random. Biometrics, 73(4), 1123-1131.
5) J. Tyrrell, J. Zheng, R. Beaumont, K. Hinton, T. G. Richardson, A. R. Wood, G. Davey Smith, T. M. Frayling, and K. Tilling (2021). Genetic predictors of participation in optional components of UK Biobank. Nature Communications, 12, 886.
6) K. A. S. Davis, J. R. I. Coleman, M. Adams, N. Allen, G. Breen, B. Cullen et al. (2020). Mental health in UK Biobank - development, implementation and results from an online questionnaire completed by 157 366 participants: a reanalysis. BJPsych Open, 6(2), E18. doi:10.1192/bjo.2019.100.
Dr. Rachel Blakey (lead), Dr. Richard Parker, Professor Kate Tilling
Health traits such as blood pressure (BP), body mass index (BMI), and depressive thoughts form continuums across the population. Epidemiological research typically describes associations between risk factors and mean measurements of a disease or trait using regression models for the mean. However, at the extremes (e.g. 10th or 90th percentiles where we typically see clinical impairment) a trait may show a unique pattern of associations. Quantile regression extends traditional regression models to estimate effects on different quantiles across a distribution.
Recently, user-contributed packages in open-source statistical software programs, like R, have made quantile regression methods more widely available (packages ‘quantreg’, ‘lqmm’ and ‘brms’) (Bürkner; Geraci; Koenker et al.). Despite increased availability and many appropriate health questions, quantile regression has rarely been applied in mainstream epidemiological research. To improve understanding and uptake of these models, there is a need to provide a review of existing methods and clear worked examples of applied quantile regression.
The aim of this study is to help facilitate appropriate application of quantile regression in epidemiology by reviewing quantile regression methods, and generating a ‘how to guide’ with a worked example (BP/BMI/depression) using R.
Depending on the background/training needs of the student, either the applied example or the statistical methodology could form a greater part of the project. I.e. the aim could be to address a specific question (e.g. “are risk factors for increased BMI different at different parts of the distribution?”) or to review the methods and show an applied example.
The student will conduct literature searches to review existing literature and applications of quantile regression methods. The student will use ALSPAC data to generate worked examples of different quantile regression methods in R. An appropriate outcome variable – such as blood pressure, body mass index or depressive traits – can be chosen in collaboration with the student.
The student will gain experience of using ALSPAC data, coding skills in R, an in-depth understanding of quantile regression methods, and the support of their supervisors to write up their work for publication.
Bürkner, Paul-Christian. "Brms: An R Package for Bayesian Multilevel Models Using Stan." Journal of Statistical Software 080.i01 (2017).
Geraci, Marco. "Linear Quantile Mixed Models: The Lqmm Package for Laplace Quantile Regression." Journal of Statistical Software 57.1 (2014): 1-29.
Koenker, Roger, et al. "Package ‘Quantreg’." Cran R-project. org (2018).
Dr Amanda Hughes (lead), Professor Laura Howe, Dr Helen Bould
In high-income countries, tackling obesity is seen as a key priority by policymakers and health inequality researchers. As a result, underweight is largely ignored, despite clear links between low body weight and worse health and higher mortality. Underweight has diverse causes: it can result from mental illnesses including eating disorders, but it may also reflect food insecurity and socioeconomic deprivation. In the UK, changes to welfare policy during the 2010s were accompanied by increased foodbank use(1) and a starkly raised risk of underweight was in this period reported among British adult jobseekers(2). More recently, the COVID-19 pandemic brought profound economic and lifestyle changes. Alongside increased food insecurity in low-income families(3) and a record rise in foodbank use(4), there was a stark rise in referrals and hospitalizations for eating disorders(5). These changes mean that the prevalence, risk factors, and social patterning of underweight must now be reassessed.
This project aims to comprehensively investigate the prevalence, social distribution, and risk factors of low body weight in the contemporary UK. To shed light on potential causes, it will explore changes between the early 2000s and the early 2020s. It will consider several stages of the life-course, including adolescence, young adulthood, and middle-age. In each historical period, it will explore the contribution of sociodemographic risk factors, physical and mental health, and genetics.
Data will come from birth cohort studies: the 1946, 1958 and 1970 Birth Cohort Studies, Avon Longitudinal Study of Parents and Children (ALSPAC), and Millennium Cohort Study (MCS). All studies contain rich information including sociodemographic factors, objective height and weight measurements, self-perception of body weight, mental health, and physical health. Repeat measurements allow investigation of change within cohorts. Between-cohort change can be examined by comparing aged-matched participants across surveys. The contribution of mental illness will be examined using measures of common mental disorder (depression/anxiety symptoms) and, where available, of eating disorders. Where genetic information is available, genetic causal inference methods will be used to further unpick the contribution of health and health-related behaviours (e.g. smoking). To capture the effects of the COVID-19 pandemic, the project will use post-pandemic information from ALSPAC, MCS, and the 1958 cohort.
1. Reeves A, Loopstra R. J Soc Policy. 2020.
2. Hughes A, Kumari M. Prev Med. 2017.
3. Parnham JC et al. Public Health. 2020.
4. Baraniuk C. BMJ. 2020.
5. Solmi F et al. Lancet Child Adolesc Heal. 2021.
Josine Min (lead), Caroline Relton, Jonathan Mill (University of Exeter Medical School) Eilis Hannon (University of Exeter Medical School) The student will also collaborate with Dr. Gibran Hemani (University of Bristol), Prof. Tom Gaunt (University of Bristol), Prof. Bastiaan Heijmans (LUMC, The Netherlands) and Dr. Jordana Bell (KCL, UK).
DNA methylation is an epigenetic biomarker that has been shown to reflect lifestyle and biological factors (smoking, alcohol use, chronological age). To date the majority of studies used to link behavioral phenotypes such as cigarette smoking, and alcohol use to health outcomes typically employ self-reported questionnaire data. Multiple DNA methylation (DNAm) sites are strongly associated with (behavioural) traits. DNAm derived scores have been used to predict (or proxy for) these traits providing greater precision and biological proximity than self-reported measures. The DNAm derived smoking score is a widely used biomarker of lifetime exposure to tobacco smoke and may explain the molecular mechanism of the long-term risk of diseases following smoking cessation. There is growing interest in conducting genome-wide association studies (GWAS) and Mendelian randomization (MR) analysis on DNAm scores to identify novel genetic and causal factors influencing behavioural traits. To date, several GWAS on DNAm derived scores of aging have been published (Lu et al. 2018, McCartney et al. 2021). The many benefits to identify novel loci and biological pathways for other phenotypes have yet to be gained. This studentship will provide cross-disciplinary training in state-of-the-art genetic and genomic epidemiological approaches (under the supervision of Dr. Josine Min and Prof Caroline Relton at the Medical Research Council Integrative Epidemiology Unit at the University of Bristol and Prof Jonathan Mill and Dr. Eilis Hannon at the University of Exeter Medical School) to address questions about the molecular mechanism underlying established disease risk factors. The student will combine epigenetic, genetic and causal inference analyses in large-scale epidemiological datasets.
The overall aim of this PhD is to identify genetic variants and biological pathways associated with disease risk factors using DNAm scores. The specific risk factors/diseases for this project would depend on the candidate's research interests, but could include cell counts, smoking or alcohol use. The Genetics of DNA Methylation Consortium (GoDMC; http://www.godmc.org.uk/) has collected genetic and DNAm data across multiple cohorts offering the student an excellent platform for these analyses.
Methods:
1) Novel methodology can be used (and potentially developed) to construct DNAm scores on disease risk factors
2) GWAS on DNAm derived phenotype datasets will be conducted followed by meta-analyses. There will be several challenges with this type of analysis including heterogeneity of datasets in age, sex and tissue type.
3) To understand what aspect of the phenotype is captured by the DNAm phenotype, GWAS meta-analysis results will be compared to GWA results of detailed (self-reported) phenotypes (eg. in UK Biobank) and methylation quantitative loci from blood and brain.
4) MR analysis will be used to investigate causal relationships between DNAm derived measures and self-reported measures and other diseases/risk factors.
5) The heritability component of DNAm derived phenotypes will be estimated.
Lu AT, Xue L, Salfati EL, et al. GWAS of epigenetic aging rates in blood reveals a critical role for TERT. Nat Commun. 2018;9(1):387. Published 2018 Jan 26. doi:10.1038/s41467-017-02697-5
McCartney DL, Min JL, Richmond RC, et al. Genome-wide association studies identify 137 genetic loci for DNA methylation biomarkers of aging. Genome Biol. 2021;22(1):194. Published 2021 Jun 29. doi:10.1186/s13059-021-02398-9
Dr. Josine Min (lead), Prof Caroline Relton, Prof Jonathan Mill (University of Exeter Medical School) Dr. Eilis Hannon (University of Exeter Medical School) The student will also collaborate with Dr. Gibran Hemani (University of Bristol), Prof. Tom Gaunt (University of Bristol), Prof. Bastiaan Heijmans (LUMC, The Netherlands) and Dr. Jordana Bell (KCL, UK).
GWAS studies have discovered many genetic associations for traits and diseases, but it has been difficult to elucidate the functional consequences of these variants. One of the reasons for this is linkage disequilibrium (co-inheritance of many variants with the disease variant). Another reason is that most GWAS signals reside in non-coding regions (outside genes), and it is likely that GWAS variants confer their effects through modulating of regulatory mechanisms. Molecular traits (such as gene expression, DNA methylation, proteins) have increasingly been used to address this question because they are often dysregulated in disease and can act as an intermediate phenotype.
DNA methylation levels can be routinely assayed at a large scale using micro-arrays. Multiple studies have identified genetic variants associated with DNA methylation (mQTL: methylation quantitative trait locus) by combining genome wide genotype information with DNA methylation levels. mQTLs can be defined as cis, which are typically nearby a DNA methylation site and can be found in small sample sizes. Trans mQTLs (defined as associations which are further away or on different chromosomes) have smaller effects and large sample sizes are needed. The Genetics of DNA methylation Consortium (GoDMC, http://www.godmc.org.uk/) brought together a large number of cohorts to identify these small effects in blood and investigated whether the mQTLs play a role in disease etiology. Using several causal inference approaches, they discovered a small number of causal relationships between trait and DNA methylation (Min JL et al.).
DNA methylation might be on the causal pathway only in the trait-relevant cell-type or context. However, it is unknown to what extent cell-type specific DNA methylation signals differ from bulk tissue (eg. blood has many cell-types and each cell-type might have a different methylation level), and whether cell-type specific DNAm signals are on a causal path to disease as genome-wide mQTL resources across multiple ancestries, tissues and cell types are not currently available. This studentship will provide cross-disciplinary training in state-of-the-art genetic and genomic epidemiological approaches (under the supervision of Dr. Josine Min and Prof Caroline Relton at the Medical Research Council Integrative Epidemiology Unit at the University of Bristol and Prof Jonathan Mill and Dr. Eilis Hannon at the University of Exeter Medical School) to address questions about the human genetic basis of DNA methylation variation. The student will combine epigenetic, genetic, cellular deconvolution and causal inference analyses in large-scale epidemiological datasets.
The overall aim of this PhD is to compile a mQTL catalogue of uniformly processed mQTLs across a wide range of tissues, cell types, ancestries and across the lifecourse. The use of the catalog for this project would depend on the candidate’s research interest. It could be used for Mendelian randomization analyses to gain insights into the regulation of GWAS signals. GoDMC has collected genetic and DNA methylation data across multiple cohorts offering the student an excellent platform for these analyses.
1) mQTL analysis will be conducted followed by meta-analyses. There will be several challenges with this type of analysis including heterogeneity of datasets (eg. age, sex).
2) Novel methodology can be used (and potentially developed) to combine mQTLs across different studies, ethnicities, tissues and timepoints.
3) Cellular deconvolution approaches and analysis of cell-types from bulk tissue will be used to understand whether mQTLs operate in a cell-type specific manner.
4) Mendelian Randomization analysis will be used to investigate causal relationships between DNA methylation and diseases/risk factors.
Min JL et al. https://www.medrxiv.org/content/10.1101/2020.09.01.20180406v1
Dr Richard Parker (lead), Dr April Hartley, Prof Kate Tilling
With evidence indicating that within-individual blood pressure (blood pressure variability; BPV) may be an independent risk factor for cardiovascular disease (CVD) above and beyond mean BP (1), a better understanding of the associations of genetic markers with BPV could provide further insight into the development of CVD. However, whilst there has been considerable work identifying the genetic markers associated with mean blood pressure (2), little is known about the associations of genetic factors with BPV.
To identify the genetic factors associated with BPV, by conducting a genome-wide association study (GWAS) with ALSPAC data on genetic variants and estimates of within-individual variability in repeatedly-measured BP.
This project will provide experience in using ALSPAC data, handling genetic data using bluecrystal, conducting a GWAS, and statistical methods for repeated measures. We will support you to write the work up for publication if you wish to do so.
Estimates of within-individual variability in BP will be derived from mixed-effects location scale (MELS) models conducted with ALSPAC data. These are multilevel models for repeatedly-measured, longitudinal data (which estimate population average and individual-specific trajectories) further extended to allow for within-individual variance – typically assumed to be constant – to instead differ between people, and to depend on other variables (e.g. age) as appropriate (3).
The associations of these model-derived estimates of BPV with genetic variants will then be explored in a GWAS.
MELS models for repeatedly-measured BP in ALSPAC have already been fitted, and code to do so in is available from the supervisors (4). GWAS of BPV can then be performed using a linear model within software such as Plink or SNPtest, with adjustment for principal components to account for potential population stratification (scripts to run these analyses will be provided).
1. Rothwell et al (2010) https://doi.org/10.1016/S0140-6736(10)60308-X
2. Evengelou et al (2018) https://doi.org/10.1038/s41588-018-0205-x
3. Hedeker et al (2008) https://doi.org/10.1111/j.1541-0420.2007.00924.x
4. Parker et al (2021) https://doi.org/10.1093/aje/kwaa224
Dr Richard Parker (lead), Dr Gareth Griffith, Dr Jon Heron, Dr Alex Kwong, Prof Kate Tilling
Depression during adolescence is associated with poorer health and socioeconomic outcomes in later life, including recurrence of depression (1). Whilst there has understandably been considerable focus on the factors associated with the mean levels of depressive symptoms, such as negative emotions, during adolescence, evidence suggests higher within-individual variability in such symptoms may be associated with poorer outcomes too (2,3). Therefore, a better understanding of the factors associated with within-individual variability in depressive symptoms from childhood through to early adulthood could provide further insight into the development of depression, broadening our understanding of the pattern of depressive symptoms during this period beyond a characterisation of their mean level.
To investigate the factors associated with within-individual variability in depressive symptoms across childhood and early adulthood, investigating data from the Avon Longitudinal Study of Parents and Children (ALSPAC).
This project will provide experience in using ALSPAC data, statistical methods for repeated measures, and handling genetic data and deriving polygenic scores. We will support you to write the work up for publication if you wish to do so.
Using data from ALSPAC, this project will build on work characterising trajectories of depressive symptoms, as measured via the Short Mood and Feelings Questionnaire (SMFQ) (4), across childhood and adolescence in this cohort (5). Multilevel growth curve models (which estimate population average and individual-specific trajectories) have been extended to allow for within-individual variance – typically assumed to be constant – to differ between people and across other measured variables, in a mixed-effects location scale (MELS) model (6). These models will be applied to these data to investigate whether factors of interest – such as parental depression, trauma and polygenic risk scores for neuroticism, for example – are associated with within-individual variability in depressive symptoms, and to explore how different people are with regard to any such within-individual variability. The supervisors have plenty of experience in these methods and will support the student in fitting these, including providing examples of code (6).
1. https://doi.org/10.1016/S0140-6736(11)60871-4
2. https://doi.org/10.1046/j.1467-8624.2003.00643.x
3. https://doi.org/10.1016/j.jecp.2010.10.007
4. https://doi.org/10.12688/wellcomeopenres.15395.2
5. https://doi.org/10.1007/s10964-018-0976-5
6. https://doi.org/10.1093/aje/kwaa224
Dr Denize Atan (lead), Dr Theresa Redaniel, Dr Tim Jones, Senior Research Associate & former cancer analyst at Public Health England, Applied Research Collaboration-West Dr Beth Stuart, Medical Statistician, University of Southampton Dr Samiel Merriel, GP with expertise in early cancer diagnosis and prevention, University of Exeter
Brain tumours affect 8 per 100,000 people in the UK each year. Brain tumours often affect vision before causing any other symptoms. Unless diagnosed early, many people with brain tumours will die or suffer long-term disabilities, like permanent sight loss.
In 2016, an 8-year-old boy called Vincent Barker was in the news. During a routine sight test, his optometrist, Honey Rose, failed to detect optic nerve swelling at the back of his eyes - a sign indicating raised intracranial pressure. He died soon afterwards. As a result, Honey Rose was convicted for gross negligence manslaughter.
Since the widespread media coverage of the Rose/Barker case, optometrists have been referring more people to hospital over concerns they might have optic nerve swelling. Because of this, we think more patients with brain tumours are diagnosed earlier and more frequently by eye specialists than 5 years ago.
Our primary aims are to find out the number of people diagnosed with brain tumours every year between 2013 to 2018 in England and the proportion who were diagnosed by eye specialists before and after the Rose/Barker case in 2016.
Our secondary aims are to determine whether patients with brain tumours were diagnosed earlier by hospital eye specialists compared with other routes-to-diagnosis and whether they lived longer and had better treatment outcomes as a result.
Public Health England and the National Cancer Registry and Analysis Service (NCRAS) routinely collect data on everyone diagnosed with benign and malignant brain tumours in England. We have National Research Ethics Committee approval to access NCRAS data linked to Hospital Episode Statistics; and we have obtained the data on all new cases of benign and malignant brain tumours diagnosed between 2013 and 2018.
Trends in the data will be investigated in the 3 years before and after exposure to the widespread media coverage Rose/Barker case in 2016 by generalised linear regression techniques. We will determine the change in:
(i) Adjusted odds ratios for the number of brain tumours diagnosed via hospital eye services
(i) Time to diagnosis
(ii) WHO tumour grade at diagnosis
(iii) Cancer stage at diagnosis
(iv) Time between diagnosis and treatment
(v) Mortality
Age, sex, ethnicity, geographical location, deprivation index, and smoking history will be used as covariates in these analyses.
1. Poostchi A, et al. Spike in neuroimaging requests following the conviction of the optometrist Honey Rose. Eye 2018.
2. Elliss-Brookes L, et al. Routes to diagnosis for cancer. B J Cancer 2012.
3. Koo MM, et al. Presenting symptoms of cancer and stage at diagnosis. Lancet Oncol 2020.
Dr Denize Atan (lead), Dr Theresa Redaniel, Dr Tim Jones, Senior Research Associate & former cancer analyst at Public Health England, Applied Research Collaboration-West Dr Beth Stuart, Medical Statistician, University of Southampton Dr Samiel Merriel, GP with expertise in early cancer diagnosis and prevention, University of Exeter
Brain tumours affect 8 per 100,000 people in the UK each year. Brain tumours often affect vision before causing any other symptoms. Unless diagnosed early, many people with brain tumours will die or suffer long-term disabilities, like permanent sight loss.
In 2016, an 8-year-old boy called Vincent Barker was in the news. During a routine sight test, his optometrist, Honey Rose, failed to detect optic nerve swelling at the back of his eyes - a sign indicating raised intracranial pressure. He died soon afterwards. As a result, Honey Rose was convicted for gross negligence manslaughter.
Since the widespread media coverage of the Rose/Barker case, optometrists have been referring more people to hospital over concerns they might have optic nerve swelling. Because of this, we think more patients with brain tumours are diagnosed earlier and more frequently by eye specialists than 5 years ago.
Our primary aim is to find out the number of people diagnosed with brain tumours every year between 2013 to 2018 in England and the proportion who were diagnosed by eye specialists before and after the Rose/Barker case in 2016.
Public Health England and the National Cancer Registry and Analysis Service (NCRAS) routinely collect data on everyone diagnosed with benign and malignant brain tumours in England. We have National Research Ethics Committee approval to access NCRAS data linked to Hospital Episode Statistics; and we have obtained the data on all new cases of benign and malignant brain tumours diagnosed between 2013 and 2018.
Trends in the data will be investigated in the 3 years before and after exposure to the widespread media coverage Rose/Barker case in 2016 by generalised linear regression techniques. We will then determine the change in adjusted odds ratios for the number of brain tumours diagnosed via hospital eye services before and after the Rose/Barker case.
1. Poostchi A, et al. Spike in neuroimaging requests following the conviction of the optometrist Honey Rose. Eye 2018.
2. Elliss-Brookes L, et al. Routes to diagnosis for cancer. B J Cancer 2012.
3. Koo MM, et al. Presenting symptoms of cancer and stage at diagnosis. Lancet Oncol 2020.
Dr Philip Haycock (lead), Prof Richard Martin,
Telomeres are DNA-protein structures at the end of chromosomes that protect the genome from damage, shorten progressively over time in most somatic tissues, and are proposed physiological markers of aging. In 2017 the Telomeres Mendelian Randomization Collaboration (TMRC) established that telomere length increases risk of several site specific cancers but reduces risk for some non-neoplastic diseases, including cardiovascular diseases. The analysis was based on a genetic instrument with 12 single nucleotide polymorphisms that together explained 2% to 3% of the variance in circulating telomere length. A recently published genome-wide association study in 472,174 participants in the UK Biobank has increased the number of independent GWAS hits for telomere length to 197 explaining 4.54% of the variance. The primary aims of this project are to: 1) update the genetic instrument for telomere length to include the newly discovered hits; 2) to conduct MR analyses of the association of telomere length with risk of cancer and non-neoplastic diseases; 3) conduct sensitivity analyses for violations of instrumental variable assumptions. Time permitting, the student might additionally: 4) assess the shape of the association between genetically instrumented telomere length and risk of neoplastic and non-neoplastic diseases; and 5) compare MR findings to observational analyses of directly measured telomere length and disease risk in UK Biobank.
The primary aims are: 1) update a genetic instrument for telomere length; 2) conduct MR analyses of the association of telomere length with risk of cancer and non-neoplastic diseases; 3) conduct sensitivity analyses for violations of instrumental variable assumptions; 4) compare MR findings to observational analyses of directly measured telomere length and disease risk in UK Biobank; 5) assess the shape of the association between genetically instrumented telomere length and risk of neoplastic and non-neoplastic diseases
Two-sample Mendelian randomization, using 197 genetic polymorphisms to instrument telomere length and summary data from genome-wide association studies of cancer and non-neoplastic diseases. Primary analyses will be based on random effects inverse variance weighted linear regression. Sensitivity analyses will include MR-Egger regression, the weighted mode estimators and weighted median estimator.
https://www.medrxiv.org/content/10.1101/2021.03.23.21253516v1
https://jamanetwork.com/journals/jamaoncology/fullarticle/2604820
Dr Duleeka Knipe (lead), Prof Laura Howe, Dr Maria Theresa Redaniel Dr Nanette Mayol Lee
The burden of mental disorders and self-harm are most acutely felt in the poorer nations of the world. Low- and middle-income countries (LMIC) account for 80% of disability-adjusted life years (DALYs -a measure of overall disease burden) to depressive disorders and self-harm (70% of all self-harm occur in LMIC, 80% of suicide deaths).
Despite this our understanding of what contributes to the development of ill mental health is largely unknown in these emerging economies and is limited by the current evidence base. Firstly, much of the existing evidence uses retrospective study designs, which make explorations of causality difficult. Secondly, there is a growing body of evidence from high-income countries that poor mental health such as depression develops as a consequence of factors that work across the life course – this is less well explored in LMIC. Lastly, many of the investigations exploring risk factors for depression and suicidal behaviour in LMIC fail to appropriately account for the different experiences (e.g. type and frequency of adverse childhood experiences) and environmental contexts. The relative importance of certain influences at various developmental stages may differ. Without better quality evidence, progress in reducing the burden of mental ill health and suicide in LMIC will continue to be hindered. An improved understanding of contributing factors to the development of depression and suicidal thoughts is urgently needed to allow for contextually relevant interventions to be developed.
This PhD will address these limitations by utilising an ongoing large cohort dataset (Cebu Longitudinal Health and Nutrition Survey - CLHNS) established in the Philippines. It is a follow-up study of 3327 Filipino women who gave birth in the early 1980s and their 3080 offspring. The offspring of the index children are also now being followed up. This rich data source includes data on physical health (e.g. stunting, morbidity),social (e.g. household composition, educational outcomes), economic (e.g. financial resources), and environmental (e.g. neighbourhood assessment) factors. There are also measures of childhood adversity (e.g. parental domestic violence, child labour). The inclusion of measures over time will allow for the exploration of potential mechanisms by which certain factors might contribute to later mental health outcomes at different developmental stages. Importantly this is one of the only established birth cohort studies which have tracked individuals through their childhood that has data on depression symptoms and suicidal thoughts in adulthood (ages 18, 21, and 35) from a LMIC.
This project will investigate the influence of factors across development on adulthood depression and suicidal thoughts in a LMIC. It will aim to identify potential high-risk groups (including modifiable risk-factors) for the development of early interventions, and modifiable mediators when the exposure is less amenable for intervention (e.g. parental death).
The study will address the following questions:
What factors across development (prenatal to young-adulthood) contribute to depression and suicidal thoughts in adulthood?
What are the mechanisms through which risk-factors impact on mental health?
Are there specific time-points at which the exposure to certain risk factors contributes to a worsening of depression and/or suicidal thoughts in adulthood?
This PhD spans the disciplines of psychiatry, psychology, and epidemiology. It provides a unique opportunity for the student to develop skills across a range of sophisticated epidemiological and statistical techniques, including mediation analysis, strategies for dealing with missing data (such as multiple imputation) and causal inference. Given the paucity of evidence and the high-quality nature of the dataset, this project has the potential to significantly contribute to the knowledge base and identify contextually relevant targets for intervention.
The candidate will learn advanced epidemiological (e.g. causal inference) and statistical (e.g. multilevel modelling) skills, as well as gaining expertise in a range of statistical software packages.
Professor Laura Howe (lead), Dr Amanda Hughes, Dr Matt Dickson, University of Bath, Institute for Policy Research Professor Frances Rice, Cardiff University, Division of Psychological Medicine and Clinical Neurosciences
The transition to higher education (HE) is a key point in a person’s life course; decisions at this time can have lifelong influences. The experience of a mental health or neurodevelopmental condition could lead some people to make decisions that undermine their academic potential, e.g. choose not to participate in HE despite receiving sufficient qualifications, choose a less prestigious university than their grades would make them eligible for, or choose to attend a local university and live at home. These differences may be exacerbated by parental mental health or by genetic factors. As such, own and parental experiences of mental health and neurodevelopmental conditions may entrench intergenerational patterns of socioeconomic (dis)advantage.
1. Assess the influence of life course trajectories of depressive symptoms and neurodevelopmental conditions with decisions about higher education.
2. Assess the influence of parental mental health with decisions about higher education, and the degree to which these are mediated by own mental health.
