Cardiometabolic & kidneyDiagnostic or prognostic study
Dynamic surveillance of the cardiovascular risk and identification of treatment responders in type 2 diabetes using a machine learning–based model
- Prognostic model development and validation study
- machine learning
- external validation
Authors' conclusion
The ML-CVD Primary model enables dynamic prediction of cardiovascular events, facilitating ongoing risk surveillance and identification of individual drug responses.
Study context
- Population
- Patients with T2DM without prior cardiovascular disease (CVD) from four clinical trials (ACCORD, CANVAS, CANVAS-R, CREDENCE)
- Participants
- 11,677 patients · 4 clinical trials
- Outcome
- CV death, nonfatal MI, nonfatal stroke
Performance
As reported in the abstract
Predictive performance of ML-CVD Primary model
Harrell's C-index in holdout and external testing datasets
0.66-0.71
Further findings
Progression of ML-CVD Primary risk score · 1 year
- Intensive interventions (intensive glycemic control and canagliflozin) significantly mitigated score progression vs control
Primary CV outcomes per SD decrease in ML-CVD Primary score
- Each SD decrease significantly associated with reduced risk
Primary CV outcomes, Responders vs Non-Responders (canagliflozin arms)
- Responders (score decrease) had lower risk vs Non-Responders (no change or increase)
- 45%
Original publication
Dynamic surveillance of the cardiovascular risk and identification of treatment responders in type 2 diabetes using a machine learning–based model. Eur J Prev Cardiol. Published 9 April 2026. doi:10.1093/eurjpc/zwag196
Summarised from the published abstract; every number is checked against it