This study utilized demographic information, laboratory data, and dietary questionnaire data from the NHANES for the years 2009–2012. A total of 20,293 participants who had completed dietary interviews, medical examinations, and health questionnaires were initially considered. The following exclusion criteria were applied: (1) missing data for TPOAb or TGAb levels (n = 16,048); (2) missing data for dietary ingested nutrients (n = 253); (3) missing education level (n = 733); (4) missing marital status (n = 1); (5) missing data on the family income-to-poverty ratio (Federal Poverty Level, FPL) (n = 283); (6) missing data for urine iodine (n = 34); (7) missing data for hypertension status (n = 4); and (8) missing data for diabetes status (n = 205). After exclusions, 2,732 participants were included in the final analysis (Figure 1).

TPOAb and TGAb levels were obtained from NHANES laboratory data. TPOAb and TGAb levels were measured using the Access immunoassay system (Beckman Coulter), which employs a sequential two-step sandwich chemiluminescent assay. According to the manufacturer’s guidelines, participants were classified as Hashimoto’s thyroiditis-positive if TPOAb > 9 IU/mL or TGAb > 4 IU/mL.

Dietary nutrient intake was evaluated based on 24-h dietary recall data collected on the first day of the NHANES dietary interview. Nutrient variables included energy, total fat, total polyunsaturated fatty acids (PUFAs), protein, carbohydrates, total sugars, dietary fiber, vitamin C, vitamin D (D2 and D3), vitamin E (as alpha-tocopherol), magnesium, and zinc. PUFA subtypes included C18:2 n-6, C18:3 n-3, C18:4 n-3, C20:4 n-6, C20:5 n-3, C22:5 n-3, and C22:6 n-3.

The following variables were considered potential confounders and were included as covariates in the regression models: sex, age, race, education level, marital status, ratio of family income to poverty, smoking status, total cholesterol, direct HDL-cholesterol, selenium, urinary iodine and alcohol consumption status. Age was categorized as 20 ~ 30, 30 ~ 50, and > 50 years. Race was classified as Mexican American, other Hispanics, non-Hispanic White, non-Hispanic Black, and other (including multiracial). Education level was grouped into four categories: less than 11th grade (including 12th grade without diploma), high school graduate/general educational development (GED), some college or associate degree, and college graduate or above. Marital status was divided into married or living with partner, divorced or separated and never married. FPL was categorized as ≤ 130, 130% ~ 350%, and > 350%. Smoking status was divided into current, former, and never smokers. Alcohol consumption status was divided into heavy, light, and never drinkers.

The sample size calculation was based on the formula: n = (Z² × p × (1- p)) /e², where Z corresponds to the standard normal deviate at the desired confidence level, p is the estimated prevalence, and e denotes the margin of error (15). For a 95% confidence level, Z was set to 1.96. According +to previous studies, the prevalence of HT is about 7.5% (16). Substituting p = 0.12 and e = 0.05 into the formula yielded a minimum required sample size of 107 patients with HT. Healthy subjects were then matched to HT cases at a 1:2 ratio, resulting in at least 54 controls.

A total of 125 HT patients and 57 healthy controls were recruited from two hospitals: 51 HT patients and 22 controls from the Affiliated Hospital of Qingdao University, and 74 HT patients and 35 controls from Zhejiang Chinese Medical University. HT patients were eligible if they were adults aged 18–60 years who had been clinically confirmed by an endocrinologist based on typical ultrasonographic features together with positive thyroid autoantibodies. Individuals in the control group were selected from routine health-check programs at the same institutions and were frequency-matched to cases by age and sex. Only participants without any history of thyroid disorders, including functional abnormalities or nodular disease, were considered as controls. General exclusion criteria were applied to both groups, including pregnancy or breastfeeding, a history of alcohol dependence, concomitant autoimmune conditions (e.g., rheumatoid arthritis, multiple sclerosis), severe chronic illnesses such as malignancies or cardiovascular disease, and recent use of immunosuppressive medications or n-3 PUFA supplementation. Demographic characteristics and lifestyle information were collected through structured interviews by trained investigators. All participants signed written informed consent prior to enrollment. The study protocol was approved by the Ethics Committee of Qingdao University (approval number: QDU-20201107-1). All procedures were following institutional ethical guidelines and the Declaration of Helsinki.

Following a 12-h overnight fast, venous blood samples were collected for lipid analysis. Samples were centrifuged at 1500 × g for 15 min at 4 °C to separate plasma and red blood cells (RBCs), which were subsequently stored at −80 °C until analysis. Total lipids were extracted from 0.5 mL of RBCs using a methanol: chloroform solution (2:1, v/v). The phospholipid fraction was isolated by thin-layer chromatography and transmethylated using methanol containing sulfuric acid (2%, v/v) at 70 °C for 2 h. After cooling to room temperature, the resulting fatty acid methyl esters (FAMEs) were extracted with hexane, dried under nitrogen, and redissolved for analysis by gas chromatography (GC-2014 system, Shimadzu) equipped with a flame ionization detector and an AOC-20i auto-injector. A DB-23 capillary column (60 m × 0.25 mm internal diameter × 0.15 μm film thickness, Agilent Technologies, USA) was used for separation. Individual fatty acids were identified by comparing retention times with a commercial standard mixture containing 32 FAMEs. Results were expressed as the relative percentage of each fatty acid out of the total fatty acids (17). The n-3 index was calculated as the combined percentage of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA).

