Preoperative triglyceride–glucose index and 30-day surgical site infection in OSCC patients without documented diabetes: a retrospective cohort study

Surgical site infection (SSI) remains an important complication after oral squamous cell carcinoma (OSCC) surgery. Although diabetes is a recognized infection risk factor, metabolic vulnerability in patients without documented diabetes is less understood. The triglyceride–glucose (TyG) index is a surrogate marker of insulin resistance, but its association with postoperative SSI in OSCC remains unclear. This retrospective cohort included 381 OSCC patients without documented diabetes who underwent curative-intent surgery between January 2017 and December 2024. The standard TyG index was calculated as ln [triglycerides (mg/dL) × fasting plasma glucose (mg/dL) / 2]. The primary outcome was 30-day SSI adjudicated according to CDC/NHSN criteria. Multivariable logistic regression used a primary surgical-adjusted model including demographic, lifestyle/comorbidity, tumor, nutritional, and surgical-complexity variables. Dose–response, sensitivity, discrimination, calibration, reclassification, and apparent decision curve analyses were evaluated. Among 381 patients, 88 (23.1%) developed 30-day SSI. TyG was higher in patients with SSI than in those without SSI (8.77 ± 0.29 vs 8.59 ± 0.29, P < 0.001). In the primary surgical-adjusted model, each 0.1-unit increase in TyG was associated with higher SSI odds (adjusted odds ratio, 1.260; 95% confidence interval, 1.144–1.388; P < 0.001). The association remained directionally consistent after adjustment for HbA1c, exclusion of possible undiagnosed diabetes, restriction to TG/FPG measurements within 7 days before surgery, and extended perioperative adjustment. Restricted cubic splines showed a significant overall association without evidence of non-linearity. Adding TyG produced statistically significant but modest apparent discrimination gains in the clinical model (AUC, 0.634 to 0.712; ΔAUC, 0.078; P = 0.005) and surgical model (AUC, 0.682 to 0.737; ΔAUC, 0.054; P = 0.018), although optimism-corrected performance was attenuated. Decision curve analysis was exploratory. In OSCC patients without documented diabetes, higher preoperative TyG was associated with increased 30-day SSI risk and provided modest incremental predictive information. Given the internally derived design, limited events per predictor parameter, attenuation after optimism correction, and lack of external validation, these findings should be considered exploratory before clinical implementation.