Early immune recovery during adjuvant anti–PD-1 as a predictor of early recurrence in resected stage II–III melanoma.

Abstract

e21545 Background: Early identification of patients at risk of recurrence during adjuvant anti–PD-1 remains an unmet need. We tested whether a longitudinal Time-in-Range (TIR) metric capturing time spent in an “immune-recovered” state during weeks 0–12 predicts early post-landmark recurrence. Methods: Consecutive AJCC 8th stage II–III melanoma patients receiving adjuvant nivolumab or pembrolizumab (n = 74) underwent routine CBC monitoring approximately every 2 weeks during weeks 0–12 (median 7 measurements/patient). The immune-recovered range was prespecified as NLR < 3 AND ALC ≥1.0×10⁹/L. TIR(0–12) was calculated using interval-based time-weighting between consecutive measurements. Secondary prespecified dynamics included ALC slope (0–12) and time-adjusted RMSSD of log(NLR). To mitigate guarantee-time bias, we applied a 12-week landmark: patients recurring before week 12 were excluded, and the primary endpoint was early recurrence from week 12 to month 12; secondary endpoint was 12-month RFS. Parsimonious models adjusted for stage and ulceration; sensitivity analyses tested alternative thresholds (NLR < 4; ALC ≥0.8×10⁹/L) and exclusion of intercurrent infection/inflammation in weeks 0–12. Results: Median age was 61 years (IQR 53–68.8); stage III 52.7%; ulceration 45.9%; nivolumab 55.4%. Intercurrent infection/inflammation in weeks 0–12 occurred in 17.6%. Overall 12-month recurrence was 24/74 (32.4%). Early recurrence < 12 months occurred in 13/74 (17.6%). One patient recurred before week 12; 73 comprised the landmark cohort, with 12/73 (16.4%) early post-landmark recurrences. Median TIR was 0.0% in early post-landmark recurrences versus 18.5% in non-recurrences. In adjusted logistic regression, higher TIR was associated with lower odds of early post-landmark recurrence (OR 0.74 per +10% TIR, 95% CI 0.57–0.95). Categorized for interpretability, the highest TIR tertile had lower early recurrence than the other tertiles combined (4.2% vs 24.0%; adjusted OR 0.06, 95% CI 0.007–0.59). Discrimination remained favorable after bootstrap optimism correction (AUC 0.83). Secondary dynamics were directionally consistent (lower ALC slope, higher NLR volatility among early recurrences). Twelve-month RFS was 78.3% overall and 95.8% in the highest TIR tertile; associations were consistent when excluding intercurrent infection/inflammation. Conclusions: A simple longitudinal metric—time spent in an immune-recovered range during the first 12 weeks of adjuvant anti–PD-1—was independently associated with early post-landmark recurrence, suggesting an actionable early window for risk-adapted surveillance. External validation is warranted.