Adjusting for confounding effects of treatment switching in a randomized phase II study of dabrafenib plus trametinib in BRAF V600+ metastatic melanoma.

Patients with BRAF V600E mutation-positive melanoma who were assigned to 150 mg dabrafenib twice daily combined with 2 mg trametinib once daily in a phase I/II study showed a median overall survival (OS) of 23.8 months, compared with 20.2 months for patients assigned to dabrafenib alone [hazard ratio (HR)=0.73, 95% confidence interval (CI) 0.43-1.24; data cutoff March 2013], on the basis of an intention-to-treat analysis. Because patients assigned to dabrafenib monotherapy were allowed to switch to combination therapy upon disease progression, we attempted to adjust for confounding effects on OS. Randomization-based adjustment methods, Rank Preserving Structural Failure Time Models and the Iterative Parameter Estimation algorithm, were used. Two analyses, 'treatment group' (assumes that treatment effect continues beyond treatment discontinuation) and 'on treatment' (assumes that the treatment effect disappears upon treatment discontinuation), were used to test assumptions on the durability of the treatment effect. A total of 45/54 (83%) patients assigned to dabrafenib monotherapy switched to the trametinib/dabrafenib combination. Adjusted OS HRs ranged from 0.47 to 0.50, depending on the analysis, compared with the unadjusted OS HR of 0.73. CIs continued to cross 1.00; thus, adjusted estimates did not provide statistically significant evidence of a treatment benefit on survival. Reduction of HRs after adjusting for the effect of treatment switching suggests that the intention-to-treat analysis underestimates the effect of dabrafenib plus trametinib on OS, although several factors, such as small trial size and methodological assumptions, affect the certainty of the conclusions.

RCT Entities:   Population Interventions Outcomes