Clinicopathologic and imaging predictors of pathologic complete response after neoadjuvant chemoimmunotherapy in resectable NSCLC: A systematic review and meta-analysis.

Abstract

e20068 Background: Neoadjuvant chemoimmunotherapy (CIT) improves pathologic complete response (pCR) rates in resectable non–small cell lung cancer (NSCLC), yet substantial variability persists across studies. Robust predictors of pCR are not well defined, limiting patient selection and translational application. We conducted a systematic review and meta-analysis to identify clinicopathologic and biologic factors associated with pCR and to assess gaps in current predictive approaches. Methods: MEDLINE, Embase, Web of Science, and Cochrane Library were searched for prospective trials and observational cohorts (2015–2025) reporting surgical resection after neoadjuvant CIT for resectable NSCLC and evaluating predictors of pCR. Candidate predictors included PD-L1 expression, clinical stage and nodal status, radiologic response (RECIST), metabolic response on FDG-PET, systemic inflammatory markers, circulating tumor DNA dynamics, tumor mutational burden, and radiomics features. Random-effects meta-analyses estimated pooled odds ratios (ORs) with 95% confidence intervals (CIs) for associations with pCR when ≥2 studies reported comparable data. Studies proposing multivariable prediction models were evaluated qualitatively using TRIPOD and PROBAST criteria. Results: A total of 14 studies including 1,842 patients were eligible (6 prospective trials, 8 observational cohorts). Higher PD-L1 expression was associated with increased odds of pCR (per 10% increase in TPS: OR 1.28, 95% CI 1.10–1.49, I²=58%). FDG-PET metabolic response (ΔSUVmax ≥50%) showed a strong association with pCR (OR 2.41, 95% CI 1.62–3.58, I²=42%) and outperformed size-based radiologic response. Absence of mediastinal nodal disease (cN0–1 vs cN2) was associated with higher pCR rates (OR 1.76, 95% CI 1.22–2.54, I²=36%). Data on tumor mutational burden, circulating tumor DNA clearance, systemic inflammatory markers, and radiomics were sparse and heterogeneous, precluding quantitative synthesis. Only a minority of studies reported multivariable analyses, and none included external validation, calibration assessment, or evaluation of clinical utility; no study met PROBAST criteria for low risk of bias. Conclusions: PD-L1 expression, metabolic response on FDG-PET, and baseline nodal status are the most reproducible factors associated with pCR after neoadjuvant CIT in resectable NSCLC. However, current evidence highlights a significant translational gap, with insufficient data and methodological rigor to support generalizable predictive models. Prospective studies integrating imaging and biologic variables with external validation are urgently needed to enable clinically actionable pCR prediction.