Background: The FDA has issued hundreds of cancer drug indications, with many new drugs, expanded indications, and biosimilars approved in recent years. While the gold standard for regulatory approval is the randomized controlled trial (RCT), RCT design including selection of control arms can differ considerably. We sought to investigate trends and patterns in RCT trial design used to support FDA approvals in oncology. Methods: We reviewed the available FDA package inserts of oncology drugs (N=258) for RCTs cited to support initial and expanded indication approvals as of January 2020; biosimilars were excluded. RCTs were linked to the HemOnc ontology, which contains trial-level metadata including publication year, endpoints, and trial design. Log-linear regression was performed to evaluate trends in approvals over time by endpoint. Study drugs were categorized as cytotoxic therapy, targeted therapy, or immunotherapy. RCTs were categorized by four designs: escalation (adding a drug or increasing the drug dose in an established regimen), in-class comparison (comparing two drugs in the same therapeutic class), out-of-class switch (comparing drugs in distinct therapeutic classes), and de-escalation (removing a drug or reducing the drug dose in an established regimen). Results: We identified 556 registration trials, 372 (67%) of which were RCTs. Approvals have been increasing exponentially over time (R² 0.9, p<0.001), both for RCTs reporting overall survival (OS) endpoints (R² 0.77, p<0.001), and non-OS endpoints (R² 0.67, p<0.001). Of the three most common trial designs (Table), in-class comparisons were least likely to report OS (28%; escalations 47%; out-of-class switches 43%, p=0.01 by Chi-squared). Class switches were common in immunotherapy trials compared to targeted or cytotoxic therapy. Conclusions: Despite growth in FDA approvals, a minority of registration trials report paradigmatic shifts in therapeutic approach (out-of-class switches), with the relative exception of immunotherapy trials. Escalation is the most common route to FDA approval, even though this design inevitably increases cost and toxicity. This suggests that new oncology drug approvals are not alone a useful metric of practice-changing innovation.

*some trials tested multiple categories