Background: ES cHL has long been classified as favorable or unfavorable by EORTC or GHSG criteria. However, these are based on dichotomized variables, and several are subjective (B symptoms) or difficult to measure (nodal sites). After integrating additional data into the HoLISTIC consortium, we further developed & validated the E-HIPI to predict 2-year (y) PFS (Evens, ASH 2023).
Methods: The model was developed in 3000 untreated patients (pts) with cHL age 18-65y with ES (I or II) cHL from 4 phase 3 clinical trials & externally validated in 1461 pts from 4 cHL registries (using TRIPOD guidelines: Moons, Ann Int Med 2015). The primary outcome of 2y PFS was estimated with a Cox model. Baseline candidate variables were sex, stage, histology, nodal sites, and continuous values of age, maximum tumor diameter (MTD), white blood & lymphocyte count, hemoglobin, albumin & erythrocyte sedimentation rate. Linearity was examined & missing data was multiply imputed. We used backward elimination to develop the model & internal validation to estimate optimism & correct for overfitting. The final prediction equation applied optimism correction to beta coefficients, hazard ratios & C-statistics. The C-statistic was reported for the external validation cohort. Model performance was compared to EORTC favorable/unfavorable status.
Results: Mean age in the development cohort was 34y; 51% were female; 81% had nodular sclerosis; 77% had stage II; mean MTD was 6.5 cm. Median follow-up was 60 months (IQR=45-75). KM estimated 2y PFS was 93.7%. Variables retained in the model were sex, MTD, hemoglobin & albumin (Figure). The optimism-corrected C-statistic in the development cohort was 0.63. Most external validation cohort characteristics were similar besides lower 2y PFS (90.2%) and longer median follow-up (108 months, IQR=63-165). The external validation C-statistic was 0.63. The E-HIPI was prognostic in both favorable (P<0.01) & unfavorable (P<0.01) EORTC subgroups. Moreover, unfavorable status was not prognostic once E-HIPI was known (P=0.36).
Conclusion: We developed & externally validated the first prediction model for ES cHL among >4400 pts, which is comprised of objective & continuous variables. Female sex and increasing hemoglobin & albumin were associated with better 2y PFS, and increasing MTD was associated with worse PFS. The E-HIPI outperformed EORTC favorable/unfavorable status and provides more robust & biologically meaningful prediction to improve decision making.
Angie Mae Rodday, John Radford, Matthew Maurer, Jenica Upshaw, Nicholas Counsell, Sara Rossetti, Ranjana H. Advani, Marc Andre, Cheryl Chang, Andrea Gallamini, Annette Hay, David Hodgson, Richard Hoppe, Martin Hutchings, Peter Johnson, Eric Mou, Stephen Opat, John M.M. Raemaekers, Kerry Savage, Qingyan Xiang, Susan K. Parsons, Andrew Evens