Bayesian methods in health technology assessment: application to overactive bladder syndrome

This thesis addresses itself to the challenges of health technology assessment (HTA) to inform healthcare policy decision making in the presence of missing or sparse data. HTA advocates the use of evidence synthesis methods. Such an approach involves the analysis of clinical trial data through to the collation, and synthesis, of all relevant information pertaining to the decision question. Motivated by an example in overactive bladder syndrome (OAB), methodological developments together with practical applications are explored throughout the evidence synthesis process. This thesis begins with a novel application of Bayesian methodology to evaluate a large randomised controlled trial (RCT) of repeat treatment in patients with interval-censored data. Performance of Bayesian prediction models were assessed for varying proportions of missing data, and misspecification of distributional form, through a series of simulation studies. Following this, all RCTs evaluating interventions for OAB were identified in a systematic review, and critically appraised. In the current literature, all cross-modality treatment comparisons were performed using pairwise meta-analyses. In this thesis, a cross-modality treatment comparisons was performed using network meta-analysis (NMA) methods in order to obtain treatment effect estimates in terms of efficacy, safety, and tolerability. Network meta-regression techniques were employed to investigate the impact of potential treatment effect modifiers including baseline effects. Building on the general NMA framework, this model was extended to account for similarities between the same interventions with different methods of administration, making use of a natural treatment hierarchy, and where appropriate, incorporating dose-response constraints. Use of hierarchical NMA models increased the precision of treatment effect estimates used for decision-making. The hierarchical NMA model was further extended to incorporate a multivariate approach. This approach borrowed information across outcomes, increasing the precision in the treatment effect estimates. Multivariate hierarchical NMA allowed for the comparison of all interventions across all outcome measures, ameliorating the impact of outcome reporting bias, and thus increasing the ability to make decisions for healthcare policy. In doing so, sacral nerve stimulation (SNS) appeared to be the most promising intervention for the management of OAB.