Bevacizumab-induced hypertension as a predictor of clinical outcome in metastatic colorectal cancer. An individual patient data-based pooled analysis of two randomized studies and a systematic review of the literature

Background: Retrospective series suggest that bevacizumab-induced hypertension (HTN) is a prognostic and potentially predictive biomarker of efficacy of the antiangiogenic drug in the upfront treatment of metastatic colorectal cancer (mCRC) patients. The immortal-time bias and the effect of pre-existing HTN might affect these findings. We conducted a pooled, post hoc analysis of 2 prospective randomized trials of chemotherapy plus bevacizumab in mCRC, and performed a systematic review of the available literature focusing on how the immortal-time bias was taken into account and how pre-existing HTN potentially requiring the use of antihypertensive drugs was managed. Methods: The pooled-analysis included patients enrolled in the phase III TRIBE and TRIBE-2 studies that compared upfront FOLFOXIRI + bevacizumab to FOLFIRI or FOLFOX + bevacizumab, respectively. Association between HTN and survival outcomes was assessed by incorporating a time-dependent Cox regression model to consider the time-dependency of the probability of HTN onset during the treatment. The systematic review was conducted according to PRISMA guidelines. Results: The systematic review retrieved 14 eligible and highly heterogeneous studies. A positive prognostic impact of bevacizumab-induced HTN was reported in the 58% of the analyses reporting Progression Free Survival (PFS) and in the 54% of the analyses reporting Overall Survival (OS) data. Immortal-time bias was incorporated in 4 studies (28%). In TRIBE and TRIBE-2 study populations (N = 1175), patients experiencing ≥ G2 HTN during first-line bevacizumab administration showed longer PFS (median: 14.7 versus 10.3 months, p < 0.001) and OS (median: 31.7 versus 24.2 months, p < 0.001). The association with OS retained statistical significance after correction for time-dependency (p = 0.003) and was confirmed in the multivariable model including HTN as a time-dependent variable (p = 0.02). Moreover, in patients with pre-existing HTN, no difference in terms of PFS and OS was observed compared with the subgroup of patients who never experienced ≥G2 HTN (HR 1.01, p = 0.86 and HR 1.02, p = 0.78 respectively. Conclusions: Bevacizumab-induced HTN during the first-line treatment of mCRC is an independent prognostic factor, also adopting a time-dependency correction. Toxicity should be interpreted as a time-dependent variable when exploring its association with clinical outcome.