Thyroid

What research found here. What was shown to help comes first, then the risks reported, most common first.

Thyroid

Strong several reviews agree · Moderate one review or consistent trials · Limited one or a few small studies · Untested no study in people, listed with what it rests on

Shown to help13

LevothyroxineStrong

Weight change when treating hypothyroidism

Shown to lower weight about 1.4 to 2.2 kg after treating hypothyroidism, in overt cases only. Weight did not change in subclinical hypothyroidism.

The numbers and sources

Effect: Across 23 studies, weight changed by -1.39 kg (95% CI -2.56 to -0.23) after levothyroxine restored normal thyroid levels. BMI changed by -0.39 kg/m², fat mass by -0.75 kg and lean mass by -0.95 kg. Weight and BMI fell in overt hypothyroidism but not in subclinical hypothyroidism, where lean mass fell [3]. Across 35 studies, weight fell 2.15 kg in overt hypothyroidism, mostly in the first 3 months, with no change in fat mass and a small fall in lean mass. In subclinical hypothyroidism, BMI, weight, fat mass and lean mass did not change [4].
Population: 763 adults with hypothyroidism [3]; 1,116 adults, 672 with overt and 444 with subclinical hypothyroidism [4].
Sources: [3], [4]
Caveats: Both reviews pooled measurements taken before and after treatment, without an untreated comparison group. They disagree on fat mass [3][4].

Thyroid peroxidase antibodies (TPO antibodies)Strong

Miscarriage and preterm birth

Linked to more miscarriage and preterm birth in women with normal thyroid function, with preterm birth in 6.6% of antibody-positive women against 4.9%.

The numbers and sources

Effect: A meta-analysis of women with normal thyroid function found an odds ratio for miscarriage with thyroid autoantibodies of 3.90 in 19 cohort studies (95% CI 2.48 to 6.12) and 1.80 in 12 case-control studies (1.25 to 2.60). The odds ratio for preterm birth was 2.07 (1.17 to 3.68) [10]. In pooled individual data on 47,045 pregnant women, preterm birth occurred in 6.6% of TPO antibody-positive women and 4.9% of antibody-negative women (odds ratio 1.33, 95% CI 1.15 to 1.56) [11]. A pooled analysis of 11,212 pregnant women found the association with premature delivery rose with the antibody concentration [6].
Population: 31 studies with 12,126 women for miscarriage and 5 studies with 12,566 women for preterm birth [10]. 47,045 pregnant women in 19 cohorts, of whom 7.5% were TPO antibody positive [11]. 11,212 pregnant women in Dutch birth cohorts [6].
Sources: [10], [11], [6]
Caveats: Observational. Each study applied its own assay and cutoff, so a positive result does not mean the same level across studies.

Sex hormone binding globulin (SHBG)Moderate

Age, body weight and thyroid status

Linked in cohort and genetic studies to age, body weight and thyroid status: higher with age in men and with an overactive thyroid, lower at a higher body mass index.

The numbers and sources

Effect: In pooled individual data from 21,074 men, SHBG was directly associated with age and inversely associated with body mass index [13]. In a population-based cohort, SHBG rose in a straight line with age in men and followed a U-shaped pattern with age in women, and people with more cardiovascular risk factors had lower levels. Total testosterone and triglycerides explained 46.2% of the variation in men. Total testosterone, triglycerides, body mass index and fasting insulin explained 31.8% in women [14]. Among women with thyroid disorders, mean SHBG differed between groups because of the group with an overactive thyroid, where thyroid hormone levels were the main correlates. In the other groups the main correlates were body mass index and age [15]. In a genetic analysis, each standard deviation of genetically higher thyroid stimulating hormone went with SHBG 1.332 nmol/L lower (95% CI 0.717 to 1.946). Genetic predisposition to an underactive thyroid went with lower SHBG and to an overactive thyroid with higher SHBG [16].
Population: 9 cohorts of community-dwelling men with individual data [13]. 3,264 men and women in a population-based cohort [14]. Women with overt or subclinical thyroid disorders [15]. Up to 194,174 women and 167,020 men in UK Biobank [16].
Sources: [13], [14], [15], [16]
Caveats: These are associations seen across people. The cohort analyses are cross-sectional and do not show that a change in weight alters a person's SHBG. The genetic analysis found no association between thyroid function and sexual function [16].

Thyroid peroxidase antibodies (TPO antibodies)Moderate

Later hypothyroidism

Linked to later hypothyroidism: over 20 years, women with thyroid antibodies alone had 8 times the odds, and 38 times with a raised TSH as well.

The numbers and sources

Effect: In a 20-year British follow-up, positive thyroid antibodies alone carried odds ratios for developing hypothyroidism of 8 in women (95% CI 5 to 15) and 25 in men (10 to 63). With a raised TSH as well, the odds ratios were 38 (22 to 65) and 173 (81 to 370) [7]. In an Australian cohort, hypothyroidism 13 years later was present in 12.0% of antibody-positive women whose baseline TSH was 2.5 mU/L or less, 55.2% at 2.5 to 4.0 mU/L and 85.7% above 4.0 mU/L [5]. In a 5-year Chinese survey, 14.44% of people positive for TPO or thyroglobulin antibodies developed thyroid dysfunction, against 3.31% of those without antibodies [8]. Among 82 women with subclinical hypothyroidism, overt hypothyroidism developed in 58.5% of those with microsomal antibodies and 23.2% of those without [9]. A US guideline puts progression from subclinical to overt hypothyroidism at 4.3% a year with TPO antibodies and 2.6% a year without [4].
Population: 2,779 adults in the Whickham Survey [7]. 1,184 adults in the Busselton Health Surveys [5]. 3,018 people in three Chinese regions with different iodine intakes [8]. 82 women followed a mean 9.2 years [9].
Sources: [7], [5], [8], [9], [4]
Caveats: Observational. The studies used different antibody assays and cutoffs, and earlier ones measured antibodies by semiquantitative methods, not automated immunoassay [5]. The US guideline also reports that about 75% of people with chronic autoimmune thyroiditis have raised antibody levels, many with normal thyroid function at the time [4].

Thyroid stimulating hormone (TSH)Moderate

Later hypothyroidism

Linked to later hypothyroidism in a graded way, starting inside the reference range and rising further when thyroid antibodies are present.

The numbers and sources

Effect: In a 20-year British follow-up, a raised TSH alone carried odds ratios for developing hypothyroidism of 8 in women (95% CI 3 to 20) and 44 in men (19 to 104). With thyroid antibodies as well, the odds ratios were 38 (22 to 65) and 173 (81 to 370). Risk rose with TSH values above 2 mU/L at the first survey [21]. In an Australian cohort followed 13 years, hypothyroidism at follow-up was present in 12.0% of antibody-positive women with baseline TSH of 2.5 mU/L or less, 55.2% with TSH between 2.5 and 4.0 mU/L and 85.7% with TSH above 4.0 mU/L [22]. Among 82 women with subclinical hypothyroidism, overt hypothyroidism developed within 10 years in 0% of those starting at 4 to 6 mU/L, 42.8% at above 6 to 12 mU/L and 76.9% above 12 mU/L [23].
Population: 2,779 adults in the Whickham Survey, with 96% of survivors followed up at 20 years [21]. 1,184 adults in the Busselton Health Surveys of 1981 and 1994 [22]. 82 women followed a mean 9.2 years [23].
Sources: [21], [22], [23]
Caveats: Observational. The Australian study defined hypothyroidism at follow-up as TSH above 4.0 mU/L or thyroxine treatment [22]. The 10-year figures come from 82 women at one clinic [23].

