An international confirmatory study of the prognostic value of early PET/CT in diffuse large B-cell lymphoma: comparison between Deauville criteria and ΔSUVmax.

PURPOSE

The role of interim PET/CT in guiding therapeutic strategies in diffuse large B-cell lymphoma (DLBCL) is debated, mainly because interpretation rules vary among centres. This study aimed to explore the reproducibility and confirm the prognostic value of early PET/CT using the Deauville criteria and ΔSUVmax.

METHODS

This international confirmatory study retrospectively evaluated 114 patients with newly diagnosed DLBCL treated with a rituximab-containing regimen. All patients underwent ¹⁸F-FDG PET/CT at baseline (PET0) and after two cycles (PET2), with no therapy change based on the latter. Scans were interpreted by three observers using the Deauville five-point scale and ΔSUVmax between PET0 and PET2 was calculated. Interpretations were evaluated for interobserver agreement and for progression-free survival (PFS) prediction.

RESULTS

Median follow-up was 39 months. Early PET/CT was predictive of outcome when interpreted with the Deauville criteria and ΔSUVmax. Using the five-point scale, the overall kappa value was 0.66 with the reference background set in the liver (score ≥4) and interobserver agreement was even better using a 66% ΔSUVmax cut-off (κ = 0.83). Moreover, the prognostic value of interim PET was slightly inferior when using a Deauville score ≥4 than when using a 66% ΔSUVmax cut-off: for the Deauville score the 3-year PFS estimate was 59% (45-73%) in PET2-positive patients vs. 81% (71-91%) in PET2-negative patients (P = 0.003); for the 66% ΔSUVmax cut-off the 3-year PFS estimate was 44% (23-65%) in PET2-positive patients vs. 79% (70-88%) in PET2-negative patients (P = 0.0002).

CONCLUSION

Although the Deauville criteria are valid for assessing the prognostic value of early PET/CT in DLBCL, computation of the ΔSUVmax leads to better performance and interobserver reproducibility, and should be preferred when a baseline scan is available.