Preoperative serum lipids as novel predictors for concomitant thyroid carcinoma in Graves’ disease

Patients

We collected the retrospective data of 577 patients with GD who underwent thyroidectomy spanning from January 2015 to December 2024 at The First Affiliated Hospital, Zhejiang University School of Medicine (depicted in Fig. 1). Studies involving human participants were reviewed and approved by the Clinical Research Ethics Committee of the First Affiliated Hospital College of Medicine, Zhejiang University (Approval number: 2024-1193). The data are anonymous, and the requirement for informed consent was therefore waived. The diagnostic criteria for GD encompassed biochemical evidence of hyperthyroidism, the presence of serum anti-thyrotropin receptor antibodies, increased diffuse uptake of ¹²³I or ⁹⁹mTc-pertechnetate on radionuclide scans, and/or the manifestation of clinical features such as goiter or Graves’ ophthalmopathy, as outlined in Ross et al. (2016). Exclusion criteria for this retrospective study encompassed: (1) patients diagnosed with anaplastic or medullary thyroid cancer; (2) those lacking comprehensive clinical data; (3) individuals with a prior history of thyroid surgical intervention; (4) patients without complete laboratory testing for serum lipid and thyroid hormone levels; and (5) patients harboring malignancies in other organ systems. The study protocol was granted approval by the Ethics Review Board of the First Affiliated Hospital, Zhejiang University School of Medicine.

Patients’ clinical information, pathological characteristics, and laboratory results were extracted from their medical records. The clinical information comprised patients’ age, gender, duration of GD, BMI, comorbidities like hypertension and diabetes mellitus, and ultrasound features of thyroid nodules, including aspect ratio (anteroposterior to transverse diameter), margin definition, shape, echogenicity, and calcification presence. Pathological data encompassed tumor size, multifocality, bilaterality, extrathyroidal extension (ETE), lymph node metastases (LNM), and staging according to the 8th edition of the International Union Against Cancer (UICC)/American Joint Committee on Cancer (AJCC) tumor node metastasis (TNM) classification. Laboratory data encompassed preoperative serum lipid levels (triglycerides (TAG), total cholesterol (TC), high density lipoprotein (HDL), low density lipoprotein (LDL), very low density lipoprotein (VLDL)) and thyroid function indicators such as thyrotropin, antithyroglobulin antibodies (TgAb), thyroid peroxidase antibody (TPOAb), thyroglobulin (Tg), and thyroid-stimulating hormone receptor antibody (TRAb). Overweight status was defined as a BMI of 25 kg/m² or greater.

Statistical analysis

Continuous variables were summarized as mean value with standard deviations (SDs), whereas categorical variables were presented as counts and percentages. Categorical variables were compared using the chi-square test. Receiver operating characteristic (ROC) curves were constructed, and optimal cutoff values for serum TAG, TC, HDL, LDL, and VLDL were determined using the Youden index. The cut-off value for TAG, TC, HDL, LDL, and VLDL was 1.185, 3.805, 1.325, 2.305, 0.515 mmol/L, respectively. Logistic regression analysis was employed to identify risk factors for thyroid carcinoma in GD patients. A nomogram was developed based on the logistic regression model to predict DTC in GD patients. Calibration plots were used to compare the actual diagnoses with the predicted probabilities. ROC curves were plotted, and area under the curve (AUC) values were computed to assess the predictive performance of the nomogram model. All statistical analyses and interaction effect evaluations in this study were conducted rigorously in accordance with statistical methodologies. Statistical computations were executed using SPSS version 26.0 (IBM Corp., Armonk, NY, USA). The nomogram was created in R software (R Foundation, version 4.0.3) utilizing the “rms” package.

