*3.7. Treatment Ranking*

The Gelman plot for checking Bayesian model assumption shows a low chain reduction over time for both ORR and PFS outcomes, and the chains seem roughly converged after maximum 10,000 iterations in chain (Figures S6 and S7). Also, cabozantinib and alectinib were found to become the first-line therapies with the highest treatment ranking probabilities of 62.5% for ORR and 87.5% for PFS, respectively (Tables S6 and S7 and Figure S8). In the k-means clustering analysis of SUCRA, ceritinib, alectinib, crizotinib, osimertinib, cabozantinib, and afatinib showed the more efficacy compared with the remaining treatment (Figure S9).

Figure S10 reports the two-dimensional graphs about RR for ORR and HR for PFS in the comparison with dummy group. DHRR indicated the decrease of HR obtained per 1 unit increase of RR for osimertinib (0.34), alectinib (0.28), bevacizumab (0.38), and vandetanib (0.14), which are higher than that for other drugs relating to EGFR, ALK/ROS1/MET, VEGF, and RET pathways.
