Angela K. Green, MD, MSc1; William A. Wood, MD, MPH1; Ethan M. Basch, MD, MSc

Use of cancer therapies for indications not approved by the US Food and Drug Administration (FDA), also known as “off-label” use, is widely practiced in oncology care.1 Examples include bevacizumab in metastatic esophageal cancer or cetuximab in metastatic prostate cancer. In the United States, off-label use of 10 common cancer drugs accounted for nearly $5 billion in costs in 2010,1 an amount that likely substantially underrepresents current off-label spending given the escalation in oncology drug pricing and number of available products, but recent comprehensive estimates have not been published.

The Centers for Medicare & Medicaid Services (CMS) uses federally designated third-party drug compendia, which are privately owned pharmaceutical reference guides, to inform coverage decisions for off-label cancer drug use in the United States. CMS is directed by the 1993 Omnibus Budget Reconciliation Act to provide coverage when the indication for an off-label cancer drug is supported by at least 1 of these compendia.2 Thus, the compendia, as the major source for federal payment decisions for off-label cancer drugs, have an influential role in the cost of cancer treatment. However, there is limited transparency about how compendia are assembled or about conflicts of interest on the part of their contributors, and there are substantial inconsistencies both between and within these resources.

There are currently 5 designated compendia: American Hospital Formulary Service (AHFS), National Comprehensive Cancer Network (NCCN) Drugs and Biologics, Clinical Pharmacology, Micromedex DrugDex, and Wolters Kluwer Clinical Drug Information Lexi-Drugs. AHFS is the only compendium remaining from the original set approved in 1993. In 2007, CMS created an annual open process to review public requests to add or remove compendia from the approved list. CMS uses a list of characteristics established by the Medicare Evidence Development and Coverage Advisory (MEDCAC) as the basis for its review for compendia, including completeness of listings, transparency in the evaluation process, descriptions of reviewed evidence, and a process for identifying (but not precluding) conflicts of interest. In 2008, CMS approved DrugDex, Clinical Pharmacology, and NCCN; in 2015, Lexi-Drugs was approved. Each compendium has a system for rating evidence and recommending therapies for indications, with no standardized approach between or among compendia (eTable in the Supplement).

Although federal legislation applies only to CMS to adhere to compendia indications, private payers also use compendia recommendations to support coverage decisions, often using different criteria than CMS or combining compendia recommendations with conclusions based on their own internal review processes. Some states require private payer coverage of drugs supported by specific compendia, leading to inconsistencies between states and within payer organizations.

A systematic review published in 20093 found that the quality of evidence cited in compendia for off-label cancer drug usage is less rigorous than the standards supporting FDA-approved drugs. This analysis of 14 off-label indications of cancer drugs found substantial limitations in the level, quantity, consistency, and timeliness of evidence among commonly used compendia. Evidence cited by the compendia was often not up-to-date and differed from evidence retrieved through an independent search by the authors. This raises concern that payers may be compelled to cover inadequately proven treatments for which the risks outweigh benefits. Despite the findings of this systematic review, this issue has not been addressed since then.

We examined the current compendia listings for the tyrosine kinase inhibitor erlotinib (which is currently FDA-approved for treatment of non–small cell lung cancer and pancreatic cancer). The off-label indications for erlotinib listed within recent versions of the compendia were compared, and the evidence cited for each indication was evaluated. Persistent inconsistencies were noted in recommendations between the compendia and methodological weaknesses in the analyses of the evidence. For example, erlotinib has no off-label indications in AHFS,4 1 off-label indication in Clinical Pharmacology (for head and neck cancer),5 8 off-label indications in DrugDex (colorectal cancer, newly diagnosed glioblastoma, recurrent glioblastoma, recurrent or metastatic head and neck cancer, renal cancer, non–small cell lung cancer, ovarian cancer, and prostate cancer),6 and 3 off-label indications in NCCN (bone cancer, leptomeningeal metastases, and renal cancer).7 Moreover, the evidence cited to support the off-label indications for erlotinib is of weak quality in several instances, such as single case reports, small case series, and a phase 1 trial (Table). In cases in which compendia disagree, generally a use is covered if at least 1 reference includes that indication.

Development of the compendia-based model for coverage decisions originated in an era when fewer drugs were available for patients, the cost of oncology drugs was lower, there was less appreciation of the potential long-term harms of treatments, and views on evidentiary standards differed from today. Given changes in the landscape of cancer treatment and costs, this model should be reassessed.

First, clear methodological standards should be specified for evaluating the evidence and generating indications for off-label coverage. Scrutiny should be of the same rigor as regulatory review for new indications or clinical practice guidelines for which evidentiary standards exist.8 Given substantial harms that oncology drugs may confer, patients deserve clear evidence that benefits outweigh risks when recommendations are made for use of off-label products. Second, the inconsistencies between existing compendia are a source of confusion. It is also a waste of resources for this work to be duplicated by various groups. The quality of listings could potentially improve if these efforts were combined into a single rigorously developed resource. Third, contributors to the compendia should have no financial relationships with the pharmaceutical industry, and this information should be transparent and made public. When names of authors or consensus committee members are provided for compendia listings (which they often are not), financial relationships with industry frequently are apparent.9,10 Moreover, drug manufacturers interact with compendia developers and can lobby for indications to be added or altered. Fourth, compendia listings should be publicly available free of charge. Currently, compendia are not widely available and are accessed through subscription services (which is often a source of revenue for their publishers), limiting the possibility of public scrutiny or availability to organizations that do not subscribe.

Toward the future, this function would be most appropriately performed under the direct oversight of the FDA, perhaps within the new FDA Oncology Center of Excellence, or alternatively by a new entity supported by a consortium of public and private payers with enforced policies to limit relationships with industry. Under such a model, new indications for existing therapies could be nominated along with accompanying evidence for impartial review and decision making through a consistent, rigorous, and transparent review process. Evidence standards might differ from full regulatory approval, with a likely role for real-world evidence. Although this work would entail substantial effort, it would undoubtedly be less than the current compendia efforts and would likely lead to more rigorous indications. This approach also could alleviate concerns about conflicts of interest among current compendia contributors and would likely engender greater transparency.

The current compendia-based approach for coverage decisions in oncology has lacked oversight and is not suitable for making safe recommendations or coverage decisions for cancer drugs. Given the clear limitations of this outdated system, a different model is warranted.

JAMA