Browse/search for people

Publication - Professor Sofia Dias

    Assessing the robustness of recommendations made in a guideline on specialist neonatal respiratory care in babies born preterm with threshold analysis

    Citation

    Phillippo, DM, Daly, CH & Dias, S, 2019, ‘Assessing the robustness of recommendations made in a guideline on specialist neonatal respiratory care in babies born preterm with threshold analysis’.

    Abstract

    Aim: To pilot a recently published threshold analysis method during development of a National Institute for Health and Care Excellence (NICE) guideline, Specialist neonatal respiratory care in babies born preterm, to assess the robustness of recommendations informed by network meta-analyses (NMAs) of two outcomes.
    Background: Concerns on the reliability of NMA results may arise if included studies are flawed in their conduct or reporting. Threshold analysis allows decision makers to assess the robustness of recommendations based on the NMA to potential biases in the included evidence.
    Methods/Process: After the guideline committee discussed the evidence including any concerns regarding bias and drafted recommendations, there was concern of bias arising from treatment switching in all studies due to lack of blinding. Threshold analysis was therefore carried out for the best- and worst-ranked treatments, to assess the robustness of positive and do-not-do recommendations, respectively.
    Results/Lessons Learnt: The recommendation to use the best-ranked treatment arising from the NMA on bronchopulmonary dysplasia was robust. However, for the mortality outcome the recommendation was sensitive to imprecision from a single study contributing to the NMA, and the committee considered this when re-evaluating the positive recommendation. The recommendation to not use the worst-ranked treatment, based on the results from the NMAs on both outcomes, was robust, providing some reassurance of this do-not-do recommendation.
    Conclusion: Results of threshold analyses lead to further scrutiny of the evidence to which the recommendations are sensitive and may placate concerns of potential biases which would not affect the treatment recommendation.

    Full details in the University publications repository