Baseline clinicopathological characteristics of patients with Graves’ disease

The baseline characteristics of patients enrolled in the study were summarized in Table 1. Among the 512 patients with GD, 299 patients were pathologically confirmed as DTC (58.4%). The average age was 45.2, most (83.6%) of patients were women. The mean disease duration was 55.3 ± 79.1 months. Most patients (90.6%) underwent total or near-total thyroidectomy. The main cause of patients undergoing surgery was preoperative diagnosis of cancer or thyroid nodule (47.1%). The mean BMI was 23.01 ± 3.39. The average level of TAG, TC, HDL, LDL, and VLDL were 1.37 ± 0.93, 4.57 ± 1.06, 1.37 ± 0.36, 2.53 ± 0.76, and 0.67 ± 0.49 mmol/L, respectively. Most patients were found to have thyroid nodules during ultrasound evaluation (91.8%).

Clinicopathological characteristics of GD patients with thyroid carcinoma

Table 2 presents the clinicopathological features of 299 individuals diagnosed with thyroid cancer. Among them, 296 patients (99.0%) were diagnosed with papillary thyroid carcinoma (PTC), whereas three patients (1.0%) were diagnosed with follicular thyroid carcinoma (FTC). Prior to surgery, 273 patients (91.3%) were diagnosed with thyroid cancer or nodules, while thyroid cancer was incidentally identified in 26 patients (8.7%). Regarding surgical procedures, 91.3% of patients underwent total or near-total thyroidectomy, and 8.7% underwent lobectomy. Fine needle aspiration biopsy of the thyroid was conducted on 213 patients (71.2%) preoperatively. The tumor size of 73.6% of the patients was less than 10 mm. Multifocal tumors were observed in 113 patients (37.8%), and bilaterality was noted in 77 patients (25.6%). A total of 127 patients (42.5%) exhibited extrathyroidal invasion, with 118 patients (39.5%) having minimal invasion and nine patients (3.0%) having gross invasion. Additionally, 275 patients (91.9%) underwent central lymph node dissection, and three patients (1.0%) underwent both central and lateral lymph node dissection. Lymph node metastasis was detected in 98 patients (32.8%). In terms of staging, 278 patients (93.0%) had stage I disease, 20 patients (7.4%) had stage II disease, and one patient (0.3%) had stage III disease, with a slight adjustment in percentage rounding for clarity.

Association of serum lipid levels and thyroid carcinoma in patients with Graves’ disease

To assess the connection between serum lipid concentrations and thyroid cancer among patients with Graves’ disease, serum lipid markers were categorized based on optimal cutoff points derived from ROC curves (see Fig. S1). Table 3 illustrates that the proportion of overweight individuals was markedly greater in the GD/DTC+ cohort compared to the GD/DTC- cohort (P < 0.001). Furthermore, both TAG and VLDL levels were significantly elevated in the GD/DTC+ group vs. the GD/DTC- group (P < 0.001). Conversely, HDL levels were notably higher in the GD/DTC- group (P < 0.001). Additionally, various ultrasonic features exhibited a significant correlation with DTC in GD patients, including nodules in thyroid (P < 0.001), aspect ratio imbalance (P < 0.001), hypoechogenicity (P < 0.001), irregular borders (P < 0.001), microcalcifications (P < 0.001) and irregular shape (P < 0.001).

Multivariate logistic regression analyses in GD patients combined with thyroid cancer

For the risk factors screened in Table 3, we selected those with P-values less than 0.01 for further univariate and multivariate regression analysis (Table 4). Univariate analysis revealed that, in patients with Graves’ disease, levels of TAG, HDL and VLDL were significantly associated with thyroid cancer (P < 0.001). The lower HDL levels and high TAG and VLDL levels presented high risks of thyroid cancer. In addition, overweight, nodules in thyroid, aspect ratio imbalance, hypoechogenicity, irregular borders, microcalcifications and irregular shape were associated with higher risk of thyroid cancer in GD patients (P < 0.001).

