What this is
- This research examines the relationship between preoperative thyroid hormone levels, specifically (), and outcomes after bariatric surgery.
- It focuses on () and () in a cohort of euthyroid adults.
- The study highlights sex differences in how levels influence these outcomes, particularly noting that lower levels are associated with reduced and , especially in females.
Essence
- Low preoperative levels are linked to less and lower rates after bariatric surgery. This association is particularly strong in females.
Key takeaways
- High baseline levels correlate with greater after bariatric surgery. In females, higher levels were significantly associated with successful , while this association weakened in males.
- rates increased with higher levels, with an odds ratio of 7.42 for females in the highest tertile compared to the lowest. This highlights the importance of in predicting postoperative glycemic control.
- Sex-specific thresholds for were identified, with levels ≤5.0 pmol/L in males and ≤4.57 pmol/L in females linked to suboptimal outcomes. These thresholds could inform preoperative assessments and risk stratification.
Caveats
- The study is observational, limiting causal interpretations of the associations found between levels and surgical outcomes.
- Resting energy expenditure was not directly measured, which is crucial for understanding the metabolic pathways linking to .
- Although patients with thyroid disease were excluded, serum thyroid antibody levels were not routinely checked, which may affect the results.
Definitions
- Free triiodothyronine (FT3): An active thyroid hormone that regulates metabolism and energy expenditure.
- Diabetes remission (DR): A state where blood glucose levels return to normal without the need for diabetes medications.
- Weight loss (WL): The reduction of body weight, typically measured as a percentage of baseline weight.
Simplified
Introduction
Type 2 diabetes mellitus (T2DM) and obesity represent major global health concerns [1, 2]. Although GLP‐1 receptor agonists have revolutionised obesity management, bariatric surgery remains the most durable intervention for severe obesity and T2DM [3]. However, predicting postoperative weight loss (WL) and diabetes remission (DR) remains challenging, as 20%–30% of patients experience inadequate weight reduction and suboptimal glycaemic control [4, 5]. Moreover, even when patients initially achieve successful WL and DR, these issues can recur [6, 7]. Identifying the baseline determinants of WL and DR is crucial for enhancing the effectiveness of bariatric surgery and reducing the risk of relapse.
Obesity is an energy imbalance disorder, with weight loss primarily depending on the extent of postoperative negative energy balance, in which thyroid hormones (TH) play a key role in regulating resting energy expenditure (REE) [8, 9, 10]. Hypothyroidism is associated with hypometabolism and weight gain, while hyperthyroidism to weight loss [10]. Low circulating levels of TH, even within normal reference concentrations, can be associated with the tendency to gain weight [11]. Free triiodothyronine (FT3) is a crucial component of TH, with low levels observed in fasting and underweight individuals, and elevated levels found in overfeeding and obesity [12, 13], suggesting that FT3 plays a central role in weight regulation and nutrient sensing. However, the impact of baseline TH levels, especially FT3, on WL after bariatric surgery remains largely unknown.
Glycaemic benefits from bariatric surgery are often attributed to sustained weight reduction [14, 15]. Hypothyroidism may lead to reduced REE and weight gain, potentially facilitating the incidence of T2DM [16, 17]. Meanwhile, thyroid dysfunction increases insulin resistance in muscle and adipose tissue, and decreases glucose transport in myocytes [18]. Improved insulin sensitivity is a key mechanism for DR following bariatric surgery. Evidence has suggested that low circulating levels of TH, even within laboratory reference ranges, may be related to an elevated risk of developing T2DM, especially among individuals with prediabetes [16, 19, 20]. Therefore, it is reasonable to deduce that baseline TH, particularly FT3, may influence glycaemic homoeostasis after bariatric surgery. However, no longitudinal studies have yet explored the associations of baseline TH levels with DR after bariatric surgery.
Considering these factors, we explored the cross‐sectional relationships of thyroid function with adiposity and metabolic parameters in patients with overweight or obesity, and longitudinally investigated the associations of baseline thyroid hormones with short‐ and long‐term WL and DR after bariatric surgery.
Materials and Methods
Study Design and Population
The data presented are from a prospective cohort study on bariatric surgery (Chinese Clinical Trial Registration number: ChiCTR1900028513; https://www.chictr.org.cn/↗). From February 2011 to November 2022, a total of 1810 patients were assessed for eligibility, with 1567 enroled in the cross‐sectional study. We conducted a longitudinal study on weight loss outcomes among 1034 patients who underwent bariatric surgery and had complete 1‐year follow‐up data, and further investigated diabetes remission in patients diagnosed with T2DM within this cohort. A detailed flowchart is presented in Supporting Information S1: Figure S1. All patients met the guidelines for bariatric surgery [21]. Inclusion criteria were as follows: age, 18–67 years; BMI, 25.0–55.0 kg/m2; and fasting C‐peptide level, > 1.0 ng/mL. Exclusion criteria included history of thyroid disease, treatment with thyroid hormone or antithyroid drugs, preoperative thyroid assessment consistent with overt hypothyroidism (elevated thyroid‐stimulating hormone (TSH) with low free thyroxine (FT4)) or overt hyperthyroidism (low TSH with high FT4 and/or high FT3), missing preoperative TSH, FT4 or FT3 data, anaemia, latent autoimmune diabetes, psychiatric or eating disorders, a history of gastrointestinal or malignant diseases, and previous gastrointestinal surgery. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Shanghai Jiao Tong University School of Medicine Affiliated Sixth People's Hospital on 24 December 2019 (approval number 2019‐KY‐049(K)). Informed consent was obtained from each patient.
