Summary: In a post hoc analysis of the randomized, open-label DIVERSITY-CVR trial (n=340 Japanese adults with type 2 diabetes), baseline alanine aminotransferase (ALT) was explored as a predictor of differential drug response. Lower baseline ALT was associated with a better response to sitagliptin, whereas higher baseline ALT favoured dapagliflozin for glycaemic control (p<0.05 across all models). No absolute between-drug efficacy difference, confidence intervals, or NNT were reported for this exploratory subgroup signal.
PICO Summary
| Element | Detail |
|---|---|
| Population | 340 Japanese adults with type 2 diabetes on metformin monotherapy or no treatment; post hoc analysis of the randomized, open-label DIVERSITY-CVR trial (Japan). |
| Intervention | Dapagliflozin (an SGLT2 inhibitor); patients stratified by baseline ALT level. Arm sizes not separately reported in this subgroup analysis. |
| Comparison | Sitagliptin (a DPP-4 inhibitor) as the active comparator from the parent trial. Arm sizes not separately reported. |
| Outcome | Co-primary endpoints were change in HbA1c and time in range at 24 weeks. Baseline ALT was significantly associated with both. Lower ALT favoured sitagliptin and higher ALT favoured dapagliflozin across all models (p<0.05). Significant ALT-BMI-HbA1c interactions were seen (p=0.02 and p=0.03), with a significant baseline ALT-to-HbA1c-change correlation in the total cohort (p=0.03) and the sitagliptin group (p<0.01). No 95% CIs, absolute risk reduction, or NNT were reported. |
Expert Commentary
This is a hypothesis-generating signal, not a treatment rule. The analysis is a post hoc, secondary look at an existing randomized trial, so it can establish association but not causation, and the result should be read as effect modification rather than proof that one agent is superior. What was observed is that baseline ALT, a routinely available liver enzyme that tracks loosely with hepatic fat, appeared to flag who responded better to which class: lower ALT leaned toward the DPP-4 inhibitor, higher ALT toward the SGLT2 inhibitor, with consistent direction across all adjustment models. The headline weakness is that p-values were reported without effect sizes, confidence intervals, or a prespecified interaction threshold, and multiple comparisons across three models inflate the chance of a spurious hit; the absolute glycaemic differences were not quantified, so clinical magnitude is unknown. The parent trial was open-label, which can bias subjective management, and the cohort was modest and exclusively Japanese, limiting generalisability. Can I use this with my patients? Not yet as a selection tool. It is reasonable to note that a patient with higher ALT and metabolic-associated fatty liver may have an additional rationale for an SGLT2 inhibitor, but that decision already rests on stronger cardiorenal and hepatic evidence. I would like to see this interaction prospectively tested with reported effect sizes before ALT is used to choose between these drugs.
References
Mochizuki K, Fuchigami A, Hirose T, Uchino H. Impact of baseline alanine aminotransferase levels on the efficacy of dapagliflozin and sitagliptin: Latent class analysis from the DIVERSITY-CVR study. Diabetes Obes Metab. 2025;27(9):5099-5107. doi:10.1111/dom.16559
