Beyond linearity and static risk: re-evaluating core prognostic factors in diffuse large B-cell lymphoma

Area of Science:

• Hematology
• Oncology
• Biostatistics

Background:

• Traditional prognostic models for diffuse large B-cell lymphoma (DLBCL), like the International Prognostic Index (IPI), assume static and linear risk factor effects.
• These assumptions may not accurately reflect the complex, time-varying nature of DLBCL patient outcomes.

Purpose of the Study:

• To investigate the time-dependent and nonlinear effects of established risk factors on survival in DLBCL patients treated with R-CHOP.
• To challenge the static nature of current prognostic indices and advocate for dynamic models.

Main Methods:

• Analysis of 664 DLBCL patients treated with R-CHOP chemotherapy.
• Utilized restricted cubic spline modeling to assess nonlinear relationships between prognostic factors and overall survival.
• Examined the temporal evolution of the predictive power of individual risk factors.

Main Results:

• The International Prognostic Index (IPI) showed overall stability, but individual components varied dynamically.
• ECOG performance score, beta-2 microglobulin (β2M), and lactate dehydrogenase (LDH) predicted early mortality but diminished over time.
• Ann Arbor stage and extranodal involvement became more important for later risk prediction, with β2M showing a resurgence after two years.
• Significant nonlinear relationships were observed between overall survival and age, LDH, and β2M (p < 0.001 for nonlinearity).

Conclusions:

• The prognostic impact of baseline factors in DLBCL is dynamic and nonlinear, not static.
• Current risk scores may oversimplify patient prognoses by not accounting for temporal changes.
• Future prognostic models for DLBCL should incorporate these time-dependent and nonlinear dynamics for improved personalized risk stratification.