Clinical and Biochemical Measurements
The data collection process was previously described [22]. In brief, anthropometric characteristics, serum biochemical parameters, and medication history at baseline and postoperative follow‐up were collected by trained investigators. As part of the inpatient study protocol, routine abdominal magnetic resonance imaging (MRI) was conducted preoperatively and during postoperative follow‐up. Patients were asked to complete an inpatient or outpatient evaluation 6 months after surgery and annually thereafter. Patients lost to follow up due to various factors (e.g., work commitments, geographical distance or changes in contact information) were contacted multiple times by telephone or mail to minimise attrition. Additionally, FT3, FT4, and TSH were measured by electrochemiluminescence immunoassays on the Cobas e601 analyser (Roche Diagnostics GmbH, Mannheim, Germany), with intra‐assay coefficients of variation < 7.0%, < 5.0%, and < 3.0% for FT3, FT4, and TSH, respectively, and inter‐assay coefficients of variation < 8.0%, < 7.0%, and < 8.0%, respectively.
Measurement of Body Composition
Body fat percentage (BFP), fat‐free mass index (FFMI), and fat mass index (FMI) were measured using a gold‐standard‐derived predictive equation. Abdominal MRI scans were performed at the third lumbar vertebra, and SliceOmatic software (version 5.0; TomoVision, Magog, Canada) was used to measure psoas muscle area (PMA), subcutaneous fat area (SFA), and visceral fat area (VFA). Detailed information on these methods and their clinical utility has been described in our recent studies [22, 23].
Definitions and Calculations of Variables
Euthyroid was defined as FT4 (11.0–25.0 pmol/L), FT3 (3.1–6.8 pmol/L), and TSH (0.27–4.20 μIU/mL) within the reference ranges. Severe obesity has been shown to be associated with increased serum TSH and FT3 levels in euthyroid adults [24]. Accordingly, in this study, patients with TSH levels above the reference range but with FT4 and FT3 not below the lower limit of normal, or with FT3 levels above the reference range but with TSH not below the lower limit of normal, were classified as having normal thyroid function. Central indices of thyroid hormone sensitivity were calculated using the following formulas: Thyrotropin index (TSHI) = ln(TSH (μIU/mL)) + 0.1345 × FT4 (pmol/L). Thyrotrophic thyroxine resistance index (TT4RI) = FT4 (pmol/L) × TSH (μIU/mL). For TSHI and TT4RI, higher values indicated lower central sensitivity to thyroid hormones. The peripheral index of thyroid hormone sensitivity was calculated as FT3/FT4 ratio = FT3 (pmol/L)/FT4 (pmol/L). Higher values of this ratio indicated higher peripheral sensitivity to thyroid hormones [25]. Percent total weight loss (%TWL) for this study was calculated as follows: 100 × (baseline weight − follow‐up weight)/baseline weight. Successful WL was defined as a TWL equal to or higher than 25%. DR was defined as a fasting glucose < 5.6 mmol/L (100 mg/dL) and haemoglobin A1c (HbA1c) level < 6.0% (42 mmol/mol) without antidiabetic medications, whereas long‐term DR was defined as remission lasting for at least 3 years after surgery. The primary outcomes were WL and DR during the 3 year follow‐up period after bariatric surgery. Diabetes, hypertension, and dyslipidemia were defined as in previous studies [26].
Statistical Analysis
Data normality was assessed using the Kolmogorov–Smirnov test. Normally distributed data, skewed data, and categorical variables were presented as mean (standard deviation), medians (interquartile range), and percentages, respectively. Group differences were analysed using the Chi‐squared or Fisher's exact test for categorical variables, and independent‐sample t‐test and Mann–Whitney U test for continuous factors. The paired t‐test and Wilcoxon signed‐rank test were conducted for follow‐up comparisons.
The correlation between thyroid parameters and clinical indicators was calculated using Spearman's rank coefficient and visualised by heatmap in Python (version 3.6.3) with Matplotlib (version 2.1.1). Breakpoint analysis was conducted using the ‘segmented, ggplot2, Hmisc, and rms’ packages in RStudio. The inflection point is one that minimises the sum residual square error of the two linear segments, above and below this point. It should be noted that this breakpoint does not represent a definitive biological threshold but rather a statistical representation of the data, aiding in understanding the segmented relationship between the variables. Participants in the longitudinal study were stratified into sex‐specific tertiles based on their baseline FT3 levels, with the lowest tertile serving as the reference group. Binary logistic regression models were used to estimate odds ratio (OR) and 95% confidence interval (CI) for successful WL and DR, while linear regression models were used to examine the independent association between baseline FT3 tertiles and the magnitude of WL following bariatric surgery. To minimise potential residual confounding, covariates in multivariable‐adjusted models with progressive adjustment were selected based on their well‐established or probable association with WL or DR after bariatric surgery. The Cox proportional hazards model was implemented using the ‘rms’ package in RStudio to analyse time‐to‐event hazard ratios (HR), incorporating restricted cubic splines (RCS) to examine and visualise the dose‐response relationship between FT3 (as a continuous variable) and successful WL and DR. The follow‐up period was precise to the day, with the primary analysis focussing on event outcomes within 1 year after bariatric surgery. Data censoring occurred at the study's conclusion. We validated the proportional hazard assumption using the Schoenfeld residual test. Kaplan–Meier curves and log‐rank tests were evaluated using the ‘survival’ package in RStudio. The cut points of baseline FT3 identified in the RCS analysis at a HR of 1.0 were stratified by sex to avoid the influence of significant sex‐related differences in serum FT3 levels.
