To Cluster or Not to Cluster? A Comparison of Approaches to Targeting Type 2 Diabetes Glucose-Lowering Therapy in the TriMaster Crossover Trial
Authors' conclusion
Clinical features are more strongly associated with differential glycemic response than PPS, and direct outcome prediction outperforms data-driven subgrouping for optimizing short-term HbA1c reduction.
Study context
- Population
- Adults with T2D who received all three medications (sitagliptin, canagliflozin, pioglitazone) in the TriMaster trial
- Participants
- 309 adults
- Comparison
- Four precision treatment strategies for targeting glucose-lowering therapy (sitagliptin, canagliflozin, pioglitazone)
- Primary endpoint
- Overall HbA1c response and within-person differential response to sitagliptin, canagliflozin and pioglitazone
- Follow-up
- 4 months
Methodological detailsthree-way crossover
- Three-way crossover
Primary result
As reported in the abstract
HbA1c reduction with treatment allocation by strategy
| Group | Value | HbA1c reduction (mmol/mol) |
|---|---|---|
| Routine-features model | 0.27 (CI 0.18–0.35) % | 2.9 [95% CI 2.0, 3.8] |
| Clusters | 0.16 (CI 0.07–0.25) % | 1.7 [95% CI 0.8, 2.7] |
Value
Routine-features model0.27 (CI 0.18–0.35) %
Clusters0.16 (CI 0.07–0.25) %
HbA1c reduction (mmol/mol)
Routine-features model2.9 [95% CI 2.0, 3.8]
Clusters1.7 [95% CI 0.8, 2.7]
- Routine-features model and cluster approach both strongly associated with overall and differential HbA1c responses
- P < 0.0001
- Clinical clusters showed benefit for only one cluster-drug comparison (severe insulin-resistant diabetes: SGLT2i vs TZD)
- P = 0.005
Original publication
To Cluster or Not to Cluster? A Comparison of Approaches to Targeting Type 2 Diabetes Glucose-Lowering Therapy in the TriMaster Crossover Trial. Diabetes Care. Published 18 June 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →