Skip to content
    Guidelines & Consensus

    Pooled analysis benchmarks radiographic discordance in solid-tumor trials

    Goldmacher et alContemporary Clinical Trials Communications

    In oncology clinical trials, progression-free survival and objective response rate assessed per Response Evaluation Criteria in Solid Tumors, version 1.1 (RECIST v1.1), rely heavily on serial radiographic evaluations of tumor burden. To establish empirical benchmarks that assist clinical trial teams in differentiating expected reader variability from data quality concerns, researchers conducted a pooled analysis of 39 phase 2 and phase 3 solid-tumor clinical trials involving pembrolizumab-based regimens. Radiographic scans were independently evaluated by treating investigators and blinded independent central review (BICR).
    Across the evaluated trials, case-level discordance between investigator and BICR assessments was 17.7% (95% CI 17.2–18.2) for progressive disease (PD; n = 20,908), 12.2% (95% CI 11.8–12.7) for confirmed objective response (OR; n = 21,520), and 14.7% (95% CI 14.1–15.3) for unconfirmed OR (n = 15,209). Substantial variation emerged across tumor types: PD discordance ranged from 11.8% in melanoma to 22.8% in hepatocellular carcinoma (HCC). Conversely, confirmed OR discordance was lowest in HCC (5.4%) and highest in cervical cancer (20.0%), with unconfirmed OR discordance following a similar pattern (6.3% in HCC to 19.7% in cervical cancer).
    Trial design characteristics also influenced discordance metrics. Studies incorporating real-time verification of progression demonstrated significantly lower PD discordance than those without verification (17.3% [95% CI 16.7–17.8] vs 20.2% [95% CI 18.8–21.7]). Study blinding did not affect PD discordance (17.6% [95% CI 16.9–18.3] in double-blind trials vs 17.8% [95% CI 17.0–18.6] in open-label trials); however, confirmed OR discordance was higher in double-blind studies compared to open-label studies (13.3% [95% CI 12.7–13.9] vs 10.9% [95% CI 10.3–11.6]). These findings establish foundational empirical benchmarks for site versus central review discordance in RECIST v1.1 assessments, although researchers noted that while the underlying causes of discordance are not unique to immunotherapy, the observed rates should be applied cautiously when evaluating non-immunotherapy clinical trials.

    Key Takeaways

    • Overall case-level discordance between investigator and blinded independent central review (BICR) was 17.7% (95% CI 17.2–18.2) for progressive disease (n = 20,908), 12.2% (95% CI 11.8–12.7) for confirmed objective response (n = 21,520), and 14.7% (95% CI 14.1–15.3) for unconfirmed objective response (n = 15,209).
    • Discordance varied significantly by tumor type, with progressive disease discordance ranging from 11.8% in melanoma to 22.8% in hepatocellular carcinoma, and confirmed objective response discordance ranging from 5.4% in hepatocellular carcinoma to 20.0% in cervical cancer.
    • Real-time verification of progression lowered progressive disease discordance compared with trials lacking verification (17.3% [95% CI 16.7–17.8] vs 20.2% [95% CI 18.8–21.7]), while progressive disease discordance remained similar between double-blind (17.6%) and open-label (17.8%) studies.
    Read the original sourceContemporary Clinical Trials Communications