A p < 0.05 (two‐tailed tests) was considered statistically significant. Because missing data were less than 5%, listwise deletion was used. All analyses were performed using SPSS version 25.0 (SPSS Inc., Chicago, IL, USA) and RStudio software v.1.3.959 (https://www.r‐project.org/↗; R, Vienna, Austria).
Results
Clinical Characteristics of the Participants in the Cross‐Sectional Study
A total of 1567 patients were enroled (Supporting Information S1: Table S1). Males did not differ significantly in age compared with females but exhibited higher BMI, waistline, blood glucose levels, and triglycerides. Body composition showed significant sex differences, with males having significantly higher PMA, VFA and FFMI (p < 0.001), while females showed greater SFA, BFP, and FMI (p ≤ 0.001). Significant differences were also observed in serum thyroid hormone profiles between sexes, with FT3 exhibiting the greatest difference, followed by TSH and FT4. Furthermore, males had significantly higher FT3/FT4 ratios but lower TSHI and TT4RI compared to females (p < 0.001).
Correlations Between Thyroid Parameters and Obesity‐Related Indices
Figure 1 showed that serum FT3 exhibited stronger correlations with obesity and metabolic indicators regardless of sex compared with other thyroid parameters. Among the variables, age and FFMI showed the strongest correlations with FT3. Furthermore, we employed segmented regression analysis to demonstrate the best‐fit relationships. Surprisingly, a nonlinear relationship existed between FT3 and age, while FT3 showed a linear correlation with FFMI. FT3 decreased until reaching the breakpoint (FT3 = 5.001 for males and 4.773 for females), after which it flattened (p < 0.01). The comparison of linear regression and segmented linear regression for the best‐fit relationship is presented in Supporting Information S1: Table S2.
Relationship between thyroid parameters and obesity‐related indices in the cross‐sectional study population (1039 females and 528 males). (A) Spearman correlations between thyroid parameters and clinical features. The heat map visualisation shows the Spearman correlation coefficients for clinical characteristics. The colour bar indicates the Spearman'svalue for each pair of variables, with darker colours representing stronger correlations. The analysis indicated that the correlations between serum FT3 and age, as well as between FT3 and FFMI, were the strongest. (B–C) The best‐fit relationships of serum FT3 with age and FFMI. (B) Serum FT3 and age. (C) Serum FT3 in relation to FFMI. Each circle represents a single participant in the study. The breakpoints, showing a sharp change in slope, are indicated by a dashed line in the corresponding colour. Segmental linear regression is applied if the correlation is significantly better than that of linear regression (< 0.05). Pearson correlation coefficients and associated‐values are shown for the female and male populations in the regression model. Breakpoints of serum FT3 were 4.773 pmol/L (females) and 5.001 pmol/L (males) for age. r p p
Baseline and Postoperative Characteristics of Bariatric Surgery Cohort
In all, 1034 patients from the cross‐sectional population were included in the longitudinal study (Table 1). Significant improvements in serum lipid profiles, glycaemic profiles, and metabolic diseases were observed after weight loss (all p < 0.001). Both FT3 and TSH levels decreased significantly after surgery regardless of sex (p < 0.001), with FT3 showing the greatest reduction. FT4 changes differed by sex, with a significant decrease in females (p < 0.001) but no change in males (p = 0.286). Additionally, the FT3/FT4 ratio, TSHI, and TT4RI all decreased significantly in both sexes after bariatric surgery (p < 0.001).