Urinary iodineModerate

Thyroid disorders in populations with high levels

Linked to more hypothyroidism in populations with excess iodine, about twice the odds of subclinical hypothyroidism in a meta-analysis of three studies.

The numbers and sources

Effect: In a meta-analysis of cross-sectional studies in adults, populations with excess iodine had higher odds of overt hypothyroidism (OR 2.78, 95% CI 1.47 to 5.27) and subclinical hypothyroidism (OR 2.03, 1.58 to 2.62) than populations with adequate iodine [7]. In a five-year study of three regions of China with median urinary iodine of 84, 243 and 651 µg/L, the cumulative incidence of subclinical hypothyroidism was 0.2%, 2.6% and 2.9%, and that of autoimmune thyroiditis was 0.2%, 1.0% and 1.3%. Overt hypothyroidism developed in 0.2%, 0.5% and 0.3% [8].
Population: Three cross-sectional studies in adults within a review of 50 articles [7]. 3,018 people in three regions of China followed from 1999 to 2004 [8].
Sources: [7], [8]
Caveats: Observational comparisons between populations defined by their median urinary iodine, not by one person's result. The meta-analysis authors describe the included studies as few and of low quality, and the source of excess iodine was mainly iodized salt or drinking water [7].

Free T4 (free thyroxine)Limited

Time of day and season

One large study found the free T4 reference interval did not shift with hour of day, time of year, age or sex.

The numbers and sources

Compared with: Samples drawn at other hours and seasons
Effect: In 112,994 free T4 measurements from people without thyroid disease, age, sex, hour of day and time of year did not affect the free T4 reference interval. Females had higher free T4 than males by no more than 0.1 ng/dL. In the same data the TSH interval did vary by hour of day [16].
Population: 112,994 free T4 measurements from people aged 1 to 104, on a single immunoassay method [16].
Sources: [16]
Caveats: A retrospective analysis of laboratory records from one health system and one assay [16].

IodineLimitedStudied in children

Cognition in iodine-deficient children

Limited evidence: iodine slightly raised cognitive scores in iodine-deficient children, with each result resting on one trial.

The numbers and sources

Effect: After 24 weeks, iodine improved 4 of 7 cognitive and motor tests against placebo, including rapid target marking, symbol search, rapid object naming and Raven's Coloured Progressive Matrices. Total thyroxine rose about 40% and low thyroxine fell below 1% [1]. A 2026 Cochrane review of 8 RCTs (2,528 children and adolescents) found a slight rise in cognitive function against placebo (SMD 0.19, 95% CI 0.04 to 0.34), from 1 study of 166 participants rated high certainty [2].
Population: 310 children aged 10 to 12 in rural Albania, median urinary iodine 43 µg/L at baseline, 87% with goiter [1]. Trials in seven countries from 1982 to 2009 [2].
Sources: [1], [2]
Caveats: Each cognition result rests on one trial in children with deficiency. The Albanian trial used a single 400 mg dose of iodized oil [1]. The Cochrane trials were old and mostly gave a single large dose [2].

IodineLimitedStudied in children

Goiter, hypothyroidism and growth in children

Limited evidence: in single trials, iodine lowered hypothyroidism in children and deaths in infants. Effects on goiter were very uncertain, and weight did not change.

The numbers and sources

Effect: In the Cochrane review, iodine likely lowered hypothyroidism (RR 0.03, 95% CI 0.00 to 0.23, 1 study, 264 participants) and all-cause death in infants (RR 0.32, 95% CI 0.14 to 0.73, 1 study, 617 infants). Weight did not differ (MD -0.20 kg, 95% CI -1.92 to 1.52, 3 studies). The effect on goiter was very uncertain (RR 0.68, 95% CI 0.15 to 3.07, 2 studies, I² 97%), and removing the study at high risk of bias gave RR 0.31 (1 study, 230 participants) [2].
Population: Children and adolescents in 8 RCTs. One study gave 150 µg a day for 28 weeks, one gave infants under 1 year a single 100 mg dose, and the rest gave 400 to 490 mg once to children aged 6 to 17 [2].
Sources: [2]
Caveats: Each outcome rests on one to three trials. No study reported adverse events related to iodine [2].

Red light therapyLimited

Thyroid function in Hashimoto's hypothyroidism

One trial found people with autoimmune hypothyroidism needed a lower levothyroxine dose 9 months after 10 sessions of near-infrared laser to the thyroid than after placebo.

The numbers and sources

Effect: Nine months after stopping and then re-titrating levothyroxine, the mean dose needed was 38.59 µg a day in the laser group and 106.88 µg a day in the placebo group (P < 0.001). Thyroid peroxidase antibodies were lower with laser (P = 0.043), and thyroglobulin antibodies did not differ [32].
Population: 43 people taking levothyroxine for hypothyroidism from chronic autoimmune thyroiditis, 23 given laser and 20 placebo [32].
Sources: [32]
Caveats: One trial at one center, using a clinical 830 nm laser applied over the thyroid by clinicians.

SeleniumLimited

Mild Graves' eye disease

One trial found better quality of life and slower eye disease progression over 6 months in people with mild Graves' eye disease.

The numbers and sources

Effect: After 6 months, 100 µg twice a day improved quality of life, reduced eye involvement and slowed progression against placebo. The Clinical Activity Score fell more on selenium. Results held at 12 months [8].
Population: 159 people with mild Graves' orbitopathy [8].
Sources: [8]
Caveats: One trial. The abstract gives P values without effect sizes. Funded by the University of Pisa and the Italian ministry for education and research [8].

Thyroid peroxidase antibodies (TPO antibodies)Limited

Thyroid inflammation after childbirth

Limited evidence: a guideline review reports postpartum thyroiditis in 33 to 50% of women who were thyroid antibody positive in the first trimester.

The numbers and sources

Effect: The American Thyroid Association reports that women who are thyroid antibody positive in the first trimester have a 33 to 50% risk of postpartum thyroiditis, and that women with the highest antibody levels have the highest risk. It reports that 10 to 20% of cases result in permanent hypothyroidism [15].
Population: A guideline summary of earlier studies, whose sizes it does not give [15].
Sources: [15]
Caveats: Secondhand figures from guideline text, not a primary study.

1 more

Urinary iodineLimited

Spot sample against a 24-hour collection

One US study found the median spot concentration did not differ from the 24-hour concentration in a group of 400 adults, while other iodine measures were not interchangeable.

The numbers and sources

Compared with: Iodine concentration in a complete 24-hour urine collection
Effect: Volunteers collected all urine for 24 hours. The median 24-hour concentration was 144.8 µg/L, and medians from timed spot samples ran from 147.3 to 156.2 µg/L, with no consistent difference by time of day. The median total excreted was 173.6 µg a day, and the median iodine to creatinine ratio in spot samples was 103.6 to 114.3 µg/g. The authors conclude that concentration, the creatinine ratio and estimated 24-hour excretion are not to be used interchangeably [11]. A European guideline gives a reference value of 100 to 300 µg per 24 hours for urinary iodine, which is an amount per day and is not placed against a result typed here [12]. In the US survey of 1988 to 1994, urinary iodine concentration in µg/L was 1.16 times the value expressed in µg per gram of creatinine [2].
Population: 400 volunteers aged 18 to 39 in the United States [11].
Sources: [11], [12], [2]
Caveats: The agreement is for the median of a group, not for one person's spot sample against that person's own 24-hour collection. A lab report in µg per 24 hours or µg per gram of creatinine is a different quantity from the µg/L on this page.

