The data analyzed in this investigation were sourced from the National Health and Nutrition Examination Survey (NHANES) database. Our study initially included 39,156 participants enrolled in the Mobile Examination Center from 2011 to 2018, with 37,117 participants excluded based on specific criteria. Ultimately, 2,039 participants were enrolled, among whom 1,346 did not use metformin and 693 were metformin users. We excluded participants with prediabetes or polycystic ovary syndrome. All participants were diagnosed with T2DM. Diabetes was diagnosed by meeting any of six criteria, including self-reported doctor-diagnosed diabetes, 2025 American Diabetes Association diagnostic criteria (fasting glucose ≥7.0 mmol/L, 2-hour oral glucose tolerance test ≥11.1 mmol/L, glycosylated hemoglobin [HbA1c] ≥6.5%, random glucose ≥11.1 mmol/L), or use of anti-diabetic drugs (excluding metformin monotherapy) or insulin (ascertained via prescription records or self-reported) [5]. Metformin use was self-reported in questionnaires. Albuminuria was defined as a urinary albumin/creatinine ratio ≥30 mg/g for logistic regression.

To control potential confounding factors, the following demographic and clinical variables were included as covariates in the analysis: sex, age, race, hypertension status, smoking status, estimated glomerular filtration rate, serum creatinine, uric acid, blood urea nitrogen, triglyceride levels, low-density lipoprotein cholesterol, and HbA1c. The inclusion of covariates was determined through a combination of prior research and consensus among clinical experts [6]. DTMR is an epidemiological approach that uses genetic variants associated with drug targets as instrumental variables to infer causal relationships between drug targets and diseases. Growth differentiation factor 15 (GDF15), a transforming growth factor β superfamily member, mediates metformin’s metabolic effects and may exert renal protection, thus serving as our Mendelian randomization (MR) instrumental variable [7]. In this analysis, nine instrumental single nucleotide polymorphisms (SNPs) (±100 kb of GDF15 locus) were selected from UK Biobank genome-wide association study (GWAS; 13,586,180 SNPs) based on genome-wide significance for HbA1c (P<5×10⁻⁸), effect allele frequency >0.01, linkage disequilibrium (r²<0.3), and exclusion of confounder-associated SNPs. Albuminuria outcomes used Chronic Kidney Disease Genetics Consortium (CKDGen) data (54,116 participants, 2,191,461 SNPs).

In the NHANES analysis, multivariable logistic regression was used to assess the association between metformin use and albuminuria, with subgroup and interaction tests to evaluate robustness. When participants were divided into metformin users (n=693) and non-users (n=1,346), logistic regression showed that metformin use was inversely associated with albuminuria across all models (model 1: odds ratio [OR]=0.599, P=0.001; model 2: OR=0.609, P=0.002; model 3: OR=0.670, P=0.025). Subgroup analysis revealed a significant sex interaction, with metformin reducing albuminuria risk in females (OR=0.53, P<0.001, interaction P=0.005) (Fig. 1) but not males. Stratified analyses demonstrated a metformin benefit in nonsmoking females only, and sex differences in smoking rates further contextualized these findings. In DTMR, inverse variance weighting was primary for causal inference, with sensitivity analyses addressing heterogeneity and pleiotropy. DTMR using GDF15 as the target showed genetically predicted metformin use reduced albuminuria risk (OR=0.299, P=0.039), consistent with regression results. MR analysis indicated that metformin has a significant protective effect on albuminuria via GDF15. However, further validation and more studies are needed to confirm these findings.

Our study suggested that metformin may be associated with reduced risk of albuminuria, and highlighted potential sex differences in its protective effects. Notably, non-smoking females gained the greatest benefits, whereas smokers, particularly males, had no notable reduction in albuminuria risk. This indicates that smoking may diminish the potential protective effects of metformin. The observed sex differences in the association between metformin use and albuminuria risk were significant and require further study, with several factors potentially driving these variations. First, males generally exhibit a higher baseline prevalence of albuminuria than females, which may diminish the detectability of significant risk reductions in males [8]. Second, sex-related differences in body composition, renal function, and hormonal profiles may contribute to variation in the pharmacokinetics and pharmacodynamics of metformin. Third, emerging evidence has indicated that females experience metformin-related side effects more frequently than males. Beyond the additional benefits females derive from metformin in reducing albuminuria, metformin has been observed to lower the incidence and mortality of colorectal cancer in females, as well as to improve recovery from coronavirus disease 2019 (COVID-19), compared to males [9]. These findings prompt a critical question: Is the more prominent side effect profile in females correlated with the enhanced therapeutic efficacy of metformin? Fourth, smoking, which is more common in males, induces inflammatory responses and oxidative stress [10]. The synergistic interaction between higher baseline smoking rates and inherent renal vulnerability may create a ‘double burden’ that overrides metformin’s renoprotective benefits.

Nevertheless, these findings remain preliminary and require validation through larger cohort studies and mechanistic investigations, particularly to address the striking sex differences that have emerged in our analysis.