Patients with type 2 diabetes initiating sodium–glucose cotransporter-2 inhibitor therapy had a modestly lower 6-year risk for diabetic foot disease compared with those starting glucagon-like peptide-1 receptor agonist treatment—11% vs 12%—corresponding to a risk ratio of 0.90 in an intention-to-treat analysis. The reduction was driven primarily by lower peripheral neuropathy risk, according to findings from a Danish population-based target trial emulation published in Annals of Internal Medicine.
The study, led by Frederik P.B. Kristensen, MD, PhD, of Aarhus University, analyzed 53,769 new sodium–glucose cotransporter-2 inhibitor (SGLT-2i) users and 30,380 new glucagon-like peptide-1 receptor agonist (GLP-1 RA) users identified through Danish national health care registries from 2013 to 2023. Outcomes were defined according to International Working Group on the Diabetic Foot criteria: peripheral neuropathy, peripheral artery disease, foot ulcers, and lower-limb amputation.
Notably, "differences did not emerge until after year 3, when 40% of SGLT-2i users and 32% of GLP-1 RA users had discontinued initial treatment," reported researchers. A timepoint when treatment adherence had substantially eroded. At 2 years, no meaningful separation was observed for neuropathy or any foot disease.
Individual Foot Disease Outcomes
Six-year risks for other foot disease components showed no statistically significant differences between SGLT-2i and GLP-1 RA users. Peripheral artery disease, foot ulcers, and all-cause mortality risks were similar between groups. Lower-limb amputation showed a numerically higher but imprecise point estimate favoring GLP-1 RAs. Specifically, at 6 years, SGLT-2i users had a neuropathy risk of 4% compared with 5% among GLP-1 RA users.
Per-Protocol Analysis Yields Divergent Signals
The per-protocol (on-treatment) analysis, which censored patients at treatment discontinuation, produced different findings for several outcomes. Foot ulcer risk was higher with SGLT-2is, and lower-limb amputation showed a more pronounced elevation. Follow-up duration was shorter for SGLT-2i users (median 1.5 years vs 1.8 years for GLP-1 RA users), reflecting the higher discontinuation rate observed in the SGLT-2i cohort—52% vs 44% at 6 years.
Research Cohort Corroborates Registry Findings
A parallel analysis in the DD2 (Danish Centre for Strategic Research in Type 2 Diabetes) research cohort—comprising 1,965 SGLT-2i users and 1,271 GLP-1 RA users with detailed clinical assessments—yielded directionally consistent but statistically imprecise estimates. The 6-year risk ratio for peripheral neuropathy was 0.64, and for any foot disease, 0.81. Additional adjustment for lifestyle behaviors (smoking status, physical activity, alcohol intake) and anthropometric measures (body mass index, waist circumference) did not meaningfully alter estimates.
Surveillance Differences Raise Detection Bias Concerns
The researchers identified differential follow-up patterns that may have influenced outcome ascertainment. "During follow-up, GLP-1RA users had more frequent diabetes-related hospital outpatient visits and podiatrist consultations than SGLT-2i users but had similar rates of neurologic examination or electrophysiologic testing and ultrasonography or angiography of the lower limbs." This surveillance asymmetry could contribute to detection bias, particularly for neuropathy—an outcome the researchers note "may be prone to detection bias because of the non–life-threatening symptoms, in contrast to late ischemic foot disease involving ulcers and amputations."
Baseline Characteristics and Confounding
Prior to inverse probability of treatment weighting, SGLT-2i users were older (median age 64 vs 61 years), more often male (63% vs 57%), had lower baseline hemoglobin A1c (7.9% vs 8.2%), less hospital-diagnosed obesity (11% vs 19%), and more cardiovascular disease (20% vs 17%). Following weighting for 45 covariates—including demographics, comorbidities, co-medications, type 2 diabetes duration, and biomarkers—all standardized mean differences were less than 0.1. In the research cohort, SGLT-2i users had lower median body mass index (31 vs 34 kg/m2) and waist circumference (107 vs 111 cm).
Methodological Considerations
The researchers acknowledge several limitations. Outcome identification relied on hospital diagnoses, with validation studies reporting positive predictive values of 74% for neuropathy and 70% for peripheral artery disease coding. The study excluded patients with prior foot disease and those with contraindications to either drug class. Denmark's homogeneous population may limit generalizability.
Addressing the broader implications, Kristensen and colleagues wrote: "The intention-to-treat effect found in our study might not be clinically meaningful because the observed pharmacologic effects depend on treatment adherence. For example, a potential adverse effect of SGLT-2is on lower-limb amputation could have been masked in our study by low adherence and frequent switching to GLP-1RA therapy."
Clinical Context
The findings emerge against a backdrop of uncertainty regarding SGLT-2i effects on foot outcomes. Canagliflozin was associated with elevated amputation risk in the CANVAS trial but not in the subsequent CREDENCE trial. The current study's active-comparator design limits interpretation: "GLP-1RAs might result in lower amputation risk than SGLT-2is, either because of antiatherosclerotic effects or if SGLT-2is increase this risk. Our study cannot distinguish between these possibilities without a placebo comparator," the researchers wrote.
The study was funded by Aarhus University and Center for Population Medicine. Disclosure forms are available with the article online at Annals.org.
Source: Annals of Internal Medicine