The present study calculated the trimester-specific reference intervals for TSH, FT4 and FT3 in a carefully selected cohort of healthy Caucasian women living in a mild iodine deficiency area, in north-eastern Italy. To the best of our knowledge, this is the first longitudinal study in pregnancy and the post-partum that has assessed the reference intervals not only of TSH and FT4 but also of FT3, carried out with the Roche platforms.

The enormous amount of work and resources needed for a longitudinal design underpins the scarce number of prospective studies published. Smaller intra-individual variation of TFP during pregnancy has been reported by several authors [21,22,23,24,25,26] who advocated the development of longitudinal TFP reference intervals, although, a few studies [27, 28] did not report significant differences when comparing the self-sequential longitudinal and cross-sectional reference interval of thyroid function tests in pregnancy.

We found that there is a modest, but significant increase in TSH concentration in the second trimester of gestation, which remains stable thereafter in the third trimester, the median values being respectively 7.0% and 6.6% higher than in the late first trimester. Conversely, there was a clearer progressive reduction during pregnancy of FT4 and FT3 concentrations, the median values being 12.1% and 11.6% lower in the second trimester, and 14.4% and 13.2% lower in the third trimester, as compared with the first trimester, respectively. However, serum TSH, FT4 and FT3 concentrations were similar in the late first trimester and after the end of pregnancy.

Thyroid dysfunction is a frequent finding during pregnancy, which may have relevant medical implications [1, 3, 7]. It is noteworthy that even small variations in the thyroid function may be associated with adverse effects on several important pregnancy outcomes, including low birthweight and miscarriage risk. However, assessment of this condition requires the availability of appropriate reference intervals for thyroid function tests.

In most cases, TSH is considered the most important parameter in the assessment of thyroid function. However, the reference intervals for TH may also be clinically relevant for distinguishing an isolated thyroxine deficiency, a phenomenon possibly associated with potential unfavourable outcomes of pregnancy, and especially for properly diagnosing and managing any conditions of glandular hyperfunction. Indeed, the latter diagnosis may be challenging, due to the interfering effects of hCG, in the initial phase of gestation, and to the changes in FT4 and FT3 reference intervals during pregnancy. It is important to underline that pharmacological treatment of thyrotoxicosis is recommended in pregnant women when TH are increased, but not when there is only a suppressed serum TSH [8].

It should be emphasized that the differences in reference intervals of thyroid function parameters can be linked both to the methods used and to population-specific factors. It is therefore important to define reference intervals that are at least population and assay (analyzer and reagent) specific. This information is urgently needed from both a theoretical and practical point of view. While guidelines recommend the assessment of TFP in pregnancy using trimester- and instrumentation/reagents-specific reference intervals, this occurs very rarely in clinical practice, where clinical laboratories and clinicians usually adopt the intervals suggested by the manufacturers.

Medici et al. [29] and McNeil et al. [30] have highlighted in 2015 the heterogeneity of available studies in terms of methods, populations, iodine sufficiency status and gestational weeks of examined subjects. Interestingly, the first trimester TSH upper reference limit (URL) of the studies reviewed by McNeil et al. [30] fell into two groups: the mean of the values reported by the authors using Abbott, Beckman and Immulite-Siemens assays were around 3.0 mU/L (respectively 3.00, 3.12 and 3.09 mU/L), whereas the mean of the values reported by the authors using Centaur-Siemens and Roche were closer to 4.0 mU/L (respectively 3.55 and 4.00 mU/L).

The results reported in available longitudinal studies, summarized in Table 4, demonstrated different trends in TSH concentration: according to some studies, TSH is stable throughout pregnancy, according to others an increasing trend is recognizable. Conversely, a decreasing trend of FT4 (and FT3) concentration is more consistently reported.

In 2021 a systematic review of published studies on TSH and FT4 reference intervals in pregnancy obtained using Abbott, Beckman, Roche and Siemens assay methods, including 139,734 pregnant women, was conducted [31]. It is noteworthy that, in the first trimester, TSH upper limits obtained with the Abbott system ranged from 2.33 to 8.30 mU/L, those obtained with Siemens from 2.83 to 4.65 mU/L, whereas FT4 higher limits ranged from 13.2 to 18.7 pmol/L with the Beckman system, and from 16.7 to 26.5 pmol/L with the Siemens method. The TSH upper limit in the first trimester differed from non-pregnant concentrations, and could not be predicted or extrapolated from non-pregnant values.

