Fertility-Sparing Treatment of Endometrial Cancer and Atypical Endometrial Hyperplasia by Molecular Classification: A Systematic Review and Meta-Analysis Accounting for Outcome-Assessment Timing and Follow-Up Duration

Background/Objectives: Molecular classification (POLE-mutated, mismatch-repair–deficient [MMRd], no-specific-molecular-profile [NSMP], p53-abnormal [p53abn]) is used to stratify women receiving fertility-sparing treatment (FST) for early endometrial carcinoma or atypical hyperplasia, but reported subtype differences are inconsistent. We estimated subtype-specific complete response (CR) and post-CR recurrence and examined how outcome-assessment timing (fixed early timepoint versus best response) and follow-up duration affect subtype comparisons. Methods: PubMed and Embase (January 2015–August 2025) were searched; overlapping reports were consolidated; proportions were pooled with binomial–normal generalised linear mixed models; late response among early non-responders was analysed in paired cohorts; recurrence was meta-regressed on follow-up; risk of bias (QUIPS) and certainty (GRADE) were assessed. Results: Twenty reports yielded 15 cohorts (10 primary). Best CR was 85% (95% CI 70–93) for POLE, 71% (57–83) for MMRd, and 85% (77–91) for NSMP. In six paired cohorts, CR rose from the early timepoint to best response in every subtype (MMRd 44→77%, NSMP 57→90%); 55% of MMRd and 76% of NSMP early non-responders later achieved CR, and within-cohort best CR was lower for MMRd than NSMP (odds ratio [OR] 0.36, 0.17–0.79). Recurrence after CR was 34% (23–46) for MMRd, 30% (17–49) for NSMP, and 19% (4–57; 0–100% across cohorts) for POLE. NSMP recurrence tended to rise with follow-up in an exploratory study-level analysis (OR 1.02/month; p = 0.06, dependent on the few long-follow-up cohorts), whereas MMRd recurrence did not (adjusted OR vs. NSMP 1.7, 0.9–3.3). p53abn tumours (n = 28) showed variable CR (17/24) and recurrence in 9/17. Occult p53abn or MMRd tumours affected 17% (12–24%; range 7–47%) of conventionally selected candidates. Certainty was low or very low. Conclusions: Subtype comparisons in FST depend on when outcomes are assessed and how long patients are followed. Molecular classification is prognostic, not predictive; its principal value is to identify occult higher-risk tumours and to calibrate counselling and surveillance, not to justify de-escalation.