Diabetes mellitus is associated with metabolic alterations, but amino-acid differences among patients with coronary artery disease (CAD) remain incompletely characterized. We screened 45 amino acids and related metabolites by diabetes status in an exploratory retrospective disease-restricted cohort (recorded diseased-vessel count ≥ 1; n = 33). Argininosuccinic acid was available for 26 patients. Comparisons used exact Wilcoxon tests, Cliff’s delta, and Benjamini–Hochberg correction. Estimated argininosuccinic acid concentrations were higher in diabetes (0.28 [0.21–0.34] vs. 0.15 [0.13–0.19] µmol/L; Cliff’s delta = 0.765; p < 0.001; q = 0.035), the only association surviving false-discovery-rate correction. All argininosuccinic acid estimates were extrapolated below the lowest calibrator and should be interpreted as semiquantitative. The diabetes coefficient remained positive in crude and age- and sex-adjusted HC3-robust models, but renal and influence analyses were imprecise. Argininosuccinic acid correlated inversely with eGFR (rho = −0.445; p = 0.023) and positively with creatinine (rho = 0.486; p = 0.012). The diabetes-by-CAD interaction was not significant (p = 0.183). These data suggest a hypothesis-generating association requiring validation with an appropriate calibration range and careful renal phenotyping.

Background/Objectives: Sodium–glucose cotransporter-2 inhibitors (SGLT2is) exert metabolic and renal effects extending beyond glycemic control, yet their relationship with circulating trace-element profiles remains poorly characterized. We investigated whether prevalent SGLT2i use was associated with differences in serum trace elements and targeted metabolites, with particular attention to renal function and the possibility of a shared multielement phenotype. Methods: In this exploratory cross-sectional study, serum trace elements and targeted amino acid metabolites were evaluated according to prevalent SGLT2i exposure. The principal clinical comparison comprised participants with diabetes mellitus (DM) receiving SGLT2i (n = 6) versus those not receiving SGLT2i (n = 10). Secondary analyses incorporated the broader three-group cohort. Because of the small exposed group and nonrandom treatment allocation, analyses were considered hypothesis-generating. Group comparisons were complemented by permutation testing, regression adjusted for recalculated uncapped 2021 CKD-EPI estimated glomerular filtration rate (eGFR), false-discovery-rate correction, inter-element correlation analysis, principal-component analysis, and influence diagnostics. Results: Among participants with DM, circulating Cu, Mg, Se, and Zn concentrations were lower in SGLT2i users, with concordant directions across complementary analyses. In the secondary analysis incorporating multiplicity correction across eight molecular outcomes, Cu (q = 0.015) and Se (q = 0.047) met the prespecified FDR threshold, whereas Mg and Zn did not (both q = 0.051). Exploratory principal-component analysis identified a first component explaining 70.8% of the variance across the four measured elements. PC1 was lower among SGLT2i-treated participants and remained associated with exposure after eGFR adjustment (β = −2.43, 95% CI −4.22 to −0.65; p = 0.008); however, this component represents statistical covariance only and should not be interpreted as evidence of a common biological mechanism or shared trace-element depletion. Conclusions: Prevalent SGLT2i use among participants with DM was associated with directionally lower circulating concentrations across several measured trace elements; however, after multiplicity correction, only Cu and Se met the prespecified FDR threshold, whereas Mg and Zn did not. The cross-sectional design, small exposed subgroup, clinical imbalance, and absence of pretreatment, dietary, urinary trace element, inflammatory, treatment duration, and adherence data preclude attribution of this pattern to SGLT2 inhibition. These findings should therefore be considered hypothesis-generating and warrant prospective validation with serial pre- and post-treatment measurements to determine whether the observed phenotype reflects pharmacological effects, treatment selection, the underlying clinical phenotype, or a combination of these factors.