Subclinical Hypothyroidism: An Evidence-Based Approach

If you order enough thyroid panels you will eventually find one: TSH slightly above the reference range, free T4 entirely normal. That combination has a name — subclinical hypothyroidism — and it accounts for a large share of thyroid prescriptions written every year.

It is also the area where the gap between what feels obvious and what the trials show is widest. The intuition is that a raised TSH means the gland is struggling, and topping up the hormone should help. Two substantial pieces of evidence say that in most cases it does not.

What the Result Actually Means

TSH is a pituitary signal, not a thyroid hormone. When the pituitary senses slightly less thyroid output than it expects, it raises TSH — and a higher TSH usually succeeds in keeping T4 in the normal range. A mildly raised TSH with normal free T4 therefore describes a compensated system, not a failing one.

Three things routinely produce that pattern without any thyroid disease at all:

1. Transient variation. A large fraction of mildly raised TSH results normalise on repeat testing weeks later. Acute illness, recovery from illness, and ordinary day-to-day variability all move the number.
2. Ageing. TSH distributions shift upward with age. A value that is abnormal for a 30-year-old may be squarely within the age-specific range for someone in their late seventies.
3. Assay interference. High-dose biotin — common in hair and nail supplements — interferes with many immunoassays and can distort thyroid results outright. Stopping biotin for a few days before testing removes that variable.

One abnormal TSH is a reason to repeat the test, not to start a prescription.

TRUST: The Trial That Tested the Intuition

The TRUST trial randomised adults aged 65 and over with persistent subclinical hypothyroidism to levothyroxine or placebo, with dose titration to normalise TSH, and followed them for a year. The primary outcomes were the ones that matter to a patient: tiredness and thyroid-related symptoms.

TSH normalised in the treatment group, as it was designed to. The symptoms did not improve. ^[1] Treating the number worked; treating the person did not.

This is a familiar and uncomfortable shape in medicine — an intervention that reliably corrects a laboratory abnormality without changing how anyone feels. It does not prove levothyroxine is useless in every case, but it removes "the TSH is high, therefore treat it" as a default in older adults.

The Meta-Analysis That Confirmed It

A systematic review and meta-analysis published in JAMA pooled randomised trials of thyroid hormone therapy in subclinical hypothyroidism and examined quality of life and thyroid-related symptoms across the literature. The conclusion matched TRUST: thyroid hormone therapy was not associated with improvements in either. ^[2]

One trial can be an outlier. A trial plus a meta-analysis of the surrounding literature pointing the same way is a finding.

When Treatment Is Still Justified

None of this makes levothyroxine the wrong answer everywhere. The evidence narrows the indication rather than abolishing it. Treatment remains appropriate when:

• TSH is above 10 mIU/L. The trials above studied mild elevation. Marked elevation carries a substantially higher rate of progression to overt hypothyroidism, and guidelines continue to support treatment.
• Pregnancy, or planning one. Fetal brain development depends on maternal thyroid status, and this is the one context where the tolerance for waiting is low. Manage it with obstetric care.
• Progressive disease. Rising TSH across serial tests with positive TPO antibodies describes a trajectory, not a snapshot, and the calculus differs from a single borderline reading.
• A younger, symptomatic patient with a clear pattern — where a time-limited trial with a pre-agreed stopping rule is a defensible experiment, provided you actually stop if nothing changes.

The population where the evidence most clearly argues against reflex treatment is older adults with mild, stable elevation and non-specific symptoms.

What to Check Before Anything Else

Fatigue, cold intolerance, weight change, and brain fog are the classic thyroid complaints and are also the classic complaints of half a dozen more common conditions. Before attributing them to a TSH of 5.8:

• Repeat the thyroid panel in 6–12 weeks, off biotin, and include free T4 and TPO antibodies.
• Check ferritin. Iron deficiency without anaemia produces the same symptom list and is far more common, particularly in menstruating women.
• Check vitamin D, which is frequently low in the same patients — see the vitamin D dosing guide.
• Audit sleep, energy intake, and training load. Sustained calorie restriction lowers T3 as an adaptive response. The remedy is eating more, not levothyroxine.

Where Nutrients Fit

If antibodies are positive, selenium is the one supplement with real trial data behind it: meta-analyses show it lowers thyroid peroxidase antibody titres. ^[4] But the 2024 Thyroid meta-analysis is explicit that evidence for patient-relevant outcomes remains limited ^[3] — the same lesson TRUST teaches, one level down. Benefit also concentrates in people whose selenium status is actually low. ^[5]

Iodine deserves the opposite emphasis. In autoimmune thyroid disease, supplemental iodine beyond dietary intake can precipitate hypothyroidism rather than correct it. ^[6] A raised TSH is not an indication to start kelp.

The nutrient layer, with doses and contraindications, is set out in the thyroid optimization protocol.

The One-Line Version

Subclinical hypothyroidism is a laboratory pattern that usually needs confirmation, context, and patience rather than a prescription. TRUST and the JAMA meta-analysis both found that normalising the number failed to change how patients felt. Repeat the test, check the antibodies, rule out iron and vitamin D, and reserve levothyroxine for TSH above 10, pregnancy, and documented progression.

Related Reading

• Selenium and Iodine: The Thyroid Mineral Pair
• The Thyroid Support Protocol
• Vitamin D Dosing: Why Most People Get It Wrong