The study included all patients (n=14,672) who underwent thyroidectomy for thyroid nodules with definite pathologic diagnosis in Chinese PLA General Hospital from January 2011 to December 2020. The exclusion criteria involved past history of thyroid surgery, iodine-131 therapy or radiofrequency ablation; age<18 years; coexistence of other cancers, acute infection, liver failure, or renal failure; pregnant or breastfeeding women; missing data on FT3, FT4 or TSH; known clinical hyperthyroidism or hypothyroidism according to the electric records system or patients’ self-reports, or with FT3, FT4, TSH out of normal ranges; subjects who had taken medications which can alter thyroid function (anti-thyroid drugs, thyroid replacement drugs, amiodarone or lithium); postoperative pathology of thyroid nodules confirmed as borderline thyroid tumor (described as “non-invasive follicular thyroid neoplasm with papillary-like nuclear features, NIFTP”, “well differentiated tumor of uncertain malignant potential, WDT-UMP”, “follicular tumor of uncertain malignant potential, FT-UMP”, “hyalinizing trabecular tumor, HTT” or general descriptions like “borderline follicular tumor” in the pathology report), medullary carcinoma, anaplastic carcinoma, or other rare types of thyroid carcinoma. Finally, 9,515 individuals were included for analysis ( Figure 1 ). The study was conducted following the Helsinki Declaration and approved by the Ethics Committee of Chinese PLA General Hospital with waived informed consent considering the collection of deidentified retrospective data.

The subjects received a physical examination and thyroid function test less than one week before the thyroidectomy. Height and weight were measured by trained nurses in a standardized process and body mass index (BMI) was calculated as weight(kg)/height²(m²). Serum FT3 (reference range: 2.76-6.30 pmol/L), FT4 (reference range: 10.42-24.32 pmol/L), TSH (reference range: 0.35-5.50 mU/L), thyroid peroxidase antibody (TPOAb, reference range:< 60 IU/mL) and thyroglobulin antibody (TgAb, reference range:< 60 IU/mL) were detected by the ADVIA centaur XP automated chemiluminescence analyzer (SIEMENS, USA) at the central laboratory of the hospital. Data regarding family history of thyroid cancer and the histories of related diseases were collected from the electric records system or patients’ self-reports.

The pathological type of thyroid nodule, nodule size, and the coexistence of Hashimoto thyroiditis (HT) were collected from postoperative pathology reports. According to the standard diagnostic workflow in the Pathology Department, after a pathologist provided an initial report of pathological findings, a senior pathologist conducted a secondary review. For cases with discrepancies, consensus was reached through collective discussion. Nodule size was defined as the maximum pathologic diameter of benign nodules for patients with benign thyroid disease and malignant nodules for patients with DTC.

where β_T  (clearance exponent for T4) = 1.1 × 10⁻⁶/s, D_T  (EC₅₀ for TSH) = 2.75 mU/L, K₄₁ (dissociation constant of T4 at thyroxine-binding globulin) = 2.0 × 10¹⁰ L/mol, TBG (standard concentration of thyroxine-binding globulin) = 300 nmol/L, K ₄₂ (dissociation constant of T4 at transthyretin) = 2.0 × 10⁸ L/mol, TBPA (standard transthyretin concentration) = 4.5 μmol/L, and α _T (dilution factor for thyroxine) = 0.1/L (20). SPINA-GT is defined as the maximum amount of T4 that the thyroid can release in a given time unit under stimulated conditions, as proposed and validated by Dietrich et al. (20). Here we calculated SPINA-GT using R software with the “SPINA” package (20, 21). No winsorizing or trimming was applied to thyroid parameters in consideration that euthyroid status inherently constrained extreme deviations.

Higher TSHI or TT4RI indicates lower central thyroid hormone sensitivity, and higher FT3/FT4 indicates higher peripheral thyroid hormone sensitivity. The value of SPINA-GT reflects the maximum secretory capacity of the thyroid gland (20).

