Interim response in diffuse large B cell lymphoma on CT: what is the optimal size reduction (ΔSPD) for predicting outcome?

OBJECTIVES: To investigate whether there was an optimal interim size reduction (iΔSPD) cutoff value that could discriminate diffuse large B cell lymphoma (DLBCL) patients with poor prognosis. METHODS: This retrospective study enrolled 265 newly diagnosed DLBCL patients with baseline and interim (after 3 cycles) contrast-enhanced computed tomographic scan (CECT) available. Two radiologists evaluated CECT images and selected target lesions according to the Lugano Response Criteria. Lymph nodes greater than 15 mm in longest diameter (LDi) and extra-nodal lesions with LDi greater than 10 mm could be chosen as target lesions and used to calculate iΔSPD. A software tool, X-Tile, was used to calculate the optimal iΔSPD cutoff value to differentiate patients with good vs. poor prognosis. Receiver operating characteristic curve analysis, Cox regression analysis, and Kaplan-Meier analyses were further used to validate the optimal cutoff value. RESULTS: The optimal cutoff value of iΔSPD calculated by X-tile was 80%. Compared with 50% and 100%, 80% cutoff value had the intermediate sensitivity and specificity (57.75% and 86.69% for overall survival (OS), 48.98% and 92.22% for progression-free survival (PFS), respectively), but the maximal Youden index (0.4744 for OS, 0.4120 for PFS, respectively) and areas under the curve (0.737 [0.680, 0.789] for OS). Cox regression analysis also revealed that iΔSPD < 80% could independently predict an inferior OS and PFS (both p < 0.001) while neither iΔSPD < 50% nor iΔSPD = 100% could. CONCLUSIONS: iΔSPD with the cutoff value 80% is an independent predictor of PFS and OS for patients with DLBCL. Results suggest that treatment should be modified for patients with iΔSPD < 80%. KEY POINTS: • The aim of interim response assessment is to identify patients whose disease has not responded to or has progressed on induction therapy. • A cutoff value of 80% in size reduction (ΔSPD) is an independent predictor of PFS and OS for DLBCL patients and is better than 50%. • In DLBCL patients with interim ΔSPD < 80%, a change to a more efficient therapy should be considered.