Second-line Treatments for Advanced Gastric Cancer: A Network Meta-Analysis of Overall Survival Using Parametric Modelling Methods

Advanced gastric cancer (AGC) is one of the most common forms of cancer and remains difficult to cure. There is currently no recommended therapy for second-line AGC in the UK despite the availability of various interventions. This paper aims to compare different interventions for treatment of second-line AGC using more complex methods to estimate relative efficacy, fitting various parametric models and to compare results to those published adopting conventional methods of synthesis.

Seven studies were identified in an existing literature review evaluating seven comparators, which formed a connected network of evidence. Citations were limited to randomised controlled trials in previously-treated AGC patients. Evidence quality was assessed using the Cochrane Collaboration’s tool. Studies were assessed for the availability of Kaplan–Meier curves for overall survival. Individual patient data (IPD) were recreated using digitisation software along with a published algorithm in R. The data were analysed using multi-dimensional network meta-analysis (NMA) methods. A series of parametric models were fitted to the pseudo-IPD. Both fixed and random-effects models were fitted to explore long-term survival prospects based on extrapolation methods and estimated mean survival.

Relative efficacy estimates were compared to those previously reported, which utilised conventional NMA methods. Results presented were consistent within findings from other publications and identified ramucirumab plus paclitaxel as the best treatment; however, all the treatments assessed were associated with poor survival prospects with mean survival estimates ranging from 5.0 to 12.7 months.

Whilst the approach adopted in this paper does not adjust for differences in trial patient populations and is particularly data-intensive, use of such sophisticated methods of evidence synthesis may be more informative for subsequent cost-effectiveness modelling and may have greater impact when considering an indication where observed data is particularly immature or survival prospects are more positive, which may then lead to more informative decision-making for drug reimbursement.