Bevacizumab maintenance therapy extended OS in patients with NSCLC

Dranitsaris and colleagues reported the results of an observational comparative effectiveness research study testing the hypothesis that maintenance bevacizumab (Avastin, Genentech) improves OS after initial treatment with bevacizumab-containing chemotherapy regimens in community-based patients with advanced nonsquamous non–small cell lung cancer. Using multiple analytical techniques to handle selection bias, the authors concluded that maintenance bevacizumab is associated with prolonged OS compared with no maintenance. The authors addressed a relevant question from the clinical and health policy perspectives, since bevacizumab is an expensive agent, with infrequent but potentially serious toxicities (including pulmonary hemorrhage and thrombotic events), and no randomized trials have yet directly evaluated the specific impact of bevacizumab in the maintenance setting. For conducting this study, Dranitsaris and colleagues should be congratulated by the timely attempt to provide an unbiased evaluation of the role of maintenance bevacizumab through the use of landmark analysis and propensity score adjusting and weighting.

Despite the authors’ laborious statistical analysis, residual selection bias probably still confounds the interpretation of their results. The study included 272 patients from a network of 23 community oncology clinics treated over 2.5 years, a number surprisingly low for what one would expect from this broad pool of patients. This small patient sample suggests that the treating oncologists were cherry-picking the fittest patients for bevacizumab-based therapies, which limits the ability to extrapolate these results to the broad community of advanced NSCLC patients. Unfortunately, the authors did not disclose the variables kept in the final models used to calculate the propensity scores. Patients treated with maintenance bevacizumab differed in many characteristics from patients not treated. Maintenance-treated patients were more likely to receive standard second-line therapies, less likely to have changes in their first-line regimen, and less likely to experience serious adverse events. If any of these characteristics were left out of the propensity score models and were associated with survival, then residual selection bias would have distorted the estimates of the benefit of bevacizumab maintenance. Performance status information was not available for one-third of patients, and therefore a residual imbalance of this well-known prognostic indicator could also have confounded results.

Other limitations apply to this study. In all propensity-score adjusted models, the 95% CIs for the HRs crossed the null effect boundary of 1.0, raising the concern of lack of power or a lack of a true benefit of maintenance bevacizumab. None of the statistical methods used fully account for the impact of nonmeasured factors, such as tumor molecular characteristics. Not accounting for nonmeasured factors could have also resulted in selection bias. Finally, the drug manufacturer that produces bevacizumab sponsored the study, highlighting the need for extra careful interpretation of these results.

The question then remains: Does maintenance bevacizumab improve survival outcomes in advanced NSCLC? Based on this study, one can conclude that maintenance bevacizumab is feasible and potentially effective in fit patients with nonprogressive disease after first-line chemotherapy. In order to provide timely and more robust evidence of effectiveness, future observational CER studies need to include larger number of patients that are actually reflective of the community patient population, and consider additional methods for handling selection bias, including instrumental variable analysis.

Platinum-doublet chemotherapy has been the standard of care for the first-line treatment of patients with metastatic non–small cell lung cancer since 1995. In 2006, the FDA approved the addition of the anti-VEGF antibody bevacizumab (Avastin, Genentech) to carboplatin and paclitaxel for first-line treatment of advanced non-squamous NSCLC based upon the ECOG 4599 trial, which showed an improved OS with the combination compared to chemotherapy alone (Sandler A. N Engl J Med. 2006;355:2542-2550). In this landmark trial, the treatment arm containing bevacizumab also included continuation maintenance with single-agent bevacizumab, and so this maintenance treatment also was included in the approved indication. Based on the study design, however, it is not possible to know to what degree — if any — that maintenance bevacizumab contributes to the efficacy of this regimen.

In a retrospective analysis that was funded by the maker of bevacizumab, Dranitsaris and colleagues studied the outcomes of 272 advanced NSCLC patients at 17 community oncology practices who received chemotherapy plus bevacizumab. The researchers compared the OS in patients who went on to receive maintenance bevacizumab to those who never received the maintenance but only received bevacizumab in the induction chemotherapy regimen. The patients who received the maintenance bevacizumab had a significantly longer OS compared with those who did not (23 months vs. 10.3 months), and the authors concluded that this is evidence supporting maintenance bevacizumab.

There are, however, a number of important problems with this analysis. First, this was not a randomized study, and so the two groups of patients were not matched for confounding variables. The authors even pointed out that there was significant selection bias: The patients who received maintenance bevacizumab were younger and healthier, received more cycles of first-line chemotherapy, responded better to that chemotherapy, and also got significantly more subsequent therapy later on, all favorable prognostic factors that likely influenced their survival. Although the authors attempted to correct for this using statistical models, the level of discordance of important prognostic variables in the two arms simply makes any conclusion about maintenance impossible. For now, maintenance bevacizumab remains the standard of care for patients who received it as part of their induction regimen, but the level of evidence supporting a benefit from this approach is still lacking.

The good news is we will have randomized trial evidence soon to answer this question. The AVAPERL study, which has been completed, randomly assigned NSCLC patients who completed four cycles of cisplatin, pemetrexed and bevacizumab to maintenance with either pemetrexed plus bevacizumab or bevacizumab alone. This has been presented in preliminary form  (Verma S. 2011. ECCO-ESMO), indicating a significantly longer PFS with the combination arm. In addition, the phase 3 ECOG 5508 trial, currently enrolling, randomly assigns NSCLC patients who have completed four cycles of carboplatin, paclitaxel and bevacizumab to maintenance with either bevacizumab alone, bevacizumab plus pemetrexed, or pemetrexed alone. This should definitively answer the question of the importance of bevacizumab maintenance.