Next, we incorporated the risk factors with a P-value less than 0.01 from the univariate regression analysis into the multivariate regression analysis. The multivariate analyses showed that higher levels of TAG (OR 2.55, 95% CI [1.39–4.71], P = 0.003) and lower levels of HDL (OR 0.41, 95% CI [0.24–0.70], P = 0.001) were independent risk predictors of thyroid cancer. Overweight (OR 2.51, 95% CI [1.34–4.71], P = 0.004), nodules in thyroid (OR 15.57, 95% CI [2.05–118.29], P = 0.008), aspect ratio imbalance (OR 16.65, 95% CI [6.21–44.62], P < 0.001), hypoechogenicity (OR 1.78, 95% CI [1.05–3.00], P = 0.031), irregular borders (OR 5.11, 95% CI [2.85–9.16], P < 0.001) and microcalcifications (OR 6.10, 95% CI [1.57–23.74], P = 0.009) also indicated high risk of thyroid cancer in patients with Graves’ disease.

Construction of nomogram for prediction of thyroid cancer in patients with Graves’ disease

To develop a clinically practical predictive model, we stratified 512 patients into a training set (n = 358) and a validation set (n = 154), maintaining a 7:3 ratio. The baseline demographics of patients in both cohorts are presented in Table S1. Utilizing multivariate Cox proportional hazards regression analysis, we constructed a nomogram to assess the risk of thyroid cancer in individuals with Graves’ disease (Fig. 2A). This nomogram highlighted aspect ratio >1 as the most influential predictor of thyroid cancer risk in Graves’ disease patients. Additionally, TAG levels, HDL levels, overweight status, thyroid nodules, hypoechogenicity, irregular borders, and microcalcifications emerged as crucial factors. Calibration plots for both cohorts demonstrated good accuracy of the nomogram in predicting thyroid cancer among Graves’ disease patients (Fig. 2B). To further validate the model’s discriminatory power, ROC curves were generated. The area under the curve (AUC) for the predictive model was 0.91 (95% CI [0.88–0.94]) in the training cohort and 0.91 (95% CI [0.87–0.96]) in the validation cohort, respectively (Figs. 2C, 2D). In the training cohort, the sensitivity/specificity of the model was 0.826/0.857. In the validation cohort, the sensitivity/specificity of the model was 0.844/0.831. In addition, to further validate the robustness of the model, we performed a 5-fold cross-validation (Fig. S2). The results showed that our model also exhibited good predictive performance under 5-fold cross-validation (average AUC: 0.906).

The association between serum lipids and clinicopathological characteristics of GD patients with PTC

To investigate the link between serum lipid concentrations and clinicopathological features in patients with papillary thyroid cancer (PTC) coexisting with Graves’ disease (GD), we stratified the 299 patients with thyroid cancer into high and low groups according to the median values of BMI, TAG, and HDL. Notably, elevated TAG levels exhibited a significant correlation with lymph node metastasis (LNM) (P = 0.033) and the American Joint Committee on Cancer (AJCC) 8th stage (P = 0.015). However, no statistically significant differences were observed in tumor size, multifocality, bilaterality, or extrathyroidal extension (ETE) between the TAG groups (Table 5). Conversely, no discernible differences in clinicopathological characteristics were noted between the high and low BMI or HDL groups (Tables S2, S3). Subsequently, a comparative analysis was conducted to investigate the potential associations between different thyroid stimulating hormone (TSH)/Trab levels and lipid profiles. Statistical analyses revealed that patients with abnormal TSH levels (either hypo- or hyperthyrotropic) exhibited significantly elevated TAG levels compared to those within the normal TSH range (Table S4). However, no significant differences were observed in other lipid parameters. Regarding Trab levels, no significant variations in any lipid markers were detected between the low- and high-Trab groups. Furthermore, we performed a subgroup analysis comparing TSH and Trab levels between patients with and without thyroid cancer (Table S5). The results of this analysis demonstrated that no discernible differences in TSH or Trab level were noted between GD patients with or without thyroid cancer (Table S6).