Early dissemination of bevacizumab for advanced colorectal cancer: a prospective cohort study

Bevacizumab, a monoclonal antibody against vascular endothelial growth factor, was the first biologic agent shown to improve median overall survival (by 4.7 months) in patients with metastatic colorectal cancer, when given with cytotoxic chemotherapy [1]. Results of the first phase III study demonstrating this benefit were released publically in June, 2003, and published in June, 2004. The United States (US) Food and Drug Administration (FDA) approved bevacizumab for first-line treatment of metastatic colorectal cancer in February, 2004. The drug was approved by the European Union a year later in January, 2005. Although effective in prolonging survival, bevacizumab is expensive,[2] can cause hypertension, cardiovascular events, and in rare instances, severe hemorrhage or gastrointestinal tract perforation [1, 3].

Little is known about the early uptake of new biologic therapies such as bevacizumab for advanced cancer. Specifically, 1) what is the initial dissemination pattern of biologic agents?; and 2) how do clinicians interpret clinical trial results for biologic agents, especially since selective trial eligibility might not represent the broad spectrum of patients treated in the community setting?[4] Registry studies for bevacizumab have examined post-marketing safety and efficacy, but these studies cannot answer our two questions since all enrolled patients received bevacizumab [5, 6].

Answering these questions is critically important in light of recent regulatory decisions related to bevacizumab. At the end of 2010, the FDA recommended removing bevacizumab's breast cancer indication due to lack of survival benefit; the European Medicines Agency has maintained its indication but only in combination with paclitaxel. In light of the high drug cost, we must understand how novel agents are used in standard care since in certain instances, rapid dissemination occurs despite lack of long-term, phase IV data.

We used data from the Cancer Care Outcomes Research and Surveillance (CanCORS) Consortium, a US population- and healthcare systems-based study of newly-diagnosed patients with colorectal cancer, to describe the uptake of bevacizumab and to identify factors associated with its use. Comprehensive medical record abstraction enabled us to evaluate bevacizumab use in relationship to comorbidity and other patient characteristics not typically available in cancer registry data [7]. Thus, we were able to examine dissemination of this new therapy in a broad range of community settings.

Over the two years following approval of bevacizumab for first-line treatment of advanced colorectal cancer, we found that 31% of a nationally-representative sample of treated patients received the agent as a component of their care. Bevacizumab use showed an initial rapid uptake in the first 6 months after FDA approval and then leveled off. Seventy-five percent of patients received bevacizumab with intravenous 5-fluorouracil (with or without other chemotherapy), in accordance with the label. Only 19% of patients received bevacizumab with irinotecan and 5-fluorouracil, as described in the pivotal clinical trial which led to FDA approval. Prior to this study, little was known about the patterns of bevacizumab use soon after FDA approval, and little was known about the factors associated with early use. Registry studies for bevacizumab examined post-marketing safety and efficacy,[5, 6] but all enrolled patients in those studies received bevacizumab; hence, those studies could not address questions related to dissemination.

The rapid increase noted in the first year was unprecedented for a cancer drug,[17] and was likely a consequence of two factors. First, when it was approved, the drug received a great deal of media attention, including being called "revolutionary."[18] This attention presumably promoted initial uptake by oncologists. Second, bevacizumab was the first anti-angiogenic agent to demonstrate improvement in overall survival for colorectal cancer-a potential paradigm shift in cancer treatment that also may have stimulated an upsurge of initial interest on the part of oncologists [19].

Why, despite the rapid rise, did early bevacizumab uptake level off at less than one-third of eligible patients? Our data suggest that clinicians may have restricted their use to a sub-segment of potentially eligible patients. We found that older age, even after adjusting for comorbidity, was associated with less bevacizumab use. The mean age of patients who received bevacizumab in our study was 58.2 years, nearly identical to the mean age of 59.5 years in the pivotal phase III trial [1]. Early adopters of the drug may have worried that its safety in elderly patients was not adequately addressed by the trial design and were hesitant to use the novel therapy in older patients without post-approval data. Hence, they limited use of the drug to patients who were similar to those treated in the definitive clinical trial. Similar findings have been noted among colorectal cancer patients eligible for adjuvant chemotherapy: the elderly receive lower doses and shorter duration of therapy than recommended by trial data [20].

Men were less likely than women to receive bevacizumab. It is unclear why gender played a role, especially since prior studies investigating the impact of gender on treatment of colorectal cancer have found no difference [21, 22]. Men are more likely than women to have subclinical cardiovascular disease [23] and to be taking aspirin [24]. In the phase III trial of bevacizumab, patients were excluded if they had significant cardiovascular disease, regular aspirin use (> 325 mg per day), preexisting bleeding disorders, or full-dose anticoagulation [1]. Although we adjusted for cardiovascular comorbidity, our measures may not have fully captured these specific conditions. Clinicians might have been more hesitant to use the drug in men with greater cardiovascular risk given the exclusion criteria in the pivotal trial. To ensure that the gender effect was not due to confounding by the largely male Veterans Administration population we did a secondary analysis excluding the Veterans Administration patients and found no change in the association between gender and receipt of bevacizumab.

