Obstructive sleep apnoea (OSA) and type 2 diabetes mellitus (T2DM) frequently co-exist, with reported OSA prevalence exceeding 50% among patients with T2DM 1. This comorbid state amplifies cardiovascular risk beyond that conferred by either condition alone, driven by shared pathophysiological mechanisms including intermittent hypoxia, sympathetic nervous system activation, insulin resistance, and systemic inflammation 1. The resultant cardiovascular burden, spanning heart failure, arrhythmias, stroke, and premature mortality, represents a major and growing clinical challenge in this population. Sodium-glucose co-transporter 2 inhibitors (SGLT-2i) have emerged as a cornerstone of cardiometabolic therapy, with landmark trials demonstrating robust reductions in cardiovascular death, heart failure hospitalization, and major adverse cardiovascular events in patients with T2DM and established cardiovascular disease 2, 3. Despite this, patients with concurrent OSA have been largely underrepresented or uncharacterized in these trials, and only a small body of literature, comprising fewer than 500 patients across all published studies, has examined the relationship between SGLT-2i and outcomes specific to the T2DM-OSA population 4. Using a large real-world federated network, we evaluated long-term cardiovascular outcomes associated with SGLT-2i versus dipeptidyl peptidase-4 inhibitors (DPP-4 is) in patients with concurrent T2DM and OSA. We conducted a retrospective cohort study using the TriNetX global federated health research network, which aggregates de-identified electronic health record data from 105 healthcare organizations across the United States. Adults aged ≥ 18 years with a diagnosis of both T2DM (ICD-10: E11) and OSA (ICD-10: G47.33) were identified between January 1, 2015, and December 31, 2023. Patients were required to have a specific diagnosis of OSA (ICD-10: G47.33); those with primary central sleep apnoea (G47.31) or other sleep apnoea (G47.39) were explicitly excluded. Cohort 1 comprised patients initiated on SGLT-2i after their T2DM and OSA diagnoses were established, while Cohort 2 comprised patients initiated on DPP-4 inhibitors as an active comparator. To reduce confounding by indication and crossover, each cohort excluded patients with recorded prior exposure to the comparator drug class within available TriNetX records, and the index treatment event was required to occur on or after the documented diagnoses of T2DM and OSA. Patients with non-OSA subtypes were excluded. To balance baseline differences between cohorts, 1:1 propensity score matching (PSM) was performed using age, sex, race, comorbid diagnoses, concomitant medications, and laboratory values as covariates (see Table 1), yielding two well-matched cohorts of 39 258 patients each. Outcomes were assessed over a 5-year follow-up window beginning 1 day after the index event. Primary outcomes were all-cause mortality and MACE, defined as a composite of acute coronary syndrome, stroke, intracranial haemorrhage, heart failure, cardiac arrest, ventricular arrhythmias, and precerebral or cerebral arterial occlusion. Secondary outcomes included heart failure exacerbation, ischemic stroke/TIA, haemorrhagic stroke, acute ischemic heart disease (IHD), atrial fibrillation/flutter, pulmonary embolism/deep vein thrombosis (PE/DVT), lower limb amputations, and Group 3 pulmonary hypertension (PH). To ensure that only incident events were captured, patients with each outcome recorded before the start of the outcome time window were excluded from that outcome's analysis. Survival analysis was performed using Kaplan–Meier estimation with the log-rank test, and hazard ratios (HRs) with 95% confidence intervals (CIs) were estimated from Cox proportional hazards models. Risk differences and risk ratios were calculated for each outcome. Following an intention-to-treat approach, follow-up began 1 day after the index treatment event and continued until the outcome of interest, death, or the last recorded observation, with time-to-event analysis accounting for unequal follow-up between cohorts. All analyses were conducted within the TriNetX platform. As TriNetX uses de-identified data, this study was exempt from formal institutional review board oversight. Following 1:1 PSM, two well-balanced cohorts of 39 258 patients each were identified. The mean age was 63.4 years in both groups, approximately 54% were male, and 67% were White. Rates of key comorbidities were similar between groups, including hypertension (84.5% vs. 84.4%), IHD (38.3% vs. 38.4%), heart failure (25.4% vs. 26.1%), and chronic kidney disease (29.9% vs. 29.8%) (Table 1). Mean follow-up after matching remained longer in the DPP-4i cohort (1134.4 vs. 910.4 days). Survival analyses accounted for this asymmetry through time-to-event modelling with censoring at the last recorded observation. SGLT-2i use was associated with a significantly lower risk of all-cause mortality compared with DPP-4i over 5 years (6.6% vs. 13.5%; HR 0.590, 95% CI 0.563–0.619; log-rank p < 0.001). SGLT-2i use was also associated with a significantly lower risk of MACE (15.5% vs. 19.5%; HR 0.914, 95% CI 0.874–0.956; log-rank p < 0.001) (Table 2). SGLT-2i use was associated with significantly lower hazards of several secondary outcomes, including ischemic stroke/TIA (HR 0.862, 95% CI 0.804–0.923), haemorrhagic stroke (HR 0.779, 95% CI 0.651–0.931), acute ischemic heart disease (HR 0.911, 95% CI 0.857–0.968), heart failure exacerbation (HR 0.902, 