Recruitment of participants in this cross-sectional study was conducted through convenience sampling and snowball sampling from September to November 2019, in the Oslo area in Norway. Participants were recruited through social media in closed Facebook groups and in online vegan and vegetarian forums. Details about the recruitment method and inclusion criteria have previously been described [23]. Study inclusion criteria were: (1) consumption of a vegan, lacto-ovo vegetarian or pescatarian diet for a minimum of 6 months; (2) age 18 years or older; (3) not currently pregnant or lactating; (4) no current use of thyroid medication.

We recruited 236 participants, and 29 people did not meet the inclusion criteria. In addition, two participants were excluded from the data analysis because of thyroid medication use and occasional meat consumption. Thus, the final sample consisted of 205 subjects, of which 115 vegans, 55 lacto-ovo vegetarians and 35 pescatarians. Participants who had provided study consent filled out an electronic questionnaire which consisted of background characteristics (age, anthropometric measures (height and weight), marital status, occupational status, educational level, smoking habits, country of birth, language, duration of adherence to vegan/lacto-ovo vegetarian/pescatarian diet) and foods included in the diet the previous 6 months (used to categorize participants into the different diet groups). The participants also conducted a 24-h dietary recall and a food frequency questionnaire (FFQ) for assessment of dietary intakes the previous 4 weeks (dietary iodine intakes are not presented in this paper, as these data have previously been published [23].

Iodine supplement use was assessed both by 24 h and habitual intake the previous 4 weeks (FFQ). The participants reported the name of the supplement, brand and amount used during the previous 24 h. By habitual intake, supplement consumption was reported as frequency per week, e.g., if a supplement contained 150 µg, and if taken four times a week, the contribution was estimated to be (50 µg × 4/7) 86 µg/day. Selenium-containing dietary supplements were also obtained from a FFQ in the electronic questionnaire in the same way as for iodine. For assessment of consumption of macroalgae, a dichotomous variable (no/yes) was used, if yes, participants reported type and amount (gram) used the previous 24 h and the habitual use the past 4 weeks. The types of macroalgae reported were Sugar kelp (Saccharina latissimi), Bladder wrack (Fucus vesiculosus), Wakame (Undaria pinnatifida), Kombu (Laminaria japonica and Saccharina japonica), Dulse (Palmaria palmata) and Laver (Porphyra spp.), with more details described in previous work [23].

All participants (n = 205) provided one non-fasting spot urine sample collected non-fasted at random times throughout the day, in a labeled 100 mL Vacuette® urine beaker (Greiner Bio-One, Kremsmünster, Austria). A subsample of urine was transferred into a 9.5 mL Vacuette® urine tube (Greiner Bio-One, Kremsmünster, Austria) and immediately put to storage at 2–4 °C before freezing at – 80 °C until analyses. The urine concentrations of iodine, selenium, and arsenic were measured at the Norwegian University of Life Science, Faculty of Environmental Science and Natural Resource Management. The analysis was performed using an alkaline sample preparation and subsequent quantitative determination using Inductively Coupled Plasma Mass Spectrometry ICP-MS [23]. An aliquot of 1 mL of urine was transferred into 15 mL pp centrifuge tubes (Sarstedt, Nümbrecht, Germany) by means of 100–5000 µL electronic pipette (Biohit, Helsinki, Finland) and diluted to 10 mL adding an alkaline solution (BENT), containing 4% (weight (w)/volume(V) 1-Butanol and 0.1% (w/V) H4EDTA, 2% (w/V) NH4OH, and 0.1% (w/V) TritonTM X-100. Method blank samples and samples of standard reference material (SRM) were prepared following the same procedures. Reagent of analytical grade or better and deionized water (18 MΩ) were used throughout. The samples were analyzed for iodine, selenium, and arsenic concentration using an Agilent 8800 ICP-QQQ (Triple Quadruple Inductively Coupled Plasma Mass Spectrometer, Agilent Technologies, Hachioji, Japan) using oxygen as a reaction gas. The concentration of iodine was determined by measuring 127I isotope. Simultaneously, the concentration of selenium in urine was determined on by measuring the mass shift from 78 to 94, while arsenic in urine was determined using a mass shift from 75 to 91. The limits of detection (LOD) and limits of quantification (LOQ) were calculated by multiplying the standard deviation of the method blank samples (n = 10) by three and ten, respectively. The obtained LOD and LOQ were 0.3 µg/L and 0.92 µg/L for iodine, 0.06 µg/L and 0.2 µg/L for selenium and 0.01 µg/L and 0.05 µg/L for arsenic, respectively. To check for method accuracy, two reference materials of urine (Sero AS, Billingstad, Norway) were analyzed; each with value assignment established in accordance with the Essential Requirements of the IVD Directive1 98/79/EC, and the ISO 17511 International standard [24]. The results were within the recommended values issued for the SeronormTM Trace Elements Urine L-1 and SeronormTM Trace Elements Urine L-2. The measurement repeatability was investigated in two different urine samples, one with visible precipitates and the other one completely transparent (each one with n = 5). The relative standard deviation (RSD) was 2.3% for iodine, 1.3% for selenium (up to 6.5% for samples with low concentrations), and < 1.0% for arsenic. UCC was measured at Fürst Medical Laboratory, Oslo, Norway.

