The Shifting Paradigm of Cardiovascular Prevention in Type 2 Diabetes
For nearly a decade, sodium-glucose cotransporter 2 inhibitors (SGLT2i) have revolutionized the management of type 2 diabetes (T2D) and cardiovascular disease (CVD). Landmark clinical trials, such as EMPA-REG OUTCOME, CANVAS, and DECLARE-TIMI 58, have unequivocally established the efficacy of SGLT2i in reducing major adverse cardiovascular events (MACE) and heart failure hospitalizations. However, much of this foundational evidence was generated in populations with established atherosclerotic cardiovascular disease or multiple high-risk factors. This has led to a clinical practice often focused on high-risk stratification to determine SGLT2i eligibility.
In contemporary practice, a significant portion of the T2D population consists of younger, working-age individuals who do not yet meet the criteria for ‘high cardiovascular risk’ based on traditional scoring systems. For these patients, the relative benefit of SGLT2i versus other second-line therapies, such as dipeptidyl peptidase 4 inhibitors (DPP4i), has remained an area of clinical uncertainty. A recent study by Mori et al., published in the European Journal of Preventive Cardiology, addresses this gap by utilizing advanced machine learning and target trial emulation to explore the heterogeneity of SGLT2i treatment effects.
Highlights
SGLT2i therapy was associated with a significant 0.38 percentage point reduction in the 3-year risk of a composite cardiovascular outcome compared to DPP4i.
Machine learning causal forest analysis revealed that 91% of individuals classified as ‘low CVD risk’ were still predicted to derive a cardiovascular benefit from SGLT2i.
The degree of cardioprotection was more strongly correlated with individual patient characteristics—specifically BMI, blood pressure, and fasting glucose—than with conventional aggregate CVD risk scores.
Study Design and Methodology
The researchers employed a robust target trial emulation (TTE) framework to minimize the biases inherent in observational research. Using a nationwide insurer-based database of working-age Japanese citizens (2015–2023), the study analyzed 150,830 individuals with T2D. The cohort had a mean age of 54 years, reflecting a younger, primary prevention-focused population than is typically seen in clinical trials.
Patients were categorized into two groups: those initiating SGLT2i and those initiating DPP4i. The primary outcome was a composite of all-cause mortality, myocardial infarction, stroke, or heart failure over a 3-year follow-up period. To move beyond the limitations of ‘average treatment effects,’ the investigators applied a causal forest model—a machine learning approach based on random forests—to estimate individual-level treatment effects (ITE). This allowed the team to assess how the benefits of SGLT2i varied according to baseline patient characteristics and conventional CVD risk scores.
Key Findings: Beyond the Average Treatment Effect
The overall results supported the use of SGLT2i, showing a 3-year risk difference of 0.38% (95% CI: 0.16–0.61) in favor of SGLT2i over DPP4i. While this absolute risk reduction might appear modest compared to high-risk secondary prevention trials, the implications for public health are substantial given the size of the low-risk T2D population.
The Heterogeneity of Benefit
The most striking finding from the causal forest analysis was the weak correlation (r = 0.287, P < 0.001) between a patient’s calculated CVD risk score and the magnitude of benefit they received from SGLT2i. This suggests that traditional risk stratification tools, which were largely developed to predict atherosclerotic events, may not fully capture the metabolic and hemodynamic mechanisms through which SGLT2i exert their cardioprotective effects.
The Low-Risk Population Analysis
Among the 107,425 individuals identified as having low CVD risk, the model predicted that 91.0% (97,757 individuals) would still experience a reduction in cardiovascular risk with SGLT2i therapy. The study identified specific phenotypic markers that predicted a higher benefit even in these low-risk individuals: higher body mass index (BMI), elevated blood pressure, and higher fasting plasma glucose levels. This suggests that the ‘metabolic burden’ of a patient may be a more sensitive indicator for SGLT2i initiation than a calculated 10-year risk of stroke or heart attack.
Expert Commentary and Clinical Interpretation
The findings by Mori et al. challenge the current ‘risk-first’ approach to SGLT2i prescription. Traditionally, clinicians have been taught to prioritize these agents for patients who have already ‘earned’ their risk through age or prior events. However, the causal forest data suggests that the physiological benefits of SGLT2i—including natriuresis, improved ventricular loading conditions, and metabolic shift—are relevant much earlier in the disease trajectory.
Mechanistic Insights
The correlation of benefit with higher BMI and blood pressure reinforces the hypothesis that SGLT2i function as more than just glucose-lowering agents. Their ability to reduce visceral adiposity and lower blood pressure through non-adrenergic pathways makes them particularly effective for patients with the metabolic syndrome phenotype, regardless of whether their aggregate CVD risk score has yet crossed a ‘high-risk’ threshold.
Study Limitations
While the target trial emulation is a sophisticated method, it remains an observational study. Residual confounding is always a possibility, although the use of causal forests and large-scale data helps mitigate this. Furthermore, the study population was restricted to working-age Japanese citizens; while the biological mechanisms of SGLT2i are likely universal, the absolute risk scales and specific lifestyle factors may differ across other ethnic and age groups.
Conclusion: A Precision Medicine Approach to T2D
The study concludes that the cardioprotective effects of SGLT2i are heterogeneous and are better predicted by individual patient characteristics than by conventional CVD risk scores. For the clinician, this means that a ‘low-risk’ label on a patient with T2D, obesity, and hypertension should not be a barrier to SGLT2i therapy. Instead, we should consider the specific metabolic profile of the patient. By expanding the use of SGLT2i to these ‘low-risk’ but ‘high-benefit’ individuals, we may prevent a significant number of cardiovascular events before the underlying disease progresses to a stage of high risk.
References
1. Mori Y, Komura T, Adomi M, Yagi R, Fukuma S, Kawakami K, Kondo N, Tsugawa Y, Yabe D, Yanagita M, Inoue K. Heterogeneous cardiovascular effects of sodium-glucose cotransporter 2 inhibitors in type 2 diabetes: a causal forest and target trial emulation study. Eur J Prev Cardiol. 2026 Jan 6;33(1):80-88. doi: 10.1093/eurjpc/zwaf539. PMID: 40889271.
2. Zelniker TA, Wiviott SD, Raz I, et al. SGLT2 inhibitors for primary and secondary prevention of cardiovascular and renal outcomes in type 2 diabetes: a systematic review and meta-analysis of cardiovascular outcome trials. Lancet. 2019;393(10166):31-39.
3. Wiviott SD, Raz I, Bonaca MP, et al. Dapagliflozin and Cardiovascular Outcomes in Type 2 Diabetes. N Engl J Med. 2019;380(4):347-357.