Thyroid function is highly sensitive to acute illness and systemic stress, and the FT3/FT4 ratio has been proposed as an integrative marker of peripheral thyroid hormone metabolism. In hospitalized patients, levels of the active hormone triiodothyronine often fall as the body diverts metabolic resources during severe disease, a pattern seen in euthyroid sick syndrome. The researchers, led by Min Peng, Jia Wang and Jinming Yu of the First Affiliated Hospital of Shenzhen University, Shenzhen Second People’s Hospital, set out to determine whether this ratio carries prognostic information in a general hospitalized population, a setting largely unexplored by earlier work that focused on community-dwelling Western populations or specific disease cohorts.

The team screened electronic medical records from Shenzhen Second People’s Hospital between October 2013 and March 2025, identifying 66,210 unique patients. They excluded 39,987 patients, most of them because thyroid function measurements were incomplete, and removed four individuals with documented thyroid disorders such as hyperthyroidism, hypothyroidism, Hashimoto’s thyroiditis or goiter. The final cohort comprised 26,223 patients with a median age of 67 years, just over half of them male. Blood samples were drawn at 5 a.m. after an overnight fast, and TSH, FT3 and FT4 were measured by chemiluminescence immunoassay. Thyroid testing was not routine; it was ordered for patients with severe illness, those screened for cardiovascular and cerebrovascular risk, or those with endocrine comorbidities.

During the index hospitalizations, 1,150 patients died from any cause. In the fully adjusted Cox proportional hazards model, each unit increase in TSH was associated with a 1.5 percent higher hazard of death (hazard ratio 1.015, 95% confidence interval 1.010 to 1.020), and each unit increase in FT4 with a 1.7 percent higher hazard (HR 1.017, 95% CI 1.002 to 1.033). In contrast, higher FT3 was associated with lower mortality (HR 0.865, 95% CI 0.806 to 0.929), corresponding to a 13.5 percent risk reduction per unit. The FT3/FT4 ratio, multiplied by 100 for analysis so that each unit represents a 0.01 increase in the raw ratio, was associated with a 2.6 percent lower hazard per unit (HR 0.974, 95% CI 0.964 to 0.983).

When the ratio was split into tertiles, patients in the middle tertile had a 19 percent lower risk of death than those in the lowest tertile (HR 0.807, 95% CI 0.688 to 0.947), and those in the highest tertile had a 36 percent lower risk (HR 0.638, 95% CI 0.529 to 0.770). Unadjusted Kaplan–Meier curves displayed through 100 days after admission showed clear separation in survival probability across the tertiles, with a log-rank P value below 0.001. Restricted cubic spline analyses found evidence of a nonlinear association only for TSH (P for non-linearity 0.028); the relationships for FT3, FT4 and the FT3/FT4 ratio were consistent with linear trends.

Several sensitivity analyses tested the robustness of these findings. Because the global Schoenfeld residual test indicated non-proportional hazards in the primary model, driven mainly by age, the authors re-fitted a Cox model stratified by age group, sex, coronary heart disease and cancer; the ratio association remained essentially unchanged (HR 0.970, 95% CI 0.961 to 0.980), and the proportional hazards assumption was no longer violated. Excluding the lowest and highest 5 percent of values for each hormone generally supported the primary results, particularly for FT3 (HR 0.854, 95% CI 0.768 to 0.949) and the ratio (HR 0.973, 95% CI 0.959 to 0.987), although the associations for TSH, FT4 and the middle tertile were attenuated and no longer statistically significant.

The most demanding test addressed selection bias. Because thyroid testing was ordered on clinical indications rather than systematically, the analyzed cohort may not represent all hospitalized patients. The authors therefore applied inverse-probability-of-selection weighting using the entire source population of 66,210 patients, estimating each patient’s probability of inclusion from measured characteristics. After weighting, covariate balance improved substantially, with the maximum absolute standardized mean difference falling from 0.153 to 0.019. The inverse association persisted (HR 0.954, 95% CI 0.945 to 0.963), and an expanded weighting model that also incorporated admission vital signs produced nearly identical results (HR 0.953, 95% CI 0.943 to 0.962).

Subgroup analyses suggested that the association was not uniform. The ratio was inversely associated with mortality among patients admitted during the day (HR 0.970, 95% CI 0.960 to 0.979) but not among those admitted at night (HR 1.019, 95% CI 0.988 to 1.051; P for interaction 0.001). It was also inversely associated among patients without kidney disease (HR 0.968, 95% CI 0.958 to 0.978) but positively associated among those with kidney disease (HR 1.029, 95% CI 1.005 to 1.053; P for interaction 0.020). Formal interaction tests additionally flagged heterogeneity by sex, coronary heart disease, diabetes, cancer, stroke and smoking. The authors caution that, given the number of comparisons and limited events in some strata, these subgroup findings are exploratory and should be interpreted cautiously.

Exploratory analyses also linked the ratio to admission vital signs: higher FT3/FT4 ratios were associated with lower heart rate, lower systolic blood pressure, higher diastolic blood pressure and a modestly lower body temperature. In multivariable Cox models, however, only admission temperature was independently associated with mortality (HR 1.002, 95% CI 1.000 to 1.004), while heart rate and blood pressures were not. The authors emphasize that because hormones and vital signs were measured nearly simultaneously, these analyses cannot establish causal mediation, and the direction of the vital-sign associations differs from reports in healthy community populations, reinforcing the idea that in hospitalized patients the ratio may reflect disease severity rather than normal thyroid function.

Mechanistically, the authors suggest that reduced FT3 may reflect impaired energy metabolism and altered substrate utilization, potentially contributing to oxidative stress and organ dysfunction, while low FT3/FT4 ratios have previously been linked to elevated inflammatory markers such as C-reactive protein and ferritin, and to enhanced sympathetic and renin-angiotensin-aldosterone system activity. Prior studies have reported U-shaped associations between the ratio and mortality in community populations and J-shaped associations with cardiovascular mortality in chronic kidney disease, making the linear inverse association observed here in acutely ill patients context-dependent rather than universal.

The study has important limitations. Excluded patients had a higher mortality rate than included patients (6.80 percent versus 4.39 percent), and although weighting balanced measured covariates, residual selection bias cannot be excluded. Medication information, including thyroid hormone preparations and antithyroid drugs, was unavailable; laboratory indicators such as CRP, albumin and renal function, body mass index, nutritional status, frailty and validated severity scores could not be adjusted for. Only in-hospital mortality was assessed, and the single-hospital design in Shenzhen may limit generalizability. The authors therefore frame the FT3/FT4 ratio as an independently associated marker that may serve as a potential indicator of outcomes among patients undergoing thyroid function testing, and they call for prospective, multicenter studies with systematic thyroid assessment to validate the findings before any clinical use.

Subject of Research: The association between the FT3/FT4 thyroid hormone ratio and in-hospital all-cause mortality in hospitalized patients

Article Title: Association between the FT3/FT4 ratio and in-hospital mortality among hospitalized patients undergoing thyroid function testing

Article References: Peng, M., Wang, J., Yu, J., Ren, L., & Xu, G. (2026). Association between the FT3/FT4 ratio and in-hospital mortality among hospitalized patients undergoing thyroid function testing. Frontiers in Medicine, 13, Article 1792939. https://doi.org/10.3389/fmed.2026.1792939

Image Credits: AI Generated

DOI: 10.3389/fmed.2026.1792939

Keywords: FT3/FT4 ratio, thyroid hormones, in-hospital mortality, retrospective cohort, hospitalized patients, TSH, FT3, FT4, proportional hazards, selection bias, Shenzhen, euthyroid sick syndrome