3. Quantify the role of mental health and neurodevelopmental conditions in the intergenerational transmission of socioeconomic (dis)advantage.
4. Evaluate the influence of genetic risk scores for educational attainment and ADHD on decisions about higher education.
The project involves statistical analysis of data from the Avon Longitudinal Study of Parents and Children (ALSPAC), including:
- Measures of depression, anxiety, ADHD and ASD, repeatedly across childhood through to adulthood
- Measures of parental depression and anxiety
- A detailed questionnaire on reasons for (not) participating in HE, reasons for choices of university and course, and details of university attended
- Linked data on educational test performance across childhood and adolescence
- Genetics
Methods will include regression modelling, techniques for modelling longitudinal data, mediation analysis, and analysis of genetic data. Where possible, we will draw on additional studies for genetic analyses to boost statistical power.
Reflecting the interdisciplinary nature of the project, this PhD involves collaborations with the Institute for Policy Research at the University of Bath and the Division of Psychological Medicine and Clinical Neurosciences at Cardiff University. The student will be based in the MRC Integrative Epidemiology Unit and the Department of Population Health Sciences at the University of Bristol, but will have the opportunity to spend time in all centres.
López-López JA et al. BJPsych Open. 2019:6(1):e6
https://www.medrxiv.org/content/10.1101/2020.05.21.20108928v2
Dardani et al. Int J Epidemiol. 2021 dyab107
Hughes A et al. NPJ Sci Learn. 2021;6(1):1.
Belfields C, et al. The impact of undergraduate degrees on early-career earnings. IFS 2020
Sullivan A, et al. Br J Sociol. 2018 Sep 1;69(3):776–98.
Dr Gemma Sharp (lead), Dr Dan Bernie, Dr Chin Yang Shapland Prof Kate Tilling
Climate change will affect human health through changes to heat stress, sanitation, access to sufficient food and safe drinking water, disease patterns, migration, and frequency of extreme weather events.
Pregnant women and the developing fetus are considered amongst the most vulnerable and marginalised members of society, and could therefore be uniquely sensitive to the effects of climate change. Previous research has shown that babies of mothers exposed to higher ambient temperatures during pregnancy are at higher risk of preterm birth and low birthweight. These perinatal outcomes are associated with greater risk of neonatal ICU admission and infant mortality, as well as long term outcomes such as neurodevelopmental delays and cardiovascular disease.
Research is required to characterise the potential effects of climate change on these outcomes, both now and in the future, and to inform strategies to reduce or manage an increase in their prevalence in line with climate change.
This exciting project aims to explore how epidemiological data and approaches can be best applied to study the effects of climate change. The project focuses on how exposure to specific aspects of climate during pregnancy relate to perinatal health outcomes: gestational age at birth and birth weight.
We will link health and demographic data from sources including UK Biobank, ALSPAC and Born in Bradford to detailed historical climate data from the UK Met Office including 1km resolution data from HadUK-Grid data, supplemented with complimentary observational station data and climate reanalysis products as necessary. We will then explore associations between climate exposures (as measured by ambient temperature, precipitation, barometric pressure, hours of sunlight, or compound measure of environmental stress) during pregnancy in relation to birth weight and gestational age at delivery. Severity and duration of exposure will be sampled on daily to monthly time scales. Results will be then used to define parameters in projection models of the impact of climate change on these outcomes going forward under different scenarios. The application to future climate projections will link to the UK Climate Projections (UKCP18) which for the first time provide national scale climate projections at a resolution comparable to weather forecasts.
In addition to working with a team of highly experienced experts in epidemiology and population health, the student will work with climate experts, including Dr Dan Bernie, who has a joint appointment with the Met office. There will also be opportunities for training and gaining experience in public and policy engagement to help translate findings and create impact.
Climate change and the potential effects on maternal and pregnancy outcomes: an assessment of the most vulnerable – the mother, fetus, and newborn child. Rylander et al. 2013 Glob Health Action https://www.ncbi.nlm.nih.gov/pubmed/23481091
Exploring associations of maternal exposure to ambient temperature with duration of gestation and birth weight: a prospective study. Li et al 2018 BMC Pregnancy Childbirth https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6311008/
Systematic review on adverse birth outcomes of climate change. Poursafa et al 2015 J Res Med Sci https://www.ncbi.nlm.nih.gov/pubmed/26109998
Dr Gemma Sharp (lead), Prof Sarah Lewis, Dr Evie Stergiakouli
Worldwide, roughly one in every 750 babies is born with a cleft of the lip and/or palate. These children face multiple issues with feeding, teeth, speech and hearing, which can last throughout their lives, even after surgery to repair the cleft(s).
There is a high degree of heterogeneity in terms of severity of outcomes. Some individuals have very few issues and require minimal clinical care or adjustment, whereas for others the impact is huge and necessitates lifelong care from multiple disciplines. This presents a challenge to health services looking to allocate finite resources, as a one-size-fits-all approach is unlikely to be economically efficient and could take resources away from those most in need. A better ability to predict which children born with a cleft are at risk of poor outcomes would help clinicians tailor care. This could ultimately result in better cleft care outcomes.
Outcomes like speech, neurodevelopment, mental wellbeing and educational attainment are likely to be influenced by a combination genetics and environment. Epigenetics, which will be influenced by genetics and the environment, could also play a role. Therefore, integrating genetic and epigenetic data with non-molecular data about a child’s cleft phenotype, family history and home environment could help predict risk of outcomes.
This project will use genetic, epigenetic, questionnaire and clinical data to develop and test new risk scores to predict health and social outcomes in children born with a cleft lip/palate. Working closely with the cleft patient and clinical community, the student’s findings will help inform development of clinically useful risk factor profiles to help tailor cleft care and ultimately improve outcomes.
In this project, the student will use information from questionnaires and medical records to identify which clinical (e.g. cleft type, feeding and speech problems) and social (e.g parent’s educational level, parent’s mental health) factors predict worse cleft-related outcomes.
They will use published genetic and epigenetic studies to identify SNPs and CpGs associated with various child health and social outcomes (e.g. obesity, ADHD, social communication, speech traits, depression, etc). They will then generate several polygenic risk scores (PRS) and DNA methylation-based risk scores (MRS) for these outcomes in data from the Cleft Collective cohort studies, which is a detailed longitudinal study of ~3000 children born with a cleft in the UK.
The student will use regression analyses and ROC-curves to assess the ability of these PRSs and MRSs to predict relevant outcomes in children born with a cleft. The student will also examine the predictive ability of non-molecular factors, such as cleft type, socioeconomic position and infant feeding, and compare predictive performance of various combinations of environmental factors, PRS and MRS in predicting different outcomes.
Finally, the student will select the highest performing combination of predictive factors and develop risk scores for each outcome. They will then test the ability of these risk scores to predict outcomes in (an) independent dataset(s) of children born with a cleft. Such datasets could include a subset of the Cleft Collective retained for testing purposes (I.e. not included in the development of the risk scores), or other cohorts of children born with a cleft with information on outcomes and genetics and/or epigenetics from our collaborators in Norway, Germany and/or the USA.
Dr Robert Thibault (lead), Prof Marcus Munafò,
BACKGROUND: Statistics training in many undergraduate courses, particularly in basic science disciplines, rarely extends beyond t-tests, ANOVAs, and basic linear regression. Without an introduction to core statistical concepts (e.g., central limit theorem, randomness) and additional, broader concepts (e.g., confidence intervals, coding), students may begin to perform ‘mindless statistics’ (Gigerenzer 2004), conducting statistical tests without understanding why they are doing so, or what the tests can (and cannot) tell us.
PROBLEM: This poor foundational training may have broader consequences—statistical misconceptions appear to extend to postgraduate researchers and even faculty (Gigerenzer, 2004; Hoekstra, Morey, Rouder, & Wagenmakers, 2014; Tversky & Kahneman, 1971). Better designed undergraduate statistics training may improve statistical knowledge and understanding, and in turn improve the quality of research outputs.
THIS PROJECT: This mini-project is one part of a larger research programme that aims to improve statistical training in the basic sciences. A previous mini-project documented and synthesised the current state of statistical training in psychology undergraduate programmes in the UK (TARG Meta-Research Group, 2020). For this project, you will develop statistical training standards in psychology by coordinating a Delphi process with various stakeholders.
The aim of this mini-project is to develop statistical training standards. The specific objectives include:
1. Assemble a panel of relevant stakeholders (e.g., instructors, researchers, learned societies, statisticians, students).
2. Conduct a Delphi process with these stakeholders.
3. Write a report that outlines the conclusions of the Delphi process.
You will lead a modified Delphi process, which is a structured method to elicit the opinions of various stakeholders and synthesize them into a meaningful conclusion.
BENEFITS FOR YOU: This project will help you gain an appreciation for the importance of good statistical practice in scientific research, including epidemiology. You will connect with researchers, instructors, and statisticians at various universities. You will learn about the field of meta-research and help prepare an article for publication which you will be an author on.
WHO TO CONTACT: If you are interested, please get in touch with me at robert.thibault@bristol.ac.uk – I can provide more details and perhaps tailor the project to your interests.
Gigerenzer, G. (2004). Mindless statistics. Journal of Socio-Economics. https://doi.org/10.1016/j.socec.2004.09.033
Hoekstra, R., Morey, R. D., Rouder, J. N., & Wagenmakers, E. J. (2014). Robust misinterpretation of confidence intervals. Psychonomic Bulletin and Review, 21(5), 1157–1164. https://doi.org/10.3758/s13423-013-0572-3
Tversky, A., & Kahneman, D. (1971). Belief in the law of small numbers. Psychological Bulletin, 76(2), 105–110. https://doi.org/10.1037/h0031322
TARG Meta-Research Group. (2020). Statistics education in undergraduate psychology: A survey of UK course content.
Dr Evie Stergiakouli (lead), Prof Laura Howe, Christina Dardani Dr Rachel Blakey
Attention Deficit Hyperactivity Disorder (ADHD) is a chronic neurodevelopmental condition, characterised by persistent difficulties in the areas of attention span/impulse control. It typically first manifests early in childhood and often persists into adulthood and has been associated with impaired education attainment and social disadvantages.
We have previously shown that ADHD genetic risks are associated with younger maternal age at birth, lower educational attainment and other indicators of social disadvantage in mothers from the general population (1). Using Mendelian randomization (MR) we have also found evidence of causal effects of genetic liability to ADHD on educational attainment, and evidence of effects of genetic liability to higher educational attainment on risk of ADHD which was independent of cognitive ability (2). Since ADHD manifests at a very young age, the causal effects of genetic liability to education on ADHD are likely to indicate parental effects. However, disentangling the individual effects of each factor as well as assessing for genetic confounding is required.
In this project, we will explore the links between educational attainment, reproductive outcomes (age at first birth, age at first sexual intercourse, number of live births), socioeconomic status and other indicators of social disadvantage on ADHD.
A. We will use two-sample Mendelian randomization (3) to investigate any causal links between genetic liability to educational attainment, reproductive outcomes (age at first birth, age at first sexual intercourse, number of live births), socioeconomic status and ADHD bidirectionally
B. We will perform Multivariable Mendelian randomization (4) to account for the exposures simultaneously
C. Apply sensitivity analyses including weighted median, weighted mode, MR-Egger regression, MR-PRESSO and colocalization analyses to assess and adjust for pleiotropy
1. Leppert et al. Association of Maternal Neurodevelopmental Risk Alleles With Early-Life Exposures. JAMA Psychiatry. 2019;76(8):834–842. doi:10.1001/jamapsychiatry.2019.0774
2. Dardani et al. Is genetic liability to ADHD and ASD causally linked to educational attainment?, International Journal of Epidemiology, 2021;, dyab107, https://doi.org/10.1093/ije/dyab107
3. Davey Smith, G. & Hemani, G. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum. Mol. Genet. 23, R89–R98 (2014)
4. Sanderson, E., Davey, G. S., Windmeijer, F. & Bowden, J. An examination of multivariable Mendelian randomization in the single-sample and two-sample summary data settings. Int. J. Epidemiol. (2018)
Dr Tim Morris (lead), Prof Kate Tilling, Prof George Davey Smith
Participants of research studies are often not fully representative of the general population at baseline. Furthermore, attrition from research studies is often patterned by a range of demographic, health and socioeconomic factors, making these studies less representative with greater elapsed time. This non-random participation has implications for the validity of results obtained from datasets [1,2], particularly where there is interaction in the selection mechanisms [3]. Non-random participation and attrition can be seen in the genomes of study participants; participation has been shown to be related to genetic risk for traits including higher BMI, neuroticism, schizophrenia, attention-deficit hyperactivity disorder (ADHD) and depression [4]. A clearer understanding of the genotypic and phenotypic patterns of study participation and attrition could help to improve interpretation of future study findings.
1. To investigate genetic signal in non-response in cohort studies such as the Avon Longitudinal Study of Parents and Children (ALSPAC), the Millennium Cohort Study (MCS), and Understanding Society (USoc).
2. To investigate the heterogeneity of genetic signal for participation across cohorts.
3. To investigate whether genetic signal for participation varies over time.
This project will provide experience in using ALSPAC data, handling genetic data, and statistical methods for analysing genetic data. We will support you to write the work up for publication if you wish to do so.
This mini project may make use of several methods outlined below; the student can initially choose which of these areas they would like to focus on. In each case, the student will be required to create indicators of non-response across all questionnaire and direct assessment timepoints for study participants. 1) Applying Genome-wide Complex Trait Analysis (GCTA-GREML) to measures of participation in cohorts to estimate the SNP heritability of participation. These results may then be meta-analysed across cohorts. 2) Conducting a GWAS of age and sex in cohorts as a negative control outcome analysis to identify selection bias. This may be supplemented by investigating associations between polygenic scores for traits with age and sex. 3) Investigating how polygenic scores for a range of traits (e.g. BMI, schizophrenia, educational attainment) associate with participation and attrition in cohorts. The results of these analyses may be meta-analysed with the above cohorts and others that the IEU has access to.
1. Griffith, G.J., Morris, T.T., Tudball, M.J. et al. Collider bias undermines our understanding of COVID-19 disease risk and severity. Nat Commun 11, 5749 (2020). https://doi.org/10.1038/s41467-020-19478-2.
2. Marcus R Munafò, Kate Tilling, Amy E Taylor, David M Evans, George Davey Smith, Collider scope: when selection bias can substantially influence observed associations, International Journal of Epidemiology, Volume 47, Issue 1, February 2018, Pages 226–235, https://doi.org/10.1093/ije/dyx206.
3. North TL, Davies NM, Harrison S, Carter AR, Hemani G, Sanderson E, Tilling K, Howe LD. Using Genetic Instruments to Estimate Interactions in Mendelian Randomization Studies. Epidemiology. 2019 Nov;30(6):e33-e35. doi: 10.1097/EDE.0000000000001096. PMID: 31469698.
4. Amy E Taylor, Hannah J Jones, Hannah Sallis, Jack Euesden, Evie Stergiakouli, Neil M Davies, Stanley Zammit, Debbie A Lawlor, Marcus R Munafò, George Davey Smith, Kate Tilling, Exploring the association of genetic factors with participation in the Avon Longitudinal Study of Parents and Children, International Journal of Epidemiology, Volume 47, Issue 4, August 2018, Pages 1207–1216, https://doi.org/10.1093/ije/dyy060.
Dr Liam Mahedy (lead), Prof. Marcus Munafò,
Growing evidence suggests that levels and changes in hand grip strength may be sensitive to subtle changes in brain health (Carson et al., 2018). Evidence from longitudinal studies have revealed mixed findings in terms of the direction of this relationship. For example, studies indicated that low hand grip strength was associated with cognitive impairment in ageing populations (Cooper et al., 2013), even when measured decades earlier in midlife (Dercon et al., 2021). Other studies found that poorer cognitive functioning was associated with low hand grip strength (van Dam et al., 2018). Findings examining the bidirectional association have also been unclear (McGrath et al., 2019, Ritchie et al., 2016). One recent genetic study that examined this association reported a positive association between genetic risk score for hand grip strength and cognitive functioning but did not examine the potential for a bidirectional relationship (Tikkanen et al., 2018).
Using genetic variants, which are fixed at conception, this project will help to determine if there is a potential causal association between hand grip strength and cognitive functioning and examine the potential for reverse causation between cognitive functioning and hand grip strength.
Summary level two-level bi-directional Mendelian randomisation.
Apply several sensitivity analyses including weighted median, weighted mode, MR-Egger regression, and MR-PRESSO to rule out pleiotropy.
Carson R.G. (2018). Get a grip: individual variations in grip strength are a marker of brain health. Neurobiol Aging. 71:189–222.
Cooper R., et al. (2014). Physical capability in mid-life and survival over 13 years of follow-up: British birth cohort study. BMJ. 348.
Dercon, Q., et al. (2021). Grip strength from midlife as an indicator of later-life brain health and cognition: evidence from a British birth cohort. BMC Geriatrics, 21, 475.
McGrath R., et al. (2019). Handgrip Strength Is Associated with Poorer Cognitive Functioning in Aging Americans. J Alzheimer’s Dis. 70:1187–96.
Ritchie S.J., et al. (2016). Do Cognitive and Physical Functions Age in Concert from Age 70 to 76? Evidence from the Lothian Birth Cohort 1936.Span J Psychol. 5; 19:E90.
Tikkanen, E., et al. (2018). Biological Insights into Muscular Strength: Genetic Findings in the UK Biobank. 8, 6451.
van Dam R., (2018). Cognitive Function in Older Patients with Lower Muscle Strength and Muscle Mass. Dement Geriatr Cogn Disord. 45,243-250.
Professor Richard Martin (lead), Dr James Yarmolinsky, Dr Philip Haycock
Circulating proteins play fundamental roles in biology and disease aetiology, including development of cancer, and are the direct targets of many drugs (1). Recent large-scale genome-wide association studies (GWAS) of circulating protein have identified thousands of protein quantitative trait loci (pQTLs) that influence protein levels (2-5). These variants can be used in a Mendelian randomization (MR) framework to estimate the causal effects on circulating proteins on cancer risk or progression (6). Robust identification of circulating proteins that causally influence risk or subsequent progression of cancer can then be used inform development of pharmacological interventions for cancer prevention and treatment.
The overall aim of this project is to identify circulating proteins that influence cancer risk and/or progression (i.e. spread or death). The specific objectives of this project are as follows:
i) Develop genetic instruments for circulating protein concentrations using recent GWAS to estimate the effect of these proteins on risk of overall and subtype-specific breast, colorectal, lung, ovarian, and prostate cancer.
ii) Examine the effect of circulating proteins on progression of breast, colorectal, lung, ovarian and prostate cancer.
iii) Examine agreement between the effect of circulating proteins on cancer risk compared with their effect on progression.
iv) Validate top findings from objectives i and ii using observational epidemiological analysis in UK Biobank or the UK wide Clinical Practice Research Datalink (where a protein discovered is also the drug target for an approved medication).
Genome-wide significant (P < 5 x 10-8) single-nucleotide polymorphisms (SNPs) associated with circulating protein concentrations will be used to develop genetic instruments for these proteins to estimate their causal effect on risk and progression of overall and subtype-specific breast, colorectal, lung, and prostate cancer using two-sample Mendelian randomization (1-5). Various sensitivity analyses will be performed to examine robustness of findings to violations of Mendelian randomization assumptions (e.g., colocalization analysis, heterogeneity testing, Steiger filtering, SloperHunter for cancer progression analyses). Concordance of effect estimates across cancer risk and progression analyses (corrected for potential index event bias) will be assessed. Top findings (i.e. those that are robust to multiple testing correction) will then be validated using observational epidemiological analysis within the UK Biobank cohort study or the Clinical Practice Research Datalink (CPRD).
This project will provide comprehensive and advanced training in genetic epidemiology and cancer, specifically Mendelian randomization applied to molecular traits and cancer biology, and experience writing up findings for publication and presentation. You will join a large cohort of fellow doctoral students and be part of a vibrant, intellectually generous and supportive Department.
1. Schmidt et al. Nature Communications, 2020. doi: 10.1038/s41467-020-16969-0
2. Folkersen et al. Nature Metabolism, 2020. doi: 10.1038/s42255-020-00287-2
3. Pietzner et al. Nature Communications, 2020. doi: 10.1038/s41467-020-19996-z
4. Gilly et al. Nature Communications, 2020. doi: 10.1038/s41467-020-20079-2
5. Sun et al. Nature, 2018. doi: 10.1038/s41586-018-0175-2
6. Zheng et al. Nature Genetics 2020. doi: 10.1038/s41588-020-0682-6
Dr Rachael Hughes (lead), Professor Kate Tilling, Dr Jon Heron Dr Jonathan Bartlett (Bath) Dr Rhian Daniel (Cardiff)
Most studies in medical research may suffer bias due to selection (participants are different to non-participants), or dropout (those who leave the study are different to those who remain). One way to investigate and address this bias is via inverse probability weighting (IPW). An important area of application is index event bias, a particular form of selection bias that occurs when examining risk factors for prognosis of a given disease. For example, IPW is of relevance in this context when studying the epidemiology of COVID-19. Currently, IPW has been less studied than other statistical approaches and there remain challenges in its application.
This project will use simulations (computer experiments where the truth is known) and applied examples (data from real-life studies) to develop IPW methods and software that will be of great relevance to many studies, especially much-used resources such as UK Biobank. First, we will investigate the best functional form of the weighting model used by IPW to generate the weights. Second, we will examine strategies for selection of variables for the weighting model and develop diagnostics for assessing the fit of the weighting model. Third, we will develop methods to carry out sensitivity analyses to examine robustness of results (e.g., sensitivity to the specification of the weighting model or stability of the weights).
Most studies in medical research may suffer bias due to selection (participants are different to non-participants), or dropout (those who leave the study are different to those who remain). One way to investigate and address this bias is via inverse probability weighting (IPW). Currently, IPW has been less studied than other statistical approaches and there remain challenges in its application.
This project will use simulations (computer experiments where the truth is known) and applied examples (data from real-life studies) to develop IPW methods and software that will be of great relevance to many studies, especially much-used resources such as UK Biobank. First, we will investigate the best functional form of the weighting model used by IPW to generate the weights. Second, we will examine strategies for selection of variables for the weighting model and develop diagnostics for assessing the fit of the weighting model. Third, we will develop methods to carry out sensitivity analyses to examine robustness of results (e.g., sensitivity to the specification of the weighting model or stability of the weights).
Luisa Zuccolo (lead), Carolina Borges, Nancy McBride
Breastfeeding is sustainable, the biological norm, and potentially life-saving, particularly for premature babies. Evidence-based strategies to support breastfeeding have been successful, but inequalities in breastfeeding rates are proving difficult to reduce, affecting the most vulnerable of mothers and babies. Successfully establishing and sustaining breastfeeding can be facilitated by both removing structural and cultural barriers, and overcoming individual challenges. Common factors such as obesity and depression/anxiety could play an important part in explaining some of the variability (and inequality) in breastfeeding duration. Conversely, maternal factors reflecting good mental and physical health could increase resilience to contexts with low systemic and cultural support for breastfeeding, such as the UK. However, the evidence on the individual determinants causally influencing successful and sustained breastfeeding is of poor quality. The identification of causal determinants of early cessation will improve breastfeeding support activities.
This project aims to establish the causal role of individual and environmental factors on successfully establishing and sustaining breastfeeding.
In particular, it will investigate:
1. Individual maternal factors as causal determinants of breastfeeding outcomes.
2. Joint and conditional effects of determinants of breastfeeding outcomes.
3. Contextual effects of determinants of breastfeeding outcomes, depending on high/low support.
Data:
Cohort studies participating in the MR-PREG consortium (MoBa, HUNT, ALSPAC, BiB...).
Maternal Exposures:
obesity, diabetes, internalising (depression/anxiety) and externalising (ADHD) problems, smoking/alcohol/caffeine use, sleep traits, eczema and skin conditions, education.
Outcomes:
successful establishment of breastfeeding (i.e. breastfeeding for 6+ weeks Vs <6 weeks); sustained breastfeeding (i.e. breastfeeding for 6+ months Vs <6 months).
Analysis:
1. 2-Samples Mendelian Randomization for each factor (exposure) in each cohort, then meta-analysis.
2. Mediation analysis to investigate to what extent the various individual factors explain education effects, thus mediating health inequalities - MultiVariable MR.
3. We will evaluate whether systemic factors underpinning high and low breastfeeding rates (level of support for breastfeeding) modify the effects attributed to the individual factors - meta-regression and cross-cohort analyses based on the summary data.
Victora CG et al. Lancet 2016; 387(10033): 474-90
Rollins NC et al. The Lancet 2016; 387(10017): 491-504
Dr Luisa Zuccolo (lead), Dr Carolina Borges, Dr Rachel Freathy
Breastfeeding is the biological norm and sustainable. It is also potentially life-saving, particularly for premature babies and those without access to clean water. Strategies to support breastfeeding have been successful, but inequalities persist and rates remain low in high-income countries such as the UK.
Although the short-term effects of breastfeeding are well documented, several questions about the epidemiology of breastfeeding remain unresolved, including which maternal and infant long-term outcomes are affected by different breastfeeding practices, and what the mechanisms behind these are.
This project will benefit mothers and babies by improving our understanding of practices and behaviours for optimal child development and long-term maternal health.
This project will improve our understanding of the effects (to the mother and the infant) of successfully establishing and sustaining breastfeeding.
Specific aims:
1. To identify genetic predictors of breastfeeding traits (to be used in genetic analyses to inform Aims 2. and 3.)
2. To estimate causal effects of breastfeeding on maternal and offspring health
3. To explore mechanisms for the long-term effects of breastfeeding
Breastfeeding traits include initiation, successful establishment, duration, exclusivity, breastfeeding problems.
This project aims to answer the above questions by combining cutting-edge methods that improve causal inference in observational studies, e.g. Mendelian Randomization and causal mediation, with classic epidemiological designs, e.g. cross-context comparisons. Results from each of these methods are likely to suffer from different biases, sometimes in opposite directions. We will exploit this using a triangulation approach, consistent results will provide stronger evidence for causality.
A key and novel component of the project will also be the identification of genetic predictors of breastfeeding traits through well-powered genome-wide association studies, to inform the Mendelian randomization analyses.
In order to fulfill the study’s objectives, an international network of collaborating cohorts has been established (N>120,000) to analyse existing data on breastfeeding traits, and putative determinants and consequences.
Victora CG et al. Lancet 2016; 387(10033): 474-90.
Dr Evie Stergiakouli (lead), Dr Gemma Sharp, Prof Sarah Lewis Christina Dardani
Cleft of the lip and/or palate (CL/P) is one of the most common congenital anomalies requiring corrective surgery within the first year of life. Children born with CL/P have 3.2 admissions and spend 13.2 days in hospital in the first two years of life. Despite treatment to repair the cleft in infancy, being born with a cleft frequently results in multiple adverse outcomes across the lifespan, including facial disfigurement, impaired speech and low intelligibility, with potentially poor educational, vocational, social, mental and physical health outcomes. We have previously shown, using data from a large cohort of children born with cleft, the Cleft Collective, that children born with CL/P have higher levels of behavioural problems than children in the general population at 5 and 10 years (1). However, we do not know whether there is genetic correlation between cleft and psychiatric disorders or whether factors associated to the cleft phenotype cause psychiatric disorders in children born with cleft. We have previously used Mendelian randomization to show that being born with cleft does not cause children to underperform at school (2).
