Association of reverse triiodothyronine and euthyroid sick syndrome with mortality in critically ill patients: a multicenter prospective cohort study

Abstract Background Euthyroid sick syndrome (ESS), a common endocrine disorder in critically ill patients, lacks reliable diagnostic methods. Reverse triiodothyronine (rT3) is a promising biomarker but requires standardized diagnostic and prognostic cutoffs. This study aimed to define reference ranges of rT3 for ESS diagnosis and evaluate its independent prognostic value for mortality. Methods This prospective multicenter cohort study was conducted between June 2022 and April 2024 in seven hospitals across three central provinces in China. A total of 782 critically ill adults were enrolled, and rT3 and other thyroid function parameters were measured at intensive care unit (ICU) admission. ESS was diagnosed based on reduced free triiodothyronine (FT3) levels in the absence of primary or secondary thyroid dysfunction. Patients were followed for in-ICU, in-hospital, and 90-day post-discharge mortality. Results Among the 782 critically ill patients (median age 68 years; 63.68% male), ESS was diagnosed in 47.31%. The 2.5th percentile (P 2.5 ) of rT3 in ESS patients was 0.64 ng/mL, and the cohort-derived cutoff for discriminating ESS was 1.07 ng/mL, with an area under the receiver operating characteristic curve [AUC] of 0.750. In the primary prognostic analysis, restricted cubic spline modeling of baseline rT3 as a continuous variable did not show a statistically significant association with in-ICU mortality, in-hospital mortality, or 90-day post-discharge mortality. Exploratory outcome-derived cutoff analyses identified a cutoff of 1.85 ng/mL for 90-day post-discharge mortality. At the 1.85 ng/mL cutoff, lower rT3 was initially associated with reduced 90-day post-discharge mortality (adjusted hazard ratio = 0.67, 95% confidence interval: 0.48–0.94), but this finding was not robust after additional adjustment and sensitivity analyses. Similarly, rT3 showed limited predictive value for in-ICU and in-hospital mortality. Exploratory machine learning analyses suggested that rT3 may contribute to mortality prediction in selected settings, with rT3 at ICU discharge ranking among the predictors for in-hospital and 90-day mortality and baseline rT3 showing possible relevance to in-ICU mortality in patients with infectious diseases. Conclusion In critically ill patients, rT3 was associated with an FT3-defined ESS phenotype but showed limited and unstable independent prognostic value for mortality outcomes. Data-derived rT3 cutoffs should be considered exploratory and require external validation before clinical use.