The FDA approved bevacizumab in 2004 to be used in combination with intravenous 5-fluorouracil-based chemotherapy,[13] based on results of the pivotal phase III study, which used bevacizumab in combination with irinotecan and 5-fluorouracil (IFL) [1]. Although the pivotal trial was designed with IFL as the chemotherapy backbone, the label considered bevacizumab in combination with any first-line 5-fluorouracil-containing regimen to be on label. The relative flexibility that the FDA provided to oncologists by labeling it for use in combination with any 5-fluorouracil-containing regimen might have been the result of rising concern over excessive rates of treatment-associated deaths with IFL [25, 26]. Hence, oncologists could use bevacizumab in combination with oxaliplatin and still remain in accordance with the FDA indication.

Despite lack of evidence at the time for use with oxaliplatin, more than twice as many patients in our study received bevacizumab with oxaliplatin than with irinotecan. Due to the concern over treatment-associated deaths with IFL, clinicians may have opted to use oxaliplatin-based regimens with the assumption that the incremental advantage of adding bevacizumab would be similar to IFL. In fact, when evidence was subsequently published in 2008 for bevacizumab given in combination with 5-fluorouracil and oxaliplatin (FOLFOX), no significant improvement in overall survival was demonstrated over FOLFOX alone [27]. Additionally, some bevacizumab-receiving patients in our study were not treated with first-line 5-fluorouracil at all, thereby receiving off-label treatment. A few bevacizumab-receiving patients were treated with capecitabine, an approach that was not supported by published evidence or FDA indication.

We observed a somewhat complex relationship between the initial evidence supporting use of bevacizumab in patients with advanced colorectal cancer and early patterns of uptake. Clinicians were apparently guided by the eligibility criteria of the trial, hesitating to treat older patients or those at risk for cardiovascular compromise. However, they chose to give the drug in combination with chemotherapy regimens not supported by the FDA or existing evidence at the time. Future studies of the dissemination of subsequently approved targeted agents will be needed to determine whether these patterns were specific to adoption of this new biologic agent or apply more broadly to other biologic treatments for cancer. We also saw no association with patient preferences or beliefs. Our findings thus suggest that oncologists, not patients, were the primary decision-makers about whether to include bevacizumab for patients treated with chemotherapy.

Our analysis revealed regional disparities, with patients living in the Atlantic region being more likely than those in other regions to receive bevacizumab. Regional variation has been well-described in relation to delivery of both surgery and chemotherapy for various cancers [28–33]. Our findings of regional variation are unlikely the result of differential reimbursement for bevacizumab since the US Centers for Medicare & Medicaid Services (CMS) approved coverage for bevacizumab retroactive to the FDA approval date. Furthermore, the regional variation is unlikely to have been the result of formulary availability within each health system, since in the US anticancer drugs are generally available on formulary at the time of FDA approval. In their review of 59 anticancer drugs (including bevacizumab), Mason et al. found that 100% of drugs were covered by CMS and the Veterans Administration health system from the time of FDA approval [34]. Additionally, at least in the Veterans Administration health system, bevacizumab was available prior to formulary listing but after FDA approval via non-formulary request (personal communication, Geraci M, 2010).

Our study has several limitations. While we assessed patient preferences and beliefs, we could not assess preferences specific to bevacizumab. Our cohort is a small sub-sample drawn from a relatively large cohort, though it is large enough to support robust analysis, and CanCORS enrollees are representative of patients included in the SEER cancer registry [9]. Additionally, longitudinal follow-up of our cohort was short.

We found a rapid uptake of bevacizumab after FDA approval, though only a third of eligible patients received the drug. Use of bevacizumab in practice differed from that described in clinical trials leading to FDA approval. Dissemination of novel agents might be improved through comparative effectiveness research measuring the benefits and safety of treatments through post-marketing (phase IV) studies that track the dissemination of new therapies and their associated outcomes.

This work and the CanCORS Consortium are supported by grants from the National Cancer Institute to the Statistical Coordinating Center (grant number U01 CA093344) and the National Cancer Institute-supported Primary Data Collection and Research Centers (Dana-Farber Cancer Institute/Cancer Research Network (grant number U01 CA093332); Harvard Medical School/Northern California Cancer Center (grant number U01 CA093324); RAND/UCLA (grant number U01 CA093348); University of Alabama at Birmingham (grant number U01 CA093329); University of Iowa (grant number U01 CA093339); and the University of North Carolina (U01 CA 093326); the Agency for Healthcare Research Quality (grant number 03-438MO-03); and by a Department of Veterans Affairs grant to the Durham VA Medical Center (grant number HSRD CRS 02-164). The study sponsors (National Cancer Institute, the Agency for Healthcare Research Quality, and the Department of Veterans Affairs) played no role in study design, data collection, analysis, interpretation, writing, or decision to submit for publication.