In this project, we will explore the links between cleft and psychiatric disorders (ADHD, autism spectrum disorder (ASD), anxiety, depression, schizophrenia).
A. We will use two-sample Mendelian randomization (3) to investigate any causal links between genetic liability to cleft of the lip and/or palate, and ADHD, ASD, anxiety, depression, schizophrenia bidirectionally
B. Apply sensitivity analyses including weighted median, weighted mode, MR-Egger regression, MR-PRESSO and colocalization analyses to assess and adjust for pleiotropy
C. We will perform Linkage Disequilibrium (LD)-score regression (4) to estimate any genetic correlation between cleft of the lip and/or palate, and ADHD, ASD, anxiety, depression, schizophrenia
1. Berman et al. Prevalence and Factors Associated with Behavioural Problems in 5-year-old Children Born with Cleft Lip and/or Palate from the Cleft Collective. medRxiv 2021.08.04.21261594; https://doi.org/10.1101/2021.08.04.21261594
2. Dardani et al. Is genetic liability to ADHD and ASD causally linked to educational attainment?, International Journal of Epidemiology, 2021;, dyab107, https://doi.org/10.1093/ije/dyab107
3. Davey Smith, G. & Hemani, G. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum. Mol. Genet. 23, R89–R98 (2014)
4. Bulik-Sullivan et al. LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat Genet 2015;47:291–95.
Dr Luisa Zuccolo (lead), Dr Gemma Sharp, Dr Doretta Caramaschi (Exeter) Dr Anita Thapar (Cardiff)
There is growing concern that paternal exposures before conception have been greatly neglected. Studying how these impact on future generations’ health could open new avenues for prevention-prospective fathers are not generally advised to change their behaviour. In animal studies, one of the paternal exposures showing the largest and most consistently reported effects relative to the prenatal period is alcohol, however convincing human evidence is lacking to date. In this project, we propose to investigate the relationship between pre-conception paternal alcohol use and offspring mental health, and in particular attention deficit hyperactivity disorder (ADHD).
1 Is paternal alcohol use before conception associated with offspring ADHD, and/or its brain morphology correlates?
2 Is this due to shared genetic influences?
3 Is this association robust to different causal inference methods of analyses such as Mendelian randomization, and the use of negative controls (eg non-biological fathers)?
4 Are offspring cord blood DNA methylation and ADHD structural brain correlates on the causal pathway from paternal alcohol exposure childhood ADHD?
To improve the chance of disentangling correlation from causation, we propose to use a paradigm similar to that of (lab studies of) rodent models of paternal effects: a pseudo-experimental study design (employing Mendelian randomization and other analytical approaches to improve causal inference), restricting to families with no intrauterine exposure to alcohol (like in the animal studies), focusing on early manifestation of the outcome occurring before the onset of own drinking to remove confounding by own drinking, and even earlier potential biological mediators of the effects (eg cord-blood DNA methylation, eliminating confounding by paternal drinking in the postnatal period, both set to ‘zero’ in animal studies).
Finegersh A, Alcohol 2015 49:461-70
Rogers JC, JAMA Psychiatry 2016 73:64-72
Chen YC, Mol Psychiatry 2017 Mar 21
Dr Liam Mahedy (lead), Dr Hannah Sallis, Dr Padraig Dixon
According to attachment theory, patterns of interpersonal relationships may be important determinants of illness behaviour, care seeking, and treatment response in patients (Ciechanowski et al. 2003). Specifically, research has shown that insecure attachment can lead to greater treatment costs, missing previously scheduled appointments, and utilisation in medical outpatients (Ciechanowski et al. 2003, Bennet et al. 2011; Ciechanowski et al., 2002; Ciechanowski et al. 2006). Previous studies have been hampered by small sample size, missing data, and a lack of control for potentially important confounders. These limitations could lead to biased findings due to inaccurate estimation of associations. This mini project will examine the association between proxy measures of attachment styles (e.g., felt loved as a child, being in a confiding relationship as an adult, and number of jobs held) and inpatient hospital costs through record linkage between UK Biobank and records of inpatient care in England and Wales.
The primary objective is to predict total admitted patient hospital costs as a function of the conditional marginal effect of measures of attachment style on healthcare costs.
1. gain a familiarity with the literature on adult attachment style and healthcare costs
2. identify suitable attachment style questions in UK Biobank
3. establish whether a relationship exists between adult attachment style and healthcare economic cost using regression-based models
4. perform a number of sensitivity models to examine the role of important confounders
5. write up the findings as a manuscript for publication
Bennett JK, Fuertes JN, Keitel M, et al. The role of patient attachment and working alliance on patient adherence, satisfaction, and health-related quality of life in lupus treatment. Patient Educ Couns. 2011;85(1):53–59.
Ciechanowski PS, Walker EA, Katon WJ, et al. Attachment theory: a model for health care utilization and somatization. Psychosom Med. 2002;64(4):660–667.
Ciechanowski P, Sullivan M, Jensen M, et al. The relationship of attachment style to depression, catastrophizing and health care utilization in patients with chronic pain. Pain. 2003;104(3):627–637.
Ciechanowski P, Russo J, Katon W, et al. Where is the patient? The association of psychosocial factors and missed primary care appointments in patients with diabetes. Gen Hosp Psychiatry. 2006;28(1):9–17.
Dr Eleanor Sanderson (lead), Dr Rhian Daniel, Professor Laura Howe
Many of an individual’s traits are observationally associated with their health outcomes. Understanding the relationships between these factors is critical to effective public health intervention. When multiple traits are potentially associated with a disease or health outcome, it is often not clear how much of the observed effect of each single trait is due to the effect of that trait or behaviour on other traits and behaviours, which then affect the outcome, and how much is “direct” in the sense that it is not mediated by the other traits being considered.
Causal mediation analysis is one approach that can be used to determine the proportion of the effect of a trait on an outcome that is via a mediating variable. However, this method relies on many strong assumptions. An alternative approach, relying on different assumptions, is Mendelian Randomisation (MR). MR is a method of instrumental variable analysis which utilises genetic variation between individuals to help understand causal effects.
The aim of this project is to conduct research on the strengths and limitations of MR when trying to understand the causal effects of multiple exposures on a health outcome. This will include investigation of how novel methods of MR analysis such as Multivariable MR relate to mediation analysis and developing and extend existing MR methods to deal with multiple mediators. This project will also explore possibilities of combining the two approaches.
In this PhD you will have the opportunity to work with leading researchers in the fields of population health, statistics, and Mendelian randomisation to further develop statistical methods for causal analysis based around MR mediation analysis.
This project will involve mathematical derivation of the properties of the extended methods and verification using simulation studies. This project will also involve analysis of both individual-level and summary-level data to apply the methods developed.
Although this project will be methodological in focus, the student will have the opportunity to develop a relevant application of these methods based on their personal research interests. Prospective applicants should have a strong quantitative background and an interest in developing methods for causal analysis within a population health setting; however, no particular background knowledge is required.
Dr Gibran Hemani (lead), Mr Matthew Lyon, Prof Tom Gaunt
Programming is an increasingly important skill to have in epidemiological research and this project will
The OpenGWAS project hosts about 200 billion genetic association records in an online database, using a technology known as ElasticSearch. The way it is currently implemented is extremely fast for users but it is not particularly efficient in terms of how much disk space it uses in the cloud, and as a consequence it is very expensive.
An alternative way to host the data would be in GWAS VCF files that are more disk space efficient. The problem here is that we need to re-write the software used to query the database to lookup data in these GWAS VCF files rather than the ElasticSearch database.
This is not the most traditional of mini projects but we think it would be a really valuable one for both the student and the community. The student will gain:
a) An opportunity to learn python programming
b) A chance to contribute to a large open source software project
c) Contribution to the next OpenGWAS paper
1. A python package already exists that can query GWAS VCF files (https://github.com/MRCIEU/pygwasvcf). Update the pyGWASVCF package to use a faster VCF parsing library
2. The OpenGWAS database is queried via an API. In order for the API to query GWAS VCF files directly we will update the API to use pyGWASVCF
3. Performance comparisons of ElasticSearch vs pyGWASVCF
The project will entail creating a fork of the OpenGWAS API (https://github.com/MRCIEU/opengwas-api/) and updating it to perform queries against a directory of GWAS VCF files. We will then compare its performance against the original API that queries the ElasticSearch database.
At the onset of the project we will consider whether we should adopt the GenomicsDB approach for hosting the GWAS VCF files, depending on its developmental maturity.
1. Elsworth et al. The MRC IEU OpenGWAS data infrastructure. Biorxiv 2020 https://www.biorxiv.org/content/10.1101/2020.08.10.244293v1
2. Hemani et al. The MR-Base platform supports systematic causal inference across the human phenome. eLife 2018. https://elifesciences.org/articles/34408
3. Lyon et al. The variant call format provides efficient and robust storage of GWAS summary statistics. Genome Biology 2021. https://genomebiology.biomedcentral.com/articles/10.1186/s13059-020-02248-0
Dr Gibran Hemani (lead), Dr Josephine Walker, Dr Neil Davies
The gender pay gap grows substantially after women become parents (Kleven et al 2019). Encouraging more equitable sharing of parental leave amongst mothers and fathers may be a critical part of reducing the gender pay gap (Johansson 2010). Though a shared parental leave scheme was introduced in the UK in 2015, only 2-3% of fathers have enrolled in the scheme. Amongst the reasons for this is that it is governed by a set of complex, inaccessible rules that vary across work places (Maternity Action, 2021). For example, the University of Bristol offers more pay to fathers than the baseline shared parental leave scheme (http://www.bristol.ac.uk/hr/policies/shared-parental-leave.html).
The Bristol branch of the Universities and Colleges Union (UCU) along with the University of Bristol HR department would like to help parents-to-be plan and understand their leave and pay options through a user-friendly calculator.
The aim of this project is to develop user friendly software to plan shared parental leave and calculate the pay and other aspects of parental leave. It is a great opportunity to learn web application development skills, whilst providing a valuable resource to the community.
An R package already exists that calculates various aspects of shared parental leave (https://github.com/explodecomputer/shared-parental-leave/). The primary task of this project will be to develop a web app that makes this R package more user friendly.
We will approach this project in a similar manner to how the USS pension modeller was developed (http://www.uss-pension-model.com/). Here, an R package was developed to perform the analysis and calculations (https://github.com/explodecomputer/USSpensions), and an R shiny app was developed as a wrapper for the calculations in the R package (https://github.com/explodecomputer/USSpensions-shiny/).
1. Kleven et al. Children and Gender Inequality: Evidence from Denmark. American Economic Journal: Applied Economics. 2019.
2. Johansson. The effect of own and spousal parental leave on earnings. Working Papers, IFAU - Institute for Evaluation of Labour Market and Education Policy. 2010. https://www.econstor.eu/handle/10419/45782
3. Maternity Action, 2021. https://maternityaction.org.uk/wp-content/uploads/Shared-Parental-Leave-briefing-May-2021.pdf
Dr Liam Mahedy (lead), Dr Emma Anderson, Dr Rebecca Richmond Prof. Marcus Munafò
Dementia is one of the leading causes of death globally and increasing longevity ensures its
prevalence will rise even further (1). As there is no known treatment that prevents the progression of dementia, there is an emphasis to focus on prevention as the best strategy to reduce incidence and prevalence. Since dementia has a long latency period, often up to 20 years before the symptoms are present (2), it is crucial to examine potential risk factors prior to this 'pre-clinical' dementia phase, as pathological accumulation could influence the risk factors. Although individuals with dementia experience sleep disruption (i.e., of duration, quality, and timing) (3), it remains unclear whether sleep disruption is also a risk factor for dementia. One of the main limitations in previous prospective cohort studies is that they do not consider the importance of sleep characteristics from early to midlife as they largely focus on adults aged >65 years at baseline and have relatively short follow-up periods (4,5). This area of research has received little attention from genetic epidemiology. A recent study (6) using a Mendelian randomisation approach found little evidence of a causal relationship between sleep traits and Alzheimer’s Disease; however, the authors recognised their study was underpowered.
The primary research question is to examine whether sleep disruption is a risk factor for dementia, and the alternative, whether dementia pathology and symptomology are a risk factors for sleep disruption. The secondary research question will examine whether critical or sensitive periods can be identified during which sleep disruption may have a greater impact on incident dementia and dementia progression, or vice versa.
This project will use secondary data sources from Dementias Platform UK (7), which contains rich questionnaire and clinic data alongside genetic information for almost 50 cohorts, comprising over 3 million individuals (e.g., the Avon Longitudinal Study of Parents and Children (Mother cohort) (8), and English Longitudinal Study of Aging) (9). Other potential data sources include The HUNT Study (10), and the International Genomics of Alzheimer’s Project (11). This project will use observational methods to examine change over time in cohort studies (i.e., latent growth modelling), and genetic methods to examine potential causal relationships (i.e., two sample Mendelian randomisation, and multivariable Mendelian randomisation).
1. Prince M et al. Alzheimer’s Dis Int. 2015;1–92.
2. Villemagne VL et al. Lancet Neurol. 2013;12,357–67.
3. Brzecka A et al. Front Neurosci. 2018;12:330.
4. Lutsey, PL et al. Alzheimers Dement. 2018;14,157–166.
5. Westwood, AJ et al. Neurology. 2017;88,1172–1179.
6. Anderson, EL et al. Int J Epidemiol. 2021;50,817-828.
7. Bauermeister, S et al. Eur J Epidemiol. 2020;35,601-611.
8. Fraser, A et al. Int J Epidemiol. 2013;42,97-110.
9. Steptoe, A et al. Int J Epidemiol. 2013;42,1640-1648.
10. Krokstad, S et al. Int J Epidemiol. 2013;42,968-977.
11. Lambert, JC et al. Nat.Genet. 2013;45,1452-U1206.
Dr Emma Vincent (lead), Dr Caroline Bull, Professor Nicholas Timpson
Growing evidence suggests that cell extrinsic factors are key in modulating tumor progression. Metabolites are small molecules that act as sources of fuel and building blocks essential for cells and tissues when present at normal levels. Many causal risk factors for cancer (such as obesity or smoking) perturb metabolite levels, meaning cells of the body are exposed to an abnormal metabolic environment in at-risk individuals. It is possible that metabolites may be involved in the causal mechanisms linking risk factors with cancer development, acting to favour the growth and survival of cancer-initiating cells.
This project will improve our understanding of the causal metabolic drivers of cancer development.
Specific aims:
1. To identify genetic predictors of cancer susceptibility and risk factors causally related to cancer
2. To estimate the causal effects of cancer susceptibility and cancer risk factors on circulating metabolites
3. To triangulate evidence to build knowledge of the causal mechanisms linking circulating metabolites with genetic susceptibility to cancer, risk factors for cancer and cancer development using epidemiological and preclinical techniques.
We will construct polygenic risk scores for cancer susceptibility and risk factors for cancer (tissues/specific risk factors will be determined by the student's interests) in mothers and children of the ALSPAC cohort to determine associations with circulating metabolites. We will use Mendelian randomization to estimate causal relationships between genetic susceptibility and risk factors for cancer with circulating metabolites and with cancer risk using summary statistics from large cancer consortia. There will also be the opportunity to work across research disciplines and to conduct mechanistic follow-up analyses in the laboratory using cell culture techniques.
Brennan, P & Davey Smith, G. Identifying Novel Causes of Cancers to Enhance Cancer Prevention: New Strategies are Needed. JNCI. https://doi.org/10.1093/jnci/djab204.
Martinez-Reyes, I & Chandel, N.S. Cancer metabolism: looking forward. Nature Metabolism. https://doi.org/10.1038/s42255-021-00478-5
Sarah Lewis (lead), Dr Kostas Tsilidis,
Diet and lifestyle are likely to play an important role in colorectal cancer risk; obesity, low levels of physical activity, red and process meat consumption and low intake of dietary fibre have all been shown to predict colorectal cancer risk. In addition, we have previously used Mendelian randomization (MR) analyses to identify physical activity (1) and iron, vitamin-B12 and selenium as potential risk factors for colorectal cancer risk (2). However, it is not clear which are the independent risk factors and which are confounded by other diet and lifestyle factors and the mechanisms via which these risk factors contribute to colorectal cancer are not well understood.
This project will; identify independent diet and lifestyle risk factors for colorectal cancer, integrate multi-omics data in a hypothesis free analysis to uncover novel risk factors and elucidate the biological mechanisms between modifiable risk factors and colorectal cancer.
This project will use many different MR methods, integrate omics datasets and perform mediation analyses to identify risk factors for colorectal cancer risk and elucidate the biological mechanisms involved in this disease.
Firstly the student will use a Bayesian framework (3), multivariable MR, co-localisation analyses and bi-directional MR methods incorporating instruments across many different potential risk factors in order to identify independent factors for colorectal cancer risk.
Once independent risk factors have been identified, the student will use two-step mendelian randomization to determine whether the risk factor-colorectal cancer link is mediated by any of the following four pathways; inflammation, growth hormones, sex hormones and insulin signalling.
Zuber V, Gill D, Ala-Korpela M, Langenberg C, Butterworth A, Bottolo L, Burgess S. High-throughput multivariable Mendelian randomization analysis prioritizes apolipoprotein B as key lipid risk factor for coronary artery disease. Int J Epidemiol. 2021 Jul 9;50(3):893-901. doi: 10.1093/ije/dyaa216. PMID: 33130851; PMCID: PMC8271202.
Tsilidis KK, Papadimitriou N, Dimou N et al. Genetically predicted circulating concentrations of micronutrients and risk of colorectal cancer among individuals of European descent: a Mendelian randomization study. Am J Clin Nutr. 2021 Jun 1;113(6):1490-1502. doi: 10.1093/ajcn/nqab003. Erratum in: Am J Clin Nutr. 2021 Jun 1;113(6):1715. PMID: 33740060; PMCID: PMC8168352.
Papadimitriou N, Dimou N, Tsilidis KK et al. Physical activity and risks of breast and colorectal cancer: a Mendelian randomisation analysis. Nat Commun. 2020 Jan 30;11(1):597. doi: 10.1038/s41467-020-14389-8. PMID: 32001714; PMCID: PMC6992637.
Dr Neil Ryan (lead), Prof Sir John Burn ,
The aim of this proof-of-concept study is to analyse urine samples, collected with Colli-Pee, in women with microsatellite instability high (MSI-H) endometrial cancers as to determine if MSI-H can be detected in their urine. This could potentially provide an alternative means of endometrial cancer surveillance in women with Lynch syndrome (LS) that would enable painless self-sampling and the triage of those who need invasive investigations.
LS is the most common inherited cancer predisposition[1]; as many as many as one in 278 individuals have LS [2]. For a woman with LS, the lifetime risk of endometrial cancer is up to 60%[3]. However, these women have limited options as to how to mitigate their risk. Risk reducing surgery has been shown to be effective[4], yet it is not suitable for all women; such surgery leads to infertility and the surgical risk may be too high for some individuals. Endometrial cancer surveillance is often offered in leu of risk reducing surgery in the form of annual endometrial biopsy[5]. Endometrial sampling can be painful, and the fear of pain can prompt some women to leave surveillance programs or opt for risk reducing surgery sooner then they may have otherwise[6]. Due to the costly and labour-intensive nature of such screening programs many centres fail to offer a comprehensive service[7]. Therefore, there is a need for alternative methods by which to detect early endometrial neoplasms in women with LS.
Endometrial cancers that arise in women with LS often display MSI-H [8]. This is an early event in the natural history of the cancer[9]. What is more, MSI-H can be detected in the urine of those with cancer [10,11]. It has also been detected in the hysteroscopic wash of patients with endometrial cancer[12,13]. It is known that those with endometrial cancer shed malignant cells that can be detected in their urine[14]. Therefore, it is theoretically possible that MSI-H could be detectable in the urine of those with MSI-H in their endometrial cancer; this has yet to be explored.
Around 20% of endometrial have MSI because of somatic events or LS[15]. As all endometrial cancers are screened for MSI in the UK, it is possible to identify women early in their cancer journey. Women with MSI-H endometrial cancer would be recruited into this study and asked to use the Colli-Pee to collect up to three morning urine samples which would be analysed for MSI. The technology to explore MSI in urine samples already exists [16–20]. These data will inform a potential future study in which those with LS could trial the use of the Colli-Pee based urine MSI analysis as to provide an alternative means of endometrial cancer surveillance in women with LS.
Aim: The overarching aim of this study is to explore the promise of a urine based endometrial cancer detection tool that could be used to provide endometrial surveillance in women with Lynch syndrome (LS). If successful, this would avoid the current need for invasive, repetitive, and costly diagnostic tests and allow women with LS to self-test.
Objectives:
Main:
To explore if microsatellite instability (MSI) can be detected in the urine of those with microsatellite high (MSI-H) endometrial cancers using the Colli-Pee device.
Specific:
• Seek and acquire ethical approval for multi-cite recruitment upon confirmation of a successful application.
• Recruit between 30-40 women with proven microsatellite instability high endometrial cancer, vaginal bleeding and before their hysterectomy from centres in the Southwest of England (Bristol, Bath and Taunton) over the course of a year.
• Recruit 5 controls without endometrial cancer but with vaginal bleeding.
• Collect up to three morning urine samples, to increase cellular yield, from each study participant.
• Within the first two months of recruitment, optimise sample preservation from collection to analysis; namely trial the use of Novosanis Urine Conservation Medium.
• Within in the first two months of recruitment, optimise material extraction for the use of a MSI Analysis System.
• Explore the use of an extended MSI panel to see if this can improve diagnostic accuracy.
• Published a peer reviewed manuscript detailing the outcomes of this study within a year of the study’s closure.
Subject identification:
Women will with microsatellite instability high (study arm) will be identified by members of the research team during the weekly multidisciplinary team meeting for their respective cancer centre. It is routine clinical practice in the Southwest of England for microsatellite analysis to be done on the diagnostic sample. This will allow for the identification of suitable study subjects before they hysterectomy. In total these centres treat around 350 endometrial cancers a year; assuming 20% [1] have microsatellite instability, there would be 70 potential women to recruit each year. Written and informed consent will be required before entry into the study. In addition, five women without cancer but vaginal bleeding will be recruited from the benign gynaecology workload as a control arm. All women will be approached during their routine clinical care.
Inclusion criteria:
• Ability to given informed and written consent
• Greater than or equal to 18 years old
• Uterus still in situ
• Histological confirmed endometrial cancer with microsatellite instability (study arm) or no clinically confirmed cancer (control arm)
• Current vaginal bleeding
• Ability to self-collect a urine sample
Exclusion Criteria
• < 18 years old
• Unable to given informed and written consent
• Concurrent urogenital or colorectal cancer
• No active vaginal bleed
• No residual cancer at hysterectomy (study arm)
• No uterus
• Significant urinary incontinence (defined as continuous urinary leak)
• Vagino-visceral fistula
• Previous pelvic radiotherapy
• Congenital urogenital abnormality/Procidentia
• Inability to self-collect urine sample.
Sample Size:
With a prevalence of disease (MSI-H endometrial cancer) of 100%, and the values of both sensitivity or specificity of the screening or diagnostic test {for both null (Ho) and alternative (Ha) hypotheses} being set at 0.6 and 0.9 respectively, a sample size of 35 is required[2].
In addition, 5 controls will be recruited in which one urine sample will be analysed. Finally, we assume some women will have no residual disease on hysterectomy and therefore 5 additional women may need to be recruited to the study arm[3].
Sample collection:
Women will be asked to produce between one to three morning urine samples; morning collection should increase the cellular yield as endometrial shed pools in the vagina overnight. These will be collected in the Colli-Pee device in combination with the Novosanis Urine Conservation Medium to maximise DNA preservation. If this leads to suboptimal DNA quality for microsatellite instability analysis other mediums will be trailed, for example the Novosanis Urine Analyte Stabiliser. For samples collected by patients at home, the Colli-Pee postal kit could be utilised for direct transportation to the genomics laboratory.
In addition, a heperinized venous blood sample will be taken from the patient as a source of reference DNA. This will be stored at below 4 degrees Celsius in registered clinical laboratory. We will aim to time this blood draw to coincide with routine clinical bloods as to minimise discomfort for the patient.
Sample analysis:
Protocols for microsatellite instability analysis in urine have been described in the literature[4–8]. These will be optimised for the detection of microsatellite instability in endometrial cancer cells found within the urine. In the first instance, the urine samples were spun at 3000 rpm for 6 minutes within 5 hours after voiding. The pellet was rinsed with 10 mL phosphate-buffered saline (PBS) and spun again at 3000 rpm for 6 minutes. The pellet then was resuspended in 800 μL PBS, brought to a 1.5-mL Eppendorf tube, and spun down at 10,000 rpm for 2 minutes. The cells were stored at −80 °C until DNA isolation. DNA extraction will follow the clinical laboratory’s current standard operating procedures. If these methods prove suboptimal, the literature will be searched for alternative protocols.
References
1 Ryan NAJ, Glaire MA, Blake D, et al. The proportion of endometrial cancers associated with Lynch syndrome: a systematic review of the literature and meta-analysis. Genet Med 2019;21:2167–80. doi:10.1038/s41436-019-0536-8
2 Bujang MA. Requirements for Minimum Sample Size for Sensitivity and Specificity Analysis. J Clin Diagnostic Res Published Online First: 2016. doi:10.7860/jcdr/2016/18129.8744
3 O’Flynn H, Ryan NAJ, Narine N, et al. Diagnostic accuracy of cytology for the detection of endometrial cancer in urine and vaginal samples. Nat Commun 2021;12:952. doi:10.1038/s41467-021-21257-6
4 Rhijn BWG van, Lurkin I, Kirkels WJ, et al. Microsatellite analysis—DNA test in urine competes with cystoscopy in follow‐up of superficial bladder carcinoma. Cancer 2001;92:768–75. doi:10.1002/1097-0142(20010815)92:4<768::aid-cncr1381>3.0.co;2-c
5 Szarvas T, Kovalszky I, Bedi K, et al. Deletion analysis of tumor and urinary DNA to detect bladder cancer: urine supernatant versus urine sediment. Oncol Rep 2007;18:405–9.
6 Utting M, Werner W, Dahse R, et al. Microsatellite analysis of free tumor DNA in urine, serum, and plasma of patients: a minimally invasive method for the detection of bladder cancer. Clin Cancer Res Official J Am Assoc Cancer Res 2002;8:35–40.