Reported risks18

CalciumLimitedVery common

Reduced levothyroxine absorption

Limited evidence: in 20 adults on levothyroxine, 1,200 mg of calcium carbonate lowered free T4 and raised TSH, which returned after stopping calcium.

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: Taking 1,200 mg of calcium as carbonate with levothyroxine for 3 months lowered free T4 from 17 to 15 pmol/L and raised TSH from 1.6 to 2.7 mIU/L. Both returned after calcium stopped. 20% of patients had TSH above the normal range on calcium [15].
Population: 20 adults with hypothyroidism on stable levothyroxine [15].
Sources: [15]
Caveats: A small uncontrolled cohort, with lab data showing T4 binds to calcium carbonate in acid [15].

IronLimitedVery common

Interaction with thyroid hormone

One small trial without a control group found TSH rose from 1.6 to 5.4 mU/L when ferrous sulfate was taken at the same time as thyroxine.

The numbers and sources

How common: Very common: more than 1 in 10 people.
Effect: In 14 patients with hypothyroidism on stable thyroxine, taking 300 mg ferrous sulfate at the same time each day for 12 weeks raised TSH from 1.6 to 5.4 mU/L. Nine had more hypothyroid symptoms. Iron and thyroxine formed a poorly soluble complex in vitro [15].
Population: 14 patients, uncontrolled trial [15].
Sources: [15]
Caveats: No control group. Free thyroxine index did not change significantly.

IodineLimitedCommon

Thyroid disease at high intake

Limited evidence: higher iodine intake was linked to more subclinical hypothyroidism, 2.6% to 2.9% against 0.2%, in a 5-year study of Chinese regions.

The numbers and sources

How common: Common: between 1 in 100 and 1 in 10 people.
Effect: Over 5 years, subclinical hypothyroidism developed in 0.2%, 2.6% and 2.9% of people with mildly deficient, more than adequate and excessive iodine intake. Autoimmune thyroiditis developed in 0.2%, 1.0% and 1.3%. People with thyroid antibodies at baseline had more raised TSH at the higher intakes [7].
Population: 3,018 people in three regions of China, followed 1999 to 2004 [7].
Sources: [7]
Caveats: Observational comparison of regions, not a trial of supplements.

Atypical antipsychoticsModerateHow often is not known

Lower thyroid hormone on quetiapine

Shown to lower thyroid hormone levels in a dose-related way on quetiapine, with thyroxine about 20% lower at the higher end of the dose range.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The label reports dose-related decreases in thyroid hormone in quetiapine trials. Total and free thyroxine fell by about 20% at the higher end of the dose range, reached a maximum in the first six weeks and reversed in nearly all cases after stopping [3].
Population: Adults in quetiapine clinical trials [3].
Sources: [3]
Caveats: An average change in a blood level. The passage read for this page gives no count of people who became hypothyroid. The evidence is for quetiapine.

Soy foodsModerateHow often is not known

Thyroid function

Shown to raise TSH slightly, by 0.25 mIU/L, in pooled trials, with no change in free T3 or free T4.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Soy raised thyroid stimulating hormone (WMD 0.248 mIU/L, 95% CI 0.001 to 0.494, I² 80.31%). Free T3 (0.027 pmol/L, 95% CI -0.052 to 0.107) and free T4 (-0.003 pmol/L, 95% CI -0.018 to 0.011) did not change. The authors call the clinical significance of the TSH rise unclear [16].
Population: Randomized trials in 18 articles [16].
Sources: [16]
Caveats: High heterogeneity, and the lower bound of the TSH interval is close to zero [16].

Alpha-GPC and choline supplementsLimitedStudied in menHow often is not known

Thyroid-stimulating hormone (alpha-GPC)

One short trial found lower TSH after 7 days on 500 mg alpha-GPC in 48 young men. Thyroid hormones were not reported.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: After 7 days, serum TSH was significantly lower in the 500 mg alpha-GPC group than in the other groups [10].
Population: 48 young men [10].
Sources: [10]
Caveats: One short trial. Thyroid hormones themselves were not reported in the abstract.

Alternate-day fastingLimitedHow often is not known

Lower metabolic rate and thyroid hormone

Small trials found alternate-day fasting lowered resting metabolic rate and triiodothyronine against control.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Over 4 weeks, alternate-day fasting, but not time-restricted eating, lowered resting metabolic rate and triiodothyronine against control [5]. Triiodothyronine was also lower after long-term alternate-day fasting [4].
Population: 76 adults aged 18 to 40 [5]; healthy middle-aged adults [4].
Sources: [4], [5]
Caveats: Short trials. The abstracts do not give the size of the changes [4][5].

DairyLimitedHow often is not known

Lower absorption of levothyroxine

One crossover study found 12 ounces of 2% milk taken with levothyroxine lowered thyroxine absorption against levothyroxine alone.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The area under the curve of total thyroxine after 1,000 µg of levothyroxine was 67.3 with milk and 73.5 without (p = 0.02) [23].
Population: 10 healthy adults with normal thyroid function, mean age 33.7, with a four-week washout [23].
Sources: [23]
Caveats: One small study using a single high dose in people without thyroid disease. Levothyroxine is on its own page.

DulaglutideLimitedHow often is not known

FDA boxed warning: thyroid C-cell tumors

FDA states dulaglutide caused thyroid C-cell tumors in rats and the human relevance is unknown. Two large GLP-1 drug studies in people disagreed on thyroid cancer risk.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The boxed warning states dulaglutide caused dose- and duration-dependent thyroid C-cell tumors in rats, and that it is unknown whether it causes them in humans. It is contraindicated with a personal or family history of medullary thyroid cancer or MEN 2. One case of medullary thyroid cancer occurred in a trial patient with calcitonin about 8 times the upper limit before treatment, and one case of C-cell hyperplasia in REWIND [1]. In French data, GLP-1 receptor agonist use for 1 to 3 years was linked to thyroid cancer (HR 1.58, 95% CI 1.27 to 1.95) and medullary thyroid cancer (HR 1.78, 95% CI 1.04 to 3.05) [9]. In Scandinavian registries, GLP-1 receptor agonists were not linked to thyroid cancer against DPP-4 inhibitors (HR 0.93, 95% CI 0.66 to 1.31) [10].
Population: Rats, and trial participants [1]; 2,562 thyroid cancer cases and 45,184 matched controls with type 2 diabetes [9]; 145,410 GLP-1 receptor agonist users and 291,667 DPP-4 inhibitor users [10].
Sources: [1], [9], [10]
Caveats: Both human studies pool all GLP-1 receptor agonists and are observational [9][10].