| Variable | Total (= 1034)n | Female (= 731)n | Male (= 303)n | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Baseline | 1 year | valuep | Baseline | 1 year | valuet/Z | valuep | Baseline | 1 year | valuet/Z | valuep | |
| Age, years | 36.4 ± 11.7 | 36.1 ± 11.5 | 37.1 ± 12.3 | ||||||||
| Sleeve/RYGB,n | 813/221 | 609/122 | 204/99 | ||||||||
| BMI, kg/m2 | 36.3 ± 6.0 | 26.1 ± 4.2 | < 0.001 | 36.1 ± 5.8 | 25.8 ± 4.1 | 76.9 | < 0.001 | 36.8 ± 6.6 | 26.9 ± 4.3*** | 41.2 | < 0.001 |
| Waistline, cm | 112.8 ± 13.9 | 90.1 ± 11.8 | < 0.001 | 110.5 ± 12.6 | 88.5 ± 11.2 | 56.5 | < 0.001 | 117.5 ± 15.2*** | 93.6 ± 12.3*** | 37.1 | < 0.001 |
| Thyroid parameters | |||||||||||
| FT3, pmol/L | 5.0 ± 0.72 | 4.2 ± 0.64 | < 0.001 | 4.9 ± 0.70 | 4.1 ± 0.56 | 24.5 | < 0.001 | 5.2 ± 0.72*** | 4.6 ± 0.65*** | 12.1 | < 0.001 |
| FT4, pmol/L | 16.4 ± 2.2 | 16.1 ± 2.2 | < 0.001 | 16.3 ± 2.2 | 15.9 ± 2.1 | 4.2 | < 0.001 | 16.8 ± 2.3*** | 16.6 ± 2.3*** | 1.1 | 0.286 |
| TSH, μIU/mL | 2.4 (1.7, 3.5) | 1.9 (1.2, 2.7) | < 0.001 | 2.6 (1.8, 3.7) | 2.0 (1.3, 2.8) | −10.9 | < 0.001 | 2.1(1.5, 2.9)*** | 1.6(1.1, 2.3)*** | −6.9 | < 0.001 |
| FT3/FT4 ratios | 0.31 ± 0.05 | 0.27 ± 0.04 | < 0.001 | 0.30 ± 0.05 | 0.26 ± 0.04 | 18.9 | < 0.001 | 0.31 ± 0.06* | 0.28 ± 0.04*** | 9.2 | < 0.001 |
| TSHI | 3.1 ± 0.70 | 2.7 ± 0.69 | < 0.001 | 3.1 ± 0.71 | 2.7 ± 0.70 | 10.6 | < 0.001 | 3.0 ± 0.66** | 2.7 ± 0.65 | 6.9 | < 0.001 |
| TT4RI | 39.0 (27.1, 56.4) | 29.6 (18.7, 43.2) | < 0.001 | 42.8 (29.6, 59.1) | 31.0 (19.0, 45.3) | −11.3 | < 0.001 | 33.6 (23.6, 48.1)*** | 27.0 (17.7, 39.2)** | −7.0 | < 0.001 |
| Lipid profiles | |||||||||||
| TC, mmol/L | 5.1 ± 1.2 | 4.7 ± 1.1 | < 0.001 | 5.2 ± 1.2 | 4.8 ± 1.1 | 6.9 | < 0.001 | 5.0 ± 1.2 | 4.5 ± 1.2*** | 6.6 | < 0.001 |
| TG, mmol/L | 1.7 (1.3, 2.5) | 0.89 (0.68, 1.2) | < 0.001 | 1.6 (1.2, 2.2) | 0.87 (0.66, 1.1) | −20.0 | < 0.001 | 2.2 (1.4, 3.2)*** | 0.97(0.71, 1.4)*** | −14.1 | < 0.001 |
| HDL‐c, mmol/L | 1.1 ± 0.27 | 1.3 ± 0.33 | < 0.001 | 1.1 ± 0.26 | 1.4 ± 0.33 | −20.9 | < 0.001 | 0.97 ± 0.26*** | 1.2 ± 0.28*** | −11.6 | < 0.001 |
| LDL‐c, mmol/L | 3.1 ± 0.84 | 2.8 ± 0.89 | < 0.001 | 3.2 ± 0.83 | 2.8 ± 0.84 | 10.8 | < 0.001 | 2.9 ± 0.85*** | 2.7 ± 0.99 | 3.5 | 0.001 |
| Glycaemic profiles | |||||||||||
| FPG, mmol/L | 7.2 ± 2.8 | 5.0 ± 1.1 | < 0.001 | 7.0 ± 2.8 | 4.9 ± 1.0 | 20 | < 0.001 | 7.6 ± 2.8** | 5.2 ± 1.2*** | 15.6 | < 0.001 |
| 2‐h PBG, mmol/L | 10.6 ± 4.4 | 5.9 ± 2.5 | < 0.001 | 10.1 ± 4.2 | 5.6 ± 2.1 | 24.7 | < 0.001 | 11.8 ± 4.7*** | 6.6 ± 3.0*** | 16.5 | < 0.001 |
| FCP, ng/mL | 3.9 ± 1.8 | 2.1 ± 0.74 | < 0.001 | 3.8 ± 1.7 | 2.0 ± 0.67 | 27.6 | < 0.001 | 4.1 ± 1.9* | 2.3 ± 0.84*** | 17.4 | < 0.001 |
| HbA1c, % | 7.1 ± 1.9 | 5.6 ± 0.78 | < 0.001 | 6.9 ± 1.8 | 5.5 ± 0.73 | 21.1 | < 0.001 | 7.5 ± 2.0*** | 5.7 ± 0.87* | 17 | < 0.001 |
| HOMA‐IR | 7.3 (4.6, 11.8) | 1.4 (0.93, 2.2) | < 0.001 | 6.9 (4.3, 11.6) | 1.4 (0.90, 2.1) | −21.1 | < 0.001 | 8.4(5.4, 12.1)*** | 1.7 (1.0, 2.5)*** | −14.3 | < 0.001 |
| HOMA‐B | 201.2 (92.0, 357.9) | 107.5 (63.9, 172.1) | < 0.001 | 205.5 (104.0.358.5) | 108.3 (67.1, 170.9) | −13.5 | < 0.001 | 181.1 (73.6, 351.8) | 106.3 (56.9, 175.7) | −7.8 | < 0.001 |
| Body composition | |||||||||||
| BFP, % | 43.4 ± 8.3 | 31.4 ± 6.8 | < 0.001 | 46.2 ± 6.9 | 34.1 ± 5.5 | 75.4 | < 0.001 | 36.4 ± 7.2*** | 24.9 ± 5.1*** | 40.4 | < 0.001 |
| FFMI, kg/m2 | 20.2 ± 2.3 | 17.8 ± 2.2 | < 0.001 | 19.0 ± 1.2 | 16.8 ± 1.5 | 51.4 | < 0.001 | 22.9 ± 1.8*** | 20.1 ± 2.0*** | 40.8 | < 0.001 |
| FMI, kg/m2 | 16.1 ± 5.5 | 8.4 ± 2.9 | < 0.001 | 17.1 ± 5.3 | 9.0 ± 2.9 | 60.3 | < 0.001 | 13.8 ± 5.3*** | 6.9 ± 2.5*** | 31.7 | < 0.001 |
| No. of patients 70228 | 369 | 238 | 131 | ||||||||
