Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1861
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-18T22:01:52.675529Z
status: published
lens: g_space
sector: health
headline: Explainable Machine Learning-Based Prediction of Progression-Free Survival in Prostate Cancer: Retrospective Cohort Study
dek: Background Progression-free survival (PFS) is a critical end point in oncology, yet real-world applications of individualized, explainable machine learning (ML) predictions remain limited. Objective This study aimed to develop and validate explainable ML models to predict PFS using retrospective data from a national prostate cancer cohort in Brunei Darussalam. Methods We analyzed a retrospective cohort of 212 patients (478 longitudinal observations) treated at the Brunei Cancer Centre (January 2018 to December 2…
gain_title: A random survival forest paired with SHAP explanations discriminated progression risk far better than a traditional Cox model in a single-centre prostate cancer cohort, supporting personalized risk stratification.
problem_title: (none)
trace_subject: (none)
gain_reading: A random survival forest paired with SHAP explanations discriminated progression risk far better than a traditional Cox model in a single-centre prostate cancer cohort, supporting personalized risk stratification.
gain_evidence: the traditional CPH model performed poorly (C-index 0.531
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers analyzed 212 prostate cancer patients treated at the Brunei Cancer Centre between 2018 and 2024, using 478 longitudinal observations. They compared four survival modeling approaches, with missing data imputed and time-based patterns captured by a recurrent autoencoder.

The random survival forest performed best, and SHAP analysis flagged laboratory values and treatment categories as drivers of progression risk. This suggests explainable models could inform individual care decisions, but the data come from one centre and the authors say wider validation is still needed before clinical use.
limitation: The authors state that external validation in a larger multi-institutional, multiomics dataset is required before routine clinical implementation.
tag: Evidence-backed gain
key_points: The cohort covered 212 patients and 478 longitudinal observations at the Brunei Cancer Centre from January 2018 to December 2024. | The random survival forest reached a C-index of 0.906 with an integrated Brier score of 0.0698. | The Cox proportional hazards model scored a C-index of 0.531, well below the tree-based ensemble. | SHAP identified sodium, alanine aminotransferase, mean corpuscular hemoglobin, platelet count, and certain treatments as key risk drivers.
rundown: The team harmonized clinical, laboratory, and treatment records and imputed missing values with extremely randomized trees. A recurrent autoencoder produced latent representations of patients' longitudinal patterns before modeling.

Four approaches were compared: Cox proportional hazards, random survival forest, gradient boosting survival, and deep neural network survival models. Evaluation used time-dependent AUC, Harrell C-index, and integrated Brier score.

Gradient boosting survival and deep survival models showed moderate performance, with C-indices of 0.844 and 0.719 respectively. The authors conclude that tree-based ensembles combined with SHAP offer high accuracy for personalized risk stratification.
sources:
- peer_reviewed | JMIR Cancer | https://doi.org/10.2196/91510 | 2026-10-05
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

ddf51bc00513faafb9666bc174b2e4a149efd957bb5c9b305cab32bda5dd9f79

previous

0000000000000000000000000000000000000000000000000000000000000000