A validated multivariable machine learning model to predict cardio-kidney risk in diabetic kidney disease
Performance · Prediction of CREDENCE primary composite endpoint by risk model (age, BMI, systolic BP, NT-proBNP, hs-cTnT, IGFBP-7, GDF-15)Result — go to the resultHazard ratio 3.43 95% CI 2.72–4.32Authors' conclusion
We describe a validated risk algorithm that accurately predicts cardio-kidney outcomes across a broad range of baseline risk.
Study context
- Population
- Individuals with diabetic kidney disease (DKD) in the CREDENCE trial; validation in separate CREDENCE cohorts and the generally lower risk CANVAS
- Comparison
- Low risk vs Medium risk vs High risk
- Outcome
- Adjudicated primary composite endpoint of CREDENCE
- Follow-up
- 3 years
Methodological detailsmachine learning, secondary analysis of CREDENCE, external validation in CANVAS
- Machine learning
- Secondary analysis of CREDENCE
- External validation in CANVAS
Performance
As reported in the abstract
Prediction of CREDENCE primary composite endpoint by risk model (age, BMI, systolic BP, NT-proBNP, hs-cTnT, IGFBP-7, GDF-15)
Hazard ratio
3.43
95% CI 2.72–4.32
P < 0.001
By group
Low risk
94%
NPV
High risk
58%
PPV
Each increase in risk level
HR 3.43 (95% CI 2.72–4.32); P < 0.001
Each increase in risk level, as printed
- In-sample C-statistic
- 0.80 (95% CI = 0.77-0.83; P < 0.001)
- Higher values associated with shorter time to event
- log rank P < 0.001
Further findings
Change in score at 1 year
- Rising values at 1 year predicted higher risk for subsequent DKD events
Effect of canagliflozin on score · 1 year
- Canagliflozin reduced score results by 1 year with consistent event reduction across risk levels
Validation
- Accuracy validated in separate cohorts from CREDENCE and in CANVAS
Original publication
A validated multivariable machine learning model to predict cardio-kidney risk in diabetic kidney disease. Cardiovasc Diabetol. Published 15 May 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →