Role of surgery in advanced epithelial ovarian cancer : the use of evidence synthesis methodology to support practice guidelines

Abstract

As Statistical Editor of the Cochrane Gynaecological, Neuro-oncology and Orphan Cancers review Group for many years, I have developed a keen interest in gynaecological cancer research, in particular in advanced stage epithelial ovarian cancer. Of women with ovarian cancer, more than 70% have epithelial ovarian cancer. The aim of this PhD was to examine the role of surgery for advanced epithelial ovarian cancer because of a lack of current firm guidelines to support clinical practice. The work described in this thesis uses evidence synthesis and meta-analysis methodology, applied in both traditional and novel ways to attempt to address this main aim. The body of work extends beyond standard approaches to develop, explore, and apply methods that aimed to raise the certainty of the evidence. I conducted two systematic review publications on the type and radicality of primary surgery. These used Cochrane Intervention Review methodology and met or exceeded the Methodological Expectations of Cochrane Intervention Reviews (MECIR). They followed what is widely considered ‘gold standard methods’ for this type of review, including the use of a standard pairwise metaanalysis approach. One publication found, with high to moderate-certainty evidence, that there may be little difference between primary debulking surgery and interval debulking surgery in survival outcomes for treatment of epithelial ovarian cancer. The other, found only very low-certainty evidence for all outcomes comparing maximal effort debulking surgery and standard surgery. To offer a different evidence perspective given the limitations of these reviews, a further prognostic factor review assessed the impact of residual disease on overall survival. This Cochrane prognostic review demonstrated the prognostic effect of debulking to no macroscopic residual disease (0 cm) in a primary debulking surgery setting (moderate-certainty evidence). Evidence for interval debulking surgery was sparse, so further work presented in the thesis focused on primary debulking surgery where there was more available evidence. I note that the body of work in the thesis did identify some evidence in an interval debulking surgery setting that if a tumour is not debulked to 0 cm, then all other residual disease thresholds may be ii sub-optimal and restricting the tumour to <1 cm may not matter. This finding has not been explored or reported in any other guidelines but needs more exploration when more studies adequately report these comparisons. Given the strong association between residual disease as a prognostic factor after primary debulking surgery and prolonged survival, the thesis then focused on methodologies that aimed to improve, if possible, the confidence in effect estimates presented in the primary analysis. This included a frequentist network meta-analysis and expert elicitation with a Bayesian network meta-analysis application that fully exploited all available evidence. These methods further consolidated the results of the primary meta-analyses reported in the Cochrane prognostic factor systematic review and provided moderate-certainty evidence, based on incorporating a more thorough and informed assessment of the publication bias domain in the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) approach and the overall certainty of the evidence judgement. These methods were applied and developed in the body of work in the thesis to demonstrate added confidence to the certainty of the evidence judgements, but more specifically that the results and conclusions of the primary prognostic factor review can be strengthened to the extent of potentially influencing policy. The evidence in the thesis suggests there is a clear benefit of achieving cytoreduction to no macroscopic residual disease. It may encourage the surgical community to attempt to increase rates of maximal effort debulking in their centres in order to achieve higher rates of cytoreduction to no macroscopic residual disease. The thesis also outlines several limitations and methodologies that require further development but could be implemented in EOC research in the future. At present, the National Institute for Health and Care Excellence may wish to consider the results of the thesis and the possible adoption of some of the proposed methods in their pending guidelines.