107P Censoring imbalance in ACIS trial for prostate cancer

In the ACIS trial, apalutamide plus abiraterone-prednisone has demonstrated longer radiographic progression-free survival (rPFS) compared with abiraterone-prednisone in chemotherapy-naïve patients with metastatic castrate-resistant prostate cancer. Despite reporting prolonged rPFS, no improvement in overall survival or quality of life was found and significant difference in rPFS was lost when assessed by blinded independent central review (BICR). Here we examined whether informative censoring might explain the discrepancy between the outcomes. Patient-level survival data were extracted and reconstructed using the semi-automated open-source tool WebPlotDigitizer and R statistical software. We utilized the reverse-Kapan-Meier method, in which the status indicator “event” and “censored” are flipped. We found that the rPFS of ACIS was associated with a significant censoring imbalance (reverse hazard ratio [HR] 1.32 [95% CI 1.08–1.61]; p=0.007), with 14% excess censoring in the intervention group. After performing sensitivity analysis, in which the balance in censoring between groups was partially restored, the difference in rPFS was lost (HR 0.86 [95% CI 0.73–1.00]; p=0·095), consistent with the non-significant results of the BICR assessment. To show this is unique to the trial we analyzed similar studies in mCRPC demonstrating anti-tumor activity and found that they were not associated with censoring imbalance. When a study lacking survival benefit relies on surrogate endpoint to claim the presence of a clinical benefit, it is essential to exclude informative censoring. The measured effect in ACIS could be a product of the inherent limitations of survival analysis when excessive censoring exists.