Inter-reader agreement of quantitative FDG PET/CT biomarkers in lymphoma: a multicentre evaluation of MTV, TLG and Dmax.

BACKGROUND

The Deauville score is a key prognostic factor in Hodgkin lymphoma (HL) and diffuse large B-cell lymphoma (DLBCL) during interim and end-of-treatment PET/CT evaluations. However, additional measurements, particularly at baseline, such as metabolic tumour volume (MTV), total lesion glycolysis (TLG), and the maximum distance between hypermetabolic lymphoma lesions (Dmax) may offer enhanced prognostic value. This study evaluates the inter-reader agreement of these metrics to assess their reliability across different physicians.

METHODS

This study included 117 patients with untreated HL or DLBCL who had baseline [18F]fluorodeoxyglucose PET/CT scans. Nine nuclear medicine physicians independently segmented lymphoma lesions using the online platform Recomia ( www.recomia.org ), without specific instructions beyond identifying lymphoma-related lesions. MTV, TLG, and Dmax were calculated from these segmentations. MTV was defined as the summed volume in cm3, TLG as MTV multiplied by SUVmean and Dmax as the distance between the centroids of the two farthest lesions, measured in the 3D reconstruction. Each patient was segmented by two physicians. Inter-reader agreement was assessed using Spearman correlation coefficients for continuous values and Cohen’s kappa coefficient (κ) for dichotomized values (above/below median).

RESULTS

The mean age of the 117 patients was 50 years (standard deviation 19), 39% female. Median (± interquartile range) values were 321 (± 597) cm3 for MTV, 2200 (± 4399) cm3 for TLG, and 35 (± 50) cm for Dmax. Spearman correlations between readers were 0.97 for MTV, 0.98 for TLG and 0.72 for Dmax (all p < 0.01). Agreement on dichotomized values was 95.7% for MTV (κ = 0.91), 97.4% for TLG (κ = 0.95), 83.8% for Dmax (κ = 0.68).

CONCLUSIONS

MTV and TLG demonstrated good inter-reader reliability, even without standardized segmentation protocols. In contrast, Dmax showed moderate variability. These findings support the robustness of MTV and TLG as quantitative biomarkers. For Dmax to be clinically reliable, clearer segmentation guidelines are essential. Especially, inconsistent inclusion of small lesions that may not contribute significantly to MTV, might affect measurement of disease dissemination.