Development and Validation of the Early-Stage Hodgkin Lymphoma (HL) International Prognostication Index (E-HIPI): A Report from the Hodgkin Lymphoma International Study for Individual Care (HoLISTIC) Consortium
Bibliographic record
Abstract
Background: There are a lack of clinical prediction tools in early-stage Hodgkin lymphoma (E-HL). We developed and validated a modern-day model, known as E-HIPI, to predict progression-free survival (PFS) within the first 5 years (y) incorporating detailed individual patient (pt) data from international clinical trials and prospective registry data that were standardized, normalized & harmonized as part of the HoLISTIC Consortium (www.hodgkinconsortium.org). Methods: Model development utilized Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines (Moons, Ann Intern Med 2015) on a cohort of 3,080 newly-diagnosed E-HL pts treated on 4 seminal phase 3 clinical trials (NCIC: Meyer, NEJM 2012; RAPID: Radford, NEJM 2015; EORTC-GELA H9U: Ferme, E ur J Cancer 2017; EORTC/LYSA/FIL H10: Andre, JCO 2017). External validation was done in an independent cohort of 462 E-HL pts from a large registry (Princess Margaret, Toronto). Pts with classic HL, stage I or II disease, and ages 18-65 y were included. The primary outcome was PFS. Cox models were utilized with follow-up truncated at 5 years. Baseline predictors were: sex, stage, B symptoms, histology, number & location of nodal sites, and continuous data values of age, maximum tumor diameter (MTD), white blood cell count (WBC), lymphocyte count, hemoglobin, albumin & erythrocyte sedimentation rate (ESR). Multiple imputation was used for missing data. To consider possible non-linear relationships for continuous variables we examined plots & cubic splines. We used backward elimination ( P<0.1) to develop the models & internal validation with 200 bootstrap samples was conducted to estimate optimism & correct for overfitting. The final prediction equations applied optimism corrections to beta coefficients & hazard ratios (HR). Sensitivity analyses included model results stratified by PET staging (vs CT) and each HL trial. The models were evaluated in the external validation cohort & compared using c-statistics. Model scores were also grouped by tertiles for visualization in Kaplan-Meier (KM) curves. Results: Median age in the development cohort was 35y; 51% were female; 81% had nodular sclerosis & 77% stage II disease; 26% had B symptoms; 32% had mediastinal bulk (≥10cm or >1/3 diameter) and the mean MTD was 6.4 cm (interquartile range (IQR), 3.9-8.2). Median follow-up was 60 months (IQR 45-75). KM estimates at 5y was 90.7% (264 events) for PFS, and 97.1% (78 events) for overall survival (OS). After backward elimination, significant variables retained were age, sex, nodal location (cervical region), MTD & albumin. Stage, ESR, or the number of nodal sites were not significant. Age was modeled with piecewise linear splines due to its non-linear relationship. See the Table for the final prediction model and optimism-corrected HRs. Optimism-corrected c-statistics in the development model was 0.63. Sensitivity analyses of stratification by type of imaging at staging or by each clinical trial demonstrated no change in model parameters or c-statistics. Most baseline characteristics and outcomes of the external validation cohort were similar besides median follow-up time (74 months, IQR 37-129). C-statistics for the E-HIPI model in external validation was 0.63. In addition, observed 5y outcomes stratified by tertile of predicted risk in the external cohort separated pts into a low & higher-risk group, which comprised 1/3 & 2/3 of pts, respectively ( Figure). Finally, model diagnostics revealed the potential for time dependency in the retained covariates. Further assessment of differential prediction between early (<2y) vs. later (2-5y) PFS events is ongoing. Conclusion: Following TRIPOD methodology, we rigorously developed and externally validated the E-HIPI among >3,500 E-HL pts.Discrete risk groups were delineated, and we identified several novel factors, including nodal location, sex & key continuous variables that predict 5y PFS. Additional model refinement is underway, including formal calibration & external validation via other HL registries (Stanford, Mayo/Iowa SPORE). A web-based calculator to simplify application of the E-HIPI will also be presented at ASH. Future planned modeling includes integration of differing treatments & response-adapted imaging via multistate modeling to enrich individualized survival estimates and simulation modeling to predict late consequences.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.068 | 0.057 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".