To demonstrate a long-term survival benefit without a new trial, Neuvivo hired a research firm to track down patients from the original study. By collecting "last date alive" information in a blinded fashion, they generated statistically significant survival data years after the trial concluded.

To frame its trial results positively, Compass Pathways used less stringent definitions for key endpoints. It defined 'clinically meaningful reduction' and 'remission' at levels below the common standard, a tactic that calls into question the true magnitude of the drug's benefit.

The most valuable lessons in clinical trial design come from understanding what went wrong. By analyzing the protocols of failed studies, researchers can identify hidden biases, flawed methodologies, and uncontrolled variables, learning precisely what to avoid in their own work.

Current bladder cancer trials often fail to differentiate between patients with primary resistance (never responded) versus acquired resistance (responded, then progressed). Adopting this distinction, common in lung cancer research, could help identify patient subgroups more likely to benefit from immunotherapy re-challenge and refine trial eligibility criteria.

In solid tumor immunotherapy, significant efficacy gains almost always correlate with increased toxicity. This study's claim of nearly doubled progression-free survival with identical toxicity rates is biologically implausible and was a primary reason for skepticism, even before analyzing the trial's methodology.

A significant criticism of the pivotal KEYNOTE-564 trial is that only half the patients in the control arm received standard-of-care immunotherapy upon relapse. This lack of subsequent optimal treatment complicates the interpretation of the overall survival benefit, raising questions about its true magnitude.

The study reported that no patients in the "before 3 PM" group ever received a dose after 3 PM over four cycles. In a busy, real-world cancer center, such perfect adherence is practically impossible due to logistical issues. This flawless data suggests the study might be a retrospective analysis of curated data rather than a truly prospective trial.

The study presented three different datasets over a short period. While efficacy endpoints like PFS and OS changed, the toxicity data remained identical. This is highly unusual, as resolving censored patient data for efficacy should also lead to updated toxicity information, suggesting a rushed or incomplete analysis process.

The study's progression-free survival (PFS) curve was unusually smooth, lacking the stepwise drops expected from scheduled scans in oncology trials. More alarmingly, the "numbers at risk" table showed more patients remaining than were represented on the graph at certain time points—a statistical impossibility suggesting a significant reporting or programming error.