| PMA, cm2 | 26.2 ± 8.4 | 22.0 ± 7.4 | < 0.001 | 21.4 ± 4.1 | 17.6 ± 3.6 | 27.2 | < 0.001 | 34.9 ± 6.9*** | 29.9 ± 5.8*** | 17.5 | < 0.001 |
| VFA, cm2 | 167.7 ± 59.0 | 61.5 ± 32.9 | < 0.001 | 159.7 ± 50.8 | 60.2 ± 28.3 | 36.5 | < 0.001 | 182.2 ± 69.4*** | 63.9 ± 39.9 | 25.1 | < 0.001 |
| SFA, cm2 | 368.6 ± 145.6 | 215.7 ± 102.3 | < 0.001 | 391.1 ± 139.6 | 231.9 ± 97.8 | 28 | < 0.001 | 327.9 ± 148.0*** | 186.3 ± 104.1*** | 18.5 | < 0.001 |
| Diabetes,(%)n | 557 (53.9) | 113 (10.9) | < 0.001 | 349 (47.7) | 61 (8.3) | < 0.001 | 208 (68.6) | 52 (17.2) | < 0.001 | ||
| OHA,(%)n | 341 (33.0) | 35 (3.4) | < 0.001 | 203 (27.8) | 25 (3.4) | < 0.001 | 138 (45.5) | 10 (3.3) | < 0.001 | ||
| Insulin therapy,(%)n | 156 (15.1) | 8 (0.8) | < 0.001 | 85 (11.6) | 2 (0.3) | < 0.001 | 71 (23.4) | 6 (2.0) | < 0.001 | ||
| Hypertension,(%)n | 563 (54.4) | 188 (18.2) | < 0.001 | 349 (47.7) | 109 (14.9) | < 0.001 | 214 (70.6) | 79 (26.1) | < 0.001 | ||
| Dyslipidemia,(%)n | 711 (68.8) | 256 (24.8) | < 0.001 | 457 (62.5) | 150 (20.5) | < 0.001 | 254 (83.8) | 106 (35.0) | < 0.001 | ||
Association Between Baseline Serum FT3 and Weight Loss After Bariatric Surgery
As shown in Table 2, significant differences were observed in the percent and absolute BMI decreases 1 year after bariatric surgery according to baseline FT3 tertiles, regardless of sex (p for trend < 0.001). The highest FT3 tertile was significantly associated with greater weight reduction in both males and females in model 3. However, after additional adjustment for baseline glycaemic parameters, this association remained highly significant in females (p for trend < 0.01) but disappeared in males. The associations between baseline FT3 tertiles and successful WL are presented in Supporting Information S1: Table S3. Consistent with the results in Table 2, high baseline FT3 was significantly associated with successful WL 1 year after bariatric surgery (p < 0.001), and this association was more pronounced in females.
| Mean changes (SD) | Additional changes after bariatric surgery (B, 95% CI) | |||||
|---|---|---|---|---|---|---|
| Unadjusted model | Model 1 | Model 2 | Model 3 | Model 4 | ||
| Females (= 731)n | ||||||
| Percent BMI, % | ||||||
| Tertile 1 (< 4.58 pmol/L) | 25.6 ± 7.0 | Reference | Reference | Reference | Reference | Reference |
| Tertile 2 (4.58–5.09 pmol/L) | 28.7 ± 7.2 | 3.1 (1.8–4.4)*** | 3.2 (1.9–4.5)*** | 1.9 (0.77–3.1)** | 1.8 (0.64–3.0)** | 1.5 (0.29–2.6)* |
| Tertile 3 (> 5.09 pmol/L) | 30.0 ± 7.1 | 4.4 (3.1–5.6)*** | 4.5 (3.2–5.8)*** | 2.7 (1.4–3.9)*** | 2.5 (1.3–3.8)*** | 2.0 (0.74–3.3)** |
| for trendp | < 0.001 | < 0.001 | < 0.001 | < 0.001 | 0.002 | |
| Absolute BMI, kg/m2 | ||||||
| Tertile 1 (< 4.58 pmol/L) | 9.1 ± 3.4 | Reference | Reference | Reference | Reference | Reference |
| Tertile 2 (4.58–5.09 pmol/L) | 10.5 ± 3.7 | 1.4 (0.76–2.0)*** | 1.4 (0.77–2.0)*** | 0.73 (0.27–1.2)** | 0.71 (0.25–1.2)** | 0.59 (0.12–1.1)* |
| Tertile 3 (> 5.09 pmol/L) | 11.2 ± 3.5 | 2.1 (1.4–2.7)*** | 2.0 (1.4–2.7)*** | 0.98 (0.49–1.5)*** | 0.95 (0.47–1.4)*** | 0.75 (0.25–1.3)** |
| for trendp | < 0.001 | < 0.001 | < 0.001 | < 0.001 | 0.004 | |
| Males (= 303)n | ||||||
| Percent BMI, % | ||||||
| Tertile 1 (< 4.86 pmol/L) | 23.2 ± 7.5 | Reference | Reference | Reference | Reference | Reference |
| Tertile 2 (4.86–5.42 pmol/L) | 26.3 ± 7.8 | 3.1 (1.0–5.2)** | 3.3 (1.2–5.4)** | 1.9 (−0.03–3.8) | 1.9 (0.02–3.8)* | 1.3 (−0.53–3.2) |
| Tertile 3 (> 5.42 pmol/L) | 28.5 ± 7.5 | 5.2 (3.1–7.3)*** | 5.4 (3.3–7.5)*** | 2.4 (0.36–4.5)* | 2.3 (0.23–4.3)* | 1.4 (−0.69–3.4) |
| for trendp | < 0.001 | < 0.001 | 0.025 | 0.034 | 0.211 | |