7 Utting M, Werner W, Muller G, et al. A Possible Noninvasive Method for the Detection of Bladder Cancer in Patients. Ann Ny Acad Sci 2001;945:31–5. doi:10.1111/j.1749-6632.2001.tb03861.x
8 Goessl C, Müller M, Straub B, et al. DNA Alterations in Body Fluids as Molecular Tumor Markers for Urological Malignancies. Eur Urol 2002;41:668–76. doi:10.1016/s0302-2838(02)00126-4
9 Mead LJ, Jenkins MA, Young J, et al. Microsatellite Instability Markers for Identifying Early-Onset Colorectal Cancers Caused by Germ-Line Mutations in DNA Mismatch Repair Genes. Clin Cancer Res 2007;13:2865–9. doi:10.1158/1078-0432.ccr-06-2174
10 Ryan NA, McMahon RF, Ramchander NC, et al. Lynch syndrome for the gynaecologist. Obstetrician Gynaecol 2021;23:9–20. doi:10.1111/tog.12706
11 Crosbie EJ, Ryan NAJ, Bosse T, et al. The Manchester International Consensus Group recommendations for the management of gynecological cancers in Lynch syndrome. Genetics in Medicine 2019.
12 Ryan N, Nobes M, Sedgewick D, et al. A mismatch in care: results of a United Kingdom‐wide patient and clinician survey of gynaecological services for women with Lynch syndrome. Bjog Int J Obstetrics Gynaecol 2021;128:728–36. doi:10.1111/1471-0528.16432
13 Auranen A, Joutsiniemi T. A systematic review of gynecological cancer surveillance in women belonging to hereditary nonpolyposis colorectal cancer (Lynch syndrome) families. Acta Obstet Gyn Scan 2011;90:437–44. doi:10.1111/j.1600-0412.2011.01091.x
14 Helder-Woolderink J, Bock G de, Hollema H, et al. Pain evaluation during gynaecological surveillance in women with Lynch syndrome. Fam Cancer 2017;16:205–10. doi:10.1007/s10689-016-9937-x
15 Yang KY, Caughey AB, Little SE, et al. A cost-effectiveness analysis of prophylactic surgery versus gynecologic surveillance for women from hereditary non-polyposis colorectal cancer (HNPCC) Families. Fam Cancer 2011;10:535–43. doi:10.1007/s10689-011-9444-z
16 Ryan NAJ, Davison NJ, Payne K, et al. A Micro-Costing Study of Screening for Lynch Syndrome-Associated Pathogenic Variants in an Unselected Endometrial Cancer Population: Cheap as NGS Chips? Frontiers Oncol 2019;9:61. doi:10.3389/fonc.2019.00061
Dr Siang Boon Koh (lead), Dr Timothy Robinson, Dr Siddhartha Kar
RAS is a family of three small GTPases deregulated in at least 30% of human cancers. Classically, RAS deregulation is associated with genetic mutations of the RAS genes. Recently, RAS regulators (namely RAS-GEFs and RAS-GAPs) have been recognised as critical players through which non-genetic RAS deregulation arises. RAS-GEFs promote RAS activity, while RAS-GAPs inhibit RAS activity. Some RAS-GEFs and -GAPs also have other oncogenic roles that remain obscure (1).
Less than 5% of breast cancer harbours RAS genetic mutations, thus RAS deregulation is not widely suspected as an oncogenic event in breast cancer. Surprisingly, our recent work identified a non-genetic RAS deregulation in triple-negative breast cancer (TNBC) (2). TNBC is an aggressive breast cancer subtype, where the only systemic treatment for most patients is chemotherapy. We discovered that a RAS-GAP called RASAL2 promotes chemoresistance (2). Remarkably, RASAL2 also drives sensitivity to RAS pathway inhibitors (2). This exciting discovery suggests that RAS regulators may represent a new class of cancer biomarkers and drug targets.
There are at least 20 RAS-GEFs and -GAPs, but little is known about their roles in tumour progression and treatment response. Studying RAS regulators that can promote cancer progression (prognostic) or predict treatment response (predictive) may lead to effective therapeutic interventions. The aims of this project are to:
1) Assess the prognostic value of RAS regulators in breast cancer
2) Assess the predictive value of RAS regulators in breast cancer
3) Assess the potential causal role of RAS regulators in breast cancer risk
We have identified at least two RAS regulators to be prognostic/predictive in breast cancer (2). Here, we will use Mendelian randomisation (MR) to validate these candidates, and evaluate the relevance of other RAS regulators in breast cancer. We have successfully applied this method in similar studies (3), and anticipate this project to support our ongoing effort in defining the emerging role of RAS regulators.
The MR approach will be applied to cancer and control cases in datasets such as TCGA and the Breast Cancer Association Consortium. Based on prior experience, single nucleotide polymorphisms (SNPs) marking expression quantitative trait loci (eQTLs) that are associated with gene expression level at genome-wide significance (P<5x10-8) can be selected as genetic instruments. To retain independent SNPs, linkage disequilibrium clumping with a threshold of r2≤0.01 based on the 1000 Genomes European ancestry reference panel data will be used. R2 and F-statistics will be calculated to assess the strength of the genetic instruments. To account for multiple testing, Bonferroni corrections will be used to establish P thresholds for evidence of a causal effect.
1. O Maertens, et al. Adv Biol Regul 55, 1–14 (2014).
2. SB Koh, et al. Clin Cancer Res 27, 4883–4897 (2021).
3. T Robinson, et al. Int J Cancer 147, 1597–1603 (2020).
Professor Abigail Fraser (lead), Dr Amy Taylor,
There is evidence that earlier age at menarche is associated with increased risk of adverse pregnancy outcomes including gestational diabetes mellitus (GDM) [1, 2], gestational hypertension [3], low birth weight [4] and preterm birth [5]. However, it is unclear the extent to which these associations are confounded by maternal body mass index (BMI). For example, adjustment for BMI attenuated associations with GDM [1, 2] and blood pressure [3] but did not remove the association with preterm birth [5]. We know from Mendelian Randomisation studies that earlier age at menarche is likely to be both a cause [6] and a consequence [7] of higher body mass index, so it is possible that the relationship between age at menarche, BMI and adverse pregnancy outcomes is complex. However, here are other plausible mechanisms through which earlier age at menarche may cause adverse pregnancy outcomes, for example exposure to higher levels of estradiol and inflammatory markers in adulthood or metabolic changes associated with earlier menarche such as insulin resistance [5].
In Mendelian randomisation, genetic variants associated with exposures can be used as proxies for measured exposures to minimise some of the problems associated with observational analyses (confounding and reverse causality) and allow for stronger causal inference [8].
The aim of this project is to perform multivariable Mendelian Randomisation [9] analyses of age at menarche, maternal BMI, and adverse pregnancy outcomes.
We will perform two sample Mendelian randomisation using data from the MR-PREG (a consortium of studies with genetic and phenotypic data on pregnant women). We will use genetic variants identified as associated with age at menarche as the exposure and look at a range of adverse pregnancy outcomes for which there are GWAS data within MR-PREG. To investigate the role of maternal BMI as a potential confounder, we will perform multivariable Mendelian randomisation, including genetic variants associated with BMI.
1. Wang, L., et al., Age at menarche and risk of gestational diabetes mellitus: a population-based study in Xiamen, China. BMC Pregnancy Childbirth, 2019. 19(1): p. 138.
2. Sun, X., et al., Age at menarche and the risk of gestational diabetes mellitus: a systematic review and meta-analysis. Endocrine, 2018. 61(2): p. 204-209.
3. Petry, C.J., et al., Age at Menarche and Blood Pressure in Pregnancy. Pregnancy Hypertens, 2019. 15: p. 134-140.
4. Coall, D.A. and J.S. Chisholm, Evolutionary perspectives on pregnancy: maternal age at menarche and infant birth weight. Soc Sci Med, 2003. 57(10): p. 1771-81.
5. Li, H., et al., Age at menarche and prevalence of preterm birth: Results from the Healthy Baby Cohort study. Sci Rep, 2017. 7(1): p. 12594.
6. Gill, D., et al., Age at menarche and adult body mass index: a Mendelian randomization study. Int J Obes (Lond), 2018. 42(9): p. 1574-1581.
7. Mumby, H.S., et al., Mendelian Randomisation Study of Childhood BMI and Early Menarche. J Obes, 2011. 2011: p. 180729.
8. Davey Smith, G. and G. Hemani, Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet, 2014. 23(R1): p. R89-R98.
9. Sanderson, E., W. Spiller, and J. Bowden, Testing and correcting for weak and pleiotropic instruments in two-sample multivariable Mendelian randomization. Stat Med, 2021. 40(25): p. 5434-5452.
Dr Gemma Sharp (lead), Dr Hannah Jones, Dr Katherine Ruth, Genetics of Complex Traits, University of Exeter Medical School Dr Anna Murray, Genetics of Complex Traits, University of Exeter Medical School Dr Arianna Di Florio, Cardiff University
Problematic menstrual symptoms such as heavy menstrual bleeding (HMB) are common and distressing, impacting substantially on quality of life and mental wellbeing of affected women. As well as this likely effect of HMB on poor mental health (MH), there could also be an effect in the reverse direction: psychological stress is also a known disruptor of menstrual cycles and can be associated with heavier bleeding. Additionally, HMB is almost always self-reported and measured subjectively (with reduced quality of life now being part of the official diagnosis), so an association between HMB and MH might be explained by women with mood disorders being more likely to assess their level of bleeding as abnormal and substantially affecting their quality of life.
Both HMB and MH are influenced by inflammation. Increasing evidence suggests that inflammation plays a causal role in the pathogenesis of psychiatric disorders, including depression, and inflammation is also known to show cyclical variation, partly influenced by fluctuations in reproductive hormones throughout the menstrual cycle. There is also some evidence that systemic and local endometrial inflammation is associated with severity of menstrual symptoms like HMB.
This project aims to explore direct effects of HMB on MH, and of MH on HMB, as well as the potential confounding or mediating role of inflammation. A better understanding of these relationships would help inform ways to predict whether women with HMB are at greater risk of MH conditions, and whether women with MH conditions are at greater risk of menstrual issues like HMB. This would help clinicians tailor care to improve outcomes for groups of women. Understanding whether these associations are likely to be causal, and whether inflammation plays a role in the manifestation of both HMB and MH conditions will help shed light on the pathogenesis of HMB (and underlying pathologies) and MH and suggest whether both could be effectively treated using anti-inflammatory medications.
This project will use a genetic epidemiology approach to examine bi-directional associations between HMB and MH, as well as the role of inflammation in explaining any relationship. The student will first examine the association between HMB and MH phenotypes in UK Biobank and the Avon Longitudinal Study of Parents and Children (ALSPAC). Then, using results from our previous genomewide association study (GWAS) of HMB in UK Biobank, the student will generate polygenic risk scores (PRS) for HMB UKB and ALSPAC and conduct reduced-phenomewide association studies (PheWAS) to identify MH variables and inflammatory biomarkers associated with genetic liability to HMB. The student will also use LD score regression (LDSR) to explore the genetic correlation between all traits of interest using our HMB GWAS results and results from previously published GWAS of MH conditions and inflammatory biomarkers. Any associations from the pheWAS and LDSR will be followed up using Mendelian Randomization (MR) to explore if relationships are likely to be causal. The student will learn to apply 2 MR approaches: 1) 1-sample MR (this will allow us to test MR assumptions using individual-level data), 2) 2-sample MR using GWAS summary statistics (this will improve power above the one sample MR). Sensitivity analyses will allow the student to explore assumptions of both methods.
The student will also explore effects in the opposite direction by using results of previous GWAS of MH conditions and inflammatory biomarkers to generate PRS in ALSPAC and UK Biobank and test for associations with HMB, as well as in 1 and 2 sample MR analyses to explore causality.
The student will triangulate findings from across the PRS and MR analyses to inform network MR analysis including HMB, MH and inflammation to explore the strength and direction of direct and indirect and combined and independent causal effects.
Dr Emma Anderson (lead), Dr Laura Howe,
Studies have previously reported both socioeconomic (e.g., low head of household social class, parental education) and psychosocial (e.g., sexual or physical abuse, emotional neglect) adversity to be associated with lower cognitive and physical capability later in life. However, very little is understood about the pathways that mediate this association. For example, are people who experience socioeconomic and psychosocial adversity in childhood more likely to go on to smoke, have higher levels of alcohol consumption, be overweight, have lower occupational social class, which then in turn go on to cause adverse health outcomes in adulthood? Which potential mediating factors have the greatest impact, and thus, would be useful targets for intervention?
This project aims to examine whether the observed association between psychosocial adversity in childhood (e.g. physical and sexual abuse, emotional neglect, parental death etc), and later physical capability is mediated by smoking, alcohol consumption, BMI, educational attainment, occupational social class (and any other potential mediators that might be of interest to the student).
This project will use data from the Avon Longitudinal Study of Parents and Children mothers’ cohort. Women answered questions retrospectively about various forms of psychosocial adversity during childhood, in questionnaires administered at enrolment into the cohort. Women have also had physical capability assessed repeatedly at research clinics from mean age 53 years. This project will enable you to gain familiarity with the ALSPAC data, to learn about the epidemiology of childhood adversity which is an important and current topic, and develop skills in mediation analysis. You will:
1. Conduct a literature search
2. Identify relevant confounders and develop an analysis plan
3. Examine the association between childhood psychosocial adversity and physical capability, and possible mediators of this association.
4. Write up the paper for publication
1. Anderson EL, Heron J, Ben-Shlomo Y, Kuh D, Cooper C, Lawlor DA, Fraser A, Howe LD. Adversity in childhood and measures of ageing in mid-life: findings from a cohort of British women. Psychol Aging. 2017 (in press: http://research-information.bristol.ac.uk/files/113238561/Adversity_in_childhood_and_measures_of_ageing_in_mid_life_Anderson_et_al.pdf).
2. Ding R, He P. Associations between childhood adversities and late-life cognitive function: Potential mechanisms. Soc Sci Med. 2021 Dec;291:114478. doi: 10.1016/j.socscimed.2021.114478. Epub 2021 Oct 9. PMID: 34649168.
Dr Evie Stergiakouli (lead), Prof Laura Howe, Dr Rachel Blakey, Christina Dardani
Neurodevelopmental and mental health problems in childhood have been linked to adverse physical health outcomes in both childhood and adult life. For example, genetic risk scores for ADHD have been associated with adverse lifestyle and physical health outcomes even in individuals from the general population who do not necessarily exhibit the disorder (1). Causal analyses have linked ADHD to childhood obesity and coronary artery disease (2). However, neurodevelopmental disorders and more specifically ADHD are also associated with adverse socioeconomic factors and impaired educational attainment (3).
In this project, we aim to investigate the effect of neurodevelopmental and mental health problems in childhood on child and adult physical health while taking into account the strong associations of neurodevelopmental problems and mental health with socioeconomic factors, education and parental psychopathology.
We will construct polygenic risk scores for childhood neurodevelopmental and mental health conditions in parents and children from the general population to investigate their associations with adult physical health and perform Polygenic Transmission Disequilibrium test (4) to asses genetic risk transmitted from parents to children. We will perform Multivariable Mendelian randomization (5) to disentangle the effects of genetic liability to neurodevelopmental problems on physical health accounting for socioeconomic factors. We will apply sensitivity analyses including weighted median, weighted mode, MR-Egger regression, MR-PRESSO and colocalization analyses as well as other causally informative designs to assess and adjust for pleiotropy. There is also the opportunity to contribute to large-scale meta-analyses including the Norwegian Mother, Father and Child Cohort Study (MoBa) and initiate collaborations with external researchers.
1) Leppert et al. Association of Maternal Neurodevelopmental Risk Alleles With Early-Life Exposures. JAMA Psychiatry 2019;76(8):834–842. doi:10.1001/jamapsychiatry.2019.0774
2) Leppert et al. The Effect of Attention Deficit/Hyperactivity Disorder on Physical Health Outcomes: A 2-Sample Mendelian Randomization Study, AJE 2021; 190 (6), 1047–1055, https://doi.org/10.1093/aje/kwaa273
3) Dardani et al. Is genetic liability to ADHD and ASD causally linked to educational attainment?, IJE 2021;, dyab107, https://doi.org/10.1093/ije/dyab107
4) Weiner et al. Polygenic transmission disequilibrium confirms that common and rare variation act additively to create risk for autism spectrum disorders. Nat. Genet. 2017 49, 978
5) Sanderson et al. An examination of multivariable Mendelian randomization in the single-sample and two-sample summary data settings. IJE 2018; 1;48(3):713-727. doi: 10.1093/ije/dyy262.
Dr Amanda Hughes (lead), Prof Laura Howe, Dr Helen Bould
Internalized weight stigma, the belief that negative obesity-related stereotypes apply to oneself(1), is associated with poorer health(2). Among higher-weight individuals, it is linked to disordered eating(3) maladaptive coping(4), and worse quality of life(5). In normal-weight and underweight individuals, it predicts disordered eating and drive for thinness(6,7), and is thus relevant to mental health across the body weight range. However, most research on internalized weight stigma has been based on small, non-representative samples (for example recruited from weight loss groups, or undergraduate psychology students). Consequently, very little is known about risk factors for weight stigma internalization, or its consequences for health, in the general population. Advancing current understanding of these relationships would help identify groups especially at risk of internalizing weight stigma, and would help inform policy interventions to mitigate the effects of weight stigma.
This project will use new data from a general population birth cohort survey (the Avon Longitudinal Study of Parents and Children, ALSPAC) to:
1. explore life-course and intergenerational predictors of internalized weight stigma in young adulthood
2. explore health and social consequences of internalized weight stigma in young adulthood
Methods such as multivariable regression and growth curve modelling will be used to explore associations with internalized weight stigma at age 31, measured using the Modified Weight Bias Internalization Scale (8). This scale is suitable for people across the bodyweight range and will be completed by all participants as part of a wider questionnaire in 2022.
Possible risk and protective factors could include socioeconomic factors, gender, trajectories of body weight through childhood and adolescence, current and former mental health, previous eating behaviours and eating disorder symptoms, and intergenerational factors (e.g. parents’ weight, body image and weight-related attitudes).
Possible consequences of internalized weight stigma could include aspects of physical and mental health such as body weight, depressive symptoms, and overall self-rated health.
1. Durso LE, Latner JD. Obesity. 2008. 2. Pearl RL, Puhl RM. Obesity Reviews. 2018. 3. O’Brien KS et al. Appetite. 2016. 4. Hayward LE et al. Obesity. 2018. 5. Latner JD et. al. Eating Behaviors. 2014. 6. Marshall RD et al. Frontiers in Psychology. 2020. 7. Schvey NA, White MA. Eating Behaviors. 2015. 8. Pearl RL, Puhl RM. Body image. 2014.
Dr Eleanor Sanderson (lead), Professor Kate Tilling, Professor George Davey Smith
Mendelian randomization (MR) uses genetic variants as instrumental variables to estimate the causal effect of an exposure on an outcome, free from bias due to unobserved confounding. E.g. the effect of BMI on cancer incidence.
Multivariable MR (MVMR) is an extension of MR that jointly estimates the causal effect of multiple exposures on an outcome. MVMR can be used to adjust for pleiotropic effects, where genetic variants are associated with multiple traits biasing MR estimates.
Many exposures, such as BMI, vary across an individual’s lifetime, however genetic variants are fixed. Recent research has focused on (1) how MR estimates should be interpreted when the exposure varies across the lifecourse (2) the degree to which it is possible to separate the causal effect of the same exposure at different stages of the lifecourse. Understanding the effect of an exposure across the lifecourse would identify windows for prevention or treatment, and shed light on the aetiology of disease.
The aim of this project is to conduct research on the estimation and interpretation of different MR methods with time varying exposures. There will be a particular focus on the estimation of the causal effects of multiple exposures that vary over time through MVMR. MVMR can be used to estimate the proportion of an effect that is mediated through other exposures and this project will also consider how MVMR for mediation can be interpreted when both the exposure and mediators vary over time.
The focus of this project is on developing and applying methods for MVMR. The project will use simulation analysis to understand the methods and interpretation considered. This project will also involve the analysis of individual and summary level data for MR and MVMR estimation to illustrate the results obtained.
The project is methodological in focus, and the student will have the opportunity to develop an application of the methods considered. This could be on a topic that is relevant to their own applied research interests, or can be developed in consultation with the supervisors.
Professor Andrew Dowsey (lead), Dr Brian Sullivan, Professor Alastair Hay, alastair.hay@bristol.ac.uk
Urinary catheters are a widely used and sometimes may lead to urinary tract infections (UTIs) and progression to urosepsis, a serious bacterial infection of the blood stream. Our goal is to understand what modifiable community factors are associated with increased risk of a UTI and urosepsis in individuals using catheters, with a focus on reducing antimicrobial resistance (AMR).
Catheters are commonly used across age groups, for many reasons including neurological problems, congenital conditions, or cancer, and may increase the risk of UTIs/Urosepsis. Using the Bristol region NHS systemwide dataset, we want to identify modifiable risk factors to help patient outcomes and improve clinical decision making.
Our dataset is novel containing anonymised patient level data on UTI history, comorbidities and other factors. The data will require curation and management. Once prepared, the dataset will be visualised (population plots and individual timelines) and statistically modelled using Bayesian logistic regression to generate odds ratios for each risk factor.
Professor Andrew Dowsey (lead), Dr Brian Sullivan, Elizabeth Beech
The Bristol region has experienced a recent increase in cases of Clostridioides difficile infection (CDI). CDI can cause diarrhoea, abdominal pain, and fever, with severe disease occasionally requiring surgery or causing death. We seek to identify risk factors for this increase as well as model historical and fluctuations in CDI.
While there a several known risk factors for CDI, it not clear what underlies the increase in the Bristol region. Using the Bristol region NHS systemwide dataset we have the ability to statistically model many potentially influential variables (e.g. history of bacterial infections, gastrointestinal comorbidities) over time.
Our dataset is novel containing anonymised patient level data on CDI history, comorbidities and other factors. The data will require curation and management. The dataset will be visualised (population plots and individual timelines) and statistically modelled using Bayesian logistic regression and Bayesian timeseries modelling.
Dr Duleeka Knipe (lead), TBC,
Suicide bereavement is a strong predictor of suicidal behaviour, with evidence that paternal suicide (especially maternal) increases risk of suicide and self-harm behaviour in offspring. There is evidence that suggests that the time since suicide bereavement also impacts on suicide and self-harming behaviour in the individual bereaved.
This evidence primarily originates from high income countries. Roughly 80% of all suicide and self-harm occurs in low- and middle-income countries (LMICs), but only 15% of research evidence originates from these settings. Given the difference in suicide and self-harm rates between countries, and the important contextual and cultural differences (including responses to suicidal behaviour), research evidence generated from high income countries may not be applicable in LMICs . Additionally in contexts, like Sri Lanka, where extended family networks are stronger, the impact of suicidal behaviour in other members of family (and community) may increase suicide or self-harm risk in the individual exposed to that behaviour.
This project will investigate the influence of exposure to self-harm behaviour and suicide in others (family and community) on subsequent suicide and self-harm risk using an established cohort dataset in Sri Lanka.
It will answer the following questions:
1) What is the risk of suicide and self-harm in individuals who are exposed to these behaviours in family and community members?
2) Does the risk vary by the sex of the individual, the kinship relationship, and the timing of the exposure (especially in young people)?
The project involves statistical analysis of data from the locked boxed trial in Sri Lanka. This will involve developing skills in survival and spatial analysis, as well as logistic regression techniques.
Pearson M, Metcalfe C, Jayamanne S, et al. Effectiveness of household lockable pesticide storage to reduce pesticide self-poisoning in rural Asia: a community-based, cluster-randomised controlled trial. Lancet 2017
Pitman A, Osborn D, King M, Erlangsen A. Effects of suicide bereavement on mental health and suicide risk. Lancet Psychiatry 2014; 1(1): 86-94.
Professor Alastair Hay (lead), Dr Ashley Hammond,
Acute otitis media (AOM) is a common, painful condition of childhood, often treated inappropriately with antibiotics.
Optical coherence tomography (OCT), considered the optical analogue of ultrasound imaging, uses a low-intensity light source to produce real-time structural images with micron-scale resolution. This technology is coupled with high-resolution digital otoscopy imaging in the OtoSightTM Middle Ear Scope. The novel OCT images produced by the reflected near-infrared light are analysed and can be used to objectively differentiate air from middle ear fluid, as well as characterize fluid properties due to scattering of the imaging signal from particulates in the fluid. Its use within this feasibility study has two purposes: (i) confirmation of the presence of middle ear fluid (and therefore the AOM diagnosis); and (ii) daily serial imaging to visualize and chronicle the natural history of middle ear fluid resolution in a primary care population of children with suspected AOM. The latter is of particular interest because it could help understand whether the mechanism by which anaesthetic/glycerol drops work is to reduce the quantity, and therefore pressure exerted by, middle ear fluid.
Serial daily microbiological evaluation also has two purposes: (i) recording the possible cause of the infection; and (ii) describing the quantitative natural history of upper respiratory tract microbiota in a primary care population of children with diagnosed AOM.
The aim of this PhD is to establish the feasibility and acceptability of daily OCT-otoscopy and microbiology in children who have presented to primary care with suspected AOM. Specifically, the objectives are:
1. Systematically review the literature for evidence regarding the use if serial OCT and microbiology
2. Conduct a prospective cohort study to
2.1 To establish parental acceptability of the study design (consent rate)
2.2 To record child and parental acceptability of daily serial middle ear fluid measures (OCT-otoscopy) and microbiology (nasal swabs)
2.3 To explore the association between quantitative OCT image signal changes for any middle ear contents present and ear pain between presentation and symptom resolution
2.4 To explore quantitative changes in bacterial/viral load and ear pain between presentation and symptom resolution.
Using methods employed successfully in the past, we propose to assemble a cohort of children with AOM who have recently presented to primary care.
For the mixed methods study, recruitment and follow up will be according to the following steps:
1. GP/practice nurse requests parental permission to notify study centre of potentially eligible children (following face-to-face or ‘COVID secure’ telephone assessment)
2. Student contacts parent to confirm interest in, and eligibility for, the study
3. E-consent taken
4. Student visits parent/child (at place of parent’s choosing – usually the family home)
5. Student confirms consent and willingness of parent to receive daily visits
6. Student conducts bilateral OCT-otoscopy and takes nasal swab, and advises parent in the use of the daily symptom diary (validated, and as used in previous studies)
7. Student visits daily until symptom resolution (usually no more than eight days)
8. Student administers brief questionnaire investigating parental acceptability of study procedures.
1. Venekamp RP, Schilder AGM, van den Heuvel M, Hay AD. Acute otitis media in children. BMJ. 2020;371:m4238.
2. Hay AD, Downing H, Francis NA, Young GJ, Clement C, Harris SD, et al. Anaesthetic–analgesic ear drops to reduce antibiotic consumption in children with acute otitis media: the CEDAR RCT. Health Technology Assessment. 2019;23:34.
3. Preciado D, Nolan RM, Joshi R, Krakovsky GM, Zhang A, Pudik NA, et al. Otitis Media Middle Ear Effusion Identification and Characterization Using an Optical Coherence Tomography Otoscope. Otolaryngol Head Neck Surg. 2020;162(3):367-74.
4. Monroy GL, Pande P, Shelton RL, Nolan RM, Spillman DR, Porter RG, et al. Non-invasive optical assessment of viscosity of middle ear effusions in otitis media. J Biophotonics. 2017;10(3):394-403.
