The thyroid result you are about to treat may already have corrected

Design

secondary analysis of the placebo arms of two parallel multicentre randomised placebo-controlled trials, with monthly thyroid testing

Population

326 pregnancies with subclinical hypothyroidism and 254 with hypothyroxinaemia, identified before 20 weeks of gestation and untreated

Primary outcome

proportion with at least one subsequent normal thyroid function test during the pregnancy, and progression to overt hypothyroidism

Effect

88.7% vs 61.4% had a later normal result (P<0.001); overt hypothyroidism in 8/326 (2.5%) and 5/254 (2.0%); no association with thyroid peroxidase antibody status

Two parallel placebo-controlled trials screened pregnancies before 20 weeks and randomised women found to have subclinical hypothyroidism (raised thyroid-stimulating hormone with normal free thyroxine) or hypothyroxinaemia (normal TSH with low free thyroxine) to levothyroxine or placebo, to see whether treatment improved offspring intelligence at five. This secondary analysis asks a different and more immediately useful question: what happened to the thyroid tests of the women who were not treated. Monthly testing was completed by 96.4% and 97.3% of the two placebo groups.

Most reverted. At least one subsequent normal thyroid function test occurred in 88.7% of the subclinical hypothyroidism group and 61.4% of the hypothyroxinaemia group (P<0.001 for the difference between groups). Progression to overt hypothyroidism during the pregnancy was uncommon in both - 8 of 326 (2.5%) and 5 of 254 (2.0%). Thyroid peroxidase antibody status, the marker most often used to decide who to treat, showed no association with subsequent results in either group.

The implication for screening is direct. A single abnormal thyroid test in early pregnancy identifies a group in which most results will normalise without treatment, which means that a screen-and-treat strategy treats a large number of women whose biochemistry was going to correct anyway. That does not settle whether treatment helps those who stay abnormal - it says the first abnormal result is a poor basis for a decision. Repeat it before starting levothyroxine, and do not let a positive antibody result substitute for the repeat.

• Repeat an abnormal early-pregnancy thyroid test before starting levothyroxine - most normalise.
• Do not use thyroid peroxidase antibody status to decide who needs treatment; it did not predict who normalised.
• Progression to overt hypothyroidism was uncommon at about 2% in both groups, so the urgency is lower than it feels.
• Hypothyroxinaemia normalised less often than subclinical hypothyroidism, 61.4% against 88.7% - the two are not equivalent.
• Anyone with known thyroid disease, previous thyroid surgery or existing replacement is outside this analysis entirely.

The statistics, in plain English

These proportions come from the placebo arms of randomised trials, which makes them a clean description of untreated natural history rather than an observational sample selected by who happened to be retested. The measure is 'at least one subsequent normal result', not 'sustained normality', so it describes fluctuation around a threshold as much as true resolution - which is exactly the argument against acting on a single value. The absence of an association with thyroid peroxidase antibodies is a null finding in 580 women; it argues against antibody status being a strong predictor, not against it having any relevance.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.