18F-FDG PET/CT radiomics across treatment timepoints for predicting response to neoadjuvant therapy in locally advanced esophageal cancer: a systematic review and meta-analysis

Abstract

Pathological complete response (pCR) after neoadjuvant therapy is achieved in only about 20%–40% of patients with locally advanced esophageal cancer (LAEC), yet it is the strongest determinant of long-term survival. 18 F-fluorodeoxyglucose ( 18 F-FDG) positron emission tomography/computed tomography (PET/CT) radiomics can enable automated, non-invasive prediction of response, but the optimal imaging time point remains undefined. Standardized uptake value (SUV) is only one of multiple biomarkers of PET/CT evaluation of LAEC. We performed a systematic review and meta-analysis evaluating the diagnostic accuracy of 18 F-FDG PET/CT radiomics for predicting pCR according to imaging time points, incorporating various feature extractions of 18 F-FDG PET/CT at baseline (pre-treatment), restaging (post-treatment), and delta (pre-to-post change). PubMed/MEDLINE, Cochrane Library, Scopus, and EBSCO were searched from inception to March 2026. Eligible studies applied 18 F-FDG PET/CT radiomics to predict strictly defined pCR after neoadjuvant chemoradiotherapy (NACRT), neoadjuvant chemoimmunotherapy (NACI), with surgical histopathological examination as the reference standard. Studies were assigned to three temporal pools, and for each study and time point, a single, pre-specified AUC was extracted. The AUCs were pooled separately within each time point, provided a minimum of 4 studies were deemed eligible per time point, using a logit-transformed REML random-effects model with the Hartung–Knapp adjustment. Methodological quality was assessed with QUADAS-2 tool and the Radiomics Quality Score (RQS). Thirteen publications were included (14 study–time-point observations; one cohort contributed to two temporal pools). The baseline pool (n = 8) yielded a pooled AUC of 0.845 (95% CI: 0.768–0.899; I 2 = 34.6%). The post-treatment pool (n = 4) yielded a pooled AUC of 0.677 (95% CI: 0.328–0.900; I 2 = 82.8%). The delta pool (n = 2) was summarised descriptively (individual AUCs 0.89 and 0.79) rather than pooled, given only two studies. Baseline 18 F-FDG PET/CT radiomics demonstrated a higher and more consistent pooled AUC than post-treatment (restaging) radiomics; however, these estimates were derived from different study sets and should not be interpreted as a direct head-to-head comparison. Currently, delta radiomics is promising but rests on few studies. Methodological heterogeneity, cohort overlap among individual centers, and limited external validation constrain certainty. Further multicenter prospective studies using standardized workflows are required before clinical translation. https://www.crd.york.ac.uk/PROSPERO/view/CRD420261351933 , identifier CRD420261351933.