Epidemiological criteria defined by WHO [5] for median UIC (not creatinine adjusted) was used to assess population iodine status; UIC < 20 µg/L severe iodine deficiency; UIC < 50 µg/L moderate iodine deficiency; UIC < 100 µg/L mild iodine deficiency; UIC in the range 100–199 µg/L adequate iodine intake; UIC in the range 200–299 µg/L more than adequate iodine intake; UIC > 300 µg/L excessive iodine intake. There is no established reference value for evaluating selenium or arsenic status based on urinary concentration. Thus, selenium and arsenic urinary concentrations were compared against measures of urinary selenium and arsenic in healthy populations [25, 26]. The urinary concentration of selenium range from 10 to 90 μg/L [26]. Regarding arsenic, in a European reference population (with no occupational exposure, no seafood consumption and drinking water concentration below 10 µg/L), mean concentrations of urinary inorganic arsenic and related metabolites are around 5–6 µg/L [27].

We adjusted UIC for creatinine for each participant by the following equation [28, 29]:

We collected a non-fasting venous blood sample and measured serum thyroglobulin (Tg), serum thyroid-stimulating hormone (S-TSH), free triiodothyronine (S-fT3), free thyroxine (S-fT4) and serum anti-TPO (S-anti-TPO). S-Tg was measured in duplicate using a sandwich serum Tg enzyme-linked immunosorbent assay (ELISA) [30]. Liquicheck Tumor Marker Control (Bio-Rad Laboratories AG, Cressier, Switzerland; LOT. 24000) was used as standard. We used laboratory-specific external quality control samples and the presented data complied with the defined criteria. The limit of detection for serum Tg assay was 2.3 µg/L [30]. For S-Tg concentrations below the limit of detection (LOD) (2.3 µg/L), we used a number generator and assigned a random value between 0.1 and 2.3 µg/L. S-TSH, S-fT4 and S-fT3 were analyzed at Fürst Medical Laboratory, Oslo, Norway using Advia Centaur XPT-instruments (Siemens Healthineers, City, Country).

The reference values from Fürst medical laboratory (Oslo, Norway) were used to evaluate S-TSH, S-fT3, S-fT4 and S-Anti-TPO levels. S-TSH reference range 0.20–4.0 mU/L (> 19 years of age); S-fT3 reference range 3.5–6.5 pmol/L; S-fT4 reference range 11.0–23.0 pmol/L and S-Anti-TPO a cutoff > 100 kU/L were used. We defined thyroid dysfunction as outlined in Supplemental Table 1. No reference values are available in adults for the used Tg assay.

IBM SPSS versions 25, 27 and 29 (IBM Corp., Armonk, NY, USA) were used for statistical analysis. Normality of the data was checked using visual evaluation of the Q–Q plots and histograms. Normally distributed data were presented as mean ± standard deviation (SD) and non-normally distributed data as median and the 25th and 75th percentiles (p25–p75) in tables. Cross-tabulation with Chi-square test was used to test differences between the dietary groups at nominal level; gender (male, female) and supplement use (yes/no). Kruskal–Wallis test with correction for multiple comparison was used to test for difference between the dietary groups, and difference in the Bonferroni post hoc test is indicated with equal superscripts. p value < 0.05 was used as significance value throughout.

Multiple linear regression analysis was used to examine the association between vegan dietary practice with Tg levels. Before performing a multivariate-adjusted analysis, univariate regression analysis was performed to examine if there was an association between the independent variable (vegan diet coded as dummy variable with pescatarian as control group) with the Tg levels (dependent). The independent variables (age, gender, smoking, supplement use, and years of dietary practice) that were significantly associated with the Tg levels in the univariate regression analysis (significance level p value < 0·005) were included in the multiple regression analysis. Two cases were identified with standardized residuals above 3. Sensitivity analysis was performed with and without the outliers. The final model was adjusted for age (years, continuous) and gender (male ref).