5. Monroy GL, Shelton RL, Nolan RM, Nguyen CT, Novak MA, Hill MC, et al. Noninvasive depth-resolved optical measurements of the tympanic membrane and middle ear for differentiating otitis media. Laryngoscope. 2015;125(8):E276-82.
6. Hay AD, Redmond NM, Turnbull S, Christensen H, Thornton H, Little P, et al. Development and internal validation of a clinical rule to improve antibiotic use in children presenting to primary care with acute respiratory tract infection and cough: a prognostic cohort study. Lancet Respir Med. 2016;4:902–10.
7. Hay AD, Sterne JA, Hood K, Little P, Delaney B, Hollingworth W, et al. Improving the Diagnosis and Treatment of Urinary Tract Infection in Young Children in Primary Care: Results from the DUTY Prospective Diagnostic Cohort Study. Ann Fam Med. 2016;14(4):325-36.
8. Watson L, Little P, Moore M, Warner G, Williamson I. Validation study of a diary for use in acute lower respiratory tract infection. Family Practice. 2001;18(0263-2136):553-4.
9. Thompson M, Vodicka T, Cohen H, Blair P, Biuckley T, Heneghan C, et al. Duration of symptoms of respiratory tract infections in children: systematic review. British Medical Journal. 2013.
Professor Matthew Ridd (lead), Dr Sarah Sullivan, Dr Ketaki Bhate
Acne is an inflammatory skin disorder comprising papules/pustules, comedones, hyper-pigmentation and scarring. Almost all teenagers are affected to some degree, with 20% being moderately-to-severely affected. There is accompanying psychosocial morbidity and the physical impairment/disfigurement caused by hyper-pigmentation or scarring can be permanent. Attendance in both primary and secondary care consume considerable NHS resources. However, there little is published on natural history and conflicting evidence surrounding the relationships between acne and diet, psychological-stress and obesity.
Further research is needed to better understand both risk factors for the development and persistence of acne; and the psychological consequences of having acne. This work could provide evidence-leading to healthcare improvements and better understanding of the link between acne and mental health in adolescence, which is a vulnerable period for mental health disorders.
This study has three aims:
1. To investigate how common acne is
2. To investigate risk factors for acne onset and persistence
3. To study the psychosocial consequences of having acne
Using data from the Avon Longitudinal Study of Parents and Children (ALSPAC) birth cohort, first the prevalence, incidence-rate and cumulative-incidence of acne between will be estimated using data from study clinics where acne was examined in detail by trained healthcare professionals. The sex, age, ethnic-group and socioeconomic distributions of young people with acne according to disease severity and comparing with those who do not have acne will be described. Persistence of acne across examinations at ages 9-13 will be described.
Data will be used to test hypotheses that: dietary factors such as dairy-rich or high glycaemic-index diets, psychological- stress and obesity early in childhood are positively associated with early onset of acne, and the progression and severity of acne; and the risk of depression, low self-esteem and time-off-school are increased in those patients who have had acne.
Bhate K, Williams H C. Epidemiology of acne vulgaris. BJD 2013; 168: 474-485.
Layton A, Eady EA, Peat M, et al. Identifying acne treatment uncertainties via a James Lind Alliance Priority Setting Partnership. BMJ Open 2015; 5: e008085 DOI: 10.1136/bmjopen-2015-008085
Prof Laura Howe (lead), Prof Abigail Fraser, Prof Alun Hughes, UCL
The response to exercise is an integrative measure of circulatory function and predicts future cardiovascular disease. Lifetime maximum aerobic capacity (VO2peak) is established by age ~30 and subsequently declines. The determinants of peak circulatory capacity are not well understood. This PhD will investigate key likely influences on peak adult circulatory capacity: adiposity/physical inactivity, and adversity/disadvantage. Measurements of Avon Longitudinal Study of Parents and Children (ALSPAC) participants at age 30 are currently underway, including assessing VO2peak using cycle-ergometry/gas- exchange, concurrent cardiac output by inert gas-rebreathing, and skeletal muscle microvascular function using near-infrared spectroscopy. The availability of multiple repeated measurements means that outcomes can be related to key antecedent exposures over the whole of fetal, child and early adult life. This will provide important insights into improvement of future cardiovascular health.
The aims of this PhD are:
1. To assess the association of life course trajectories of adiposity and physical (in)activity with VO2peak.
2. To assess socioeconomic inequalities in VO2peak, and the degree to which these are explained by adiposity and physical (in)activity.
3. To assess the association between adverse childhood experiences and VO2peak, and the degree to which these are explained by adiposity and physical (in)activity.
In this PhD, you will draw on the detailed longitudinal data available in ALSPAC, to look at the life course determinants of VO2peak. In particular, you will draw on repeated measures of adiposity (including BMI and DXA-determined fat mass) and physical (in)activity (including accelerometry-based measured), and detailed data from two generations on socioeconomic position (occupation, education, financial difficulties) and adversity (child maltreatment and measures of family dysfunction).
You will have the opportunity to develop skills in advanced statistical methodology, including multilevel models and structural equation models to estimate trajectories, techniques for mediation analysis, and methods for dealing with missing data. You will join a supportive and collaborative research team, in the vibrant environment of a leading centre for epidemiology, and will have the opportunity to lead on publishing the results of your research.
Professor Stafford Lightman (lead), Dr Thomas Upton, Prof Wuge Briscoe (Department of Chemistry, University of Bristol) Prof Krasi Tsaneva-Atanasova (Mathematics, University of Exeter) Collaborators: Mr Robin Crossley (DesignWorks Windsor), Prof Colin Dayan (Cardiff University), Prof Ido Kema(University Medical Centre Groningen), Dr Suvi Ruuskanen (University of Turku, Finland)
Rhythms characterise all living things, and our physiology can be a considered as a state of continuous dynamic
equilibrium. Despite this, almost all clinical tests of human health consist of single time point measurements, which inevitably do not reflect normal and inherent daily or even hourly variation. To overcome this, we have developed a novel microdialysis-based ambulatory technology which allows 24-hour ambulatory, minimally invasive, blood free sampling (URHYTHM, www.designworks.studio/ultradian-u-rhythm, www.u-rhythm.co.uk/).
Using the technique we have successfully demonstrated the dynamics of adrenal hormones including the stress hormone cortisol in hundreds of human participants (www.ultradian.eu).
To broaden the use and impact of the technique we now wish to investigate the use of U-RHYTHM to understand dynamics of other hormones crucial to normal growth and development, in particular sex and thyroid hormones that exhibit differential effects across tissues and the lifespan.
The student will undertake a multidisciplinary programme of work to test the hypothesis that sex and thyroid hormone dynamics can be measured in subcutaneous tissue. This will involve learning and applying the technique of U-RHYTHM microdialysis, using state of the art physical chemistry methods to describe the interaction of hormones with the U-RHYTHM microdialysis system and conducting a proof-of-principle clinical trial in human participants.
The project will be based at the University of Bristol within the Labs for Integrative Neuroscience and Endocrinology and the Department of Chemistry. The student will learn techniques for the analysis and interpretation of dynamic data under supervision of the Department of Mathematics for Healthcare at the University of Exeter. The project will be supported by clinical experts at the University of Cardiff and Bristol.
Stafford Lightman, Professor of Medicine. (lead), Dr Thomas Upton, Clinical Research Fellow., Wuge Briscoe, Professor of Chemistry
Rhythms characterise all living things, and our physiology can be a considered as a state of continuous dynamic
equilibrium. Despite this, almost all clinical tests of human health consist of single time point measurements, which inevitably do not reflect normal and inherent daily or even hourly variation. To overcome this, we have developed a novel microdialysis-based ambulatory technology which allows 24-hour ambulatory, minimally invasive, blood free sampling (URHYTHM, www.designworks.studio/ultradian-u-rhythm, www.u-rhythm.co.uk/).
Using the technique we have successfully demonstrated the dynamics of adrenal hormones including the stress hormone cortisol in hundreds of human participants (www.ultradian.eu).
The project will be based at the University of Bristol within the Labs for Integrative Neuroscience and Endocrinology and the Department of Chemistry. The student will also learn techniques for the analysis and interpretation of dynamic data
This project will develop the use and impact of our novel technology to understand the importance of dynamic changes in hormone levels over the day on human health. In particular the stuent will investigate the use of U-RHYTHM to understand dynamics of the hormones crucial to normal growth and development, in particular sex and thyroid hormones that exhibit differential effects across tissues and the lifespan. The student will undertake a multidisciplinary programme of work to test the hypothesis that sex and thyroid hormone dynamics can be measured in subcutaneous tissue, and that abnormalities in these rhythms is a cause of ill health.
This will involve learning and applying:
1) The technique of U-RHYTHM microdialysis in human subjects.
2) The use of state of the art physical chemistry methods to describe the interaction of hormones with the URHYTHM microdialysis system.
3) Conducting a proof-of-principle clinical trial in human participants.
4) the use of mathematical techniques to evaluate time series measurements of dynamic changes of hormone levels over time, together with the use of artificial intelligence and machine learning
www.u-rhythm.co.uk
www.ultradian.eu
Bhake R, et al J Clin Endocrinol Metab. 2020 Apr 1;105(4):dgz002. doi: 10.1210
Bhake et al J Clin Endocrinol Metab. 2019 Dec 1;104(12):5935-5947. doi: 10.1210
Russell GM, et al Clin Endocrinol (Oxf). 2014 Aug;81(2):289-93. doi: 10.1111
Bhake RC, et al J Med Eng Technol. 2013 Apr;37(3):180-4. doi: 10.3109
Kalafatakis et al Proc Natl Acad Sci U S A. 2018 Apr 24;115(17):E4091-E4100.
Dr Lavinia Paternoster (lead), Prof George Davey Smith, Dr April Hartley
Typical epidemiological studies aim to identify causal risk factors for onset of disease by comparing diseased cases with disease free controls. However, if the aim is to identify effective treatments for use after the onset of disease, then factors that explain why some patients progress through disease stages quickly whilst others recover or progress more slowly may be the more pertinent comparison.
To identify genetic and non-genetic causal risk factors for disease progression that may be useful for identification of new treatments
You will use existing in-house datasets for a disease of your interest.
You will use state-of-the-art methods developed in the Integrative Epidemiology Unit to identify factors associated with progression of disease, taking account of collider/selection bias that can be introduced in such analyses.
You will use Mendelian Randomization to determine if relationships observed are causal.
doi.org/10.1371/journal.pgen.1006944
Prof Nicky Welton (lead), Dr. David Phillippo,
Network meta-analysis (NMA) is a method to pool published summary treatment effects from randomised controlled trials (RCTs) to obtain estimates of relative treatment effects between multiple treatments. NMA is routinely used to inform decisions as to which treatments are effective or cost-effective, but requires that the RCT evidence forms a connected network of comparisons (the map of comparisons made in RCTs is a connected network). Covariates such as age, biomarker status, or disease severity can be classified into (i) those that interact with relative treatment effects (Effect Modifiers), and (ii) those that predict outcomes but don’t interact with treatment effects (Prognostic Factors). NMA assumes that, if effect modifiers are present, their distribution is the same, or similar, in all the included trials. However, this may not hold. Recently a multi-level network meta-regression (ML-NMR) method has been developed that relaxes this assumption, as long as individual patient data is available from one or more RCTs. ML-NMR fits a model for individual-level treatment effects in studies where there is individual patient data, and then for each of the studies with aggregate-level data integrates the individual-level likelihood over the joint distribution of effect modifiers in each study to obtain an aggregate-level likelihood. Estimates can then be obtained in any population of interest (for example, the population represented by one of the included studies, or the UK population) by integrating over the relevant joint distribution of effect modifiers. To date, the method has only been developed for the case where networks of evidence are connected (there is a path of RCT evidence joining any two treatments in the network). However, it is becoming more common that health care policy makers are confronted with disconnected networks of evidence which may include single-arm (non-randomised) studies. Population adjustment with disconnected networks of evidence requires that not only the effect modifying covariates are accounted for, but also all prognostic factors since absolute outcomes rather than relative effects are modelled.
The aim of this project is to extend the multi-level network meta-regression (ML-NMR) method for disconnected networks of evidence for a range of likelihoods, including likelihoods for survival outcomes, and to explore methods to assess the validity of the ML-NMR method in the context of disconnected networks of evidence.
Multi-level network meta-regression (ML-NMR) works by using copulae to approximate the joint distribution of effect modifiers and quasi-Monte Carlo integration to obtain the aggregate-level likelihood. This project will include: (i) formulating extensions to the ML-NMR model for modelling absolute outcomes alongside relative effects, without introducing bias into the estimated relative treatment effects (ii) assessing the performance and properties of such approaches in a simulation study (iii) developing in-sample methods for assessing the validity of assumptions, such as cross-validation to estimate the proportion of variation explained by the model so that any unexplained variation will be largely due to missing prognostic variables or effect modifiers that have not been accounted for; and (iv) developing out-of-sample methods that aim to estimate prediction error by identifying external studies in a given population and comparing the absolute outcomes predicted by ML-NMR in this external population with the observed outcomes. The project work will be conducted using the STAN package for Bayesian statistics within R, and the methods developed will be incorporated into an existing package for ML-NMR.
Phillippo DM, Ades AE, Dias S, Palmer S, Abrams K, Welton NJ. Methods for population-adjusted indirect comparisons in health technology appraisal. Medical Decision Making. 2018. 38:200-211.
Phillippo DM, Dias S, Ades AE, Belger M, Brnabic A, Schacht A, Saure D, Kadziola S, Welton NJ. Multilevel Network Meta-Regression for population-adjusted treatment comparisons. JRSSA 2020. https://doi.org/10.1111/rssa.12579.
Phillippo DM, Dias S, Ades AE, Welton NJ. Assessing the performance of population adjustment methods for anchored indirect comparisons: A simulation study. Statistics in Medicine. 2020 https://doi.org/10.1002/sim.8759
Dias S, Ades AE, Welton NJ, Jansen JP, Sutton AJ. Network Meta-analysis for Comparative Effectiveness Research. Wiley. Hoboken NJ. 2018.
Prof Nicky Welton (lead), Dr. David Phillippo,
In the UK, decisions as to which medical treatments and interventions to make available on the NHS take into consideration both the costs and benefits of the treatment over a patients lifetime. This is usually based on a cost-utility analysis which identifies the treatment with the highest expected net benefit, based upon a cost-effectiveness model. Cost-effectiveness models are typically non-linear functions of the model input parameters, such as a Markov models which track patient movements between health states over time. Key inputs to cost-effectiveness models are estimates of the relative impact of different treatments on transitions between health-states. It’s common for there to be multiple treatment options that policy makers need to compare (the decision set). Randomised controlled trials (RCTs) provide the most robust evidence of the relative efficacy of different treatment options. However, when there are multiple treatment options it is unlikely that a single RCT has compared all treatment options, and instead there are multiple RCTs that have each compared different sets of treatments within our decision set (and can even include additional treatments outside of the decision set). When the RCT evidence forms a network of treatment comparisons, then the evidence can be pooled in a network meta-analysis (NMA) (Dias et al. 2018), to obtain pooled relative treatment effect estimates that can be used as inputs to a cost-effectiveness model. The validity of NMA relies on the consistency assumption, where direct estimates from RCTs that have compared two treatments (eg AvB) head-to-head are not in conflict with indirect estimates obtained from the rest of the network (eg AvC and BvC). One reason why there may be inconsistency is due to a lack of methodological rigour in some of the included RCTs, leading to biased estimates. Studies included in an NMA are assessed for risk of bias, however this provides no indication of the impact of potential bias on a decision based on the NMA. For example, the AvB study may overestimate the effect of B against A, but if C clearly has the highest expected net benefit, then adjusting for the bias in the AvB study is unlikely to have any impact on the treatment recommendation. A threshold analysis method has previously been developed to address this question when the net benefit function is a simple linear function of a single measure of treatment efficacy (Phillippo et al 2018).
The aim of this project is to develop algorithms to assess the impact of bias in relative treatment effect inputs to cost-effectiveness models, where the relative estimates have come from a network meta-analysis (NMA), accounting for (i) non linear relationships between the NMA estimates and the net benefit function and (ii) the complex relationship between RCT evidence and resulting NMA estimates via the hierarchical NMA model. The methods will answer the question: “how biased would the RCT evidence have to be before a rational risk neutral decision maker changed its recommendations?”
This project will extend the work of Phillippo et al 2018 to decisions based on non linear relationships between the NMA estimates and the net benefit function. The project will also consider alternative decision rules (such as recommend all treatments within a minimally clinically significant margin), and decisions that are based on multiple outcomes via a multi-criteria decision analysis (MCDA). The project will draw on recent methodological developments for value of information analysis (Heath et al. 2018), such as generalised additive models, integrated nested Laplace approximations, and multi-level Monte-Carlo methods. These meta-modelling techniques have the potential to obtain fast and accurate computational tools that can evaluate the thresholds of input parameters for which decisions change, based on maximising any kind of net benefit function.
An important output of the project will be the creation of software tools (using R and R Shiny) to allow easy application for health economists without advanced statistical training. The methods will be applied to examples from the National Institute of Health and Care Excellence (NICE) guidelines. Case studies may include social anxiety, non-small cell lung-cancer, headaches, and atrial fibrillation.
Phillippo DM, Dias S, Ades AE, Didelez V, Welton NJ. Sensitivity of treatment recommendations to bias in network meta-analysis. JRSSA. 2018. 181:843-867. https://doi.org/10.1111/rssa.12341
Heath A, Manolopoulou I, Baio G. A Review of Methods for Analysis of the Expected Value of Information. Medical Decision Making. 2017. 37: 747-758 https://doi.org/10.1177/0272989X17697692
Dias S, Ades AE, Welton NJ, Jansen JP, Sutton AJ. Network Meta-analysis for Comparative Effectiveness Research. Wiley. Hoboken NJ. 2018.
Dr Gibran Hemani (lead), Dr Josephine Walker,
Recent work has shown a massive bias in terms of the ancestries who are represented in genetic studies (https://gwasdiversitymonitor.com/). This has known major consequences, in that the results from genetic studies conducted in one ancestry may not transport to other ancestries, and this could lead to widening national and global health inequalities. What is not currently known is how this impacts the choice of disease areas that are studied. National and global health priorities are published by the World Health Organisation as well as annual trends in disease burden. Is the genetics community conducting research that tracks with global health needs?
In this project we pose the following questions:
1. Are the health issues that have formed the focus of genome-wide association studies (GWAS) representative of the health issues that are in most need of attention at a national or a global level?
2. Relatedly, we will ask which population demographics have been best served by the GWASs that have been performed to date.
We will use the NHGRI-EBI GWAS catalog (https://www.ebi.ac.uk/gwas/) to calculate the priority that GWAS has given to different health-related phenotypes, in terms of number of studies and number of samples. We will then contrast these trends against global disease burden (https://www.healthdata.org/gbd/2019).
Following on from this we will stratify by nation, to identify if for example genetic studies conducted or led by UK institutes track against UK disease burden.
Finally we will analyse whether the focus of GWAS is more closely associated with the disease burden experienced by particular demographics within the UK (eg. age, gender, socio-economic status, region, ethnicity).
Mills, M.C and Rahal, C., (2020). 'The GWAS Diversity Monitor Tracks diversity by disease in real time'. Nature Genetics, 52, 242-243. doi: 10.1038/s41588-020-0580-y
The Global Burden of Disease Study 2019. The Lancet. https://www.thelancet.com/journals/lancet/issue/vol396no10258/PIIS0140-6736(20)X0042-0
Buniello A et al. The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019. Nucleic Acids Research, 2019, Vol. 47 (Database issue): D1005-D1012.
Ali et al. Ethnic disparities in the major causes of mortality and their risk factors – a rapid review. https://www.gov.uk/government/publications/the-report-of-the-commission-on-race-and-ethnic-disparities-supporting-research/ethnic-disparities-in-the-major-causes-of-mortality-and-their-risk-factors-by-dr-raghib-ali-et-al#leading-causes-of-mortality
Márquez-Luna C et al. Multiethnic polygenic risk scores improve risk prediction in diverse populations. Genet Epidemiol. 2017;41(8):811-823
Prof Sarah Lewis (lead), Dr Miguel Renteria, Dr Hannah Jones
Sleep has long been known to play a role in physical and mental wellbeing and sleep disruption has been observed to co-occur with a number of health conditions including: depression, chronic pain, diabetes, chronic kidney disease, ADHD, impulsivity, anxiety, and bipolar disorder (1). Recent large genome wide association studies have uncovered genetic correlations between sleep traits and health conditions. For instance, there is evidence of genetic correlation at the whole-genome level between snoring and schizophrenia( 2), and between sleep apnea and and self-harm. However, the precise molecular pathways, causal relationships and mechanisms are not clear. Does sleep disturbance cause health problems or do health problems lead to disrupted sleep or are the two linked by shared biological pathways early in development?
Mendelian randomization and state-of-the-art statistical genetics methods will be used to disentangle the genetic relationships between these complex traits.
This project will investigate the causal pathways between sleep traits and physical and mental health using publicly available genome wide association study data, in order to identify risk factors for sleep apnea and downstream effects of sleep apnea on health and wellbeing.
During this project the student will:
1. Conduct a systematic review of the published literature to determine the strength of the evidence for the association between sleep and mental health traits of interest.
2. Refine genetic instruments for snoring, sleep apnea and mental health traits.
3. Conduct Mendelian randomization analyses of the effect of snoring and sleep apnea on a series of mental health outcomes.
4. Conduct sensitivity analyses to test whether the results could be influenced by the underlying assumptions of Mendelian randomization analyses.
5. Use analytical methods such as multi-trait analysis of genome wide association studies (GWAS), genomic structural equation modelling or co-localisation analysis and functional annotation to characterize the shared molecular pathways between sleep and physical and mental health problems.
1. Tarokh L et al. Sleep in adolescence: Physiology, cognition and mental health. Neurosci Biobehav Rev. 2016;70:182-188.
2. Campos AI et al. Insights into the aetiology of snoring from observational and genetic investigations in the UK Biobank. Nat Commun. 2020 Feb 14;11(1):817.
3. Davey Smith G, Ebrahim S. 'Mendelian randomization': can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol. 2003;32:1-22.
Professor Sarah Lewis (lead), Dr Evie Stergiakouli, Professor Heather Cordell (University of Newcastle).
Orofacial clefts (OFCs) are the most prevalent craniofacial congenital disabilities globally, occurring at a frequency of 1 out of 700 newborns. Children born with a cleft usually require more than one surgery and ongoing support for speech, psychological wellbeing and educational need. Thus individuals and families affected by OFCs suffer from a significant burden on physical and psychological health, socioeconomic well-being, and quality of life).
Generally, OFCs can be divided into two main types with different etiologies, cleft palate only (CPO) and cleft lip or without cleft palate (CL/P). Most OFCs are isolated non-syndromic (nsOFCs), while OFCs with other clinical signs are classified as syndromic. The syndromic form is often related to genetic mutations and chromosomal aberrations, but the occurrence of nsOFCs is complex and involves genetic susceptibility, environmental factors (e.g., smoking, alcohol drinking and dietary factors), and gene-environment interactions. Thus, a better understanding of the genetic architecture of nsOFCs may contribute to identifying populations at high risk, providing prevention strategies, and reducing the incidence of the anomaly. Previous genome-wide association studies (GWAS) have identified over 40 loci associated with nsOFCs risk(1), but most studies have focused on the overall nsCL/P and have not had sufficient power to study subtypes. In addition previous studies have not been able to consider the contribution of parental genotypes to cleft and the interaction between child and parent genotypes.
To carry-out trio analysis of genome wide association study data to investigate the genetic architecture of nsOFCs of European ancestry and identify genetic variants that are associated with its subtypes, CLO, CPO, CLP.
To study the contribution of parental genotype to cleft, to determine whether interactions between parent and child genotypes contribute to cleft and to investigate parent of origin effects.
The current study will conduct a genome wide association study using already genotyped data from case-parent trios of European ancestry recruited from the Cleft Collective and the Bonn Cohort using a transmission disequilibrium test design (2) to identify the new common variants that are associated with nsOFCs and its subtypes. The study will carry-out statistical analyses using the trio data to determine whether interactions between parent and child genotypes contribution to cleft phenotypes and investigate parent of origin effects for those genetic variants which are implicated as risk factors.
1. Yu Y, Zuo X, He M, Gao J, Fu Y, Qin C, Meng L, Wang W, Song Y, Cheng Y, et al. Genome-wide analyses of non-syndromic cleft lip with palate identify 14 novel loci and genetic heterogeneity. Nat Commun 2017;8:14364. doi: 10.1038/ncomms14364.
2. Ruiz-Narvaez EA, Campos H. Transmission disequilibrium test (TDT) for case-control studies. Eur J Hum Genet 2004;12(2):105-14. doi: 10.1038/sj.ejhg.5201099
Professor Jenny Donovan (lead), Dr Jelena Savovic,
Problems with recruitment are always identified as a critical challenge to the conduct and completion of RCTs. Many RCTs require expensive extensions, end without sufficient power, or close prematurely. Over 100 SWATs (randomised studies within trials) have tested simple interventions to improve recruitment and findings need to be systematically reviewed
(a) identify published SWATs; (b) categorise published SWATs into types of recruitment improvement interventions; (c) extract data relating to recruitment interventions and outcomes; (d) identify robust strategies to improve recruitment as evaluated in SWATs (and ineffective strategies); (e) produce a meta-analysis or systematic review of outcomes for publication.
Systematic searching and review of published SWATs, with evidence synthesis and the possibility of meta-analysis.
Professor Carol Joinson (lead), Dr Jon Heron,
Incontinence and constipation are common paediatric health problems, and most cases arise from functional impairments in the bladder and/or bowel. Chronic constipation is a cause of both urinary and faecal incontinence in children. Our recent research found that pre-school children who were exposed to maternal anxiety and depression have an increased risk of incontinence and constipation at age 7. However, the causal pathways that underlie this relationship are yet to be identified. Maternal psychopathology is linked to maladaptive parenting and offspring emotional/behavioural problems, both of which are associated with an increased risk of incontinence and constipation in children. Negative parenting behaviours around the time of potty training could disrupt the process of learning bladder and bowel control, which could lead to children developing toilet anxiety and stool withholding/constipation. Exposure to maternal psychopathology in the early years has also been linked to fussy eating in children, which could increase the risk of constipation.
This project will examine mediators of the relationship between exposure to maternal psychopathology in the early years and offspring incontinence and constipation at school age
The project will use ALSPAC data on maternal postnatal anxiety and depression, childhood incontinence, and a range of mediators. There is the possibility of using linked health data on childhood incontinence and constipation. Data are available on a range of possible mediators including maladaptive parenting, children’s behaviour/emotional problems, dietary factors (e.g. fussy eating), age at initiation of toilet training, and toilet anxiety.
The analysis will use causal mediation methods based on counterfactual theory, to examine mechanisms underlying the relationship between maternal psychopathology and childhood incontinence and constipation. These methods provide a powerful set of novel and innovative techniques for understanding causal pathways and make explicit assumptions about confounders, including intermediate confounding (i.e. mediators affected by the exposure); permit consideration of exposure-mediator interactions, and enable examination of effects of altering the mediator-status. Confounders will include a range of socioeconomic factors, indices of family discord, stressful life events, difficult temperament in early childhood, and delayed child development.
