Prediction of pathological response to neoadjuvant immunochemotherapy with baseline and post-treatment 18F-FDG PET imaging biomarkers in patients with locally advanced gastric cancer

Abstract

Background

Neoadjuvant immunochemotherapy (NICT) has shown promising therapeutic benefits in patients with locally advanced gastric cancer (LAGC). Our study aimed to predict the pathological response to NICT in LAGC before surgery by correlating the metabolic parameters of baseline and post-treatment ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) positron emission tomography/computed tomography (PET/CT) of the primary lesion with the pathological response following radical surgery.

Methods

Thirty-six LAGC patients who received three cycles of NICT (combination of sintilimab and CapeOx), followed by radical surgery, were included in this study. Both baseline ¹⁸F-FDG PET/CT (bPET) and post-treatment ¹⁸F-FDG PET/CT (pPET) were conducted, the metabolic parameters derived from the PET/CT scans, including the maximum standardized uptake value (SUV_max), metabolic tumor volume (MTV), and total lesion glycolysis (TLG) on bPET and pPET (bSUV_max and pSUV_max, bMTV and pMTV, bTLG and pTLG), as well as their reductions post-treatment (ΔSUV_max, ΔMTV, and ΔTLG), were assessed for their correlation with treatment efficacy and tumor regression grade (TRG) following NICT.

Results

Out of the 36 patients, 13 patients had a good response (GR), which included 5 cases with TRG 0 and 8 cases with TRG 1. Conversely, 23 patients exhibited a poor response (PR), with 20 patients having TRG 2 and 3 patients having TRG 3. Univariate analysis revealed that pMTV and pTLG in the GR group were significantly lower compared to the PR group (all p < 0.05). The identified cutoff values of pMTV and pTLG were 1.68 cm³ (area under the cure (AUC) = 0.683) and 4.71 cm³ (AUC = 0.683) for the GR and PR groups, respectively. On receiver operating characteristic (ROC) curve analyses, these values corresponded to sensitivity, specificity, and accuracy of 68.8%, 80.0%, and 73.1%, respectively, with no statistically significant differences between them after the DeLong test and McNemar test (all p > 0.05). Furthermore, bSUV_max, bMTV, bTLG, ΔSUV_max, ΔMTV, and ΔTLG in the TRG 0 group were significantly higher than those in the TRG 1 group (all p < 0.05). Upon performing ROC curve analyses for the TRG 0 group, the thresholds for bSUV_max, bMTV, bTLG, ΔSUV_max, ΔMTV, and ΔTLG were determined to be 7.8 (AUC = 0.916), 36.76 (AUC = 0.768), 105.55 (AUC = 0.819), 4.82 (AUC = 0.923), 22.64 (AUC = 0.807), and 104.7 (AUC = 0.845), with no statistically significant differences between them after the DeLong test (all p > 0.05). These thresholds demonstrated high sensitivity (80% for bMTV and 100% for others), specificity (83.9%, 71.0%, 67.7%, 83.9%, 61.3%, and 71.0%), and accuracy (86.1%, 66.7%, 72.2%, 86.1%, 66.7%, and 75.0%) in predicting TRG 0 after NICT, with no statistically significant differences between them after the McNemar test (all p > 0.05).

Conclusions

Imaging biomarkers from the combination of baseline and post-treatment ¹⁸F-FDG PET/CT showed potential in predicting pathological response to NICT in LAGC patients before surgery.