The impact of trial inclusion criteria on outcomes in DLBCL patients treated with R-CHOP in the first line: a Danish nationwide study

Up to 50% of diffuse large B-cell lymphoma (DLBCL) patients are ineligible for participation in clinical trials. Ineligible patients have inferior outcomes, but less is known about the impact of commonly used organ-function-based inclusion criteria on drug efficacy estimates. Data on DLBCL patients treated with CHOP+/-rituximab were retrieved from the Danish Lymphoma Registry. Trial inclusion criteria were extracted from four international DLBCL trials (REMoDL-B, GOYA, POLARIX, and HOVON-84). Differences in overall survival (OS) and 5-year restricted mean survival differences (5 y-RMSDs) between trial eligible and ineligible patients were computed. The effectiveness of adding rituximab to CHOP was quantified by the 5 y-RMSD between CHOP and R-CHOP-treated patients and the impact of individual trial criteria on estimated effectiveness was quantified by Shapley-values. In total, 4,083 R-CHOP-treated and 890 CHOP-treated DLBCL patients were included. Across the trials, 18.6-29.3% of the included R-CHOP-treated patients were deemed ineligible for trial based on organ function and performance status alone. Ineligible patients had significantly worse survival, with adjusted absolute differences in 5-year OS of 9-15%. The impact of individual criteria on the estimated effectiveness of adding rituximab to CHOP was small (Shapley-value range, -2.74-0.31). Using a smaller set of criteria derived from a data-driven approach, the number of eligible patients increased by 16-38% and the 5 y-RMSD increased by 0.7-3.1 months. In conclusion, OS among trial ineligible DLBCL patients is inferior as expected, but relaxing trial criteria would have increased the number of trial participants without making major changes in estimated efficacy for a hypothetical CHOP versus R-CHOP trial. This does not necessarily imply that trial findings based on selected patients are unreliable, as the estimated effectiveness of adding rituximab to CHOP was only slightly affected by omitting selected inclusion criteria.