Assessing trends in conditional survival for non-Hodgkin lymphoma and chronic myeloid leukemia via novel extensions to the joinpoint survival model

Abstract Background: Five-year relative survival for non-Hodgkin lymphoma (NHL) and chronic myeloid leukemia (CML) has improved, reflecting therapeutic advances, but prognosis beyond five years post-diagnosis is less understood. We introduced a method to estimate trends in conditional survival, the probability of further survival given survival to a specific time. Methods: We developed a Landmarked joinpoint survival (JPSurv) model that employs data beginning at a pre-specified follow-up interval. We apply this approach to NHL (overall and by major subtype) and CML survival data from the Surveillance, Epidemiology, and End Results (SEER) Program. We estimated calendar trends in both 5-year relative survival and 5-year conditional relative survival (given survival to 5 years) using annual absolute changes in survival represented by percentage points (pp) over a time frame. Results: The largest annual increases in 5-year relative survival were observed for diffuse large B-cell lymphoma (+2.44 pp 1995-2002) and CML (+2.52 pp 1996-2011). Five-year conditional relative survival improved most for CML (+1.79 pp 1985-2016), chronic lymphocytic leukemia/small lymphocytic lymphoma (+0.80 pp 1988-2016), and follicular lymphoma (+80 pp 1990-2016). For patients diagnosed after 2010, 5-year conditional relative survival was approximately 90% for all cancers studied. Conclusions: Both 5-year relative survival and 5-year conditional relative survival improved substantially, reflecting advances in therapy and long-term patient outcomes. Impact: The Landmarked JPSurv model is a novel framework for conditional survival analysis that can be used to inform survivorship research, offering insight into lasting treatment effects and the likelihood of patients obtaining a sustained remission.