Novel and existing flexible survival methods for network meta-analyses - Supplementary materials

  

Figure 1. A) Overall survival hazard plots in evidence network for the pooled Checkmate 017-057 trials comparing nivolumab with docetaxel. B) Keynote 010 trial comparing pembrolizumab with docetaxel, C) POPLAR trial comparing atezolizumab with docetaxel, D) OAK trial comparing atezolizumab with docetaxel  

  

Figure 2. Network of evidence

  

Figure 3. Log cumulative hazard plots  and Schoenfeld plots for A) pooled checkmate 017 and 057 trials, Keynote 010; C) POPLAR D) OAK

  

Figure 4 Predicted survival best fitting reduced models and corresponding 95% credible interval on therapy with longest predicted survival; A) loglogistic standard parametric, B) loglogistic piecewise, C) loglogistic spline, D) Loglogistic non-mixture cure, E) loglogistic mixture cure and F) Weibull mixture, G) second order Fractional polynomial.

  

Table 1 Overview of the percentiles related to placements of knots and the corresponding uncensored survival times in the considered NMA dataset. 

 Table 2. LOOICs of fully specified (treatment effects on all parameters) survival network meta-analysis