Truace
Subscribe

AI predicts bladder cancer drug side effects in 542 patients, but it is not ready for clinics

Study title: Machine learning integrated explainable artificial intelligence in predicting toxicities of enfortumab vedotin in urothelial carcinoma: an exploratory study

In a proof-of-concept study published September 16, 2026, investigators used real-world data from 542 patients with advanced urothelial carcinoma to train machine learning models to predict toxicities from enfortumab vedotin, including severe adverse events and specific events like diarrhea and cutaneous toxicity.

Read the study abstract

Background Enfortumab vedotin (EV) therapy for advanced urothelial carcinoma is limited by adverse events (AEs). Early identification of high-risk patients is needed. This proof-of-concept study evaluated whether machine learning (ML) with explainable AI (SHAP) could predict EV toxicities using real-world data. Research design and methods Data from 542 patients (51 centers, 24 countries) were analyzed. Six outcomes were predicted including grade 3-4 AEs. Four ML algorithms were trained on an 80% split and tested…

By the Truvace desk1 min read1 sourceEvidence: Moderate · 60/100 — Observational study · n = 542 · 1 independent sourcePeer-reviewedTRV-2026-1121· cite & verify
Machine learning integrated explainable artificial intelligence in predicting toxicities of enfortumab vedotin in urothelial carcinoma: an exploratory study
"Barriers to conducting replications of experiment in cancer research" by Authors of the study: Timothy M Errington, Alexandria Denis, Nicole Perfito, Elizabeth Iorns, Brian A Nosek is licensed under CC BY 4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/.

In brief

The work matters because early identification of high-risk patients could help manage a therapy currently limited by adverse events, but the authors emphasize the results are exploratory and not ready for clinical use without prospective validation and larger event counts.

Main points

  1. Analysis included 542 patients from 51 centers across 24 countries with 80% training and 20% testing split.
  2. Six toxicity outcomes were modeled; Random Forest led for diarrhea, severe AEs and dose skipping, XGBoost for cutaneous toxicity and diabetes, LASSO for neuropathy.
  3. SHAP interpretability identified age as most important variable, followed by prior immunotherapy and ECOG performance status, with liver and lung metastases influencing specific toxicities.

The problem

The models are exploratory, built on small event counts with no external validation, so they cannot yet be used in clinics to reduce side effects from enfortumab vedotin.

The rundown

Researchers trained four algorithms on data from 542 patients treated at 51 centers in 24 countries, evaluating performance on a 20% holdout for six outcomes including grade 3-4 AEs, diarrhea, cutaneous toxicity, neuropathy, diabetes, and dose skipping.

Interpretability analysis highlighted age, prior immunotherapy, and ECOG performance status as top associated variables, with SHAP showing atezolizumab/nivolumab linked to lower cutaneous risk and female sex to higher risk, while liver metastases influenced diabetes and cutaneous toxicity.

Sources

  1. Peer-reviewedExpert Review of Anticancer Therapy2026-09-16

The debate