IodineLimitedHow often is not known

Thyroid nodules and thyroid cancer

Limited evidence: thyroid nodules were linked to iodine deficiency, not high intake. High urinary iodine was linked to thyroid cancer, while dietary intake was not.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: In 25 cross-sectional studies (54,621 people), thyroid nodules were more common with iodine deficiency than adequate status (OR 1.28, 95% CI 1.09 to 1.50), and did not differ with more than adequate (OR 1.02) or excessive intake (OR 1.13, 95% CI 0.98 to 1.30) [9]. A second review of 31 studies found the same pattern, with OR 1.24 for deficiency and no significant link for higher intake [10]. In 17 studies (273,651 people), high urinary iodine was linked to thyroid cancer (OR 6.43, 95% CI 2.72 to 15.22), while dietary iodine intake was not (OR 0.75, 95% CI 0.37 to 1.52) [11].
Population: Population and clinical studies of iodine status [9][10][11].
Sources: [9], [10], [11]
Caveats: Observational studies, mostly cross-sectional for nodules, which cannot show cause [9][10]. Urinary iodine is a single measure of recent intake. None of these studies tested supplements.

LevothyroxineLimitedHow often is not known

Toxicity at weight-loss doses (boxed warning)

The label warns thyroid hormones are not for weight loss, and doses beyond daily needs may produce serious or even life-threatening toxicity.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The label's boxed warning states that thyroid hormones are not for obesity or weight loss, and that doses beyond daily hormonal needs may produce serious or even life-threatening toxicity. Adverse reactions are listed as those of hyperthyroidism from overdose, including arrhythmias, heart attack, tremor, insomnia, muscle weakness and heat intolerance [10]. In obesity studies, thyroid hormone therapy induced subclinical hyperthyroidism [9].
Population: Label [10]; obesity studies [9].
Sources: [10], [9]
Caveats: Label statements, not trial rates.

LiraglutideLimitedHow often is not known

Thyroid C-cell tumors (boxed warning)

The label warns of thyroid C-cell tumors in rats and mice, and states it is unknown whether liraglutide causes them in humans.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The label carries a boxed warning. Liraglutide caused dose- and duration-dependent thyroid C-cell tumors in rats and mice. The label states it is unknown whether it causes them in humans, and that postmarketing reports of medullary thyroid carcinoma are insufficient to establish or exclude a causal link. It is contraindicated with a personal or family history of medullary thyroid carcinoma or MEN 2 [13].
Population: Animal (rats and mice), plus postmarketing case reports [13].
Sources: [13]
Caveats: The tumor finding is from animals. No human trial measured it as an outcome.

6 more

MagnesiumLimitedHow often is not known

Lower levothyroxine absorption

One trial found magnesium aspartate lowered levothyroxine absorption 12% in healthy adults, while the 7% drop with magnesium citrate was not significant.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Taken with 1 mg of levothyroxine, magnesium aspartate lowered thyroxine exposure (AUC) by 12% (GMR 0.88, 95% CI 0.81 to 0.95). Magnesium citrate lowered it by 7%, which was not statistically significant (GMR 0.93, 95% CI 0.86 to 1.01) [17].
Population: 15 healthy euthyroid adults, open-label randomized crossover [17].
Sources: [17]
Caveats: Single doses over 6 hours in people without thyroid disease. Effects on TSH in treated patients were not measured.

OrforglipronLimitedHow often is not known

Label warnings: thyroid tumors, pancreatitis and gallbladder

The label warns of thyroid C-cell tumors seen with related drugs in rodents, with human relevance not determined. Pancreatitis and gallstones were slightly more common.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The label carries a boxed warning about thyroid C-cell tumors. Such tumors occurred in rodents with GLP-1 drugs that are active in rats and mice, while orforglipron is not active in rats or mice and did not produce tumors in them. The label states that the human relevance has not been determined, and the drug is contraindicated with a personal or family history of medullary thyroid carcinoma or MEN 2. In the weight trials, acute pancreatitis was confirmed in 6 people on orforglipron (0.14 per 100 years of exposure) against 1 on placebo (0.04), gallstones in 1% against 0.7%, acute kidney injury in 0.2% against 0.05%, low blood pressure in 2% against 0.5%, and hair loss in 4 to 5% against 2% [6].
Population: Label pool of the two weight trials and nonclinical studies [6].
Sources: [6]
Caveats: The thyroid warning rests on animal findings for the drug class. Pancreatitis events were few.

Paleo dietLimitedStudied in womenHow often is not known

Iodine

One trial found urinary iodine fell by half within 6 months in postmenopausal women, while it held steady on a Nordic guideline diet.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: Median 24-hour urinary iodine concentration fell from 71.0 to 36.0 μg/L at 6 months (p 0.001), and 24-hour iodine excretion from 134.0 to 77.0 μg/d. Levels did not change on the comparison diet. FT4 and TSH did not differ, and FT3 was lower at 6 months [8]. In a 4-week trial, iodine intake was 47.9 μg a day lower than on the Australian national guide diet (95% CI 16.5 to 79.2) [7].
Population: 70 postmenopausal women with overweight or obesity over 2 years [8]; 39 women over 4 weeks [7].
Sources: [7], [8]
Caveats: The diet excludes table salt and dairy, which the authors call two of the largest iodine sources [8]. Small trials in women.

Soy foodsLimitedHow often is not known

Lower absorption of levothyroxine

The levothyroxine label warns soybean flour may decrease its absorption. A case report agrees, while a crossover study of a soy isoflavone supplement found no change.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The Synthroid label states that soybean flour, cottonseed meal, walnuts and dietary fiber may bind and decrease levothyroxine absorption, and its instructions for crushed tablets exclude soybean-based infant formula as a vehicle [17]. A woman taking a soy protein drink right after levothyroxine needed unusually high doses, and lower doses sufficed once the two were taken apart [18]. In a crossover study, 60 mg of soy isoflavones taken with levothyroxine did not change its absorption (ratio 0.99 for total exposure) [19]. A review found the data mainly case reports and the clinical impact, if any, small [20].
Population: People taking levothyroxine [17]. One 45-year-old woman [18]. 12 postmenopausal women on levothyroxine [19].
Sources: [17], [18], [19], [20]
Caveats: The crossover study tested an isoflavone supplement, not soy protein or soy flour [19]. Levothyroxine is on its own page.

TirzepatideLimitedHow often is not known

Thyroid C-cell tumors

The label warns that tirzepatide causes thyroid C-cell tumors in rats, and that it is unknown whether it causes them in humans.

The numbers and sources

How common: How often is not known: the sources give no rate.
Effect: The label carries a boxed warning that tirzepatide causes thyroid C-cell tumors in rats at clinically relevant exposures, and that it is unknown whether it causes them in humans, including medullary thyroid carcinoma. It is contraindicated with a personal or family history of medullary thyroid carcinoma or with MEN 2 [25]. In pooled trials, calcitonin rose more on 10 and 15 mg than placebo, and no papillary thyroid carcinoma was reported [24].
Population: Rats (animal) [25]; 13,761 trial participants [24].
Sources: [25], [24]
Caveats: The tumor finding is from animals. Human relevance has not been determined [25].

Sea mossLimited

Thyroid overactivity in Graves' disease

Case reports: a woman with untreated Graves' disease had thyroid overactivity attributed to iodine from a sea moss and bladderwrack supplement, easing after she stopped.

The numbers and sources

Effect: A 28-year-old woman with fatigue, palpitations and weight loss had TSH below 0.01 mIU/L and raised free T4 (2.46 ng/dL). She had used a store-bought Irish sea moss and bladderwrack supplement on and off for 2 years and was diagnosed with Graves' disease. The authors attributed her thyrotoxicosis to excess iodine (the Jod-Basedow phenomenon). One week after stopping sea moss, free T4 was 1.4 ng/dL [3].
Population: One case report [3].
Sources: [3]
Caveats: A conference abstract. The product also contained bladderwrack, a brown seaweed, and the patient had untreated thyroid disease [3].