Joinson C, Grzeda MT, von Gontard A, Heron J. A prospective cohort study of biopsychosocial factors associated with childhood urinary incontinence. Eur Child Adolesc Psychiatry. 2019;28:123-130.
Joinson C, Grzeda MT, von Gontard A, Heron J. Psychosocial risks for constipation and soiling in primary school children. Eur Child Adolesc Psychiatry. 2019; 28: 203–210.
Tyler J VanderWeele. Mediation Analysis: A Practitioner's Guide. Rev Public Health. 2016;37:17-32.
Professor Carol Joinson (lead), Professor Abigail Fraser,
Urinary incontinence (UI) affects around 423 million adults (≥ 20 years old) worldwide and is estimated to be around three times more common in women than in men. The comorbidity between UI and affective disorders including depression and anxiety is well-established, but the precise nature of this relationship is unknown. It is often assumed that the elevated rates of mental health problems are due to the adverse consequences of incontinence on the daily lives of affected people. An alternative explanation, that mental health problems are a cause of incontinence, is highly plausible, but is insufficiently studied. Much of the existing research examining relationships between UI and mental health problems is cross-sectional and is, therefore, unable to disentangle whether poor mental health is a cause or a consequence of incontinence. A small number of prospective studies have found evidence for bidirectional associations between depression/anxiety and UI, but residual confounding is a problem.
The aim of the project is to examine whether depression and anxiety are causes or consequences of UI or whether there is a true bidirectional causal effect.
This project will use bidirectional two-sample Mendelian randomisation (MR) analysis to estimate causal effects of depression and anxiety on UI and to test if the effects are bidirectional. Two-sample MR analysis uses genetic variants that are robustly associated with the exposure as instrumental variables (IVs) to test causal effects. By using MR analysis, the present study will improve causal inference because, at a population level, genetic variants should not be associated with genetic or environmental confounding factors that can distort observational studies. The project will use summary statistics from genome wide association studies (GWAS) of depression, anxiety and UI. The analysis will conducted using the TwoSampleMR package in R.
Felde G et al. Anxiety and depression associated with urinary incontinence. A 10-year follow-up study from the Norwegian HUNT study (EPINCONT). Neurourol Urodyn. 2017;36(2):322-328.
Hartwig FP et al. Two-sample Mendelian randomization: avoiding the downsides of a powerful, widely applicable but potentially fallible technique. Int J Epidemiol. 2016;45(6):1717-1726.
Dr Emma Anderson (lead), Dr Neil Davies,
There is evidence that lower levels of cognitive function in midlife are associated with higher risk of Alzheimer’s disease from both observational and Mendelian Randomization (MR) studies1 2. There is also evidence that educational attainment protects against Alzheimer’s disease risk, and multivariable MR suggests that this is potentially because of the positive effect that educational attainment has on later cognitive function1. This offers support to the ‘cognitive reserve hypothesis’3, which is thought to enable people to compensate for greater levels of neurodegeneration before experiencing symptoms. Understanding whether educational attainment might offer some protection against developing Alzheimer’s disease (or delaying the age at onset), specifically in people at higher risk of Alzheimer’s disease (e.g. carriers of the APOE e4 genotype), has the potential to inform public health prevention strategies.
This project aims to understand whether educational attainment can reduce risk of Alzheimer’s disease by improving cognitive function in midlife, in those who are at high risk (e.g. APOE ε4 carriers).
The student will use data from the UK Biobank; one of the world’s largest population health databases of over 500,000 people. The student will learn how to generate polygenic risk scores for Alzheimer’s disease and educational attainment, using summary data from the latest genome-wide association studies (GWAS). The association between genetic risk for Alzheimer’s disease (high vs middle/low polygenic risk score for AD) and two outcomes will be examined: cognitive function in midlife and incident AD. An interaction effect with the polygenic risk score for educational attainment will then be added, to interrogate whether educational attainment alters the effect of AD genetic risk on midlife cognitive function or risk of AD.
1. Anderson EL, Howe LD, Wade KH, et al. Education, intelligence and Alzheimer's disease: evidence from a multivariable two-sample Mendelian randomization study. Int J Epidemiol 2020;49(4):1163-72. doi: 10.1093/ije/dyz280 [published Online First: 2020/02/01]
2. Novotny JS, Gonzalez-Rivas JP, Medina-Inojosa JR, et al. Investigating cognition in midlife. Alzheimers Dement (N Y) 2021;7(1):e12234. doi: 10.1002/trc2.12234 [published Online First: 2022/01/11]
3. Stern Y. Cognitive reserve in ageing and Alzheimer's disease. Lancet Neurol 2012;11(11):1006-12. doi: 10.1016/S1474-4422(12)70191-6
Emma Anderson (lead), Kaitlin.wade@bristol.ac.uk,
Dementia is the leading cause of death in England and Wales, and it is the only leading cause of death globally without any treatments or a cure. The immune system has long been considered as an important contributor to the development of dementia. However current research is hindered by the fact that dementia itself (especially preclinical dementia, i.e., before a person develops symptoms), can affect our immune function. The human gut microbiome is integral to the development of the innate and adaptive immune system, and thus may be an intervenable risk factor for dementia. Many of the current studies examining the effect of the microbiome of dementia risk have either involved animal models or small observational studies of cross-sectional or case control designs[1,2]. Therefore, limitations such as confounding, reverse causation make inferring causality between the gut microbiome and dementia difficult.
There has been a growth in genome-wide association studies (GWASs) characterising the human genetic contribution to the gut microbiome, most recently with the MiBioGen consortium[3,4]. Whilst the host genetic influence on the gut microbiome is likely outweighed by that of the environment, these GWAS results allow the opportunity to apply Mendelian randomization (MR) analyses to support or challenge the causal role of the gut microbiome on human health and disease[5].
1. Perform a review of the current literature focusing on the relationship between the gut microbiome and dementia and cognitive function.
2. Appraise causality in relationships observed in the literature, using two-sample MR analyses.
3. Perform sensitivity analyses assessing the validity of MR assumptions (namely, the “exclusion restriction” criteria describing horizontal pleiotropy).
4. Write in the required format for the target journal.
To appraise causality in relationships between the gut microbiome, dementia & cognitive function, two-sample MR analyses will be conducted using data from the most recent GWASs of the gut microbiome, Alzheimer’s disease (ref) and fluid intelligence (a marker of cognitive function) (ref). Genetic variants associated with the gut microbiome will be obtained from the GWAS of the Flemish Gut Flora Project (FGFP), Food Chain Plus (Focus) and PopGen (N=3,980) and the MiBioGen consortium (N=18,340) and used as instrumental variables in MR analyses [3,4]. Briefly, these studies assessed the host genetic contribution to microbiome phylogenetic profiles that were characterised with gene amplicon sequencing of hypervariable regions in the 16S ribosomal RNA. Harmonization of the summary-level data and the two-sample MR will be completed with MR-Base functionality.
1) Hill JM, et al. Front Aging Neurosci 2014; 6:127.
2) Saji N. et al. Sci Rep 2020;10(1):8088.
3) Hughes DA, et al. Nature Microbiology 2020; 5: 1079-1087.
4) Kurilshikov A, et al. Nature Genetics 2021; 53; 156-165.
5) Davey Smith G & Hemani G. Human Molecular Genetics 2014; 23: R89-R98.
Dr Sarah Watkins (lead), Dr Matthew Suderman,
For a number of years governments around Europe have discussed using molecular biomarkers as a potential tool to confirm the age of some individuals seeking asylum. This is now a measure being actively considered by the UK government. Although there have been a number of commentaries published addressing issues with the accuracy of such biomarkers, including DNA methylation age, little attention has been paid to how social, psychological, and other external factors associated with individuals needing to seek asylum might impact these biomarkers and therefore bias age estimates.
This project will review the literature to determine how external factors might impact age estimates of individuals seeking asylum obtained from molecular biomarkers. It could involve a simulation to illustrate how good estimates would need to be before they could be reliably used to estimate the age of individuals.
The first step will be to identify from the literature the most important exposures and experiences that impact individuals seeking asylum, particularly children and young people. This includes exposures such as trauma, poor access to health services, and extreme poverty. Then the molecular biomarker literature will be reviewed to assess the current evidence (and the extent to which it exists) on the impact of the identified exposures and experiences on epigenetic age. There will be a possible meta-analysis of results, and it may be possible to conduct a simulation.
https://post.parliament.uk/research-briefings/post-pn-0666/
Sauer PJ, Nicholson A, Neubauer D; Advocacy and Ethics Group of the European Academy of Paediatrics. Age determination in asylum seekers: physicians should not be implicated. Eur J Pediatr. 2016 Mar;175(3):299-303. doi: 10.1007/s00431-015-2628-z. Epub 2015 Sep 18. PMID: 26385241.
Dupras, C., Beck, S., Rothstein, M. A., Berner, A., Saulnier, K. M., Pinkesz, M., ... & Joly, Y. (2019). Potential (mis) use of epigenetic age estimators by private companies and public agencies: human rights law should provide ethical guidance. Environmental Epigenetics, 5(3), dvz018.
Dr Sarah Watkins (lead), Dr Matthew Suderman,
Epigenetic clocks are algorithms that utilise DNA methylation (DNAm) data to predict a person’s age. Over 10 epigenetic clocks are frequently used in the literature, containing between 10 and 514 DNAm sites. As technology to measure DNAm changes, so too do the sites that are measured; this means that sites used as part of clock algorithms may be missing in some datasets. Because of this, the age predictors may be biased, but this has been inadequately assessed in previous literature, and so the robustness of the predictors to losing sites is unclear.
The aim of this project is to calculate epigenetic age with ten commonly used epigenetic clocks, in multiple publicly available datasets that have measured chronological age. We will utilise datasets measured on the 450k array (as this is typically the set of probes used to derive the predictors) and test the robustness of the age predictors by removing sites. This will include a demonstration of how clock estimates change depending on the technology used to measure DNA methylation.
Using R you will calculate the ten epigenetic clocks, and then will re-calculate them with iterations of missing sites. This will be done across multiple datasets. You will then use those iterations to build a picture of how robust the clock estimates are to missing features of the predictor. This project will introduce you to using R (if you are not already familiar with it), analysis of DNA methylation data, calculating epigenetic age, and writing up and presenting findings.
BERGSMA, T. & ROGAEVA, E. 2020. DNA Methylation Clocks and Their Predictive Capacity for Aging Phenotypes and Healthspan. Neurosci Insights, 15, 2633105520942221.
HORVATH, S. 2013. DNA methylation age of human tissues and cell types. Genome Biol, 14, R115.
Prof Celia Gregson (lead), Dr Anya Burton, Dr Hannah Wilson
Analysis and data source: Secondary quantitative - data cleaned and available
What is the source of these data? Cross-sectional study conducted in Harare in 2000-21; PI = Celia Gregson. Data already cleaned. Data dictionary prepared. Data stored on RDSF at University of Bristol.
Background: Across Africa, the scale-up of HIV treatment has dramatically improved survival, such that increasing numbers of women with chronic HIV are now reaching the menopause. However, research has seldom focussed on African woman at this stage of their lives. We need to understand morbidities in mid-life women and their impact. This cross-sectional study collected data on 400 such women living in Harare, 200 of whom were living with HIV.
Aims/Objectives:
1. To quantify differences in muscle parameters: muscle mass, strength, performance, by HIV status, and associations with self-reported falls
2. To determine potential risk factors for muscle deficits, e.g. obesity, physical inactivity, food insecurity, joint pain.
3. To determine, in women with HIV, HIV-associated risk factors for muscle deficits, e.g. CD4/viral load, ART regime/duration.
Methods: Quantitative analyses. Data have already been cleaned. Construct of hypothesised causal diagram. Descriptive statistics initially (mean [SD], median [IQR], histograms, t tests, Chi squared tests), then univariate and multivariate linear and logistic regression. Generation of tables and graphs to show results. Analyses using Stata. These analyses could lead to a conference abstract and/or published paper.
Additional learning: Background reading on HIV pandemic in Africa, ART regimes, muscle measurements.
Prof Celia Gregson (lead), Dr Anya Burton, Dr Hannah Wilson
Analysis and data source: Secondary quantitative - date are available
What is the source of these data? Cross-sectional study conducted in Harare in 2000-21; PI = Celia Gregson. Data already cleaned. Data dictionary prepared. Data stored on RDSF at University of Bristol.
Background: Populations living in sub-Saharan Africa are ageing more rapidly than in any other region globally. Hence age-related diseases are expected to rise. Furthermore, the scale-up of HIV treatment has dramatically improved survival, such that increasing numbers of women with chronic HIV are now reaching older age. However, research has seldom focussed on African woman at this stage of their lives. We need to understand morbidities in mid-life women and their impact. This cross-sectional study collected data on 400 such women living in Harare, 200 of whom were living with HIV.
Aims/objectives:
1. To determine the prevalence of self-reported arthritis, joint pain, and back pain and associations with Health-related quality of life (using EQ5D5L)
2. To quantify use of analgesia according to arthritis, joint pain, and back pain
3. To determine whether the prevalence of arthritis, joint pain, and back pain differs by HIV status.
4. To understand potential risk factors associated with arthritis, joint and back pain e.g. physical inactivity, obesity, mental health, diabetes.
Methods: Quantitative analyses. Data have already been cleaned. Construct of hypothesised causal diagram. Descriptive statistics initially (mean [SD], median [IQR], histograms, t tests, Chi squared tests), then univariate and multivariate linear and logistic regression. Generation of tables and graphs to show results. Analyses using Stata. These analyses could lead to a conference abstract and/or published paper.
Additional learning: Background reading on HIV pandemic in Africa, ART regimes, osteoarthritis and lower back pain.
Dr Zoe Reed (lead), Dr Angela Attwood, Professor Marcus Munafò
Emotion recognition is an important part of social interaction, and difficulties in recognising others’ emotions can have a negative impact on this (Denham et al., 2015; Ferretti and Papaleo, 2018). In addition, these difficulties may subsequently impact social function and mental health (Trevisan and Birmingham, 2016; Wells et al., 2021). It is also possible that difficulties in this area e.g., social anxiety, could disrupt school attendance in children (Finning et al., 2019). Previous studies which have examined these relationships tend to be small and cross-sectional. We propose using a large existing cohort with data at multiple timepoints to overcome these limitations and allow the direction of association to be examined as well. It is important to further understand these relationships particularly given that poorer emotion recognition is observed in autistic individuals, or those with more autistic traits, and therefore these relationships will be particularly relevant for autistic individuals with emotion recognition difßficulties.
This mini project will examine the bidirectional association between emotion recognition and outcomes related to social traits, wellbeing/mental health and school attendance. Data from the Avon Longitudinal Study of Parents and Children (ALSPAC) will be used for this. Data will be used from different time points to attempt to understand the direction of any associations found. This work would be useful in informing what downstream effects there may be with interventions targeting emotion recognition, for example, in autistic individuals.
The aim of this project is to examine whether associations exist between emotion recognition and social, wellbeing and school attendance outcomes and if so, what direction this may be in.
1. Gain an understanding of the current literature around emotion recognition and social, wellbeing and school outcomes (e.g., social anxiety, social skills, wellbeing and mental health, school attendance)
2. Identify relevant measures in ALSPAC
3. Conduct appropriate regression analyses to examine the relationship between emotion recognition and these outcomes and vice versa using data from different time points
4. Write up the results for a potential publication
Denham, S.A., Bassett, H.H., Brown, C., Way, E., Steed, J., 2015. “I Know How You Feel”: Preschoolers’ emotion knowledge contributes to early school success. Journal of Early Childhood Research 13, 252–262.
Ferretti, V., Papaleo, F., 2018. Understanding others: Emotion recognition in humans and other animals.
Finning, K., Ukoumunne, O.C., Ford, T., Danielson-Waters, E., Shaw, L., Romero De Jager, I., Stentiford, L., Moore, D.A., 2019. Review: The association between anxiety and poor attendance at school – a systematic review. Child Adolesc Ment Health 24, 205–216.
Trevisan, D.A., Birmingham, E., 2016. Are emotion recognition abilities related to everyday social functioning in ASD? A meta-analysis. Res Autism Spectr Disord 32, 24–42.
Wells, A.E., Hunnikin, L.M., Ash, D.P., van Goozen, S.H.M., 2021. Improving emotion recognition is associated with subsequent mental health and well-being in children with severe behavioural problems. Eur Child Adolesc Psychiatry 30, 1769.
Dr Elinor Curnow (lead), Dr Paul Madley-Dowd, Dr Rachael Hughes
Around 19,000 young people, born in the UK in 2000-02, initially agreed to take part in the Millennium Cohort Study. Over time, many cohort members have been lost to follow-up (“non-responders”). Missing data due to non-response can lead to bias in estimates of exposure-outcome associations.
The aim of this project is to assess whether non-response, and predictors of non-response, are different for male and female cohort members, and to explore the possible impact on estimates of associations that vary by sex.
• Assess and compare predictors of non-response by cohort member sex.
• Assess the extent of bias in estimates of the association between maternal education and birth weight (or another question of interest to the student) using complete case vs. weighted analysis.
https://cls.ucl.ac.uk/wp-content/uploads/2020/01/MCS7_Technical_Report.pdf
Dr Anya Skatova (lead), Dr Philip Newall,
Gambling is now seen as a public health issue in the UK, with recent government estimates suggesting a societal cost of £1.2 billion a year (Public Health England, 2021). It is generally considered that disadvantaged groups bear a disproportionate share of this burden, as supported by the observation for example that bookmakers tend to cluster in areas of relative socioeconomic deprivation (Newall, 2015; Wardle et al., 2017). However, as with much gambling research, these observations are limited by methodological factors, for example via the selection of nonrepresentative retrospective samples (Muggleton et al., 2021).
These previous observations would be backed-up by confirmation in prospective representative samples, such as via Avon Longitudinal Study of Parents And Children (ALSPAC), a dataset which has already been used in several pieces of gambling research (Emond et al., 2022; Hollén et al., 2020).
Since many gambling measures are available in the ALSPAC dataset, and has already been used in previous gambling research (Emond et al., 2022; Hollén et al., 2020), we expect that a significant amount of work can be conducted during this mini project. Specifically, we expect that a descriptive piece of work on the distribution of gambling-related harm across distinct socioeconomic groups can be conducted and written-up, and submitted to a field journal such as Journal of Gambling Studies.
Descriptive statistics, correlation and regression analysis using previously collected data within ALSPAC.
Emond, A., Nairn, A., Collard, S., & Hollén, L. (2022). Gambling by young adults in the UK during COVID-19 lockdown. Journal of Gambling Studies, 38(1), 1–13.
Hollén, L., Dörner, R., Griffiths, M. D., & Emond, A. (2020). Gambling in young adults aged 17–24 years: A population-based study. Journal of Gambling Studies, 36(3), 747–766. https://doi.org/10.1007/s10899-020-09948-z
Muggleton, N., Parpart, P., Newall, P., Leake, D., Gathergood, J., & Stewart, N. (2021). The association between gambling and financial, social, and health outcomes in big financial data. Nature Human Behaviour, 5, 319–326. https://doi.org/10.1038/s41562-020-01045-w
Newall, P. W. S. (2015). How bookies make your money. Judgment and Decision Making, 10(3), 225–231.
Public Health England. (2021). Landmark report reveals harms associated with gambling estimated to cost society at least £1.27 billion a year. GOV.UK. https://www.gov.uk/government/news/landmark-report-reveals-harms-associated-with-gambling-estimated-to-cost-society-at-least-1-27-billion-a-year
Wardle, H., Asbury, G., & Thurstain-Goodwin, M. (2017). Mapping risk to gambling problems: A spatial analysis of two regions in England. Addiction Research & Theory, Journal Article.
Dr. Paul Yousefi (lead), Professor. Richard Martin, Dr. Matthew Suderman Dr. Athene Lane Dr. Sam Merriel
Prostate Cancer (PCa) is a leading cause of male mortality, with 336,000 deaths worldwide each year (1). Although most PCa cases are indolent, slow-growing, and tend not to progress, a subset of PCa cases are more aggressive and will progress to metastases, treatment resistance and death. Aggressive cases are a major driver of PCa mortality, which is the second most frequent cause of UK male cancer deaths (2). Currently, performance of even NICE recommended post-diagnostic clinical prediction models of PCa progression (based on clinical variables and prostate biopsy) is limited, with C-index at 10 years ranging from 0.73 to 0.81. Tools that improve discrimination of PCa progression and reduce harm due to both under- and over-treatment are required to guide treatment clinically.
DNA methylation (DNAm) is an epigenetic modification that regulates tissue-specific gene expression and is disrupted in cancer development where it’s a hallmark of oncogenesis and progression pathophysiology (3). Attempts to identify DNAm differences in PCa have been underpowered and featured substantial design flaws (e.g. insufficient treatment adjustment, inappropriate follow-up window, etc.). However, blood DNAm is increasingly being appreciated as valuable for predicting cancer progression and prognosis (4), which merits further exploration with adequate sample size and sufficient methodology.
Given the role of DNAm in tumorigenesis, the potential for DNAm to capture signal of early disease progression, and the poor performance of existing biomarkers, this project aims to improve discrimination of the rate and severity of PCa progression by:
(1) Identifying genome-wide blood DNAm patterns that differ prospectively between aggressive and indolent forms of PCa
(2) Developing a DNAm signature using machine learning techniques for predicting aggressive PCa among patients with confirmed localised disease that could inform the need for radical therapeutic intervention
(3) Evaluating whether such a DNAm signature can improve upon existing NICE-recommended clinical progression methods
This project will use participants from the Prostate Testing for Cancer and Treatment (ProtecT) trial which included 1,600 UK men with confirmed, localised PCa and prospectively evaluated two forms of radical PCa treatment (prostatectomy and radiotherapy) to active monitoring. Median follow-up was 10-years and PCa mortality was the primary outcome. Using DNA isolated from whole blood samples collected at baseline, genome-wide DNAm levels will be quantified by Illumina HumanMethylationEPIC BeadChip for N = 850 (125 aggressive cases, 725 indolent controls) age and treatment matched ProtecT participants.
An Epigenome Wide Association Study (EWAS) will be performed to identify where DNAm is differentially methylated between aggressive and indolent PCa cases at over >850k CpG sites measured on the Illumina HumanMethylationEPIC BeadChip. Analysis for identification of differentially methylated regions (DMRs) will also be performed. To determine the CpGs and CpG combinations most predictive of PCa progression, we will apply feature selection and engineering approaches, and a library of pre-specified supervised machine learning methods (e.g. elastic-net regression, tree-ensembles, etc.). All models will be evaluated by outcome-stratified repeated k-fold cross-validation to robustly assess out-of-sample predictive performance and tune relevant hyperparameters. C-statistic/AUC will be the primary performance metric, but several alternatives will be considered.
1. Pernar CH, Ebot EM, Wilson KM, Mucci LA. The Epidemiology of Prostate Cancer. Cold Spring Harb Perspect Med TA - TT -. 2018;8(12):a030361.
2. Prostate Cancer incidence statistics: Cancer Research UK (CRUK) [Internet]. 2017. Available from: https://www.cancerresearchuk.org/health-professional/cancer-statistics/statistics-by-cancer-type/prostate-cancer/incidence
3. Zhang J, Huang K. Pan-cancer analysis of frequent DNA co-methylation patterns reveals consistent epigenetic landscape changes in multiple cancers. BMC Genomics. 2017. Available from: https://pubmed.ncbi.nlm.nih.gov/28198667/
4. Yousefi PD, Suderman M, Langdon R, Whitehurst O, Davey Smith G, Relton CL. DNA methylation-based predictors of health: applications and statistical considerations. Nat Rev Genet. 2022 Mar 18;1–15.
Dr Christina Dardani (lead), Dr Jasmine Khouja, Professor Marcus Munafò
Tobacco smoking, has been found to be causally linked to psychiatric and neurological outcomes, including depression, schizophrenia, Parkinson’s disease (1,2). The mechanisms underlying these links are unclear. It has been hypothesised that tobacco smoke has effects on neuropsychiatric outcomes because of its most widely studied constituent, nicotine(3). Stimulation of nicotine receptors leads to increased dopamine release in the brain, a neurotransmitter that has a central role in brain structure and function. However, investigating the causal influence of nicotine exposure on psychiatric and neurological outcomes has not identified clear effects so far, particularly due to confounding by exposure to tobacco smoke and its other (than nicotine) constituents(4). Considering that an RCT of nicotine exposure among non-nicotine users would be unethical, Mendelian randomisation (MR) approaches can be a powerful alternative to aid causal inference(5). Disentangling the effects of nicotine on neuropsychiatric conditions can inform current perspectives on smoking cessation policies and offer unique insights into potential intervention targets.
We will assess the direct, independent of other constituents of tobacco smoke, causal effects of nicotine exposure on psychiatric and neurological outcomes by employing MR approaches.
We will use single nucleotide polymorphisms (SNPs) robustly associated with nicotine metabolite ratio (NMR), which is the ratio of 3'hydroxycotinine (3HC)/cotinine (COT) and indicates how quickly a person metabolises and clears nicotine(6). We will separately use SNPs robustly associated with smoking heaviness (cigarettes per day among smokers)(7). We will use two-sample MR(8) to assess their total causal effects on three psychiatric (depression, schizophrenia, bipolar disorder) and two neurological (Parkinson’s, Alzheimer’s disease) conditions. Furthermore, we will use Multivariable MR(9) to assess the, independent, direct effects of nicotine versus other constituents of tobacco smoke on the outcomes of interest, by entering in the models SNPs robustly associated with smoking heaviness(7). We will perform sensitivity analyses to assess the robustness of the causal effect estimates.
1. Larsson SC, Burgess S. Appraising the causal role of smoking in multiple diseases: A systematic review and meta-analysis of Mendelian randomization studies. EBioMedicine. 2022;82:104154.
2. Wootton RE, Richmond RC, Stuijfzand BG, Lawn RB, Sallis HM, Taylor GMJ, et al. Evidence for causal effects of lifetime smoking on risk for depression and schizophrenia: a Mendelian randomisation study. Psychol Med. 2020;50(14):2435–43.
3. Quik M. Smoking, nicotine and Parkinson’s disease. Trends Neurosci. 2004;27(9):561–8.
4. Khouja JN, Sanderson E, Wootton RE, Taylor AE, Munafò MR. A multivariable Mendelian randomisation study exploring the direct effects of nicotine on health compared with the other constituents of tobacco smoke: Implications for e-cigarette use. medRxiv. 2021;
5. Davey Smith G, Ebrahim S. “Mendelian randomization”: can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol. 2003;32(1):1–22.
6. Buchwald J, Chenoweth MJ, Palviainen T, Zhu G, Benner C, Gordon S, et al. Genome-wide association meta-analysis of nicotine metabolism and cigarette consumption measures in smokers of European descent. Mol Psychiatry. 2021;26(6):2212–23.
