Whether or not toxicity predicts clinical outcomes has long been a question regarding cancer treatments. While prior studies have focused on specific cancers, therapies, and toxicities, no comprehensive evidence exists on whether treatment toxicity predicts favorable outcomes.

We abstracted treatment toxicity and clinical outcome data from a sample of phase III oncology randomized clinical trials (n = 99 trials). We investigated whether treatments with relatively greater toxicity compared with their controls had relatively higher, lower, or equivocal rates of clinical efficacy, measured by progression-free survival (PFS) and overall survival (OS). Several toxicities were assessed (all grades, grades III/IV, cutaneous rash, gastrointestinal toxicity, and myelosuppression).

Toxicity and efficacy were greater among treatments than controls (e.g. 3.5 instances of all-grade toxicity per patient in treatment arms versus 2.8 instances in controls, P < 0.001; mean PFS of 9.1 months across treatment arms versus 7.1 months across controls, P < 0.001; mean OS of 18.6 months across treatment arms versus 16.9 months across controls, P < 0.001). Across trials, greater relative treatment toxicity was strongly associated with greater PFS in treatments versus controls (P < 0.001), but not OS (P = 0.44). Although higher relative rates of myelosuppression and cutaneous rash among treatments were not associated with greater treatment efficacy, greater relative gastrointestinal toxicity among treatments was associated with greater relative PFS compared with controls (P = 0.007).

Across trials, treatments with relatively greater all-grade toxicity compared with controls are associated with relatively greater PFS but not OS.