Mixed results14

Free T4 (free thyroxine)Moderate

Pregnancy and child outcomes

Mixed results: low free T4 in pregnancy was linked to lower child IQ and preterm birth, while a trial that treated low free T4 found no change in IQ.

The numbers and sources

Effect: In a Dutch birth cohort, maternal free T4 in early pregnancy had an inverted U-shaped association with child IQ, with both low and high concentrations corresponding to a 1.4 to 3.8 point lower mean IQ [11]. In pooled data on 47,045 pregnant women, preterm birth occurred in 7.1% of women with a low free T4 and normal TSH against 5.0% of women with normal results (odds ratio 1.46, 95% CI 1.12 to 1.90) [12]. In a trial of 526 pregnant women with free T4 below 0.86 ng/dL and normal TSH, median child IQ was 94 with levothyroxine and 91 with placebo (P = 0.30), with no significant differences in other pregnancy or neonatal outcomes [6].
Population: 3,839 mother and child pairs, with IQ tested at a median age of 6 [11]. 47,045 pregnant women in 19 cohorts [12]. 526 women randomized at a mean of 17.8 weeks of pregnancy [6].
Sources: [11], [12], [6]
Caveats: A marker that tracks an outcome is not always one whose correction changes it. The cohort findings are observational, and the trial began treatment between 8 and 20 weeks of pregnancy [6].

IodineModerateStudied in pregnancy

Thyroid function in pregnancy

Mixed results: reviews agree iodine lowered maternal thyroglobulin in pregnancy, but disagree on whether TSH changed.

The numbers and sources

Effect: Most studies found no effect on maternal or infant TSH or free thyroxine. Most RCTs found lower maternal thyroglobulin, and 3 found less growth in thyroid volume [3]. Cochrane found no clear difference in hypothyroidism in pregnancy or postpartum and postpartum hyperthyroidism fell (RR 0.32, 95% CI 0.11 to 0.91, 3 trials, 543 women) [5]. A 2026 meta-analysis found lower TSH (SMD -0.14, 95% CI -0.23 to -0.05) and lower thyroglobulin (SMD -0.22, 95% CI -0.32 to -0.13) with supplementation, with larger TSH drops at 200 to 300 µg a day [6].
Population: 37 publications in mildly to moderately deficient women [3]; 14 trials, 11 contributing data from over 2,700 women [5]; iodine-deficient pregnant women in English and Chinese studies [6].
Sources: [3], [5], [6]
Caveats: Cochrane rated the evidence low or very low quality and found the data insufficient for meaningful conclusions [5]. The reviews agree on lower thyroglobulin and differ on TSH [3][6]. The 2026 abstract does not give the number of studies or women [6].

LevothyroxineModerate

Pregnancy outcomes with subclinical hypothyroidism or thyroid antibodies

Mixed results: fewer preterm and low birth weight births in one review, lower miscarriage in women with thyroid antibodies, and no difference in an NIH trial.

The numbers and sources

Effect: In randomized trials of women with subclinical hypothyroidism in pregnancy, levothyroxine was associated with less preterm birth (RR 0.56, 95% CI 0.41 to 0.77) and fewer low birth weight infants (RR 0.56, 95% CI 0.35 to 0.89). Small for gestational age and macrosomia did not differ [5]. In euthyroid pregnant women with thyroid peroxidase antibodies, miscarriage was lower (RR 0.78, 95% CI 0.63 to 0.98), while preterm birth (RR 0.78, 95% CI 0.55 to 1.13) and live birth (RR 1.05, 95% CI 0.99 to 1.12) did not differ [6]. In subclinical hypothyroidism, gestational hypertension (OR 1.03, 95% CI 0.85 to 1.25) and pre-eclampsia (OR 1.02, 95% CI 0.66 to 1.58) did not differ from controls [7].
Population: 30 RCTs and cohort studies with 18,568 pregnant women with subclinical hypothyroidism [5]; 8 RCTs with 1,645 women [6]; trials and observational studies in pregnant women with subclinical or overt hypothyroidism or thyroid autoimmunity [7].
Sources: [5], [6], [7]
Caveats: [5] pooled RCTs with cohort studies, and the figures above are its RCT estimates. Preterm birth fell in [5] but not in [6], which studied a different group. The NIH trial of treatment from mid-pregnancy found no difference in pregnancy outcomes [8].

SeleniumModerate

Hashimoto's thyroiditis

Mixed results: thyroid antibodies fell in people with Hashimoto's in both reviews, while reviews disagree on whether TSH changed.

The numbers and sources

Effect: Thyroid peroxidase antibodies fell in both reviews. In people on levothyroxine, the weighted mean difference after 3 months was -271 (95% CI -366 to -175) [5]. Across 29 cohorts, the SMD was -0.96 (95% CI -1.36 to -0.56) [6]. On TSH, sources disagree. In people not on thyroid hormone, a 2017 review found no change in TSH, quality of life or ultrasound findings [7]. A 2024 review found TSH lower (SMD -0.21, 95% CI -0.43 to -0.02, 869 people) [6].
Population: Adults with chronic autoimmune thyroiditis, 16 trials [5], 9 trials [7] and 35 studies [6].
Sources: [5], [6], [7]
Caveats: Antibody levels are a lab marker. The 2016 review rated evidence quality low [5], and the 2024 review moderate, with I² of 90% for antibodies [6].

Thyroid peroxidase antibodies (TPO antibodies)Moderate

Live birth and miscarriage in trials of levothyroxine for antibody-positive women

Mixed results: two large trials of levothyroxine in antibody-positive women with normal thyroid function found no change in live births or miscarriage, after small earlier trials reported fewer miscarriages.

The numbers and sources

Effect: In a placebo-controlled trial of 952 women with TPO antibodies, normal thyroid function and a history of miscarriage or infertility, the live-birth rate was 37.4% with levothyroxine 50 micrograms a day and 37.9% with placebo (relative risk 0.97, 95% CI 0.83 to 1.14), with no significant differences in pregnancy loss or preterm birth [12]. In an open-label trial of 600 antibody-positive women having in vitro fertilization, miscarriage rates were 10.3% with levothyroxine and 10.6% without, and live-birth rates 31.7% and 32.3% [13]. Earlier, a pooled estimate from two small randomized studies gave a relative risk of miscarriage of 0.48 (0.25 to 0.92) [10]. In one earlier study, 115 antibody-positive pregnant women were divided into a levothyroxine group and an untreated group, with miscarriage in 3.5% and 13.8% [14].
Population: 952 women at 49 UK hospitals, drawn from 19,585 tested [12]. 600 women at one hospital in China [13]. 115 antibody-positive pregnant women in Italy [14].
Sources: [12], [13], [10], [14]
Caveats: These trials treated women who had the antibodies and did not set out to lower the antibody level. The largest trial started treatment before conception and the Chinese trial was open label [12][13].

Thyroid stimulating hormone (TSH)Moderate

Pregnancy and child outcomes

Mixed results: raised TSH in pregnancy was linked to more preterm birth, while a levothyroxine trial found no difference in child IQ or pregnancy outcomes.