A large variation in reference limits within the same assay, and the lower FT4 reference intervals using Beckman assay compared to the other assays were confirmed in the most recent systematic review and meta-analysis carried out by Osinga et al. [32] including 63,198 pregnant women. These authors also reported that about 15% of the studies included in their systematic review narrowed the 2.5th to 97.5th percentile reference intervals (mostly to the 5th to 95th percentile) and observed that in this case the upper limit of TSH decreased substantially (− 0.63, − 0.65, and − 0.73 mU/L in the first, second, and third trimester, respectively), with a considerable potential increase in the number of women diagnosed with subclinical hypothyroidism. However, in our opinion, the 2.5th to 97.5th percentile reference interval, which is adopted by most laboratory professionals, manufacturers and clinicians, and is recommended by current CLSI standards, should still be preferred to the 5th to 95th percentiles, to avoid risk of overdiagnosis and overtreatment of pregnant women, in the present absence of clear evidence that this change may lead to advantages in the clinical setting. Interestingly, the robust analysis of our data (Table 1) showed a minor effect on the upper limit of TSH (+ 0.14, − 0.05, and − 0.06 mU/L, in the first, second, and third trimester, respectively). These small differences may be due to the distinctive characteristics of our cohort, entirely constituted by carefully selected healthy women. Anyway, coupling robust and non-parametric methods could be an effective way to assess the effect of TSH right skewed distribution, without increasing dishomogeneity of the TFP reference interval studies.

In the present study TSH URL was around 4 mU/L throughout all pregnancy. As shown in Table 4, this finding is consistent with other studies carried out using Roche.

It is noteworthy that iodine intake has been poorly investigated in the past. Indeed, urinary iodine has not been measured in many studies summarized in Table 4 [14,15,16, 19] or, when measured, has not been assayed with reference technology [13, 20]. In studies reported in the review by Medici et al., it was sufficient only in two of the investigated cohorts and mild-moderately insufficient in all other studies [29]. In our cohort a mild iodine deficiency was found. It should be noted that the WHO-recommended thresholds of urinary iodine concentrations can only be used on a population basis, whereas these values are hardly applicable to assess the iodine status of the individual, due to the large variability of urinary iodine excretion. However, according to recent reports, reference limits are not significantly affected by iodine deficiency, when including mild to moderate iodine-insufficient participants (urinary iodine secretion 50–149 µg/L) [32].

In the present study, TSH lower reference limits in the first trimester were higher than those reported by other authors (Table 4). This may be accounted for by the relatively late phase of the first sampling in subjects enrolled in our study (14–16 gestational weeks), which differed from other studies. A reduction of serum TSH might be expected in very early pregnancy under the effect of high hCG levels. Consistently with this phenomenon, it should be noted that in our study the 90% CI of the low reference limit for first trimester TSH ranged between 0.07 and 0.550. The relatively late recruitment of pregnant women in our study could also potentially explain the differences between the TSH reference intervals calculated in our study and those reported in another longitudinal study carried out using Roche in a Caucasian population, in Poland [22], although the gestational weeks in which blood samples were obtained were not detailed in the latter study. However, it is noteworthy that, in this study, the lower reference limits of TSH remained unusually low throughout pregnancy (0.05 and 0.11 mU/L in the second and third trimester, respectively) despite the corresponding FT4 and FT3 reference limits being similar to our findings. Differences in TSH values between these studies are not easily explained. Unfortunately, this study did not report the reference intervals obtained in these women after the end of pregnancy.

Apart from TSH, changes during pregnancy in TH concentrations must be considered in order to avoid the potentially harmful misinterpretation of clinical findings. Indeed, in our study both FT4 and FT3 concentrations showed a progressive reduction during physiological pregnancy, the median values being about 11.5–12% lower in the second trimester, and 13–14% lower in the third trimester, as compared with values measured in the first trimester and in non-pregnant subjects.

The strengths of our study are the careful selection of subjects (healthy women with a physiological pregnancy and healthy newborns), the longitudinal collection of blood samples during the different trimesters of pregnancy in the same individuals, and the comparison of pregnancy data with hormone concentrations after the end of pregnancy, in a subgroup of these women. It should be underlined that studies investigating TFP reference interval in pregnancy have been rarely carried out longitudinally in carefully selected women, and only a few of them were methodologically accurate and complete. Notwithstanding the TSH not normal distribution of values, the reference limits obtained in the three trimesters using non-parametric analysis were substantially confirmed using robust method, scarcely affected by skewness.

A limitation of the study is the small size of the sample. However, the CLSI EP28-A3c document endorses for nonparametric method to collect samples from a number of qualified reference individuals sufficient to yield a minimum of 120 samples. A further limitation, as it is for any published studies in this field, is that the reported reference limits could be appropriate in laboratories that serve the Caucasian population by using Roche analyzers, but not in laboratories that serve a different population or the same population using analyzers of other manufacturers.