Key result
The ISCHEMIA-PREDICT mortality risk score contains significant discrepancies in baseline characteristics, including hypertension and diabetes, compared to original trial data.
Answers are grounded in this paper and cite their sources.
The authors raise critical concerns regarding data accuracy and cohort composition in the ISCHEMIA-PREDICT mortality risk score model, highlighting major discrepancies in baseline characteristics compared to the original ISCHEMIA trial.
We read with interest the report by Navarese et al.1 describing ISCHEMIA-PREDICT, a machine learning model for mortality risk in chronic coronary syndrome derived from the ISCHEMIA trial data deposited in the NIH/NHLBI data repository. As investigators in the ISCHEMIA trial, we have several questions regarding data accuracy, cohort composition, missing data handling, and model performance that warrant clarification before this score is adopted in clinical practice. Several baseline characteristics reported in Supplemental Table 1 are difficult to reconcile with the published ISCHEMIA trial data. Hypertension prevalence was 47% in the supplement vs. 73% reported previously.2 Diabetes was 58% in the supplement compared with 42% reported previously. Discrepancies of this magnitude are not readily explained by the exclusion of 123 participants. We cannot reproduce various other summary data for baseline characteristics e.g. degree of ischemia, number of diseased vessels by CCTA, and coronary stenosis ≥70%.
No takes yet. Share an insight, caveat, or question.
O’Brien et al. (2026) conducted a letter in chronic coronary syndrome. ISCHEMIA-PREDICT mortality risk score was evaluated. The ISCHEMIA-PREDICT mortality risk score contains significant discrepancies in baseline characteristics, including hypertension and diabetes, compared to original trial data.
Synapse has enriched 10 related papers. Consider them for comparative context: