Subclinical Thyroid Dysfunction in the First Trimester of Pregnancy

Disease' Versus Physiological (Pulsatile) Variation in TSH Concentrations

Krzysztof C. Lewandowski; Karolina Garnysz; Wojciech Horzelski; Joanna Kawalec; Karolina Budzen; Mariusz Grzesiak; Andrzej Lewinski


Clin Endocrinol. 2020;93(6):739-745. 

In This Article


Our study clearly demonstrates that regardless of the 'cut-off point' defining subclinical thyroid dysfunction, significant number of woman might be labelled, as having some form of 'thyroid disease', purely as a result of physiological variation in TSH concentrations. This discrepancy is less striking for subclinical hyperthyroidism [eg five vs six women (out of 110) for TSH <0.1 mIU/L, for aTPO-negative women], but is highly pronounced for otherwise healthy pregnant women, who might be labelled as having 'subclinical hypothyroidism'. Hence, in aTPO-negative women, without previous history of thyroid disease, the prevalence of 'subclinical hypothyroidism' would vary from about 21% to 8% for TSH 2.5 mIU/L cut-off. During pregnancy, in contrast to nonpregnant subjects, there is a tendency to instigate treatment, if some laboratory parameter is outside the reference range and treatment is relatively cheap and easy available (such as levothyroxine). This implies that if a 2.5 mIU/L cut-off point is applied, then almost 8.2% of pregnant women might be subjected to unnecessary treatment, purely as a result of pulsatile nature of TSH secretion. This percentage would be even higher (about 13%) for aTPO-negative women during first trimester of pregnancy. Given the total number of deliveries per year in countries, like the UK or Poland, this would mean that tens of thousands of otherwise healthy women would be subjected to unnecessary levothyroxine treatment, several additional clinic visits and blood tests, if only a single TSH measurement is taken into account. We can speculate that determination of 'normality' would not be easier for aTPO-positive women, who in our study had slightly higher TSH concentrations, but due to their small number (n = 11), it was not possible to draw any valid conclusions. Application of mean TSH value out of five measurements may be potentially useful, as it helps to determine how many pregnant women remained either below or above the selected cut-off point for most of the investigated time span. This was particularly visible for upper TSH cut-off of 4.0 mIU/L (ATA 2017 guidelines[4]) or 4.2 mIU/L (upper assay reference range) for aTPO-negative women (Table 2), but such approach remains to be validated.

As mentioned above, there is no consensus regarding the TSH reference range in pregnancy. For instance, in a recent study from China, based on a data from 46 262 pregnant women[9] the authors suggested a TSH reference range of 0.03–3.52 mIU/L, for aTPO-negative women, while among aTPO-positive women (n = 4628, ie about 10% of the total), first-trimester TSH 'reference range' was 0.05–4.57 mIU/L. Similar observations suggesting TSH reference range of 0.005–3.65 mIU/L are based on data from Turkey (n = 1258).[10] Such a reference range would also effectively include all our women with subclinical hyperthyroidism, with exception of a single patient with TSH <0.005 mIU/L and raised anti-TSH receptor antibodies, who had incidentally diagnosed Graves's disease. Data from the Czech population (n = 4965), that is ethnically similar to Polish, suggest first-trimester TSH reference range of 0.16–3.43 mIU/L and 0.02–2.95 mIU/L for singleton and twin pregnancies, respectively.[11] For these reasons, some authors point out that previously suggested 'fixed' TSH cut-off point of 2.5 mIU/L seems to be too low.[12] Available Polish data, though based on a smaller number of women (n = 172), suggest first-trimester TSH reference range of 0.009–3.177 mIU/L.[13] For completeness, it should be mentioned, however, that certain unfavourable processes, such as increased lipid peroxidation[14] or lower concentrations of mannan-binding lectin,[15] can be observed for TSH concentrations above 2.5 mIU/L, even in nonpregnant female population. This issue is, however, further complicated by recent data[16] showing that levothyroxine treatment improved miscarriage rate in the first trimester for TSH 2.5–4.08 mIU/L, but later resulted in doubling the rate of gestational diabetes. On the other hand, our data demonstrate that even if we adopt a higher TSH 'cut-off' point (ie >4.0 mIU/L), then still five out of six aTPO-negative women would have TSH either above or below 4.0 mIU/L depending on which of five consecutive TSH values are taken into account. It should be remembered, however, that according to Recommendation 1 of ATA 2017 guidelines,[4] 'reference range determinations should only include pregnant women with no known thyroid disease, optimal iodine intake, and negative TPOAb status', while recent data[17] show a significant increase in obstetric complications for women with TSH >4.0 mIU/L.