The numbers and sources

Effect: In pooled data on 47,045 pregnant women, preterm birth occurred in 6.1% of women with a raised TSH and normal free T4 against 5.0% of women with normal results (odds ratio 1.29, 95% CI 1.01 to 1.64). Each standard deviation higher TSH had an odds ratio of 1.04 (1.00 to 1.09) [27]. In a trial of 677 pregnant women with TSH of 4.00 mU/L or more and normal free T4, the median child IQ score was 97 with levothyroxine and 94 with placebo (P = 0.71), with no significant differences in other pregnancy or neonatal outcomes [28].
Population: 47,045 pregnant women in 19 cohorts, median 12.9 weeks at blood sampling [27]. 677 women randomized at a mean of 16.7 weeks of pregnancy [28].
Sources: [27], [28]
Caveats: The cohort finding is observational. The trial began treatment between 8 and 20 weeks of pregnancy, so it does not describe earlier treatment [28].

Urinary iodineModerate

Child IQ and reading when iodine was low in pregnancy

Mixed results: one UK cohort linked low iodine in early pregnancy to lower child verbal IQ, a pooled analysis found a weaker link, and one supplement trial found no effect.

The numbers and sources

Effect: In a UK cohort, children of women whose first-trimester urinary iodine to creatinine ratio was below 150 µg/g were more likely to score in the lowest quarter for verbal IQ at age 8 (OR 1.58, 95% CI 1.09 to 2.30) and for reading accuracy at age 9 (OR 1.69, 1.15 to 2.49) than children of women at 150 µg/g or above. The group's median urinary iodine concentration was 91.1 µg/L [4]. A pooled analysis of three European birth cohorts found that a ratio below 150 µg/g was not significantly associated with nonverbal IQ (difference -0.6 points, 95% CI -1.7 to 0.4) or verbal IQ (-0.6 points, -1.3 to 0.1). It found a curved relation with verbal IQ that was present only up to 14 weeks of pregnancy [5]. In a randomized trial in India and Thailand, 200 µg of iodine a day from about 11 weeks of pregnancy did not change children's verbal IQ at age 5 to 6 (89.5 against 90.2 on placebo, difference -0.7, 95% CI -2.9 to 1.5), performance IQ or executive function [6].
Population: 1,040 mother and child pairs in the United Kingdom [4]. 6,180 mother and child pairs in the Netherlands, Spain and the United Kingdom [5]. 832 pregnant women with a median urinary iodine of 131 µg/L, and 313 to 315 of their children at follow-up [6].
Sources: [4], [5], [6]
Caveats: The cohort studies are observational and measured iodine as a ratio to creatinine in µg/g, a different quantity from the concentration on this page. The trial enrolled women with mild deficiency and followed fewer than half of the children to age 5 to 6. All of these studies compared groups of women. None tested acting on one woman's urine result.

Free T3 (free triiodothyronine)Limited

Biotin supplements and the assay

One small study found 10 mg a day of biotin gave falsely high results in 5 of 8 biotin-based competitive immunoassays, across analytes that included free T3.

The numbers and sources

Compared with: The same person's result before taking biotin
Effect: After 7 days of 10 mg a day of biotin, interference appeared in 9 of 23 assays built on biotin and streptavidin (39%) and in none of 14 assays built without them. Five of 8 biotin-based competitive immunoassays read falsely high and 4 of 15 biotin-based sandwich immunoassays read falsely low. The 11 analytes included free T3, free T4 and TSH, measured on 4 diagnostic systems [11].
Population: 6 adults described as healthy, mean age 38 [11].
Sources: [11]
Caveats: Six people. The abstract does not list which free T3 assays were affected, and many assays showed no interference [11].

Free T4 (free thyroxine)Limited

Biotin supplements and the assay

One small study found 10 mg a day of biotin gave falsely high results in 5 of 8 biotin-based competitive immunoassays, across analytes that included free T4.

The numbers and sources

Compared with: The same person's result before taking biotin
Effect: After 7 days of 10 mg a day of biotin, interference appeared in 9 of 23 assays built on biotin and streptavidin (39%) and in none of 14 assays built without them. Five of 8 biotin-based competitive immunoassays read falsely high and 4 of 15 biotin-based sandwich immunoassays read falsely low. The 11 analytes included free T4, free T3 and TSH, measured on 4 diagnostic systems [17].
Population: 6 adults described as healthy, mean age 38 [17].
Sources: [17]
Caveats: Six people. The abstract does not list which free T4 assays were affected, and many assays showed no interference [17].

GroundingLimited

Blood chemistry and thyroid tests

One set of small experiments found grounding during night rest lowered serum iron, free T3 and several electrolytes and raised free T4, TSH and ferritin.

Compared with a group not in contact with the earth [9]

The numbers and sources

Effect: Earthing during night rest lowered serum iron, ionized calcium and inorganic phosphorus, and lowered urinary calcium and phosphorus. It decreased free T3 and increased free T4 and TSH. Continuous earthing decreased blood glucose in patients with diabetes. Earthing decreased sodium, potassium, magnesium, total protein and albumin, and increased transferrin, ferritin and globulins [9].
Population: Five experiments of 12 to 84 people each [9].
Sources: [9]
Caveats: The abstract gives no sizes, no durations for each experiment and no clinical meaning for the changes. Nobody uses grounding to shift these tests, and the direction is neither a benefit nor a harm that the authors define. One author is a co-author of a review of grounding's proposed benefits [10].

Thyroid stimulating hormone (TSH)Limited

Repeat testing after an out-of-range result

One large study found that without thyroid treatment, 62% of mildly raised and 51% of low TSH results were inside the laboratory range on repeat testing.

The numbers and sources

Compared with: The same person's repeat TSH measurements over 5 years
Effect: In a health organization database, TSH returned to the laboratory's limits of 0.35 to 5.5 mIU/L in 62.1% of people who started above 5.5 and up to 10 mIU/L, 27.2% of those who started above 10 mIU/L and 51.2% of those who started below 0.35 mIU/L. Of people with a first result inside the limits, 98% stayed inside them [29]. The US Preventive Services Task Force states that multiple tests over 3 to 6 months are used to confirm or rule out an abnormal finding [2].
Population: 422,242 outpatients with no thyroid disease or pregnancy, of whom 346,549 took no thyroid medication during follow-up [29].
Sources: [29], [2]
Caveats: One database from one health organization. The study gives no threshold for agreement, so the result is given as numbers only [29].

Thyroid stimulating hormone (TSH)LimitedStudied in men

Variation within one person

One small study found a single TSH result placed a man's usual level only within about 50%, across monthly tests for a year.

The numbers and sources

Compared with: The same person's results on 12 monthly draws
Effect: In 16 men sampled monthly for 12 months, one test result described the individual's set point with a precision of plus or minus 50% for TSH. The difference needed to be 95% confident of a real change between two results averaged 0.75 mU/L, with a range of 0.2 to 1.6 mU/L across the men. Each man's own spread of results was about half as wide as the group's [30]. A US guideline reports that TSH varies through the day by up to about 50% of mean values [3].
Population: 16 men described as healthy [30].
Sources: [30], [3]
Caveats: Sixteen men at one center. The study states no threshold for agreement [30].

2 more

Thyroid stimulating hormone (TSH)Limited

Differences between assays

One comparison found the upper reference limit differed by up to 2.13 mIU/L between manufacturers' TSH assays run on the same samples.

