An omics-based machine learning approach to predict diabetes progression: a RHAPSODY study
Authors' conclusion
Using machine learning approaches, we show that insulin requirement risk can be modestly well predicted by predominantly clinical variables. Inclusion of molecular markers improves the prognostic performance beyond that of clinical variables by up to 5%. Such prognostic models could be useful for identifying people with diabetes at high risk of progressing quickly to treatment intensification.
Study context
- Population
- People with T2D in the Hoorn Diabetes Care System (DCS) and Genetics of Diabetes Audit and Research in Tayside Scotland (GoDARTS) cohortsDCS 585, GoDARTS 571
- Comparison
- Clinical-variable models vs models adding molecular markers
- Outcome
- Prediction of time to insulin requirement (Harrel's C statistic)
Methodological detailsIMI-RHAPSODY, development in one cohort, replication in a second cohort, lasso, ridge, GRridge, random forest
- IMI-RHAPSODY
- Development in one cohort, replication in a second cohort
- Lasso, ridge, GRridge, random forest
Performance
As reported in the abstract
Base model, DCS discovery cohort: C statistic
0.71
95% CI 0.64, 0.79
Base model, GoDARTS replication cohort: C statistic
0.71
95% CI 0.69, 0.75
Extended model, DCS: C statistic
0.74
95% CI 0.67, 0.81
Extended model, GoDARTS: C statistic
0.73
95% CI 0.69, 0.77
Further findings
Insulin requirement during follow-up · DCS 1.0-11.4 years; GoDARTS 0.3-11.8 years
- Required insulin, DCS (of 585)
- 69
- Required insulin, GoDARTS (of 571)
- 175
Most frequently selected clinical variables
- HbA1c (18 of the 36 models)
- 50%
- Age (15 models)
- 41.2%
- C-peptide (15 models)
- 41.2%
Original publication
Composed by AI from the published abstract and checked against it. How Briefs are made →