95% CI 0.861–0.946), and PE/DVT (HR 0.839, 95% CI 0.781–0.902). No statistically significant difference was observed for atrial fibrillation/flutter (HR 0.952, 95% CI 0.896–1.011), lower limb amputation (HR 0.860, 95% CI 0.736–1.005), or Group 3 PH (HR 1.055, 95% CI 0.828–1.343) (Table 2). In this large real-world study, SGLT-2i use was associated with lower risks of all-cause mortality, MACE, and several other cardiovascular outcomes compared with an active DPP-4i comparator among patients with concurrent T2DM and OSA. Our expanded MACE definition differs from the conventional 3P-MACE (cardiovascular death, nonfatal myocardial infarction, and nonfatal stroke) used in landmark RCTs, limiting direct comparability; however, individual components, including ischemic stroke/TIA and acute ischemic heart disease, are reported separately. The differences in mean follow-up periods likely reflect the more recent clinical adoption of SGLT-2i relative to DPP-4i; survival analyses accounted for this asymmetry through time-to-event modelling with censoring at last recorded observation. The observed association with lower all-cause mortality (HR 0.590) exceeds the 11%–14% reductions reported in broader T2DM and heart failure randomized trials 5, and may reflect the particularly high baseline cardiometabolic risk of patients with concurrent T2DM and OSA, characterized by cyclical intermittent hypoxia, heightened sympathetic tone, systemic inflammation, and insulin resistance, all pathways through which SGLT-2i exert meaningful pleiotropic benefit 6. These findings are broadly consistent with real-world evidence from unselected T2DM populations, where SGLT-2i have demonstrated similar cardiorenal protective associations, though the magnitude of benefit observed in our OSA-enriched cohort may reflect the additive cardiovascular burden conferred by concurrent OSA 7. The breadth of outcomes favourably impacted in this cohort extends well beyond mortality. Significant reductions in heart failure exacerbation, acute ischemic heart diseases, ischemic and haemorrhagic stroke, and PE/DVT collectively reflect the multi-organ cardiovascular protective profile of SGLT-2i. These benefits align with SGLT-2i's pleiotropic effects, natriuresis, sympathetic attenuation, and anti-inflammatory activity, which may act synergistically with OSA-specific pathophysiology to confer broader cardiovascular protection 8. Additionally, SGLT-2i-induced weight loss may independently attenuate OSA severity over time, representing a potential mediating pathway that warrants prospective evaluation. The reduction in acute ischemic heart events and ischemic strokes further reinforces the microvascular and macrovascular benefits of SGLT-2i previously demonstrated in dedicated trials, now extended to this high-risk subgroup 7. No statistically significant association was observed for atrial fibrillation/flutter, lower limb amputation, or Group 3 pulmonary hypertension. These null findings may reflect smaller effect sizes, limited power for less frequent outcomes, or differences in the mechanisms underlying these events compared with the outcomes that showed clearer benefit. The null finding for atrial fibrillation/flutter may additionally reflect differential ascertainment, whereby more intensive cardiac monitoring in SGLT-2i users could increase AF detection, biasing the estimate toward null. They do not diminish the overall signal, but rather are consistent with the heterogeneity of SGLT-2i effects observed across outcome domains in broader randomized trial analyses 5. This study has several limitations. As a retrospective observational study, residual confounding, including potential calendar-time and channelling bias, cannot be excluded despite robust PSM. OSA severity, CPAP adherence, specific SGLT-2i agent, and treatment duration were not captured; patients on SGLT-2i with better CPAP adherence may represent a healthier, more care-engaged subgroup, potentially overestimating the observed benefit. Additionally, TriNetX relies on coded EHR data, making outcome misclassification and unmeasured confounding inherent constraints. OSA diagnosis relies on ICD-10 coding within participating institutions, and undiagnosed cases would not be captured, potentially selecting for more healthcare-engaged individuals. Finally, formal sensitivity analyses were not performed, and as this cohort was restricted to patients with OSA specifically, findings may not generalize to patients with other sleep-disordered breathing phenotypes. In this large propensity-matched real-world analysis of patients with T2DM and OSA, SGLT-2i use was associated with significantly lower risks of all-cause mortality, expanded MACE, heart failure exacerbation, ischemic stroke/TIA, haemorrhagic stroke, acute ischemic heart disease, and PE/DVT compared with DPP-4i. While the magnitude of the mortality association exceeds that observed in broader RCTs and may reflect residual confounding, healthy-adherer bias, or calendar-time confounding, these findings are consistent with and potentially extend the cardiovascular benefits of SGLT-2i to a high-risk and underrepresented patient population, warranting confirmation in prospective studies with more granular characterization of OSA severity and treatment adherence. The authors have nothing to report. The authors declare no conflicts of interest. The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer-review/10.1111/dom.70909.
Odai et al. (Thu,) studied this question.