The numbers and sources

Compared with: A statistically derived all-assay target, since TSH has no reference measurement method
Effect: When 14 manufacturers measured the same panels, the assay-specific estimates of the 97.5th percentile spanned 2.13 mIU/L, the medians 0.89 mIU/L and the 2.5th percentiles 0.27 mIU/L. After the assays were recalibrated to a common target those spans fell to 0.77, 0.29 and 0.12 mIU/L, and the spread of assay means fell from a coefficient of variation of 9.5% to 4.2%. After recalibration, 14 of 15 assays met the bias specification [32].
Population: Clinical serum panels measured by 14 manufacturers in a study by the IFCC Committee for Standardization of Thyroid Function Tests [32].
Sources: [32]
Caveats: A proof of concept. The abstract does not say whether laboratories have adopted the recalibration, so a result may still depend on the assay used [32].

Thyroid stimulating hormone (TSH)Limited

Biotin supplements and the assay

One small study found 10 mg a day of biotin gave falsely low results in 4 of 15 biotin-based sandwich immunoassays, across analytes that included TSH.

The numbers and sources

Compared with: The same person's result before taking biotin
Effect: After 7 days of 10 mg a day of biotin, interference appeared in 9 of 23 assays built on biotin and streptavidin (39%) and in none of 14 assays built without them. Four of 15 biotin-based sandwich immunoassays read falsely low and 5 of 8 biotin-based competitive immunoassays read falsely high. The 11 analytes included TSH, free T4, free T3, total T4 and total T3, measured on 4 diagnostic systems [33].
Population: 6 adults described as healthy, mean age 38 [33].
Sources: [33]
Caveats: Six people. The abstract does not list which TSH assays were affected, and many assays showed no interference [33].

Not shown to help13

LevothyroxineStrong

Symptoms and quality of life in subclinical hypothyroidism

Not shown to improve quality of life or thyroid symptoms in adults with subclinical hypothyroidism, including those 65 or older, though it brought TSH into range.

The numbers and sources

Effect: Thyroid hormone lowered mean TSH into the reference range (0.5 to 3.7 mIU/L against 4.6 to 14.7 on placebo) but did not change general quality of life (SMD -0.11, 95% CI -0.25 to 0.03) or thyroid-related symptoms (SMD 0.01, 95% CI -0.12 to 0.14) [1]. In adults aged 65 or older, the change in Hypothyroid Symptoms score at 1 year was 0.2 on both levothyroxine and placebo (difference 0.0, 95% CI -2.0 to 2.1). Tiredness scores did not differ (difference 0.4, 95% CI -2.1 to 2.9), and no secondary outcome improved [2].
Population: 21 RCTs with 2,192 nonpregnant adults, 3 to 18 months [1]; 737 adults with mean age 74.4 and TSH 4.60 to 19.99 mIU/L [2].
Sources: [1], [2]
Caveats: Most participants had mildly raised TSH. [1] rated the evidence moderate to high quality. [2] was funded by the European Union FP7 and others.

Thyroid stimulating hormone (TSH)Strong

Symptoms and quality of life in trials that lowered a mildly raised TSH

Not shown to improve quality of life or thyroid symptoms when levothyroxine lowered TSH into range in 21 trials of adults with subclinical hypothyroidism.

The numbers and sources

Effect: In a trial of 737 adults aged 65 and older with TSH of 4.60 to 19.99 mIU/L, levothyroxine left TSH at 3.63 mIU/L at 1 year against 5.48 on placebo. The Hypothyroid Symptoms score did not differ (between-group difference 0.0, 95% CI -2.0 to 2.1), nor did the Tiredness score (0.4, -2.1 to 2.9) [24]. Across 21 trials, thyroid hormone brought mean TSH to 0.5 to 3.7 mIU/L against 4.6 to 14.7 mIU/L on placebo, with no benefit for general quality of life (standardized mean difference -0.11, 95% CI -0.25 to 0.03) or thyroid-related symptoms (0.01, -0.12 to 0.14) [25]. A guideline panel that reviewed the same trials issued a strong recommendation against thyroid hormones for adults with subclinical hypothyroidism, and stated that it does not apply to women trying to become pregnant or to people with TSH above 20 mIU/L [26].
Population: 737 adults, mean age 74.4, mean baseline TSH 6.40 mIU/L [24]. 2,192 nonpregnant adults in 21 randomized trials lasting 3 to 18 months [25].
Sources: [24], [25], [26]
Caveats: A marker that tracks an outcome is not always one whose correction changes it. The pooled trials lasted 3 to 18 months. The guideline panel judged effects on cardiovascular events and deaths to be uncertain, measured in one trial with few events [26].

AshwagandhaModerate

Thyroid hormones

Not shown to change TSH or T3 in adults across 23 trials. T4 rose slightly, by 0.61 µg/dL.

The numbers and sources

Effect: Across 23 placebo-controlled trials, TSH and T3 did not change, and T4 rose by 0.61 µg/dL (p = 0.02) [6].
Population: 1,706 adults [6].
Sources: [6]
Caveats: The authors note heterogeneity and limited data on some endpoints [6].

Free T3 (free triiodothyronine)Moderate

Symptoms in trials that added T3 to thyroxine for hypothyroidism

Not shown to change pain, mood, fatigue, quality of life, weight or cholesterol when T3 was added to thyroxine in 11 trials of adults with hypothyroidism.

The numbers and sources

Effect: A meta-analysis of 11 randomized trials found no difference between combined T4 and T3 treatment and T4 alone for bodily pain (standardized mean difference 0.00, 95% CI -0.34 to 0.35), depression (0.07, -0.20 to 0.34), anxiety (0.00, -0.12 to 0.11), fatigue (-0.12, -0.33 to 0.09) or quality of life (0.03, -0.09 to 0.15). Body weight and blood lipids did not differ, and adverse events did not differ between regimens [10].
Population: 1,216 adults with clinical hypothyroidism in 11 randomized trials [10].
Sources: [10]
Caveats: The trials compared treatments in people already treated for hypothyroidism, and the abstract does not report the free T3 levels reached. Searches ran to September 2005 [10].

Free T4 (free thyroxine)Moderate

Immunoassays against the reference method

Shown to read low against the reference method: 15 of 17 free T4 procedures tested had negative biases, by as much as 42%.

The numbers and sources

Compared with: Equilibrium dialysis with isotope dilution mass spectrometry
Effect: An international working group compared 15 immunoassays and 2 dialysis-based laboratory procedures with a candidate reference procedure. Of the 17, 2 had a mean bias within 10% and 15 had negative biases of up to 42%. Correlation with the reference procedure ranged from 0.92 to 0.78, and total imprecision from 1.5% to 14.1% [13]. In a later study, recalibrating 13 manufacturers' assays to the reference procedure cut the spread of assay means from a coefficient of variation of 13% to 5%, and 12 of 13 recalibrated assays met a 10% bias specification. One met the 13% total error specification [14].
Population: Panels of single-donor sera measured by 9 manufacturers [13]. Standardization and reference interval panels measured by 13 manufacturers [14].
Sources: [13], [14]
Caveats: The second study was a proof of concept, and its abstract does not say whether laboratories have adopted the recalibration. Until they do, a result and its reference interval belong to the assay that produced them [14].

Free T3 (free triiodothyronine)Limited

Immunoassays against the reference method

One comparison found 9 of 14 free T3 procedures read below the reference method, by as much as 30%, and one read 22% above it.

