Significance of marrow complete remission (mCR) in higher-risk MDS: A meta-regressive assessment of surrogate endpoint validity.

Abstract

e18580 Background: mCR is frequently used in trials evaluating therapies for higher-risk MDS (HR MDS) as an indicator of treatment efficacy and is often incorporated into composite response endpoints. Although mCR increases response rates, evidence suggests that patients achieving mCR derive inconsistent benefit, and its association with improved overall survival (OS) remains unclear. If mCR lacks prognostic value, its inclusion in composite endpoints may overestimate treatment efficacy and misguide clinical interpretation, trial design, and regulatory decisions. We therefore evaluated the association between mCR and median OS (mOS) in prospective trials of hypomethylating agents (HMA) for HR MDS. Methods: PubMed and Embase were systematically searched from July 2006 through August 2025. We identified prospective phase I–III trials of HMA enrolling adults with HR MDS per IPSS/IPSS-R that reported mCR (IWG 2006) and survival outcomes. Retrospective studies, case series, conference abstracts, duplicate datasets, and post-transplant trials were excluded. Five reviewers independently screened 2,273 records in Covidence, with inclusion determined by consensus of ≥2. Data were abstracted in accordance with PRISMA guidelines. For studies lacking 95% CIs for mOS (n=2), individual patient data were reconstructed from Kaplan–Meier curves using the Guyot algorithm. Quantitative synthesis included random-effects meta-analysis (REma) of mOS and a primary mixed-effects meta-regression with random intercepts for study and treatment arm, evaluating the association between arm-level mCR proportion and mOS. An unadjusted random-effects meta-regression was performed as sensitivity. Results: Twelve trials (16 arms; 859 pts) were included (1,562 unique records; 105 full texts assessed). Across arms, mean mCR was 23.5% (SD, 12.5%). REma of 11 arms estimated a log-transformed pooled mOS of 12.8 months (I²=95.0%). Knapp–Hartung–adjusted, disease-setting estimates were 16.1 (frontline), 12.1 (mixed), and 8.9 months (relapsed/refractory, R/R). In the primary mixed-effects meta-regression (MR, final analytic dataset: 9 arms with complete mCR/mOS data), each 10% increase in mCR was associated with an estimated +1.02 months in mOS (95% CI, 0.30–2.16; P=.60; I²=97.7%), a non-significant association. In unadjusted MR, the estimate was +0.22 months per 10% mCR (95% CI, −0.49 to 0.93; P=.55; I²=98.2%). Covariate adjustments (R/R status, TP53 mt) attenuated effects and did not yield significance. Conclusions: Across trials, mCR did not demonstrate a robust association with longer mOS at the study-arm level. Apparent correlations were attenuated after adjustment for adverse biology and R/R status, suggesting that the relationship may reflect confounding rather than true surrogacy. Reliance on mCR risks overstating therapeutic benefit and misdirecting late-phase testing.