NHANES uses a complex, multistage probability sampling design. To account for this, we applied sampling weights (WTDRD1), primary sampling units (SDMVPSU), and strata (SDMVSTRA) as recommended in the NHANES analytic guidelines. Continuous variables were presented as weighted means and standard deviations (SDs), while categorical variables were expressed as weighted percentages (%).

Dietary nutrient intake levels were categorized into quartiles, with the lowest quartile (Q1) serving as the reference group. Weighted multivariable logistic regression models were used to determine the associations between dietary nutrient intake and the risk of HT. Three models were constructed: Model 1, unadjusted; Model 2, adjusted for sex, age, race, education level, marital status, and ratio of family income to poverty; Model 3, additionally adjusted for smoking status, total cholesterol, direct HDL-cholesterol, selenium, urinary iodine, and alcohol consumption status. Additionally, sensitivity analyses were performed by further adjusting for hypertension and diabetes to examine the robustness of the conclusions from Model 3. The nonlinear associations of total PUFAs intake and C18:3 n-3 intake with the risk of HT were assessed using restricted cubic splines (RCS).

In the case-control study, continuous variables are expressed as mean and SD, and categorical variables as percentages. Logistic regression models were used to determine the associations between erythrocyte phospholipid PUFA levels and HT status. Model 1, unadjusted; Model 2, adjusted for age and sex; Model 3, further adjusted for marital status, family history of HT, body mass index, educational level, total erythrocyte n-6 PUFA, and total saturated fatty acids. A two-tailed p-value less than 0.05 was considered statistically significant. All statistical analyses were performed using R software (version 4.3.2).

After applying inclusion and exclusion criteria, a total of 2,732 participants were included in this study. Among them, 1,333 (52%) were female, 1,255 (42%) were over 50 years of age, 1,262 (70%) were non-Hispanic White, 643 (30%) had attained a college degree or higher, 1,608 (62%) were married or living with partner, 835 (42%) had a higher income level, 963 (31%) had hypertension, 327 (9.4%) had diabetes, 1,486 (55%) were never smokers, and 2,033 (80%) were heavy drinkers. Statistically significant differences were observed between the HT-positive and HT-negative groups in sex, age, race and total cholesterol levels (p < 0.05; Table 1).

Univariate analysis showed that females had a significantly higher risk of HT compared to males (OR = 1.862, 95% CI: 1.368–2.535, p < 0.001). With individuals aged 20–30 years as the reference, those over 50 years (OR = 2.323, 95% CI: 1.437–3.755, p = 0.001) exhibited significantly increased HT risk (Table 2).

Compared to the HT-negative group, energy (1,889.667 ± 753.040 kcal vs. 2,212.200 ± 1,051.583 g, p < 0.001) and total fat (67.181 ± 36.684 g vs. 82.990 ± 47.093 g, p < 0.001) were significantly lower in the HT-positive group. Total PUFA intake was also significantly lower in the HT-positive group (16.007 ± 10.276 g vs. 19.248 ± 12.686 g, p = 0.002). Specifically, intakes of linoleic acid (C18:2 n-6; 14.195 ± 9.444 g vs. 17.065 ± 11.504 g, p = 0.002), α-linolenic acid (C18:3 n-3; 1.417 ± 0.942 g vs. 1.701 ± 1.192 g, p = 0.009), arachidonic acid (C20:4 n-6; 0.117 ± 0.112 g vs. 0.153 ± 0.146 g, p = 0.004), and docosapentaenoic acid (DPA, C22:5 n-3; 0.018 ± 0.030 g vs. 0.025 ± 0.051 g, p < 0.001) were all significantly lower among HT patients. Protein (74.376 ± 34.562 g vs. 83.987 ± 42.939 g, p = 0.004), carbohydrate (232.981 ± 97.791 g vs. 264.785 ± 131.062 g, p = 0.003), and total sugars intake (98.785 ± 60.853 g vs. 118.017 ± 80.378 g, p = 0.002) were also significantly lower in the HT-positive group. Eicosapentaenoic acid (C20:5 n-3; 0.038 ± 0.130 g vs. 0.030 ± 0.094 g, p = 0.014) and vitamins C (89.934 ± 80.980 g vs. 81.647 ± 81.647 g, p = 0.042) were significantly higher in the HT-positive group. No significant differences were observed for C18:4 n-3, C22:6 n-3, dietary fiber, vitamins D, E, magnesium, or zinc (all p > 0.05; Table 3).

Associations between dietary nutrient intake and the risk of HT are shown in Table 4. In the unadjusted model (Model 1), higher intake of energy, total fat, total PUFAs, C18:2 n-6, C18:3 n-3, C20:4 n-6, protein, carbohydrates, and total sugars were all associated with decreased risk of HT. However, after adjustment for covariates in Models 2 and 3, only energy, total fat, total PUFAs and C18:3 n-3 remained significantly associated with lower HT risk (p < 0.05).