The numbers and sources

Compared with: Equilibrium dialysis with isotope dilution mass spectrometry
Effect: An international working group compared 13 immunoassays and 1 dialysis-based laboratory procedure with a candidate reference procedure. Four had a mean bias within 10%, 9 had negative biases of up to 30% and 1 was positively biased by 22%. Correlation with the reference procedure ranged from 0.88 to 0.30, and total imprecision from 2.4% to 10.0%. Recalibration to the reference procedure removed the assay-specific biases, while sample-related effects remained [5].
Population: A panel of single-donor sera measured with assays from 9 manufacturers [5].
Sources: [5]
Caveats: Published in 2010, so the assays may since have changed. The abstract names no assay [5].

Free T4 (free thyroxine)Limited

Immunoassays in pregnancy

Limited evidence: a guideline review reports that free T4 immunoassays read lower in late pregnancy by an amount that depends on the assay.

The numbers and sources

Compared with: Equilibrium dialysis, ultrafiltration or mass spectrometry measurement
Effect: The American Thyroid Association reports that automated immunoassays show a significant, assay-dependent fall in measured free T4 in the third trimester, although direct measurement by more precise methods does not show a similar degree of reduction. It attributes this to the rise in thyroxine-binding globulin and fall in albumin during pregnancy [2]. An earlier US guideline states that many pregnant women have values below the nonpregnant range in the third trimester, and that ranges specific to the method and trimester have not been generally established [15].
Population: Guideline summaries of earlier studies, whose sizes they do not give [2][15].
Sources: [2], [15]
Caveats: Secondhand statements from guideline text, not a primary study.

LevothyroxineLimited

Child IQ when treated in pregnancy

One trial found no difference in child IQ at 5 years when women with subclinical hypothyroidism or low free T4 were treated from mid-pregnancy.

The numbers and sources

Effect: Children of women treated for subclinical hypothyroidism had a median IQ of 97 against 94 on placebo (P = 0.71). In a parallel trial for low free T4, median IQ was 94 against 91 (P = 0.30). No other neurocognitive or pregnancy outcome differed [8].
Population: 677 women with subclinical hypothyroidism and 526 with low free T4, starting at a mean 16.7 to 17.8 weeks of pregnancy, with children followed for 5 years [8].
Sources: [8]
Caveats: One trial. Treatment started at 8 to 20 weeks, so it does not describe treatment from conception. Funded by the National Institutes of Health [8].

LevothyroxineLimited

Weight loss in obesity

Limited evidence: small studies could not establish a consistent effect of T3 or T4 on weight loss in obesity, and the therapy induced subclinical hyperthyroidism.

The numbers and sources

Effect: During calorie restriction in people with obesity, T3 lowered TSH and T4, but consistent effects of T3 or T4 on weight loss, protein breakdown, metabolic rate or heart rate could not be established. The review found the therapy induced subclinical hyperthyroidism [9].
Population: RCTs and prospective controlled studies of T3 or T4 in obesity [9].
Sources: [9]
Caveats: Few small studies of poor quality, with variable endpoints [9].

Thyroid peroxidase antibodies (TPO antibodies)Limited

Differences between assays

One study found upper reference limits for TPO antibodies from 6.4 to 28.7 IU/mL across five automated assays run on the same men.

The numbers and sources

Compared with: Four other automated immunoassays run on the same blood samples
Effect: Using the reference group that the National Academy of Clinical Biochemistry guideline specifies, the upper reference limits in men and women were 28.7 and 29.0 IU/mL on Immulite 2000 XPi, 24.6 and 25.4 on Maglumi 2000 Plus, 6.4 and 6.9 on Kryptor Compact Plus, 8.3 and 10.0 on Phadia 250 and 14.2 and 17.9 on Liaison XL. These were lower than the limits stated by the manufacturers, except for Liaison XL in women, by 11.3% to 66.8%. The five methods did not compare well with each other, although all were standardized against the same international reference preparation [2].
Population: 120 men and 120 women under 30 with TSH of 0.5 to 2.0 mIU/L, a normal thyroid ultrasound and no thyroid or other autoimmune disease [2].
Sources: [2]
Caveats: One Italian study of 240 people. The authors attribute the differences to the TPO antigen preparations the assays use [2]. A result is read against the cutoff of the assay that produced it.

Thyroid peroxidase antibodies (TPO antibodies)Limited

Thyroid disease when only thyroglobulin antibodies are positive

One large US survey found thyroglobulin antibodies without TPO antibodies were not significantly associated with thyroid disease, while TPO antibodies were.

The numbers and sources

Effect: In the NHANES III survey, thyroglobulin antibodies were positive in 10.4% and TPO antibodies in 11.3% of people who reported no thyroid disease. TPO antibodies were significantly associated with hypothyroidism or hyperthyroidism and thyroglobulin antibodies were not. Thyroglobulin antibodies alone, in the absence of TPO antibodies, were not significantly associated with thyroid disease [16].
Population: 17,353 people aged 12 and older representing the US population from 1988 to 1994, of whom 16,533 reported no thyroid disease [16].
Sources: [16]
Caveats: One cross-sectional survey, using the assays of its time. Positive antibodies were more common in women, rose with age and were less common in Black participants (4.5%) than in white participants (12.3%) [16].

Thyroid stimulating hormone (TSH)Limited

Time of day

One large study found the upper limit of the TSH reference interval shifted with the hour of the blood draw, because TSH rises overnight.

The numbers and sources

Compared with: Samples drawn at other hours of the day
Effect: In 465,593 TSH measurements from people without thyroid disease, the reference interval varied significantly by hour of day, age, sex and ethnicity, and not by time of year. The lower limit stayed constant, while the upper limit rose from 6.45 to 7.55 mIU/L with age, mainly through a larger night-time surge [31]. A review of the reference range debate states that early morning values are greater than later ones, and that the pattern is accentuated by sleep deprivation, strenuous exercise and night or evening shift work [10].
Population: 465,593 TSH measurements from people aged 1 to 104, on a single immunoassay method [31].
Sources: [31], [10]
Caveats: A retrospective analysis of laboratory records from one health system and one assay. The abstract does not give the size of the difference between hours [31].

1 more

Urinary iodineLimited

One sample against a person's usual level

Small studies found a spot result varied about 38% within one person, so 10 or more samples were needed to place an individual within 20%.

The numbers and sources

Compared with: The same person's results across repeated collections over 13 to 15 months
Effect: In women who collected repeated samples for 15 months, the within-person coefficient of variation was 38% for spot urinary iodine concentration and 32% for iodine measured in 24-hour collections, and 10 spot or 24-hour samples were needed to estimate one woman's iodine status with 20% precision [9]. In men sampled monthly for 13 months, single spot results ranged from 10 to 260 µg/L while each man's own average ranged from 30 to 87 µg/L. Twelve or more samples were needed for 20% precision in an individual, and the authors conclude that fewer than ten samples in an individual may be misleading [10].
Population: 22 women aged 52 to 77 in Switzerland, with 177 spot and 341 24-hour samples [9]. 16 men living in an area of mild to moderate iodine deficiency [10].
Sources: [9], [10]
Caveats: Two small studies, of 22 women and 16 men. Both conclude that a single sample does not establish an individual's iodine status, which matches WHO's statement that the test is for populations [1]. WHO attributes part of the variation between samples to hydration [3].