Safety Signals in Older Diabetes Care
In this observational matched cohort, SGLT2 inhibitor use was associated with lower hazards for several safety and vascular outcomes but higher hazards for genital candidiasis and heart-failure hospitalization versus DPP-4 inhibitors; these associations are not causal estimates or treatment recommendations.
> Research explainer: This briefing examines verified primary research published 58 days before the briefing date. It is not a same-day research update and does not provide medical advice.
Evidence
This research explainer examines a retrospective, new-user, active-comparator cohort study of adults aged 80 years or older with type 2 diabetes. Using TriNetX electronic-health-record data, the investigators matched new users of sodium-glucose cotransporter-2 inhibitors (SGLT2i) and dipeptidyl peptidase-4 inhibitors (DPP-4i) one-to-one with propensity scores that incorporated demographics, comorbidities, frailty proxies, and concomitant medications. The matched analysis contained 65,119 people in each treatment cohort; mean age was 82.2 years, 48.7% of participants were female, and mean follow-up was 269 days. [pmid:42366424]
Relative to DPP-4i, SGLT2i exposure was associated with lower hazards of falls (hazard ratio [HR] 0.90, 95% confidence interval [CI] 0.85–0.95), hip fracture (0.78, 0.69–0.89), acute kidney injury (0.82, 0.78–0.85), urinary tract infection (0.79, 0.75–0.83), hypoglycemia (0.74, 0.67–0.82), and volume depletion (0.91, 0.86–0.97). The same comparison was associated with lower hazards of stroke (0.88, 0.82–0.95) and all-cause mortality (0.73, 0.70–0.76). [pmid:42366424]
Not all observed associations pointed in the same direction. SGLT2i exposure was associated with a higher hazard of genital candidiasis (HR 1.70, 95% CI 1.50–1.93) and a modestly higher hazard of heart-failure hospitalization (1.07, 1.02–1.12). The study reported no statistically significant difference between treatment cohorts for hypotension, syncope, or myocardial infarction. [pmid:42366424]
The falls association was reported as consistent in subgroups using insulin, benzodiazepines, or loop diuretics. A sensitivity analysis that used fixed 250-day follow-up confirmed all primary findings. These checks add information about the stability of the reported comparison within the study’s defined data and design; they do not turn the comparison into a randomized experiment. [pmid:42366424]
Analysis — Comparing Associations Across Safety Domains
The useful signal here is not a single verdict about either drug class. It is the pattern of associations across outcomes in a population that is often difficult to study: adults at least 80 years old initiating one of two diabetes drug classes. In the matched cohort, several outcomes commonly relevant to safety monitoring—falls, fracture, acute kidney injury, hypoglycemia, volume depletion, and urinary tract infection—had hazard ratios below 1.00 for SGLT2i relative to DPP-4i. Stroke and all-cause mortality were also lower in this comparison. At the same time, genital candidiasis had a materially higher reported hazard and heart-failure hospitalization had a slightly higher reported hazard. [pmid:42366424]
Hazard ratios describe relative differences in the timing of observed events between groups; they do not state how many events occurred in either group. For example, an HR below 1.00 is evidence of a lower observed hazard in this analysis, but without absolute event rates it cannot establish the size of any absolute difference. The contrast between lower volume-depletion and acute-kidney-injury hazards, no significant hypotension difference, and higher genital-candidiasis hazard underscores why outcome-by-outcome reporting matters more than a broad safety label. [pmid:42366424]
The active-comparator, new-user design and propensity-score matching are meaningful design choices because they seek to make the two observed cohorts more comparable at treatment initiation. Still, they address measured characteristics and recorded proxies, not every possible difference in health status, clinical decision-making, care access, or outcome capture. The appropriate reading is therefore comparative and bounded: within this electronic-health-record cohort and follow-up window, SGLT2i use was associated with a mixed set of outcome differences versus DPP-4i. It does not establish that the medication class caused those differences, nor does it supply a universal prescribing conclusion. [pmid:42366424]
Limitations
This was a retrospective observational study, so residual confounding and other biases may remain after matching; the reported associations do not prove causation. Frailty was represented through electronic-health-record proxies rather than directly captured, validated frailty scales. [pmid:42366424]
The available study information does not provide absolute event rates, so the hazard ratios should not be read as absolute risk reductions or increases. The cohort was limited to adults aged 80 years or older with type 2 diabetes who met the study’s eligibility and exclusion criteria, including the specified treatment-initiation design; applicability outside that population is uncertain. [pmid:42366424]
Finally, mean follow-up was 269 days and the sensitivity analysis used 250 days. The findings therefore characterize associations over those study windows rather than long-term outcomes, individual experiences, or patient-specific decisions. [pmid:42366424]