KEY TAKEAWAYS
- The study aimed to explore whether ML could surpass current models in predicting OS and PFS.
- ML-based prognostic model for aHL showed significant enhancement over IPS models, with limited improvement compared to A-HIPI.
Historically, patients with advanced-stage Hodgkin lymphoma (aHL) have been stratified based on the International Prognostic Score (IPS) for lymphoma.
Rasmus Rask Kragh Jørgensen and the team aimed to explore whether a machine learning (ML) method could surpass current models in predicting overall survival (OS) and progression-free survival (PFS).
Patient data from the Danish National Lymphoma Register served as the development cohort for model creation. Employing stacking, which integrates various predictive survival models (including Cox proportional hazard, flexible parametric model, IPS, principal component, and penalized regression), the ML model was developed and compared against two IPS versions (IPS-3 and IPS-7) and the newly devised aHL international prognostic index (A-HIPI).
Internal validation was conducted via nested cross-validation, while external validation utilized patient data from the Swedish Lymphoma Register and Cancer Registry of Norway as the validation cohort.
The results revealed that in the development cohort, the ML model demonstrated a concordance index (C-index) of 0.789 for OS, outperforming IPS-7, IPS-3, and A-HIPI with C-index values of 0.608, 0.650, and 0.768, respectively.
In the validation cohort, the corresponding C-index values were 0.749, 0.700, 0.663, and 0.741. For PFS, the ML model exhibited the highest C-index in both cohorts, with values of 0.665 and 0.691 in the development and validation cohorts, respectively.
Additionally, the time-varying area under the curves (AUCs) for the ML model and A-HIPI consistently surpassed those of the IPS models within the initial 5 years post-diagnosis in both cohorts.
The application of ML techniques in developing a new prognostic model for aHL showcased significant enhancement compared to IPS models. However, its predictive performance showed only marginal improvement compared to A-HIPI.
Funding for the research was provided by Celgene, Genmab, Roche, Takeda, Novartis, Janssen, Merck, AbbVie, and AstraZeneca.
Source: https://pubmed.ncbi.nlm.nih.gov/38608215/
Rask Kragh Jørgensen R, Bergström F, Eloranta S, et al. (2024) “Machine Learning-Based Survival Prediction Models for Progression-Free and Overall Survival in Advanced-Stage Hodgkin Lymphoma.” JCO Clin Cancer Inform. 2024 Apr;8:e2300255. doi: 10.1200/CCI.23.00255. PMID: 38608215.
