Trial rescue delivers a clean, on-time readout

The story

A late-stage biotech had a global Phase 3 in solid tumors that was drifting off course. Enrollment lagged by 40%. Screen failures topped 35%. Minor protocol deviations piled up. ePRO compliance slid below 70%. The primary endpoint risked dilution because several regions permitted rescue meds that confounded assessment. The CRO had turned over key staff twice. The DMC signaled a rising futility risk if operational noise continued.

We ran a 30-day diagnostic that combined blinded data forensics, site funnel plots, and a country-by-country feasibility refresh. Three issues drove most of the variance: ambiguous inclusion language that widened the population, uneven endpoint ascertainment, and a long tail of underperforming sites.

Interventions were narrow and fast. We issued a targeted protocol amendment that clarified biomarker thresholds, tightened rescue-med rules, and standardized timing windows for assessments. We closed the bottom quartile of sites and activated 42 pre-qualified centers with proven enrollment in the last 12 months. We replaced the regional CRO in two countries and stood up a single command PMO with weekly decision rights. Risk-based monitoring shifted SDV from 100% to targeted verification aligned to a central-statistical signal plan. ePRO adherence moved to automated reminders with nurse follow-ups for misses beyond 24 hours. Imaging moved to a single central lab with rapid adjudication. We cleaned data in “war rooms” and retired 60% of open queries in four weeks.

Statistics and governance matched the operational reset. We ran a blinded sample size re-estimation based on observed variability and set a prespecified alpha-spending plan to keep Type I error intact. We defined tipping-point analyses for missing data and established a clean chain of custody for all changes in the TMF. The team rehearsed inspection responses and closed CAPAs tied to prior monitoring findings.

Results followed within two quarters. Median enrollment rate rose from 0.17 to 0.44 patients per site per month. Screen failures fell from 38% to 22%. Protocol deviations per patient dropped 46%. ePRO compliance rose to 93%. Data queries per patient fell by half. The DMC endorsed continuation at the second look with no further warnings. The study locked 8 weeks earlier than the revised plan with no critical findings in sponsor audits. The final analysis produced a hazard ratio of 0.79 with p=0.021 and aligned sensitivity results. The sponsor completed a Type C meeting on schedule and advanced to submission with a coherent, defensible story of trial rescue and data integrity.