Continuous variables were expressed as mean ± standard deviation (SD) for normally distributed data and median (interquartile range) for skewed distributed data. Categorical variables were expressed as numbers (percentages). Independent samples t-test and Mann-Whitney U test were used for between-group comparisons of normally and skewed distributed variables, respectively. The chi-square test was used for categorical variables. TSH, TT4RI, FT3/FT4, and SPINA-GT were natural log (ln) transformed in further analyses given their skewed distributions. Different logistic regression models were constructed to estimate odds ratios (OR) with 95% confidence intervals (95%CI) for DTC across quartiles or per SD increase of thyroid parameters. Quartile-based analysis focused on evaluating categorical trends, while SD-change analysis quantified associations across continuous distributions. These approaches analyzed the data from distinct analytical dimensions. The crude model was without adjustment; Model 1 was adjusted for age and sex; Model 2 was further adjusted for BMI, family history of thyroid cancer and Hashimoto thyroiditis confirmed by pathology. The covariates included in the models were potential confounding and/or mediating factors for the relationship between thyroid function and DTC. As the proportion of missing data for each covariate was less than 5% ( Supplementary Table S1 ), missing data were not processed.

Stratified logistic regression analyses were performed in subgroups stratified by age 55 (age for the TNM staging of DTC (22)), sex, BMI (Chinese BMI classification (23): BMI<24, underweight and normal weight; 24≤BMI<28, overweight; BMI≥28, obese), diabetes, hypertension, Hashimoto thyroiditis, the status of thyroid autoantibodies, nodule size and surgery periods, with their interactions tested. Generalized additive model and smoothed curve fitting were used to provide more detailed information about the relationship between thyroid hormone sensitivity indices (not ln-transformed for clearer and more intuitive visualization) and the probability of DTC. When significant nonlinearity was detected via likelihood ratio testing, the inflection point was calculated using a recursive algorithm and the threshold effects were analyzed using segmented regression models. 95% CIs for inflection points were estimated using bootstrap resampling with 1,000 iterations. The degrees of association between DTC and various thyroid parameters were assessed by comparing the standardized regression coefficient β of each thyroid parameter in the multivariable-adjusted model and receiver operating characteristic (ROC) curve analysis was conducted. All analyses were performed with SPSS version 26 and R software version 4.3.1. P values<0.05 (2-sided) were considered significant statistically.

The study included 9,515 individuals, of whom 7,447 suffered from DTC. The average age of the overall population was 45.47 ± 11.38 years, and 28.9% were male. Compared to subjects with benign disease, subjects with DTC exhibited younger age, higher BMI, TgAb and TPOAb levels, smaller nodule size, and higher prevalences of Hashimoto thyroiditis confirmed by pathology and family history of thyroid cancer (all P<0.05), with no statistical difference in sex distribution, history of external neck irradiation, hypertension, or diabetes. In terms of thyroid parameters, FT4, TSH, TSHI, and TT4RI were higher, while FT3, FT3/FT4, and SPINA-GT were lower in subjects with DTC (all P<0.001) ( Table 1 ). Vice versa, the proportions of DTC increased across quartiles of TSH, TSHI, and TT4RI but decreased with rising quartiles of FT3, FT3/FT4, and SPINA-GT (all P<0.001). Differences existed in the proportions of pathology for FT4 quartile groups (P<0.001) but not concentration-dependently ( Supplementary Figure S1 ).