| Absolute BMI, kg/m2 | ||||||
| Tertile 1 (< 4.86 pmol/L) | 8.4 ± 3.9 | Reference | Reference | Reference | Reference | Reference |
| Tertile 2 (4.86–5.42 pmol/L) | 9.9 ± 4.1 | 1.5 (0.39–2.6)** | 1.6 (0.53–2.8)** | 0.67 (−0.06–1.4) | 0.69 (−0.38–1.4) | 0.47 (−0.25–1.2) |
| Tertile 3 (> 5.42 pmol/L) | 11.2 ± 4.0 | 2.8 (1.6–3.9)*** | 2.9 (1.7–4.0)*** | 0.81 (0.02–1.6)* | 0.78 (−0.01–1.6) | 0.47 (−0.32–1.3) |
| for trendp | < 0.001 | < 0.001 | 0.049 | 0.059 | 0.264 | |
Preoperative Serum FT3 Is Associated With Diabetes Remission at 1 Year After Bariatric Surgery
Significantly higher FT3 levels were observed in the DR group at baseline and even 1 year after surgery compared with the non‐DR group (Supporting Information S1: Table S4). As shown in Table 3, the probability of DR increased with higher FT3 tertiles, regardless of sex (p for trend < 0.001). After adjusting for preoperative thyroid function and serum lipid profiles at baseline, the adjusted OR for females in the highest tertile was 7.42 (p < 0.001), while for males it was 2.96 (p = 0.008), compared with the lowest tertile. In model 5, each unit increase in baseline FT3 level was associated with a 1.81‐fold increase in the probability of DR in females (OR 2.81 [1.53–5.18], p = 0.001), whereas this association was not observed in males.
| Serum FT3 levels in females (= 349)n | for trendp | Per‐unit FT3 increase | |||
|---|---|---|---|---|---|
| Tertile 1 (< 4.36 pmol/L) | Tertile 2 (4.36–4.91 pmol/L) | Tertile 3 (> 4.91 pmol/L) | |||
| Median, pmol/L | 4.11 | 4.67 | 5.37 | ||
| No. of patients | 117 | 119 | 113 | ||
| No. of diabetes remission (%) | 55 (47.0) | 79 (66.4) | 97 (85.8) | ||
| Unadjusted | 1 | 2.23 (1.32–3.77)** | 6.83 (3.60–13.0)*** | < 0.001 | 3.88 (2.48–6.09)*** |
| Model 1 | 1 | 2.24 (1.32–3.81)** | 7.04 (3.61–13.7)*** | < 0.001 | 4.15 (2.58–6.67)*** |
| Model 2 | 1 | 2.45 (1.41–4.27)** | 7.42 (3.75–14.7)*** | < 0.001 | 4.38 (2.67–7.21)*** |
| Model 3 | 1 | 2.25 (1.26–4.02)** | 5.53 (2.71–11.3)*** | < 0.001 | 3.51 (2.09–5.90)*** |
| Model 4 | 1 | 2.16 (1.20–3.90)* | 5.94 (2.86–12.3)*** | < 0.001 | 3.63 (2.14–6.15)*** |
| Model 5 | 1 | 2.10 (1.06–4.18)* | 4.01 (1.74–9.24)** | 0.001 | 2.81 (1.53–5.18)** |
The Dose‐Response Relationship of Baseline Serum FT3 Levels With Successful Weight Loss and Diabetes Remission
As shown in Figure 2, due to the uneven distribution of sample sizes and increased uncertainty in model estimation at the margins, we observed that the confidence intervals of all RCS curves widened even with slight deviations near HR = 1. Furthermore, the p values for the tests of linearity hypotheses in all models were > 0.05, confirming a linear relationship of baseline FT3 levels with WL and DR. As baseline FT3 levels increased, the probability of achieving successful WL and DR significantly increased in the thyroid function‐adjusted model, regardless of sex (p‐overall < 0.05). However, these dose‐response relationships disappeared in males after further adjustments. The optimal FT3 thresholds were 4.75 for females and 5.05 for males in achieving successful WL, and 4.57 for females and 5.00 for males in DR. When preoperative FT3 levels were below these thresholds, the risk of failing to achieve ideal glycaemic control and weight loss significantly increased. We stratified the entire cohort according to the FT3 threshold identified in the RCS analysis, and constructed Kaplan‐Meier survival curves. Compared with patients with higher baseline FT3 levels, those with lower FT3 levels had significantly lower probabilities of both successful WL and DR during the 3 year follow‐up (log‐rank p < 0.001).
