Identifying Minimum Threshold Level of Trial Crossover Rates Yielding Nonsignificant Overall Survival Benefit Associated with Cancer Treatments: A Systematic Literature Review of HTA Submissions.

Background: Crossover distortion in oncology trials occurs when control-arm patients switch to agent(s) in the experimental arm, potentially attenuating treatment effects in intention-to-treat (ITT) overall survival (OS) analyses. This complicates health technology assessment (HTA) decision-making by masking true treatment benefits. Understanding crossover thresholds compromising ITT OS statistical significance is critical for interpreting trial outcomes. Objective: We aimed to identify minimum threshold beyond which crossover rates are associated with loss of ITT OS statistical significance in advanced/metastatic breast, colorectal, gastrointestinal stromal, lung, prostate, and renal cancers. Methods: A systematic literature review of HTA reports (2013-2024) from NICE-UK, CDA-Canada, HAS-France, G-BA/IQWiG-Germany, PBAC-Australia, AIFA-Italy, AEMPS-Spain, and ICER-US was conducted. Searches were augmented with information from clinical practice guidelines and product labels from regulatory agencies. Submissions reporting crossover rates, with/without crossover-mitigation strategies applied, were analyzed to quantify the relationship between crossover rates and ITT OS statistical significance. Results: Fifty unique trials from 114 submissions reported crossover rates (NICE = 25, HAS = 19, CDA = 15, PBAC = 13, G-BA/IQWiG = 32, AEMPS = 9, ICER = 1), with crossover-mitigation strategies applied in 26 trials. A hypothesis-generating minimum benchmark of crossover rate (the ratio of control-arm participants switching to agent(s) in the experimental arm) ~42.7% was identified above which ITT OS lost statistical significance. Trials with high crossover rates (42.7%-88%) showed nonsignificant OS improvements, whereas lower rates (<42.5%) demonstrated significant OS benefits. Higher crossover rates were observed in trials with built-in vs natural crossover (52.9% vs 8.1%). In first-line setting, crossover rates >42.5% typically diluted OS benefit. In later-line and refractory diseases, OS significance often persisted beyond 42.7% threshold due to limited confounding by post-progression treatment options and faster death event accrual. Conclusion: Crossover impact varied by tumor type, treatment line, crossover type, and trial maturity. Crossover >42.7% generally compromised ITT OS significance; however, OS maturity attained in the trial or sustained efficacy of the intervention beyond progression may preserve statistical significance despite high crossover. A crossover rate exceeding 42.7% was associated with nonsignificant ITT OS outcomes in most trials, particularly in first-line settings. OS benefits may persist in later-line settings with fewer post-progression treatment alternatives. This evidence supports contextual interpretation of crossover-affected OS during HTA decision-making.