Lower triglyceride-glucose index is independently associated with multidrug-resistant organism infection in patients with pyogenic liver abscess: a retrospective cohort study

Background The triglyceride-glucose (TyG) index, a surrogate marker of insulin resistance, has been linked to various infectious outcomes. However, no prior research has explored its independent association with multidrug-resistant organism (MDRO) infection among patients with pyogenic liver abscess (PLA). This retrospective cohort study aimed to investigate the relationship between TyG index and MDRO infection identified at or during the index admission in hospitalized PLA patients. Methods We conducted a retrospective cohort study of 440 PLA patients. The association between the TyG index and MDRO infection was evaluated using univariate and multivariate logistic regression, adjusting for a comprehensive panel of confounders, including severity of illness proxies. Receiver operating characteristic (ROC) curves were constructed to determine the predictive value of the TyG index, and restricted cubic splines (RCS) were used to examine the dose–response relationship. Interactions were examined via subgroup analysis, and adjusted odds ratios (ORs) accompanied by 95% confidence intervals (CIs) were depicted in a forest plot. Results MDRO infection was identified in 40 (9.1%) patients, including one patient that was simultaneously infected with both ESBL-KP and E. coli . After full multivariate adjustment (Model III), continuous TyG index exhibited a significant inverse independent association with MDRO infection: per standard deviation (SD) elevation of TyG index, the adjusted OR (aOR) was 0.470 (95% CI: 0.298–0.742). Relative to the lowest TyG tertile (T1, reference group), patients in the second (T2) and third (T3) tertiles presented markedly reduced MDRO odds, with aORs of 0.244 (95% CI: 0.095–0.628) and 0.241 (95% CI: 0.089–0.655), respectively. RCS modeling showed a declining relationship without statistically significant non-linearity (non-linear p -value = 0.233); because a non-significant non-linearity test does not establish linearity, and MDRO rates were similar in T2 (5.4%) and T3 (6.7%), the pattern is best characterized as threshold-like, with risk concentrated in the lowest TyG range. The TyG index yielded modest predictive accuracy for MDRO infection, with an area under the ROC curve (AUC) of 0.629. Subgroup interaction analyses detected no significant effect modification across all stratified clinical subgroups (all interaction p -values ranged from 0.146 to 0.864 in adjusted models). Conclusion Among PLA patients, reduced TyG index independently correlates with elevated risk of MDRO infection. These novel findings reveal an inverse association of uncertain mechanism between host metabolic status and MDRO infection risk in PLA.