Analysis of the associations of baseline serum FT3 levels with successful weight loss and diabetes remission. (A–D) Restricted cubic splines for successful weight loss and diabetes remission 1 year after bariatric surgery among females and males. Four knots were set at the 5th, 35th, 65th and 95th percentiles. The red line and shading represent the HR and its 95% CI for females, while the blue line represents the same for males. A vertical dotted line indicates the threshold value of baseline serum FT3, while a horizontal dotted line represents an HR of 1.0. For Panels (A) and (B), the upper half of each panel shows the model adjusted for preoperative thyroid function (TSH and FT4), while the lower half is further adjusted for age, baseline BMI, type of surgery, and presence/absence of diabetes at baseline. For Panels (C) and (D), the upper half of each panel represents the model adjusted for preoperative thyroid function (TSH and FT4), and the lower half is further adjusted for age, baseline BMI, type of surgery, diabetes duration, insulin usage, fasting C‐peptide, HbA1c, HOMA‐IR at baseline. (E–F) Kaplan–Meier plots of long‐term successful weight loss and diabetes remission during the 3 year follow‐up. Follow‐up ranged from 0.5 to 3.0 years for both successful weight loss (mean [SD] follow‐up: 2.0 [0.9] years; median [IQR] follow‐up: 2.0 [1.0–3.0] years), and diabetes remission (mean [SD] follow‐up: 2.0 [0.9] years; median [IQR] follow‐up: 2.1 [1.1–3.0] years). The Kaplan–Meier plots were based on the sex‐specific threshold value of baseline serum FT3 identified in the RCS analysis. The number at risk indicated the number of patients who achieved successful weight loss or diabetes remission at each time point after bariatric surgery. F, female; M, male.
Discussion
In this study, we found that a nonlinear relationship existed between FT3 and age, whereas FT3 showed a linear correlation with FFMI. In cohort analysis, high baseline FT3 levels (rather than TSH and FT4) were significantly associated with successful WL and DR at 1 and 3 years after bariatric surgery, exhibiting a dose‐dependent relationship. Notably, the pattern of these associations was similar in females and males, but stronger in females. To our knowledge, these findings have not been reported in the literature.
Obesity was associated with elevated serum levels of TSH and FT3, but not with FT4 levels [24]. Furthermore, obesity was linked to increased peripheral sensitivity and decreased central sensitivity to thyroid hormones. FT3 is the active form of thyroid hormone, involved in stimulating REEs [10, 27]. Therefore, it can be inferred that changes in thyroid hormone sensitivity associated with obesity aim to elevate FT3 levels, thereby increasing energy expenditure and limiting weight gain. This helps to explain why FT3 shows the strongest associations with adiposity and metabolic parameters compared with other thyroid indicators. Previous studies have emphasised the role of age and fat‐free mass in determining levothyroxine requirements in adults with hypothyroidism [28, 29]. Similarly, FT3 showed the strongest correlation with age and FFMI in this study. In fact, serum FT3 was shown to negatively correlate with age [30]. However, our study first demonstrates that FT3 exhibits a breakpoint effect with age independent of sex. A recent study suggested that as TSH levels increase, FT3/FT4 ratios increase until age 40, but this increase does not occur in older age groups [31]. Therefore, the breakpoint effect of FT3 with age is likely related to the cliff‐like decline in TSH action or deiodinase activity with age. Given that human skeletal muscle is the major site of type II iodothyronine deiodinase activity and thermogenesis [32], the parallel increase between FT3 and FFMI is understandable. Furthermore, the underlying mechanisms of elevated TSH and FT3 levels in individuals with obesity remain unclear [33]. Unlike the hypothesis that adipose tissue expansion mediates increased TSH secretion [34], our study found that muscle mass, particularly FFMI, was significantly positively correlated with both FT3 and TSH levels. This suggests that the elevation of TSH and FT3 in individuals with obesity may be related to increased muscle mass.
Consistent with our findings, most studies have shown that bariatric surgery is associated with a decrease in TSH and FT3 levels [35]. However, this change is probably mediated by weight loss rather than an intrinsic effect of bariatric surgery [36]. Similar phenomena have been observed in patients with obesity after non‐surgical weight loss [27, 37]. On the other hand, the impact of bariatric surgery on FT4 levels is more controversial [34, 38, 39]. In our study, FT4 significantly decreased in females but showed no significant change in males after surgery. Given the increased thyroxine requirement, obesity could increase the risk of hypothyroidism [17]. Interestingly, the associations between obesity and hypothyroidism also show sex differences, with obesity significantly increasing the risk of hypothyroidism in females but not in males [40]. Taken together, our results suggest that females may have a more severe thyroid burden before surgery and experience greater improvements in thyroid function after bariatric surgery, highlighting significant sex differences in thyroid health and outcomes from bariatric surgery. Furthermore, consistent with existing literature [27, 41], the elevated peripheral and decreased central sensitivity to thyroid hormones in individuals with obesity normalised following bariatric surgery. FT3 plays a central role in the adaptive changes in thyroid hormone sensitivity, likely explaining why it undergoes the most significant changes after surgery compared to other thyroid indicators.