7. Liu M, Jiang Y, Wedow R, Team 23andMe Research. Association studies of up to 1.2 million individuals yield new insights into the genetic etiology of tobacco and alcohol use. Nat Genet. 2019;51(2):237–44.
8. Burgess S, Scott RA, Timpson NJ, Smith GD, Thompson SG, Consortium E-I. Using published data in Mendelian randomization: a blueprint for efficient identification of causal risk factors. Eur J Epidemiol. 2015;30(7):543–52.
9. Sanderson E, Davey Smith G, Windmeijer F, Bowden J. An examination of multivariable Mendelian randomization in the single-sample and two-sample summary data settings. Int J Epidemiol. 2019;48(3):713–27.
Dr Ashley Budu-Aggrey (lead), Dr Lavinia Paternoster,
Several traits such as cardiovascular disease[1] and mental health disorders[2] have been reported to be observationally associated with atopic dermatitis (AD). The most recent genome-wide association study for AD has also uncovered genetic correlations[3]. However, the causal relationships and direction of effect is yet to be determined. Establishing causality will aid the early detection of AD or later health outcomes and determine whether intervention on one condition will affect the other.
To investigate causal risk factors and outcomes of AD using Mendelian Randomization (MR)
1. The literature will be screened to identify hypotheses for examination with MR.
2. Observational analysis will be performed using datasets available in-house such as the UK Biobank
3. Suitable datasets will be identified, and genetic instruments for AD and other traits will be refined
4. 2 sample MR analyses will be performed to investigate causal relationships
5. Sensitivity analyses will be performed to ensure the assumptions of the MR analyses have not been violated
1. Standl, M. et al. Association of Atopic Dermatitis with Cardiovascular Risk Factors and Diseases. Journal of Investigative Dermatology 137, 1074–1081 (2017).
2. Budu-Aggrey, A. et al. Investigating the causal relationship between allergic disease and mental health. Clin Exp Allergy 51, 1449–1458 (2021).
3. Budu-Aggrey, A. et al. European and multi-ancestry genome-wide association meta-analysis of atopic dermatitis highlights importance of systemic immune regulation. Submitted to Nature Communications (2022).
Dr Ashley Budu-Aggrey (lead), Dr Lavinia Paternoster,
An observational relationship has been reported between atopic dermatitis (AD) and BMI, where evidence of a causal relationship has also been found for higher BMI increasing the risk of AD. Up until now, this causal relationship has not been investigated with respect to childhood and adulthood BMI separately especially given the AD mostly presents in early childhood.
To investigate the causal relationship with AD and childhood and adulthood BMI separately using Mendelian Randomization (MR)
Two-sample MR will be used to investigate causality between AD and childhood and adulthood BMI separately, and also determine the direction of effect (AD -> childhood/adulthood BMI or childhood/adulthood BMI -> AD). Genetic instruments for the MR analyses will be derived from the most recent genome-wide association studies (GWAS) for AD and childhood/adulthood BMI. Sensitivity analyses will also be applied to ensure the assumptions of the MR analyses have been met.
- Budu-Aggrey, A. et al. European and multi-ancestry genome-wide association meta-analysis of atopic dermatitis highlights importance of systemic immune regulation. Submitted to Nature Communications (2022).
- Budu-Aggrey, A. et al. Assessment of a causal relationship between body mass index and atopic dermatitis. Journal of Allergy and Clinical Immunology 147, 400–403 (2021).
- Richardson, T. G., Sanderson, E., Elsworth, B., Tilling, K., & Smith, G. D. (n.d.). Use of genetic variation to separate the effects of early and later life adiposity on disease risk: mendelian randomisation study. https://doi.org/10.1136/bmj.m1203
Prof Julian Higgins (lead), Prof Kate Tilling, Prof Marcus Munafò
Triangulation, in which multiple methods are strategically used to answer a single question, is a currently developing area. Lawlor, Tilling and Davey Smith (2016) explained how causal inferences can be strengthened by integrating results from several approaches with different key sources of potential bias. The statistical methods for combining the results from multiple sources of evidence within a triangulation framework are, however, underdeveloped. This PhD seeks to develop, illustrate and evaluate such methods.
The project seeks to develop and implement quantitative methods for triangulation of multiple lines of evidence addressing the same underlying epidemiological question
Work is expected to focus on three key areas as follows.
1) At its simplest, triangulation involves comparison and combination of studies of the same exposure-outcome effect that use different designs or analytic methods. For example, randomized trials, Mendelian randomization studies and traditional multivariable regression analyses of observational evidence might all tackle a question relating to the same exposure-outcome effect. The studies may produce different effect estimates because they are (i) asking subtly different questions (e.g. in relation to the period or patterns of exposure), (ii) compromised by different biases and/or (iii) subject to chance. Triangulation combines these issues in a statistical model and assesses the extent to which the observed data fit together – an approach known as multiparameter evidence synthesis. Methods for producing these models, assessing coherence and drawing conclusions about causal effects of the exposure on the outcome will be developed. The project will primarily explore Bayesian methods, because they are flexible and allow incorporation of external information through prior distributions.
2) Another form of triangulation arises when some (or all) studies address only a component of the underlying question. For example, if the exposure-outcome effect occurs through an intermediate, then studies of the exposure-outcome effect might be triangulated with a combination of studies (i) of the effect of exposure on the intermediate and (ii) of the effect of the intermediate on the outcome. Methods will be developed to synthesise these three sets of studies, and account for true differences, biases and chance.
3) In addition to working on novel statistical methods, the student may explore other methodological questions. First, how should we define and identify studies suitable for a triangulation exercise? Automation tools may help here, such as MELODI (http://melodi.biocompute.org.uk), which we have developed to identify studies examining intermediates between exposure and outcome. Second, how should we evaluate the risk of bias in studies for which formal frameworks (such as RoB 2 and ROBINS-I; http://riskofbias.info) have not been developed? Third, what sources of information are available about biases, to inform prior distributions, and how can more information be generated?
Methods developed in these three areas will be illustrated through application to important causal questions in epidemiology.
Lawlor DA, Tilling K, Davey Smith G. Triangulation in aetiological epidemiology. Int J Epidemiol. 2016 Dec 1;45(6):1866-1886. doi: 10.1093/ije/dyw314.
Munafò MR, Davey Smith G. Robust research needs many lines of evidence. Nature. 2018 Jan;553(7689):399-401. doi: 10.1038/d41586-018-01023-3.
Munafò MR, Higgins JPT, Davey Smith G. Triangulating Evidence through the Inclusion of Genetically Informed Designs. Cold Spring Harb Perspect Med. 2021 Aug 2;11(8):a040659. doi: 10.1101/cshperspect.a040659.
Turner RM, Spiegelhalter DJ, Smith GC, Thompson SG. Bias modelling in evidence synthesis. J R Stat Soc Ser A Stat Soc. 2009 Jan;172(1):21-47. doi: 10.1111/j.1467-985X.2008.00547.x.
Ades AE, Welton NJ, Caldwell D, Price M, Goubar A, Lu G. Multiparameter evidence synthesis in epidemiology and medical decision-making. J Health Serv Res Policy. 2008 Oct;13 Suppl 3:12-22. doi: 10.1258/jhsrp.2008.008020.
Dr Evie Stergiakouli (lead), Prof Laura Howe, Alexandra Havdhal, University of Oslo Department of Psychology
Attention Deficit Hyperactivity Disorder (ADHD) is a chronic neurodevelopmental condition, characterised by persistent difficulties in the areas of attention span/impulse control. Approximately 65% of children diagnosed with ADHD have symptoms and impairment that persist into adulthood and ADHD can lead to educational, social, and occupational difficulties (1).
We have previously shown that higher genetic risk for ADHD is associated with younger maternal age at birth, lower educational attainment and other indicators of social disadvantage in mothers from the general population (2). Using Mendelian randomization (MR) we have also found evidence of higher genetic liability to ADHD causing lower educational attainment, and evidence of genetic liability to lower educational attainment increasing risk to ADHD independent of cognitive ability (3). Since ADHD manifests at a very young age, the causal effects of genetic liability to education on ADHD are likely to indicate parental effects. The causal link between parental education and ADHD could be mediated by optimal lifestyle and general health factors during pregnancy and/or socioeconomic factors linked to better access to educational resources. Disentangling the individual effects of each factor as well as assessing for genetic confounding is required.
In this project, we will explore the links between educational attainment, reproductive outcomes (age at first birth, number of live births), prenatal factors, socioeconomic status and other indicators of social disadvantage on ADHD.
Our aims are: 1. To assess the contribution of parental educational attainment on trajectories of ADHD traits in offspring and 2. To disentangle it from the offspring own genetic background and investigate the causal pathways linking parental educational attainment and offspring ADHD.
This is an exciting opportunity for a student to perform advanced genetic epidemiological analyses on large multigenerational longitudinal cohorts from two countries: the Avon Longitudinal Study of Parents and Children in the UK and the Norwegian Mother and Child Cohort Study (MoBa) in Norway.
For aim 1, we will compare the associations of maternal and paternal genetic liability of educational attainment on ADHD trajectories from two general population samples with very different educational systems and social structures. Polygenic Transmission Disequilibrium tests will be used to assess if genetic liability to lower educational attainment is overtransmitted to offspring with ADHD (4).
For aim 2, we will use within-families MR (5) in MoBa to investigate causal effects of the offspring’s own genetic liability to education attainment while adjusting for parental genetic liability. We will also perform Multivariable Mendelian randomization (6) to account for multiple exposures (educational attainment, reproductive outcomes, prenatal factors, socioeconomic status) simultaneously. Finally, we will apply sensitivity analyses including weighted median, weighted mode, MR-Egger regression, MR-PRESSO and colocalization analyses to assess and adjust for pleiotropy.
Thapar A, Cooper M. Attention deficit hyperactivity disorder. Lancet. 2016 Mar 19;387(10024):1240-50. doi: 10.1016/S0140-6736(15)00238-X.
Leppert B et al. Association of Maternal Neurodevelopmental Risk Alleles With Early-Life Exposures. JAMA Psychiatry. 2019;76(8):834–842. doi:10.1001/jamapsychiatry.2019.0774
Dardani et al. Is genetic liability to ADHD and ASD causally linked to educational attainment?, Int. J. Epidemiol. 2021;, dyab107, https://doi.org/10.1093/ije/dyab107
Weiner D et al. Polygenic transmission disequilibrium confirms that common and rare variation act additively to create risk for autism spectrum disorders. Nat Genet 49, 978–985 (2017). https://doi.org/10.1038/ng.3863
Brumpton, B et al. Avoiding dynastic, assortative mating, and population stratification biases in Mendelian randomization through within-family analyses. Nat Commun 11, 3519 (2020). https://doi.org/10.1038/s41467-020-17117-4
Sanderson E et al. An examination of multivariable Mendelian randomization in the single-sample and two-sample summary data settings. Int J Epidemiol. 2019 Jun 1;48(3):713-727. doi: 10.1093/ije/dyy262.
Dr Kaitlin Wade (lead), Professor Golam Khandaker, Dr Christina Dardani
Low-grade systemic inflammation is associated with risk of neuropsychiatric conditions such as depression, schizophrenia, autism, with emerging genetic evidence suggesting a potentially causal role of inflammatory cytokines such as interleukin 6 (IL-6) in these conditions (1, 2). Gut microbiome alterations are thought to be an important source of systemic low-grade inflammation, and disruptions in the “gut-brain axis” are implicated in neurodevelopment, cognitive function, and risk of neuropsychiatric conditions (3). However, it is unclear whether gut microbiome alterations play a causal role in neuropsychiatric conditions, and whether immune activation represents a potential mechanism through which gut microbiome alterations influence neuropsychiatric risk. Uncovering causal pathways between the gut microbiome, immune response and neuropsychiatric conditions may offer unique insights into novel intervention targets.
1. Summarize current evidence on the associations of the gut microbiome and neuropsychiatric conditions (depression, schizophrenia, and autism).
2. Examine evidence for causality in these associations.
3. Assess whether inflammatory cytokines represent a potential mechanism through which gut microbiome alterations influence neuropsychiatric conditions.
4. Assess the predictive utility of the gut microbiome (directly or through genome-wide predictors) and neuropsychiatric conditions.
This PhD will provide training in the following important research methods:
1. Systematic review and meta-analysis of studies on the gut microbiome and neuropsychiatric conditions (depression, schizophrenia, and autism).
2. Multivariable linear/logistic regressions assessing these associations in available individual-level data (e.g., Flemish Gut Flora Project and those in the MiBioGen consortium) (4, 5).
3. Bi-directional Mendelian randomization analyses testing potential causality (and direction) in relationships between gut microbiota and neuropsychiatric conditions (6).
4. Mediation analysis to test whether inflammation (e.g., levels/activity of immune proteins such as C-reactive protein [CRP], IL-6, tumour necrosis factor alpha [TNF-a]), mediate the association between gut microbiome alterations and neuropsychiatric conditions.
5. Prediction models of the association between gut microbiota and neuropsychiatric conditions, outside causal framework analyses.
1. Khandaker et al. JAMA Psychiatry 2014;71:1121.
2. Perry et al. Brain Behaviour and Immunity 2021;97:176.
3. Carabotti et al. Ann Gastroenterol. 2015;28:203.
4. Hughes et al. Nat Microbiol. 2020;5:1079.
5. Kurilshikov et al. Nature Genetics 2021;53:156.
6. Wade and Hall. Wellcome Open Res 2019;4:199
Professor Nicholas Timpson (lead), Professor Paul Martin,
Repair of adult tissues involves a complex interplay of several key cell lineages and inevitably leads to formation of a fibrotic collagenous scar. We use human phenotypic variation to examine genetic correlates of scarring and to better understand biological contributions to wound healing. As an introduction into using population approaches to identify scar-associated genes, we have already assessed a sample of mothers from the Bristol-based Avon Longitudinal Study of Parents and Children (ALSPAC, www.bris.ac.uk/alspac) for their scarring response to a standard lesion that they all will have received as teenagers. The first natural experiment in this case was that of receiving a BCG vaccination in early life. Approximately 800 scars were measured and these exhibited a normal, bell shaped distribution of scar diameters (mean 6mm SD 3.37). Genome-wide association analysis of this phenotype revealed a low frequency single nucleotide polymorphism (SNP) in a regulatory intronic region of the GPCR, LGR4, that we find associates with the tendency to scar less.
1-Complete current ALSPAC studies of scarring genetics – across human association studies and lab-based model.
2-Combine existing recall by genotype (RbG) and GWAS results, developing bioinformatic analysis of established and possibly new signals for scarring traits.
3-Extend results from the ALSAC to other data sets and other pertinent phenotypes – specifically measured BCG and C-section scars in the Pelotas cohort, Brazil.
4-Work with a collaborator consortium focused on fibrosis traits more generally to widen the analysis of genetic contributions to scarring.
Extension of existing population health data will be to combine the existing evidence available from GWAS and the RbG undertaken in ASPAC in order to provide best estimates of association between known loci and scarring. Work will expand on the initial GWAS analysis through the incorporation of the Pelotas 1982 birth cohort where we examine BCG scar size at the current whole cohort clinic and exapand this to include the study of C-section scars which – in Brazil – are far higher in frequency than in the UK and provides an excellent opportunity to examine both the observational factors and genetic factors associated with differential scarring. Lastly, as part of a broad and exploratory portfolio of approaches to the assessment of genetic contributions to scarring events which may flag loci of interest for mechanistic studies in model organisms (below), we propose to incorporate the systematic followup of new findings for pulmonary fibrosis through a GWAS study that is currently being collaboratively led by colleagues at the University of Leicester.
[1] Hopkinson-Woolley J HD, Gordon S and Martin P. Journal of Cell Science
107, 1159-1167 (1994). 1994;107:1159-67.
[2] Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D, et al. Elife. 2018;7.
[3] Victora CG, Barros FC. Int J Epidemiol. 2006;35(2):237-42.
[4] Horta BL, Gigante DP, Goncalves H, dos Santos Motta J, Loret de Mola C,
Oliveira IO, et al. Int J Epidemiol. 2015;44(2):441, a-e.
Professor Nicholas Timpson (lead), Dr Laura Corbin, Dr Kaitlin Wade
There is strong evidence that BMI causally influences a wide range of health outcomes, but there is little understanding of the mechanisms driving BMI effects and at the level of the population we are limited in our ability to alter BMI. The main theme of this work is to dissect the causal effects which have been shown to lie between BMI and health outcomes; i.e. to better understand how body mass index (BMI) exerts an effect on human health. The aim is to do this using multiple-omic data sets in complementary study designs and through applied genetic epidemiology. This is a unique opportunity to study in a research team dedicated to this and to get access to a series of mature data sets able to address this question. This is a PhD which will be coordinated with a Wellcome Trust funded programme of research that explicitly seeks to use the most contemporary and powerful study designs, multi-omic data (including that from the metabolome, microbiome and proteome) and analytical techniques to explore BMI as a risk factor.
This work aims to identify omic intermediates important in the link between BMI and disease. The work sets out to better understand how body mass index (BMI) exerts an effect on human health using in complementary study designs and through applied genetic epidemiology. Proposed work will aim to integrate data from multi-omic data collections in efforts to assess these as mediating routes between BMI variation and disease.
The proposed project will focus on multiple, overlapping omic measurements and how they can be used to understand the mechanisms by which BMI contributes to disease. This work will involve the use of standard epidemiological analysis approaches (such as linear regression and applied approaches like Mendelian randomisation), but will also lean on data driven approaches to rich omic data sets collected across trials of BMI/weight targeted interventions to describe shared and unique relationships between change and omic intermediate read-out (principal component and variable analysis along with systematic approaches to omic profile effects). There will be an opportunity to develop new methods across these data sets and study designs. This will include a causal mapping of the human faecal microbiome onto the blood metabolome (using data from the Flemish Gut Flora Project, which includes ~3,000 individuals with questionnaire, clinic, genetic and microbiome data), but also other sources of metabolomic and proteomic data available across multiple studies with marked examples of variation (induced and natural) in body composition and specifically BMI.
[1] doi.org/10.1371/journal.pmed.1003786
[2] doi: 10.2337/db21-0397
[3] doi: 10.1038/s41564-020-0743-8
[4] doi.org/10.1038/s41366-021-00896-1
[5] doi.org/10.1101/2022.07.15.22277671
[6] doi.org/10.1002/oby.23441
Professor Nicholas Timpson (lead), Dr Laura Corbin (University of Bristol), Dr Brian Lam (University of Cambridge)
The hypothalamus is central to controlling energy homeostasis and reproduction1. The leptin-melanocortin pathway in the hypothalamus senses hormonal signals, namely leptin and insulin from the body and regulates energy2. Recently we showed that loss-of function (LoF) mutations in MC4R are associated with increased BMI, which is driven by both elevated linear growth and fat mass accrual3. MC3R, is also implicated in linear growth puberty4.
To understand the neuro-architecture of the hypothalamus we aggregated single-cell studies and created an integrated atlas5. Cell populations identified are known to regulate energy balance (e.g. POMC and GLP1R neurons) and growth (GHRH neurons) and gene markers characteristic for these populations are druggable and are targeted for treatments of obesity T2D6.
Working across the MRC Integrative Epidemiology Unit and the Wellcome-MRC Institute of Metabolic Science (Professor Sir Stephen O’Rahilly), we will systematically characterise genes expressed in hypothalamic cells, identify associated human genetic variation and study potential effects on anthropometric and reproductive development.
1.Spatially characterise gene expression in the mouse and human hypothalamus using state-of-the-art spatial transcriptomic techniques
2.Identify rare, non-synonymous Human hypothalamic genetic variations in the 9,000 exomes of Avon Longitudinal Study of Parents and Children (ALSPAC7) and investigate their association with anthropometric measurements (cross-sectional and trajectories) and reproductive development (e.g. Tanner stages, age at menarche, peak height velocity)
3.Experimental studies on the effects of selected genetic mutations in-vitro cellular models
The methods deployed within a PhD nested in this initiative will be tailored to the specific project work undertaken. Two main of focus areas will be the characterisation of locus contributions within model studies and the analysis of specific genetic variation in population-based studies:
• Histological techniques including tissue processing and sectioning
• Spatial transcriptomics
• Bioinformatics analysis and integration of single-cell and spatial datasets
• Tissue cell culture and in-vitro functional assays
• Molecular biology and immunoblotting
• Confocal microscopy and imaging
• Preparation of population-based longitudinal phenotype data.
• Pooled sample and whole population exome sequence data analysis.
• Association testing within population-based studies with longitudinal phenotypic data based on genomewide array data using basic genome-wide association approaches.
• Similarly focused whole exome and targeted sequencing data using rare variant and structural variant approaches.
1. Lowell, N. Engl. J. Med. 380, 459–471 (2019).
2. Yeo, Mol Metab, 48 101206.3. Wade, Nature medicine 27, 1088-1096 (2021).
4. Lam, Nature 599, 436-441
5. Steuernagel, Nat Met 4, 1402–1419 (2022).
6. Vilsbøll BMJ (2012) 7771(January):1–11. doi: 10.1136/bmj.d7771
7. Boyd, IJE 42 (1), 111–127 (2013)
Dr Kushala Abeysekera (lead), Prof Matthew Hickman , Prof Nicholas Timpson
Historically, cirrhosis (irreversible liver scarring), often due to alcohol-related liver disease (ARLD) and non-alcoholic fatty liver disease (NAFLD), presents in the 5th and 6th decade. 1 in 6 inpatients die during hospital admission (1), with over 70% of new liver disease presenting acutely, many dying without the chance to change their lifestyle (2).
Liver disease presents is now one of the leading causes of death in 35-49 year olds, overtaking suicide (3). Early detection of liver disease is crucial to prevent progression to advanced disease and facilitate treatment. Young adults represent the next frontier in tackling the public health crisis of liver disease (4). In this regard, the Avon Longitudinal Study of Parents and Children (ALSPAC) is uniquely placed to provide normative data and evaluate the rising burden of liver disease.
When aged 17 years, ALSPAC participants were offered liver scans to assess NAFLD, revealing a 2.7% prevalence (5). When aged 24 years, 20.7% of participants had steatosis and 1 in 40 had liver scarring, the precursor to cirrhosis (6). Our participants are now 30 and over 4000 are set to receive further liver scan measurements.
1. Estimating prevalence of NAFLD and ARLD in young adults in the general population aged 30 years.
2. Studying early life course determinants to understand why some participants are developing liver scarring earlier than others.
3. Evaluating if existing polygenic risk scores can be used to detect and predict individuals with subclinical liver scarring.
Examining and presenting normative data from the 30 year liver clinic to determine the prevalence of NAFLD and ARLD amongst ALSPAC participants. Use of linear and logistic regression to explore the strength of associations of common exposures associated with liver disease e.g. alcohol use, body mass index and sociodemographics. Mapping progression of disease from participants who had fibrosis aged 24 years.
Using the existing data within ALSPAC with outcomes of steatosis and fibrosis across ages 17, 24 and now 30 years, evaluate if there are sensitive periods for an impact of change in, or accumulation of, exposures including adiposity, cardiometabolic factors, and alcohol on liver steatosis/fibrosis at 30 years and its progression from 17 to 30 years. This will involve linear and logistic regression modelling and interaction analyses.
Finally, polygenic risk scores for detection of chronic liver disease exist (7). Using existing genomic data from ALSPAC, work will be done to explore if such scores are useful in this different ages (17, 24, 30 years), and whether they can be combined with clinical markers to improve detections and prediction of liver fibrosis.
1. TORCH-UK. Gut. 2022;71.
2. Williams et al. The Lancet. 2014;384(9958):1953-97.
3. Trust. The alarming impact of liver disease in the UK: Facts and statistics. 2019.
4. Williams et al. Lancet. 2018;391(10125):1097-107.
5. Lawlor et al. J Clin End Met. 2014;99(3):E410-7.
6. Abeysekera et al. The Lancet G & H. 2020.
7. De Vincentis et al. Clin G H. 2021.
Prof Tom Gaunt (lead), Dr Maria Sobczyk,
Mendelian Randomization (MR) is a genetic epidemiology method which utilises variants sourced from genome-wide association studies (GWAS) to assess causality between risk/protective factors and disease outcomes in a manner less biased to observational studies(1).
Recently, applications of MR to drug target prioritization gained a lot of interest. One approach is to use expression or protein quantitative trait loci (QTL) for a drug target gene/protein as exposures with the aim of establishing the effects of perturbing the intended target directly (“on-target”)(2). However, most drugs will have broader molecular consequences, either in parallel (“off-target”) or downstream of the intended target which can be exploited to repurpose drugs for novel indications. One way to identify such off-targets is to mine high-throughput expression datasets of various cell lines exposed to small molecule drugs and genetic perturbations(3).
Using a variety of high-throughput genomics/other -omics resources, the project will aim to discover and triangulate with Mendelian Randomization:
1) Drug off-target side effects
2) Drug repurposing opportunities
You will use data on transcriptional responses to drugs from expression perturbation databases and other resources to identify additional genes and proteins for off-target side effect prediction, using multivariable MR to identify direct effects. In addition, you will use protein-protein interaction data collated in our locally developed resource: EpiGraphDB(4) to further refine gene sets to instrument.
You will also use drug perturbation data sources to evaluate the potential to repurpose drugs with a previously-published approach(5) that compares disease-associated transcriptomic profiles to in vitro drug transcriptomic profiles to identify profiles that may reverse the effect of disease on gene expression. You will consider tissue-specific and pathway-specific transcriptomic profiles for a variety of diseases and explore whether this identifies additional repurposing opportunities. You will validate drug repurposing and off-target side effect predictions using observational data from UK Biobank and electronic health records, in addition to MR.
1. Sanderson, E. et al. Mendelian randomization. Nat. Rev. Methods Prim. 2, 6 (2022).
2. Gill, D. et al. Mendelian randomization for studying the effects of perturbing drug targets. Wellcome open Res. 6, 16 (2021).
3. Keenan, A. B. et al. The Library of Integrated Network-Based Cellular Signatures NIH Program: System-Level Cataloging of Human Cells Response to Perturbations. Cell Syst. 6, 13–24 (2018).
4. Liu, Y. et al. EpiGraphDB: a database and data mining platform for health data science. Bioinformatics 0–0 (2020).
5. Wu, P. et al. Integrating gene expression and clinical data to identify drug repurposing candidates for hyperlipidemia and hypertension. Nat. Commun. 13, 46 (2022).
Dr Eleanor Sanderson (lead), Dr Kaitlin Wade,
Mendelian randomization (MR) uses genetic variants to estimate the causal effect of an exposure on an outcome of interest in a way that aims to overcome from bias from unobserved confounding. Under a set of standard assumptions, the estimates obtained from MR can be interpreted as ‘population average causal effects’ (i.e., the average causal effect for the population studied). However, for many exposure-outcome relationships of interest, the causal effect of the exposure on the outcome may differ for different groups of individuals. For example, the causal effect of body mass index (BMI) on blood pressure (BP) may differ by whether an individual is male or female. There are different approaches that could be used to estimate whether there is a difference in the causal effect across different groups, but which approach is the most appropriate is unknown.
