Integrated bone marrow and splenic [18F]FDG PET/CT uptake for predicting bone marrow biopsy positivity in diffuse large B-cell lymphoma — a retrospective single-centre study

Abstract Background Bone marrow biopsy (BMB) and 2-deoxy-2-[ 18 F]fluoro-D-glucose ([ 18 F]FDG) positron emission tomography/computed tomography (PET/CT) provide complementary information in staging diffuse large B-cell lymphoma (DLBCL). We evaluated the diagnostic performance and inter-reader reproducibility of a liver-referenced visual classification integrating bone marrow (BM) and splenic uptake. Survival associations were secondary exploratory objectives. Methods This retrospective single-centre study included a primary diagnostic cohort of 107 patients with newly diagnosed DLBCL and adequate pretreatment BMB, and a separate prognostic cohort of 141 patients. Two nuclear medicine physicians independently classified BM and splenic uptake as High or Low relative to liver, incorporating focal uptake. Any High indicated that one or both organs were High. Diagnostic performance was evaluated in the full cohort and in a sensitivity cohort excluding focal BM uptake ( n = 94, 12 BMB-positive). Progression-free survival (PFS) and overall survival (OS) were analysed using Cox regression. Results BMB was positive in 18 of 107 patients. Any High had 83.3% sensitivity (95% confidence interval [CI], 58.6–96.4%), 80.9% specificity (71.2–88.5%), 96.0% negative predictive value (88.8–99.2%), and 81.3% accuracy (72.6–88.2%). Compared with BM-High alone, Any High had higher sensitivity and lower specificity, with similar accuracy of 81.3% versus 82.2% and an exact McNemar p value greater than 0.999. Three of 75 Both Low patients were BMB-positive, and estimates were robust when two indeterminate biopsies were reclassified as negative or excluded. Inter-reader agreement was substantial to almost perfect for the principal criteria (κ 0.750–0.843) but was lower for background splenic uptake (κ 0.648) than for background BM uptake (κ 0.773). In secondary exploratory analyses, 66 PFS events and 49 deaths occurred. Any High was associated with shorter PFS after adjustment for the International Prognostic Index (IPI, hazard ratio [HR] 2.04, 95% CI 1.08–3.84) or the National Comprehensive Cancer Network IPI (HR 1.96, 1.05–3.69), but was not independently associated with OS. Conclusions The combined liver-referenced classification improved sensitivity for BMB positivity over BM assessment alone and showed good inter-reader reproducibility, although agreement was lower for background splenic uptake and BMB positivity occurred in some Both Low patients. Its association with PFS was exploratory. Prospective external validation is required before clinical use.