Obesity is associated with increased FT3 levels, which decrease after bariatric surgery [13, 35, 37]. High preoperative FT3 increases energy expenditure and limits weight gain, while low postoperative FT3 enhances energy conservation and promotes weight regain. This indicates the presence of a specific body weight set‐point, above which FT3 levels increase to promote weight loss, and below which FT3 levels decrease to restore weight. The ‘physiological’ FT3 signal helps maintain weight balance. From this perspective, the higher the pre‐operative FT3, the greater the potential to reach the body weight set‐point after surgery. Since improvements in total body insulin sensitivity are weight‐dependent and WL is a strong predictor of DR after bariatric surgery [15, 19], it follows that the association between baseline FT3 and DR is likely a result of its positive effect on weight reduction. Given that the difference in FT3 levels before surgery is more pronounced, baseline FT3 significantly impacts postoperative WL and DR.
Preoperative FT3 was associated with short‐ and long‐term WL and DR after bariatric surgery. Although previous studies have suggested that preoperative FT3 levels may be associated with the extent of excess weight loss after surgery [42], our data further demonstrate that sex has a significant influence on these associations. At lower serum FT3 levels in females, both WL and DR increased with rising FT3 levels. In contrast, the associations with WL and DR were weaker at higher serum FT3 levels in males, suggesting a threshold effect. As baseline FT3 increases, both WL and DR tend to reach a plateau. In the FT3 versus WL and DR plots, females predominantly occupy the linear part of the curve on the left, while males occupy the flat part on the right. Ultimately, the substantial difference in serum FT3 levels between males and females may be a critical factor explaining the sex differences in the association of baseline FT3 with WL and DR. Of note, our hypothesis aligns with a recent important study, which found a significant association between lean body mass and cardiovascular parameters in females but not in males [43]. Our findings further validate this sex‐difference pattern in the field of obesity and indicate that sex‐specific effects arising from differences in body composition and metabolic levels may be more widespread. Together with the FT3 inflection points identified in the cross‐sectional study, our exploratory analysis suggested FT3 levels ≤ 5.0 pmol/L for males and ≤ 4.57 pmol/L for females as potential thresholds associated with suboptimal outcomes. Because hypothyroidism is approximately tenfold more prevalent in females than in males [44], these cut‐points, which require external validation, may be particularly relevant to females. Furthermore, individuals with low preoperative FT3 levels may warrant intensified postoperative monitoring. However, given the observational nature of this study, whether preoperative thyroid function optimization (e.g., thyroid hormone supplementation) could improve postoperative outcomes remains speculative and requires evaluation in prospective interventional studies.
Our study had several limitations. First, this was an observational cohort study, and the associations identified cannot be interpreted as causal. Secondly, REEs were not directly measured. Because REE is central to the proposed mechanism linking FT3 to postoperative weight loss, this limitation precluded direct verification of the metabolic pathway linking preoperative FT3 to observed surgical outcomes. Finally, although the participants with thyroid disease were excluded, serum thyroid antibody levels were not routinely checked. Nonetheless, given the large sample size, this limitation is unlikely to affect the interpretation of our results.
In conclusion, serum FT3 shows a nonlinear age‐related breakpoint and a linear correlation with FFMI in euthyroid adults with obesity. Low preoperative FT3 at or below sex‐specific exploratory thresholds (≤ 5.0 pmol/L in males and ≤ 4.57 pmol/L in females) was associated with suboptimal short‐ and long‐term WL and DR after bariatric surgery, particularly in females. These thresholds may support preoperative risk stratification and inform individualised treatment selection as obesity therapies expand, including bariatric surgery and GLP‐1 receptor agonists. Future randomized controlled trials should confirm these thresholds and determine whether any strategy targeting patients with low preoperative FT3 can improve surgical outcomes.
Author Contributions
S.L., Y.T., Y.M., H.Y., and F.L. conceived and designed the study. Y.T. and S.L. contributed to data acquisition. S.L. and H.Y. performed the literature search and statistical analysis. S.L. wrote the manuscript. Y.B., J.Y., H.Y., and F.L. supervised the study and critically revised the manuscript for important intellectual content. H.Z. contributed to standardized surgical procedures. Y.X. contributed to the standardized abdominal MRI examination. All authors contributed to the article and approved the submitted version.
Funding
This work was supported by the National Natural Science Foundation of China (82601612, 82470881), the Noncommunicable Chronic Diseases‐National Science and Technology Major Project (2023ZD0509205), the Shanghai Municipal Health Commission of General Programme (202440117), and the Shanghai Municipal Education Commission—Gaofeng Clinical Medicine Grant Support (20172025).
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Acknowledgements
The authors express their gratitude to all clinicians, nurses, and technicians at the Shanghai Clinical Center for Diabetes for their contributions to this study, as well as to all participating patients.
Contributor Information
Yujin Ma, Email: mayujin126@126.com.
Haoyong Yu, Email: yuhaoyong@shsmu.edu.cn.
Fengjing Liu, Email: liufengjing@shsmu.edu.cn.
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.
References
Associated Data
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.