BackgroundGrowing studies have shown that insulin resistance (IR) is associated with poor prognoses among patients with diabetes, whereas the association between IR and mortality has not been determined. Hence we aimed to evaluate the associations between IR and all-cause, cardiovascular diseases (CVDs) and cancer-related mortality in patients with diabetes.MethodsWe enrolled 2,542 participants with diabetes with an average age of 57.12 ± 0.39 years and 52.8% men from the 1999–2014 National Health and Nutrition Examination Survey (NHANES 1999–2014). A novel metabolic score for insulin resistance (METS-IR) was considered as alternative marker of IR. Mortality data were obtained from the National Death Index records and all participants were followed up until December 31, 2015. Cox proportional hazards regression, restricted cubic spline and Kaplan-Meier survival curves were performed to evaluate the associations between METS-IR and all-cause and cause-specific mortality in patients with diabetes.ResultsDuring 17,750 person-years of follow-up [median (months), 95% CI: 90, 87–93], 562 deaths were documented, including 133 CVDs-related deaths and 90 cancer-related deaths. Multivariate Cox regression showed that compared with Quintile 1 (METS-IR ≤ 41), METS-IR in Quintile 2, 3, and 4 was all associated with all-cause mortality (Q2 vs. Q1: HR 0.65, 95% CI 0.49–0.87, P = 0.004; Q3 vs. Q1: HR 0.69, 95% CI 0.50–0.96, P = 0.029; Q4 vs. Q1: HR 0.57, 95% CI 0.36–0.91, P = 0.019; respectively). Restricted cubic spline indicated that METS-IR was non-linearly associated with all-cause and CVDs-related mortality. Threshold effect analyses determined that threshold values of METS-IR for all-cause and CVDs-related mortality were both 33.33. Only METS-IR below the threshold was negatively associated with all-cause and CVDs-related mortality (HR 0.785, 95% CI 0.724–0.850, P < 0.001; HR 0.722, 95% CI 0.654–0.797, P < 0.001; respectively). Sensitivity analyses showed that when excluding participants who died within 1 years of follow-up, the results of threshold effect analyses remained consistent, whereas excluding participants with CVDs, METS-IR below the threshold was only negatively correlated with all-cause mortality. Subgroup analyses indicated that for all-cause mortality, the results were still stable in all subgroups except newly diagnosed diabetes, but for CVDs-related mortality, the association persisted only in participants who were ≤ 65 years, male, White, non-White, already diagnosed diabetes, or uesd oral drugs, insulin, insulin sensitizing drugs.ConclusionMETS-IR was non-linearly associated with all-cause and CVDs-related mortality in patients with diabetes, and METS-IR below the threshold was negatively associated with all-cause and CVDs-related mortality.