• To estimate whether the causal effect of BMI on BP by differs by sex.
• To determine which method is the most appropriate to test for differences in the causal effect of BMI on BP by sex.
This project will use individual-level data from UK Biobank to apply different approaches to estimating whether there is a difference in the causal effect of BMI on BP by sex. These approaches will include
1. Conducting sex-specific MR analyses and testing for a difference in the causal effect between these groups
2. Including a sex interaction term in a single MR estimate using data from the whole population
3. Any other approaches identified that can be applied in individual-level MR estimation.
The student may also have the opportunity to conduct a small simulation study to identify any bias in the approaches used; however, this part of the project is optional and will depend on available time.
1. Davies, Neil M., Michael V. Holmes, and George Davey Smith. "Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians." bmj 362 (2018).
2. Sanderson, Eleanor, et al. "Mendelian randomization." Nature Reviews Methods Primers 2.1 (2022): 1-21.
Professor Laura Howe (lead), Dr Annie Herbert, Associate Professor Jon Heron Dr Alison Teyhan
Around 1 in 3 young adults in the UK have experienced intimate partner violence and abuse (IPVA; psychological, physical, or sexual abuse from an intimate partner) by the time they are 21. Effective interventions delivered in healthcare settings (e.g. GP practices or Accident & Emergency) aim to prevent further abuse, and address current social or health problems (e.g. access to housing or poor mental health). Such interventions have been trialled in older adults. However, little is known about the population of young adult IPVA survivors who come into contact with healthcare settings, and whether and how these health service contacts alter their health trajectories. Such understanding can indicate whether interventions trialled in adults may also be effective in young adults, for which outcomes, and what can be done to better tailor interventions towards this age group. This project will use rich birth cohort data linked to administrative primary and secondary healthcare data, combined with sophisticated longitudinal statistical methods, to build a picture of the life course health and health care trajectories of young adult survivors of IPVA.
Aim: The overall aim of this project is to describe health and healthcare service use trajectories of young adult IPVA survivors, and the impact of service use on subsequent health outcomes
Objectives:
1) To describe the health trajectories of young adults who do and do not experience IPVA, using both cohort and administrative data.
2) To describe the patterns of health service use amongst young adults who do and do not experience IPVA.
3) To assess the degree to which interaction with the health service may mitigate the health trajectories of people who experience IPVA, or conversely, how patterns of health service use may be a red flag for health problems or continued IPVA exposure.
4) To investigate whether the associations of IPVA with health trajectories and patterns of health service use differ by sex, socioeconomic circumstances, early life adversity, or type, timing, and severity of IPVA.
This project will use data from the Avon Longitudinal Study of Parents and Children (ALSPAC; Children of the 90’s). Participants reported on their current and past IPVA experiences at age 21, including IPVA subtypes (psychological, physical, sexual), frequency, and impact. Rich longitudinal health data is collected through questionnaires and research clinics. The ALSPAC data have been linked to primary and secondary care data, which will be used to capture healthcare contacts and referrals until age 28. READ codes in primary care data and ICD-10 codes in secondary care data, supplemented by ALSPAC clinic and survey data, will be used to capture health and health care trajectories. The PhD will provide training in standard regression models, life course and longitudinal modelling techniques (e.g. multi-level and latent growth models, difference-in-difference), and in the use of both cohort and administrative data sources. These models will incorporate interaction analyses, and multiple imputation to deal with missing data.
- Herbert A, Heron J, Barter C et al. Risk factors for intimate partner violence and abuse among adolescents and young adults: findings from a UK population-based cohort [version 3; peer review: 2 approved]. Wellcome Open Res 2021, 5:176 (https://doi.org/10.12688/wellcomeopenres.16106.3)
- Ogbe E, Harmon S, Van den Bergh R, Degomme O. A systematic review of intimate partner violence interventions focused on improving social support and/ mental health outcomes of survivors. PLoS One. 2020 Jun 25;15(6):e0235177. doi: 10.1371/journal.pone.0235177. PMID: 32584910; PMCID: PMC7316294.
Feder G. Responding to intimate partner violence: what role for general practice? Br J Gen Pract. 2006 Apr;56(525):243-4. PMID: 16611508; PMCID: PMC1832227.
- Kristoffer S. Berlin, PhD, Gilbert R. Parra, PhD, Natalie A. Williams, PhD, An Introduction to Latent Variable Mixture Modeling (Part 2): Longitudinal Latent Class Growth Analysis and Growth Mixture Models, Journal of Pediatric Psychology, Volume 39, Issue 2, March 2014, Pages 188–203, https://doi.org/10.1093/jpepsy/jst085
Professor Abigail Fraser (lead), Professor Tom Gaunt, Prof Laura Howe; Dr Emma Hart
Hypertension (defined as systolic/diastolic pressures >140/90 mmHg) is the leading risk factor for cardiovascular disease (CVD) and events, including stroke, myocardial infarction and heart failure. Both systolic and diastolic blood pressures (BP) are associated with CVD end points, albeit differentially across different stages of the life course.
Systolic and diastolic BP reflect different aspects of the cardiovascular system, and likely reflect different underlying pathologies. The physiological determinants of diastolic BP are more complex than determinants of systolic BP and include 1) the level of the systolic BP, 2) the level of stiffness of the arteries and arterioles and 3) the resistance of arterioles to blood flow. It is also thought to reflect different pathologies at different life stages: in young adults, higher diastolic BP is associated with greater CVD risk in later life but in older populations, lower diastolic BP (for a given systolic BP) is associated with greater mortality risk.
Diastolic BP has been understudied compared to its systolic counter-part, with the latter more commonly used in CVD prediction models and as a clinical end point in RCTs. However diastolic BP at different stages of the lifecourse, and in relation to the level of systolic BP, may help improve risk stratification, and have translational impact.
To better understand diastolic blood pressure changes across the life course, its determinants, and if/how it relates to cardiovascular outcomes independently of systolic blood pressure.
The student will use population data (from ALSPAC, UK Biobank and other population based cohorts) and epidemiological methods, as well as experimental approaches and bioinformatics (should they wish). The student will be encouraged to identify specific objectives within the general scope of the project that are of interest to them.
We envisage that the project will include some, not necessarily all, of the following. The student will be supported to tailor the project to their interests.
1. Literature review to identify determinants of DBP at both a mechanistic (e.g. microvascular stiffness) and population level (e.g. sodium intake).
2. Assembling life course trajectories of diastolic blood pressure using multilevel models, equivalent to those constructed by Wills et al Plos Medicine 2011 and investigating potential sex-differences, as well as effects of puberty and menopause on diastolic blood pressure in women.
4. Using Mendelian Randomization approaches, including multivariable MR to identify
a. causal risk factors for diastolic blood pressure (independent of systolic blood pressure) and
b. the cardiovascular consequences of variation in diastolic blood pressure across the life course.
5. Use bioinformatics to interrogate the molecular pathways mediating the effects of genetic variants uniquely associated with diastolic blood pressure.
Gibran Hemani (lead), George Davey Smith,
Estimating the influence of genotypes and traits on fitness is a fundamental question in biology. While the question requires causal inference, many analytical frameworks (e.g. Price equation) are embedded in a statistical framework that are unable to separate cause from correlation. Mendelian randomisation is an analytical framework that is widely used in epidemiological studies, and we have shown it can be used to make causal inference of traits on fitness. In this project we will investigate the theoretical properties of this approach and apply it to a vast array of genetic data to understand the landscape of the genotype-phenotype map on fitness. This mapping will allow us to make inference about critically important questions, such as the degree to which pleiotropy modulates directional selection, and the degree to which stabilising selection maintains natural genetic variation.
1. One approach to estimate the influence of traits on fitness is to estimate the causal influences on reproductive traits such as fecundity, age of first menarche, age of first birth and age of menopause. We will investigate the way in which these traits approximate fitness, using multi-generation samples and latent modelling through genomic structural equation modelling
2. We will perform an exhaustive scan of traits on measures of fitness determined from (1), to build a profile of the genotypes-trait-fitness landscape
3. We will use (2) to model and infer the degree to which important mechanisms such as pleiotropy and trait network effects either accelerate or constrain directional selection, and to what degree they maintain natural genetic variation.
Throughout the project we will use a mixture of genome-wide association studies, a suite of methods developed for Mendelian randomisation, as well as theoretical and simulation analyses of selection processes.
Evershed RP, Davey Smith G et al. Dairying, diseases and the evolution of lactase persistence in Europe. Nature. 2022 Aug;608(7922):336-345. doi: 10.1038/s41586-022-05010-7 – Example of how, in principle, Mendelian randomization could be used in future to infer fitness relationships
Hemani G, Knott S, Haley C. An evolutionary perspective on epistasis and the missing heritability. PLoS Genet. 2013 Feb;9(2):e1003295. doi: 10.1371/journal.pgen.1003295 – Exploration of the mechanisms maintaining genetic variation in the context of widespread additive genetic variation
Mills, M.C., Tropf, F.C., Brazel, D.M. et al. Identification of 371 genetic variants for age at first sex and birth linked to externalising behaviour. Nat Hum Behav 5, 1717–1730 (2021). https://doi.org/10.1038/s41562-021-01135-3 - A GWAS on reproductive traits
Okasha, S. & Otsuka, J. The Price equation and the causal analysis of evolutionary change. Philos. Trans. R. Soc. B Biol. Sci. 375, 20190365 (2020) – The relationship between causal inference and evolutionary modelling
Dr Nabila Kazmi (lead), Prof Sarah Lewis,
Prostate cancer is the second most common male cancer worldwide, but there is substantial geographical variation, suggesting a potential role for modifiable risk factors in prostate carcinogenesis.
Mendelian randomization (MR) analysis is a method which uses genetic variation as instrumental variables to investigate the casual relationship between exposure and outcome. In our previous MR analysis, we found that serum iron was protective against overall prostate cancer, where, OR = 0.92 (95% CI =0.86–0.98 and p-value = 0.007).
Now, it is known that blood cell characteristics are associated with prostate cancer risk. We know that iron is essential for haemoglobin production and has an impact on other blood cell characteristics. The purpose of this study is to determine whether effects of iron on blood cell characteristics could be responsible for the protective effect of iron on prostate cancer risk. To investigate this hypothesis, student will perform a Two-step MR project, with iron as the exposure, blood cell counts are the intermediate and prostate cancer as the outcome.
Students choosing this project must have taken the Molecular Epidemiology unit.
The aim of this study is to apply two- sample Mendelian randomization analyses to evaluate the evidence of a causal link between serum iron and red blood cell count and red cell distribution width (RDW).
Objective:
1. To find genetic variants that satisfy the instrumental variable assumptions and to test their associations with the outcome in the largest available dataset that is relevant to the causal question.
2. Using Inverse variance weighted (IVW) method to evaluate the primary causal effect estimate.
3. Using MR-Egger regression, weighted median method (WME), mode-based simple estimation, Mode-based weighted estimation to test the reliability and stability of the results.
Using data from large-scale genome-wide association study (GWAS) of exposure and intermediates and a GWAS of prostate cancer, we will investigate the causality between iron and intermediate biomarkers including red blood cell count and RDW using IVW method. Sensitivity tests will be performed to evaluate the robustness of the estimated results. IVW will be used as the primary causal effect estimate. We will adopt another four methods to test for pleiotropy, these are MR-Egger regression, weighted median method (WME), mode-based simple estimation, Mode-based weighted estimation. If the above five different MR models produce similar estimates of causal effects, we consider that serum iron level has a causal effect on the pathways investigated.
Bell S et al. (2021) A genome-wide meta-analysis yields 46 new loci associating with biomarkers of iron homeostasis. Communications Biology. 4(1):156.
Watts E et al. (2020) Hematologic Markers and Prostate Cancer Risk: A Prospective Analysis in UK Biobank. Cancer Epidemiology, Biomarkers and Prevention. 29(8):1615-1626.
Kazmi N et al. (2020) Appraising causal relationships of dietary, nutritional and physical-activity exposures with overall and aggressive prostate cancer: two-sample Mendelian randomization study based on 79,148 prostate cancer cases and 61,106 controls. International Journal of Epidemiology. 49(2), 587-596
Davey Smith G ES. (2003) “Mendelian randomisation”: can genetic epidemiology contribute to understanding environmental determinants of disease? International Journal of Epidemiology. 32, 1-22
Prof Sarah Lewis (lead), Dr Nabila Kazmi, Dr Rebecca Richmond
Worldwide there were over 4.5 million newly diagnosed cancers of the breast, colorectum, prostate and endometrium in 2018 representing the 2nd, 3rd, 4th and 14th most common cancers respectively. The World Cancer Research Fund as part of their continuous update project reviewed all the evidence from human studies on physical activity and cancer risk. They found that increased physical activity levels may protect against prostate, breast, colorectal and endometrial cancers.
We have previously applied two-sample Mendelian randomization (MR) to appraise the evidence for a causal link between lifestyle and anthropometric risk factors with overall and aggressive prostate cancer risk. The analyses showed that physical activity was inversely associated with overall prostate cancer. Separately we have conducted MR studies which show that physical activity can reduce the risk of colorectal and breast cancers, and others have performed a similar study to show that physical activity can reduce endometrial cancer risk. However, the biological mechanisms for this effect are not currently known.
Two-step MR analyses have been used to identify important biological pathways to provide evidence of their involvement in the disease process. The two-step approach first uses a genetic proxy for the exposure of interest (i.e., physical activity) to assess the causal relationship between exposure and an intermediate (i.e., a CpG site) which represents a biological pathway in the disease process. A second step then utilizes a genetic proxy for the same intermediate to interrogate the causal relationship between this intermediate and the outcome (i.e., cancer risk).
The aim of this project is to identify differentially methylated CpG sites associated with both physical activity and cancer risk which might indicate potential mechanisms between physical activity and cancer risk using Two-step MR approach.
The overall aim of this project is to identify differentially methylated CpG sites associated with both physical activity and cancer risk and to determine whether these CpG sites are on the causal pathway between physical activity and cancer risk.
Objectives:
1. The differentially methylated CpG sites in relation to physical activity have been previously reported by a study conducted in the Melbourne Collaborative Cohort study (MCC). Take all differentially methylated CpG sites identified to be associated with at least 1 measure of physical activity (p<0.001).
2. Identify differentially methylated CpG sites to be associated with breast, prostate, colorectal and endometrial cancers using published or publicly available data from epigenome wide association studies (EWAS).
3. Identify overlapping CpG sites between physical activity and cancer risk.
4. Using genetic instruments for physical activity and for overlapping CpG sites exploit publicly available genome wide association study (GWAS) data to carry-out a two-step MR analyses to determine whether these CpGs are on the casual pathway between physical activity and cancer risk.
This project aims to identify differentially methylated CpG sites which are associated with both physical activity and cancer risk in order to identify potential mechanisms between physical activity and cancer risk.
1. The student will use publicly available methylation datasets to identify differentially methylated CPG sites associated with both physical activity and one of the following cancer sites; breast, prostate, colorectal and endometrium.
2. The student will identify genetic instruments for physical activity and the CpG sites of interest to use in a two-step MR analysis.
3. The student will then perform a two-sample MR analyses to determine whether physical activity influences cancer risk via changes in methylation at the specific sites identified.
Bray F et al. (2018) Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA: a Cancer Journal for Clinicians.68(6), 394–424
Roekel V et al. (2019) Physical Activity, Television Viewing Time, and DNA Methylation in Peripheral Blood. Medicine and Science in Sports and Exercise.51(3), 490–498
Davey Smith G ES. (2003) “Mendelian randomisation”: can genetic epidemiology contribute to understanding environmental determinants of disease? International Journal of Epidemiology. 32, 1-22
Natalia Lewis (lead), Dr Elizabeth Cook, Dr Estela Capelas Barbosa Dr Sally McManus
Intimate partner violence has a negative impact on social and economic outcomes. However, it has not been established how long those social and economic consequences last and how the duration varies with the type of violence. Duration of such effects is needed to cost the social and economic harms from violence.
To quantify the average length of time for which social and economic outcomes are affected by intimate partner violence, by type of violence (sexual, physical, emotional, financial).
A focused systematic review of studies reported in peer reviewed literature that recruited adults, had multiple time points, a social and/or economic outcome(s) and where intimate partner violence was a predictor, independent variable, or inclusion criterion.
Support will be provided by supervisors and subject librarian.
Training on quantitative evidence synthesis is available via short course and guided self-learning.
Patton SC, Szabo YZ, Newton TL. Mental and Physical Health Changes Following an Abusive Intimate Relationship: A Systematic Review of Longitudinal Studies. Trauma Violence Abuse. 2022 Oct;23(4):1079-1092. doi: 10.1177/1524838020985554. Epub 2021 Jan 20. PMID: 33468040.
Page, M.J., McKenzie, J.E., Bossuyt, P.M. et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. Syst Rev 10, 89 (2021). https://doi.org/10.1186/s13643-021-01626-4
Prof Sarah Lewis (lead), Prof Richard Martin, Dr Richard Pulsford (University of Exeter) Prof Russ Jago Dr Freddie Bray (International Agency for Research on Cancer)
Around 40% of all cancers are thought to be avoidable by modification of lifestyle factors. Obesity has been found to be a risk factor for several cancers and was estimated to have caused around 3.6% of all new cancers which occurred in 2012. There is now an emerging evidence base which shows that low levels of physical activity can increase the risk of cancer. The World Cancer Research Fund (WCRF), as part of their continuous update project, have concluded that there is strong evidence that high levels of physical activity decreases the risk of cancers of the breast, endometrium and colon and rectum. In addition, we have recently shown using common genetic variation (a technique called Mendelian randomization) that physical activity protects against prostate, colorectal and breast cancer risk. In our analyses we found larger protective effects of physical activity on cancers of the breast, prostate, colon and rectum than were estimated by observational studies. It is possible that cancers at other sites are similarly influenced by physical activity but they have not yet been investigated in this way.
The aims of the project are to:
1. Identify the patterns of physical activity which are most likely to be causing cancer using wrist-worn accelerometer data and cancer outcomes in UK Biobank.
2. Test whether low levels of physical activity is a causal risk factor for cancer at several sites using Mendelian randomization and data from large cancer consortia
3. Use the best publicly available data to estimate the global prevalence of low levels of physical activity by country/region/ethnicity.
4. Estimate the global burden of cancer attributable to low levels of physical activity using cancer surveillance data compiled by IARC broken down by country/region and cancer type.
The student will be access wrist worn accelerometry data from UKBiobank or other similar population based cohorts. The student will characterised in a variety of different ways in order to investigate different features or patterns in behaviour and will use epidemiological techniques to understand the relationship between patterns of activity and cancer.
The student will be trained on the theory and application of Mendelian radomization and will use this to determine which cancers are caused by low levels of physical activity (and other patterns of physical activity).
The student will receive training in the interpretation of cancer registry data and will use this data alongside global data on physical activity to determine the proportion of cancers which can be attributed to insufficient physical activity worldwide.
https://bjsm.bmj.com/content/52/13/826.long
https://www.wcrf.org/dietandcancer/exposures/physical-activity
https://www.nature.com/articles/s41467-020-14389-8
Dr Qian Yang (lead), Dr Carolina Borges,
Previous observational studies suggested (1) maternal antenatal depression was associated with higher risks of preterm birth, low birthweight and intrauterine growth restriction; (2) antidepressants use during pregnancy were associated with higher risks of miscarriage, stillbirth and preeclampsia, and low Apgar scores. However, these findings may be vulnerable to residual confounding.
(1) To explore effects of maternal genetic liability to depression on pregnancy and perinatal outcomes using two-sample MR;
(2) To write-up methods and findings and co-author a paper of this study.
Genetic instruments and their associations with depression will be obtained from Psychiatric Genomics Consortium GWAS. Instrument-outcome associations will be extracted from MR-PREG GWAS, combining UKB, ALSPAC, BiB, MoBa and FinnGen. In ‘TwoSampleMR’ R package, inverse variance weighted will be used for main analyses with sensitivity analyses to assess MR assumptions.
Dr Qian Yang (lead), Prof Deborah Lawlor, Dr Charlie Hatcher
Hypertensive disorders of pregnancy (HDP) influences ~10% of pregnancies. MR found Bifidobacterium protected pre-eclampsia using FinnGen. However, it is important to replicate this independently. It might be useful to explore whether there is evidence for causal effects on gestational hypertension, and small-for-gestational age which is increased in women with HDP.
(1) To explore effects of gut microbiome reported in previous MR (PMID: 36380372) on pre-eclampsia (for replication), gestational diabetes, and small-for-gestational age using two-sample MR in an independent (to FinnGen) sample.
(2) To learn about additional methods for testing results validity.
We will follow the previous MR to extract genetic instruments and their associations with gut microbiome from MiBioGen consortium GWAS. Instrument-outcome associations will be extracted from MR-PREG GWAS, combining UKB, ALSPAC, BiB and MoBa, or InterPregGen GWAS. Two-sample MR will be conducted using ‘TwoSampleMR’ R package.
Josine Min (lead), Gibran Hemani, Johann Hawe (Illumina)
Genome wide associations studies (GWASs) have discovered many genetic associations with a large range of human traits, but the functional consequences of GWAS signals often remain elusive, as most GWAS signals reside in non-coding genomic regions. However, GWAS signals are enriched in DNA regulatory elements and cell type specific annotations, and thus it is likely that GWAS signals confer their effects through modulating gene regulatory mechanisms.
Genetic factors for molecular traits (DNA methylation, gene expression, protein levels) are being discovered at an astonishing rate. A major hope for these genetic factors is that they can be used to identify causal mechanism of complex traits.[1] Fascinatingly, the dimensionality of molecular phenotyping is bound to surpass the density of human genetic variation, meaning that genetic pleiotropy (where one variant influences multiple phenotypes) is a necessary feature amongst molecular phenotypes. This has critical downstream implications for being able to use genetics to make valid causal inference of putative molecular targets on disease incidence and progression.
This project will build a resource for storing and querying harmonized molecular QTL data in a computational efficient manner, and then use that resource to build pleiotropy maps of human molecular phenotypes. These maps will subsequently be used in evolutionary modelling and in collaboration with Illumina using machine learning and artificial intelligence approaches to understand the basis of molecular pleiotropy. This will include a research visit to Illumina AI lab in Germany.
1. Develop a computational framework for storing and querying molecular QTLs that will integrate with the OpenGWAS project
2. Generate pleiotropy maps using fine mapping and colocalization
3. Use evolutionary models to understand the impact of pleiotropy on natural selection processes
4. Use deep learning to predict disease mechanisms and disease progression from molecular pleiotropy maps
Currently summary statistics are stored for each GWA dataset separately. However this is not sustainable for QTL summary statistics with millions of molecular features. Therefore a new framework will be developed to store complete molecular QTL statistics for each dataset. Fine mapping and colocalization analysis will be used to integrate methylation QTL statistics from the Genetics of DNA Methylation Consortium, expression QTL statistics from eQTLGen and protein QTL statistics from SCALLOP and ALSPAC. This will result in maps of colocalized molecular traits. We will investigate biological models of pleiotropy, for example by using evolutionary models and gene-environmental interactions. We will use deep learning to identify molecular pleiotropy maps that correspond to distinct phenotypic patient subgroups.
1. Neumeyer S, Hemani G, Zeggini E. Strengthening Causal Inference for Complex Disease Using Molecular Quantitative Trait Loci. Trends Mol Med. 2020;26(2):232-41.
Josine Min (lead), Gibran Hemani, Johann Hawe (Illumina)
DNA methylation (DNAm) plays a central role in gene regulation. However, it is unknown how DNAm patterns change. For example, through genetic factors or physiological states.
Longitudinal birth cohort studies such as ALSPAC provide an unique opportunity to study DNAm patterns over time and to link it to varying physiological states. Quantitative traits comprising your physiological state (such as BMI, glucose and inflammation levels) have varying patterns over time. Identifying changes in DNAm preceding a change in physiological state may lead to identification of a marker that predicts disease outcome.(1)
Multiple studies have identified genetic variants associated with DNAm (mQTL: methylation quantitative trait locus) by combining genome wide genotype information with DNAm levels.(2) The Genetics of DNA methylation Consortium brought together a large number of cohorts to identify mQTLs in blood and investigated whether the mQTLs play a role in disease etiology.(3) Modelling DNAm trajectories with genetic variation could improve our understanding of biological mechanisms.
Aims: In this project we will identify longitudinal mQTL associations and examine whether these associations are involved in change of physiological state.
Objectives:
1) To identify longitudinal mQTLs across 4 timepoints in children and 2 timepoints in mums
2) To examine whether varying DNAm patterns predict varying disease/trait patterns
3) To examine whether longitudinal mQTLs are involved in disease progression
A longitudinal mQTL resource will be developed through genome-wide association analysis on DNAm levels in ALSPAC. Longitudinal trajectories of inflammation proteins and physiological traits (glucose, BMI) will be examined. To examine the relationship between longitudinal mQTL trajectories and these age-dependent health outcomes, we will use machine learning methods and mathematical modelling. Colocalization methods will be used to examine whether GWA loci for disease progression are shared with longitudinal mQTLs.
1. Chen R, Xia L, Tu K, Duan M, Kukurba K, Li-Pook-Than J, et al. Longitudinal personal DNA methylome dynamics in a human with a chronic condition. Nat Med. 2018;24(12):1930-9.
2. Gaunt TR, Shihab HA, Hemani G, Min JL, Woodward G, Lyttleton O, et al. Systematic identification of genetic influences on methylation across the human life course. Genome Biol. 2016;17:61.
3. Min JL, Hemani G, Hannon E, Dekkers KF, Castillo-Fernandez J, Luijk R, et al. Genomic and phenotypic insights from an atlas of genetic effects on DNA methylation. Nat Genet. 2021;53(9):1311-21.
Dr Eleanor Sanderson (lead), Dr Rebecca Richmond,
Multivariable MR (MVMR) is an extension of Mendelian randomization that allows for multiple exposures to be included in a single estimation. MVMR can be implemented with Genome-wide association study summary statistics. This enables the causal effect of multiple correlated traits on a single outcome, such as the causal effect of multiple lipids traits on coronary heart disease, to be estimated with existing publicly available data. However with many traits, such as lipids, it is not possible to include all potential exposures in a single model. This occurs because high correlation between the traits leads to low power and weak instruments. One potential approach to deciding how to include different traits is ‘stepwise’ MVMR where exposures are selected sequentially to include as many as possible while avoiding weak instrument bias.
The aim of this project is to explore different approaches to stepwise MVMR to establish the strengths and weaknesses of each approach.
This project will use an exemplar application of the effect of lipid traits on coronary heart disease to explore the application of different approaches to stepwise MVMR. We will use summary-data MR and MVMR methods applied through the ‘TwoSampleMR’ and ‘MVMR’ packages in R.
1. Sanderson, E., et al., An examination of multivariable Mendelian randomization in the single-sample and two-sample summary data settings. International journal of epidemiology, 2019. 48(3): p. 713-727.
2. Sanderson, E., et al., Mendelian randomization. Nature Reviews Methods Primers, 2022. 2(1): p. 6.