In the fully adjusted model (Model 3), each 1 g increase in dietary total PUFAs intake was associated with a 2.2% reduction in HT risk (OR = 0.978, 95% CI: 0.957–1.000, p = 0.049), and each 1 g increase in dietary C18:3 n-3 intake was associated with a 21.4% reduction (OR = 0.786, 95% CI: 0.642–0.963, p = 0.025). Quartile analyses showed that participants in the fourth quartiles of C18:3 n-3 intake had significantly lower HT risk compared to the lowest quartile (Q4: OR = 0.507, 95% CI: 0.265–0.969, p = 0.042). Furthermore, the sensitivity analyses incorporating additional adjustments for hypertension and diabetes yielded consistent results (Supplementary Table 1). Nevertheless, no significant curvilinear association were observed between total PUFAs intake (p = 0.58 for non-linearity) or C18:3 n-3 intake (p = 0.57 for non-linearity) and HT risk (Figure 2).

To evaluate the consistency of associations across population subgroups, subgroup analyses were performed based on age, sex, and race. Dietary total PUFAs intake was significantly and inversely associated with HT risk in males (Model 3: OR = 0.998, 95% CI: 0.996–0.999, p = 0.041), individuals aged > 50 years (Model 3: OR = 0.997, 95% CI 0.994–0.999, p = 0.037) and non-Hispanic White individuals (Model 3: OR = 0.997, 95% CI 0.995–0.999, p = 0.022) (Figure 3A). The inverse association between C18:3 n-3 intake and HT remained significant in several strata. Specifically, C18:3 n-3 intake was inversely associated with HT in males (Model 3: OR = 0.980, 95% CI: 0.963–0.997, p = 0.028), females (Model 3: OR = 0.962, 95% CI: 0.927–0.999, p = 0.047), individuals aged 30–50 years (Model 3: OR = 0.962, 95% CI 0.929–0.997, p = 0.034) and non-Hispanic White individuals (Model 3: OR = 0.965, 95% CI 0.942–0.989, p = 0.008) (Figure 3B).

To explore whether differences in HT risk across demographic groups could be attributed to variation in dietary C18:3 n-3 intake, we compared the intake of C18:3 n-3 across sex, age and race (Table 5). The results revealed that males had significantly higher C18:3 intake than females (Kruskal-Wallis test: p < 0.001; logistic regression: OR = 0.748, 95% CI: 0.692–0.809, p < 0.001). Given that females exhibited higher HT risk in univariate analysis, this finding further supported that insufficient intake of C18:3 n-3 may contribute to HT development.

Given that the NHANES-based analysis revealed a significant inverse association between dietary polyunsaturated fatty acids (particularly C18:3 n-3) and the risk of HT, a case-control study was conducted to validate these findings. The study included 125 HT patients and 57 healthy controls. No significant differences were observed in sex, age, or BMI between groups (p > 0.05), supporting the comparability of the matched design (Table 6). However, compared to the healthy control group, the HT group had a significantly higher proportion of participants with less than a college education, those who were married, and those with a family history of HT (p < 0.05).

As shown in Table 7, erythrocyte levels of total PUFAs (35.075 ± 6.760% vs. 39.867 ± 9.905%, p = 0.041), total n-3 PUFAs (3.277 ± 1.720% vs. 5.516 ± 6.613%, p = 0.038), EPA (C20:5 n-3; 0.635 ± 0.167% vs. 0.891 ± 0.289%, p = 0.048), DHA (C22:6 n-3; 1.428 ± 1.190% vs. 2.272 ± 0.974%, p = 0.006), and the n-3 index (2.116 ± 0.913% vs. 3.123 ± 0.821%, p = 0.033) were significantly lower in HT patients than in healthy controls, suggesting a possible link between systemic n-3 PUFA deficiency and HT pathogenesis.

Adjusted associations between erythrocyte phospholipid PUFA levels and HT risk are presented in Table 8. In the unadjusted model (Model 1), higher levels of total n-3 PUFAs, C18:3 n-3, EPA, DHA and the n-3 index were significantly associated with reduced HT risk (p < 0.05). In both Model 2 and Model 3, the inverse associations of total n-3 PUFAs (Model 2: OR = 0.571, 95% CI: 0.322–0.987, p = 0.039; Model 3: OR = 0.616, 95% CI: 0.443–0.992, p = 0.041), C18:3 n-3 (Model 2: OR = 0.597, 95% CI: 0.465–0.823, p = 0.034; Model 3: OR = 0.627, 95% CI: 0.558–0.874, p = 0.037), EPA (C20:5 n-3: Model 2: OR = 0.651, 95% CI: 0.412–0.964, p = 0.011; Model 3: OR = 0.736, 95% CI: 0.581–0.975, p = 0.042), and the n-3 index (Model 2: OR = 0.674, 95% CI: 0.525–0.973, p = 0.028; Model 3: OR = 0.790, 95% CI: 0.531–0.984, p = 0.046) with HT risk remained statistically significant.