Systematic review and meta-analysis of interim FDG PET in large B-cell lymphoma patients

Our rationale was to investigate the potential of interim [F]FDG PET (iPET) in predicting large B-cell lymphoma (LBCL) patients' outcomes, including the end-of-treatment response, progression-free survival (PFS), and overall survival (OS). A systematic search was conducted on the 3 main medical literature databases until January 8, 2025. Published original articles that evaluated iPET for prognosticating LBCL patients and provided crude data for meta-analytic calculations were considered eligible. The hierarchic method was used to pool the random-effect model's end-of-treatment prediction performance measures of sensitivity, specificity, positive likelihood ratio (LR), and negative LR. The bivariate model was used to determine the corresponding 95% CI. Furthermore, PFS and OS were pooled across studies, including survival proportions and hazard ratios (HRs). All analyses were conducted using STATA 16 software. The initial search resulted in 939 studies. After the full-text review, 44 studies were considered eligible. The pooled sensitivity and specificity of iPET for predicting end-of-treatment response were 0.78 (95% CI, 0.69-0.84) and 0.84 (95% CI, 0.78-0.88), respectively. On the basis of the pooled positive and negative LRs, iPET was not a reliable modality to exclude or confirm the end-of-treatment [F]FDG PET results. The pooled PFS HR across all available datasets ( = 27) was 2.88 (95% CI, 2.35-3.52) for positive iPET (Deauville score [DS] 4-5). In datasets with HRs for positive iPET versus DS 5 alone, the pooled HRs were 2.36 (95% CI, 1.74-3.21) versus 5.59 (95% CI, 4.35-7.19). Furthermore, when datasets were limited to those with data on change in SUV versus DS 5 ( = 6), the pooled HRs were 3.41 (95% CI, 2.67-4.35) versus 5.59 (95% CI, 4.35-7.19), respectively. The pooled OS HR across all available datasets ( = 22) was 3.71 (95% CI, 2.93-4.69) for positive iPET. In datasets with HRs for positive iPET versus DS 5 alone ( = 5), the pooled HRs were 3.31 (95% CI, 2.09-5.24) versus 8.10 (95% CI, 6.04-10.85). Moreover, when datasets were limited to those with data on change in SUV versus DS 5 ( = 5), the pooled HRs were 4.54 (95% CI, 3.57-5.78) versus 8.10 (95% CI, 6.04-10.85). On the basis of our comprehensive systematic review and meta-analysis, DS 5 is a highly reliable predictor of unfavorable response to therapy on iPET. The utility of this threshold to escalate or modify therapy needs to be evaluated in future prospective trials.