Development and validation of a prediction model for 1-year mortality among older adults with Hodgkin Lymphoma who receive dose-intense chemotherapy

Background: Older adults with Hodgkin Lymphoma (HL) have poorer outcomes than younger patients. There are little data about which baseline patient and disease factors inform prognosis among older patients. We sought to create a prediction model for 1-year mortality among older patients with new HL who received dose-intense chemotherapy.

Methods: We included adults ≥65 years old with a new diagnosis of classical HL between 2000-2013 from the Surveillance, Epidemiology, and End Results (SEER)-Medicare dataset who received full-regimen chemotherapy. Through a non-random 2:1 split, we created development and validation cohorts. Multiple imputation was used for missing data. Using stepwise selection and logistic regression, we identified predictive variables for 1-year mortality. The model was applied to the validation cohort. A final model was then fit in the full cohort.

Results: We included 1315 patients. In the development cohort (n = 813), we identified significant predictors of 1-year mortality including age, Charlson comorbidity index (CCI), B symptoms at diagnosis, and advanced stage at diagnosis. The c-statistic was 0.70. When this model was applied to the validation cohort (n = 502), the c-statistic was 0.65. Predictors of 1-year mortality in the final model were CCI (OR = 1.41), B symptoms (OR = 1.54), advanced stage (OR = 1.44), and older age at diagnosis (OR = 1.33).

Conclusion: We present a prediction model for use among older adults with HL who receive intensive chemotherapy. We identify risk factors for death within 1 year of diagnosis. Future work will build upon prognostication and shared decision-making after diagnosis for this population.