Reduced FT3, FT3/FT4, or SPINA-GT was associated with DTC: After adjustment for age, sex, BMI, family history of thyroid cancer and Hashimoto thyroiditis confirmed by pathology, for each SD change in FT3, FT3/FT4, and SPINA-GT, the ORs (95% CI) for DTC were 0.79 (0.75-0.83), 0.81 (0.77-0.86), and 0.78 (0.74-0.82), respectively ( Table 2 ). Analysis comparing extreme quartiles (Q1 vs. Q4) showed similar associations ( Supplementary Table S2 ), with subjects in the lowest quartiles (Q1) of FT3, FT3/FT4, and SPINA-GT demonstrating a significantly higher risk of DTC compared to those in Q4. Increased FT4 was associated with DTC in the crude model, which was no longer significant after adjustment for confounders ( Table 2 ). However, subjects with the highest quartile of FT4 still had a higher prevalence of DTC than those with the lowest quartile in all models (all P values < 0.05) ( Supplementary Table S2 ). Increased TSH, TSHI or TT4RI was associated with DTC: After adjustment for confounders, for each SD change in TSH, TSHI, and TT4RI, the ORs (95% CI) for DTC were 1.33 (1.26-1.40), 1.34 (1.27-1.41), and 1.35 (1.28-1.42), respectively ( Table 2 ). Quartile analysis showed analogous results ( Supplementary Table S2 ), with subjects in the highest quartiles (Q4) of TSH, TSHI, and TT4RI demonstrating a significantly higher risk of DTC compared to those in Q1.

The ORs (95%CI) of per SD increase in thyroid parameters for DTC were calculated in subgroups after adjustment for potential confounders ( Table 3 ). In subjects stratified by age, sex, BMI, diabetes or hypertension, TSHI and TT4RI were positively associated with DTC, while FT3/FT4 and SPINA-GT were negatively associated with DTC (all P values <0.05). The association of DTC with TSHI, TT4RI or SPINA-GT was stronger in females than in males (P for interaction <0.05). In addition, there was a statistically significant interaction between FT3/FT4 and BMI with DTC (P for interaction = 0.038).

In subjects with HT confirmed by pathology, the association of TT4RI or SPINA-GT with DTC was no longer significant. Additionally, there was a statistically significant interaction between FT3/FT4 and HT with DTC, similarly for SPINA-GT. However, in both thyroid autoantibody-positive and negative individuals, all thyroid hormone sensitivity indices and SPINA-GT were significantly associated with DTC. No statistical heterogeneity in this association was observed in subgroups of different statuses of autoantibodies (all P for interaction >0.05). The contribution of thyroid hormone sensitivity indices remained in subgroups of nodule size (all P for interaction >0.05). And the association of SPINA-GT with DTC was no longer statistically significant in subjects with nodule size ≤1cm. Subgroup analysis between the surgical periods of 2011–2015 and 2016–2020 demonstrated results consistent with those in the total population.

Adjusted smoothed curves suggested a non-linear relationship between thyroid hormone sensitivity indices and the probability of DTC ( Figure 2 ). Threshold effect analysis showed that the inflection points were 2.72 (95% CI: 2.02-2.89), 18.76 (95% CI: 16.40-24.28) and 0.27 (95% CI: 0.25-0.39) for TSHI, TT4RI and FT3/FT4, respectively. The probability of DTC increased with TSHI up to the inflection point (2.72) (OR 2.125, 95% CI 1.850-2.441, P<0.001). When TSHI was >2.72, the relationship between TSHI and the probability of DTC was not significant (OR 1.024, 95% CI 0.834-1.264, P= 0.820). The probability of DTC increased with a more prominent trend when TT4RI was below the inflection point (18.76) (OR 1.087, 95% CI 1.067-1.108, P<0.001). No significant association was found when FT3/FT4 was <0.27 (OR 10.213, 95% CI 0.072-1237.818, P=0.350); however, the probability of DTC decreased with FT3/FT4 when FT3/FT4 was >0.27 (OR 0.003, 95% CI 0.001-0.011, P<0.001). These results suggested a strong and significant association of higher central and peripheral thyroid hormone sensitivity with lower risk of DTC in the thyroidectomy population.

The association of DTC with central thyroid hormone sensitivity indices (TT4RI and TSHI) was stronger than with other thyroid parameters such as TSH ( Table 4 ). Based on the ROC curves for the distinction between DTC and benign disease ( Supplementary Figures S2 , S3 ), among various thyroid parameters, the highest sensitivity was 74.43% in TSH, and the highest specificity was 82.40% in SPINA-GT. The area under the curve (AUC) for TSHI or TT4RI was higher than that for TSH; nevertheless, there was no single thyroid parameter with the coexistence of high sensitivity and specificity ( Table 5 ).