In our opinion, interpretation of any 'borderline' TSH concentrations should also account for physiological variation of TSH before a diagnosis of a 'disease' is made. Lowest TSH concentrations are observed during the afternoon, followed by a rise during the evening and a maximum in the first part of the night. In fact, diurnal pattern of TSH secretion slightly resembles the pattern of prolactin, with the difference that peak TSH concentrations occurs at the earlier phase of the night than prolactin.[18] Nocturnal concentrations of TSH might be more than doubled in comparison with daily values. In addition to this diurnal pattern, repeated variably sized bursts of TSH release occur during the 24-hour period, while the mean amplitude of these short-term TSH bursts was reported to oscillate about 13 per cent of the mean level.[19] It is not fully known how TSH pulses are generated though it was demonstrated that constant TRH infusions at various doses did not alter TSH pulse frequency in normal or treated hypothyroid subjects, although pulse amplitude was increased.[20] Unfortunately, there are hardly any data regarding TSH pulsatility in pregnancy. Erikson et al[21] measured TSH, prolactin and cortisol at 30-minute intervals, and demonstrated that TSH diurnal pattern was preserved in pregnancy with maximal TSH concentrations around midnight. The study, however, included only six pregnant women, where only two of them had TSH measured at earlier stages of pregnancy (week 11 and 17). It should be noted, however, that we did not observe any significant difference in TSH variability between pregnant first-trimester women and nonpregnant controls. Adriaanse et al[7,8] demonstrated that in healthy individuals (n = 16), there were 7.8 ± 3.1 TSH pulses between 20.00 and 8.00 hours, and 2.0 ± 1.7 pulses between 8.00 and 20.00. The amplitude of nocturnal pulses (0.48 ± 0.23 mIU/L) was, however, higher than during the day (0.29 ± 0.21 mIU/L). In our study, we selected to measure TSH concentrations between 7.00 and 9.00 hours, that is at the junction between the 'night' and 'day' TSH pattern. The average maximal increase in TSH concentrations (0.38 ± 0.51 mIU/L) fell in between the 'night' and 'day' TSH amplitude as described by Adriaanse et al[7,8] Furthermore, some women displayed much wider variation in TSH concentrations (see Figure 1A-D). In contrast to other immunoassays, such as total testosterone, where coefficients of variation may vary as much as between 17.3% and 34.9%,[22] TSH assay variation is much smaller, for example 4.1% for subclinical hypothyroidism (TSH 6.6 mIU/L), based on a data from a large UK study,[23] and up to 3% for intra-assay variation, as quoted by manufacturer for our assay, so most of the observed variation (23.74% above the mean; see Table 5) can be attributed to TSH pulsatility rather to assay inaccuracy.

We have deliberately chosen the hours between 7.00 and 9.00, as this is the time where blood samples are taken in real-life conditions. In countries, such as Poland, the majority of pregnant women during the first trimester still go to work (or to the University), and fasted blood samples (not only for TSH, but also for other investigations) are taken precisely at that time, that is typically before work begins.

Despite the plethora of papers pertaining to thyroid hormone reference ranges in pregnancy, available studies are effectively based on single thyroid hormone measurement.[9–11,13,16,17] This applies also to affluent European countries, such as Denmark, where thyroid function and aTPO status were recently assessed in 14 323 pregnant women in the first trimester of pregnancy.[24]

From our experience (though limited to our local area only), the majority of our Obstetrics & Gynecology colleagues are not aware of the above-mentioned diurnal variation in TSH concentrations. This raises the question, whether diagnosis of subclinical thyroid dysfunction in pregnancy could be really made just on a single TSH measurement. If we take an example of diabetes mellitus, it should be noted that if glucose concentrations are close to the 'cut-off point', then it is universally accepted that the diagnosis of diabetes should not be based on a single measurement, due to effects of diet or stress hyperglycaemia.[25,26] We postulate that similar approach should be also applied to diagnosis of subclinical thyroid dysfunction in pregnancy, and all 'borderline' results should be confirmed on repeated testing.

In summary, we have demonstrated that physiological variation in TSH concentrations may have a significant impact on the diagnosis of subclinical thyroid dysfunction in pregnancy. As a result, significant number of otherwise healthy women may be subjected to unnecessary treatment, where instead of a 'disease' they would indeed undergo treatment of physiological variation in TSH concentrations. We postulate that a diagnosis of subclinical thyroid dysfunction in pregnancy should not be based on a single measurement, though the number of tests remains yet to be optimized.