Diagnostic performance of 18F-FDG PET/CT metabolic parameters for early prediction of pathological response in NSCLC treated with neoadjuvant immuno(chemo)therapy: A systematic review and meta-analysis

Early and accurate prediction of pathological response in non-small cell lung cancer (NSCLC) receiving neoadjuvant immune(chemo)immunotherapy is essential for guiding treatment strategies. This meta-analysis aims to assess the predictive performance of PET/CT-derived metabolic parameters for identifying major pathological response (MPR) and pathological complete response (pCR). The PubMed, Web of Science, Embase, and Cochrane Library were searched for related studies from inception to January 2025. Studies assessed evaluated predefined 18F-FDG PET/CT parameter (PERCIST, PERCIST-max, ΔSUVmax, ΔSULpeak, ΔSULmax, ΔMTV, ΔTLG, pre/post-treatment SUVmax, post-treatment SULmax/SULpeak) and reported pathological outcomes were included. Methodological quality was evaluated using QUADAS-2. Sensitivity, specificity and heterogeneity were analyzed via HSROC models. Publication bias was assessed via Deeks’ test. Sixteen studies met inclusion criteria. For MPR, PERCIST showed excellent sensitivity (1.00) but limited specificity (0.67), while PERCIST-max achieved both high sensitivity (0.92) and superior specificity (0.94). ΔSUVmax demonstrated balanced accuracy (sensitivity: 0.92; specificity: 0.77; AUC: 0.94), and ΔSULpeak also yielded high diagnostic performance (sensitivity: 0.94; specificity: 0.91). Volumetric metrics such as ΔMTV and ΔTLG showed variable results, with ΔMTV performing moderately (sensitivity: 0.77; specificity: 0.83), and ΔTLG exhibiting lower sensitivity (0.63) despite good specificity (0.81). For pCR, ΔSUVmax retained high sensitivity (0.94) but showed reduced specificity (0.57; AUC: 0.86). SUVmax-Pre showed moderate accuracy (sensitivity: 0.78; specificity: 0.81). Subgroup analyses indicated that the diagnostic performance of ΔSUVmax varied depending on treatment regimen. The study support ΔSUVmax as a reliable indicator of MPR, although its specificity for pCR remains limited. PERCIST’s high sensitivity aids in excluding non-responders. However, limited study numbers for parameters like ΔSULpeak, SULpeak-Post, and ΔSULmax reduce generalizability, necessitating larger validation studies to confirm their diagnostic utility.