Sex Ratio at Birth Is Associated with First-Trimester Maternal Thyrotropin in Women Receiving Levothyroxine

Introduction

The human sex ratio at birth, defined as the proportion of male to total live births, is historically set at 0.515 (1,2), but it is affected by multiple factors, both environmental and medical. Environmental factors described to date include latitude (1), maternal weight and parity, periconceptional smoking (3), pollutants, and disasters such as exposure to dioxins after the chemical plant explosion in Seveso (2) or the September 11 attacks in the United States (4).

Many medical factors are known to affect the human sex ratio. On one hand, the ratio increased in the early 1900s as the result of improvements in obstetric care. This increase probably occurred because male fetuses, being more vulnerable to adverse conditions, benefited more from this better care (5). On the other hand, the sex ratio appears to decrease in the presence of parental diseases such as hepatitis or lymphoma (6). Within endocrinologic disorders, the sex ratio seems to be lower in mothers with congenital adrenal hyperplasia (7), and also appears to be influenced by the type of assisted reproduction (8,9). Data concerning sex ratio and diabetes, however, are conflicting (10), and data regarding thyroid conditions are scarce. Bernstein (11) reported in the 1960s that sex ratio was decreased in families with one or more children with Graves' disease. More interesting is the article of Haddow et al. (12) addressing neuropsychological outcomes in offspring of pregnant women with thyrotropin (TSH) >6 mUI/L in the second trimester; in their data, the sex ratio displayed a trend toward being lower in mothers with hypothyroidism (0.417 vs. 0.526 in the remainder of the cohort), an aspect that was not further discussed by the authors.

The aim of the current study was to assess whether thyroid autoimmunity and/or the degree of compensation of maternal hypothyroidism were associated with the sex ratio at birth. Our hypothesis was that inadequate thyroid compensation and/or positive autoimmunity would be associated with a decreased ratio.

Materials and Methods

We studied a cohort of 167 women treated during pregnancy at the Endocrinology and Pregnancy Clinic of the Hospital de la Santa Creu i Sant Pau in Barcelona who had received treatment with levothyroxine already before pregnancy (delivery between 1 January 1986 and 31 December 2010). We performed a retrospective cohort analysis using the ongoing database of the clinic that is prospectively collected. The Ethics Committee of the center approved the study. STROBE guidelines have been used for preparing this report.

Inclusion criteria were available information on maternal autoimmunity status, first-trimester TSH, and fetal sex. Exclusion criteria were secondary/tertiary hypothyroidism, pregestational diabetes, multiple pregnancies, and fetal losses (miscarriages, stillbirths, ectopic pregnancies, and pregnancy terminations). TSH was measured at the first visit and at 6–8-week intervals depending on whether the results were abnormal or normal, respectively. When autoimmunity status before pregnancy was not known, antibodies against thyroid peroxidase and thyroglobulin were measured.

Different TSH and antibody methods have been used during the observation period. The TSH methods are, in chronological order, as follows: a double-antibody radioimmunoassay (Diagnostic Products Corp., Los Angeles, CA; reference value <1–4.7 mUI/L), an immunoradiometric assay (RIA-gnost hTSH; Behringwerke AG, Marburg, Germany; reference value 0.3–4 mUI/L), an automated immunoassay system (ACS:180; Ciba Corning Diagnostics, Medfield, MA; reference value 0.25–5 mUI/L), an electrochemiluminescent immunoanalysis (Roche Modular Analytics E 170; Roche Diagnostics, Basel, Switzerland; reference value 0.3–5.0 mUI/L), and a chemiluminescent microparticle immunoanalysis assay (Abbott Architect c/16200; Abbott Laboratories, Abbott Park, IL; reference value 0.3–5.0 mUI/L).

We considered the following variables as potential predictors of newborn sex: maternal age, body mass index, gravidity, parity, prior treatment with radioactive iodine, smoking status at the beginning of gestation, autoimmunity status, and first-trimester TSH.

Autoimmunity was classified as positive or negative. Mean maternal first-trimester TSH was calculated and classified into five groups: I <0.10, II 0.10–2.49, III 2.5–4.99, IV 5.00–9.99, and V ≥10.00 mUI/L, cutoffs corresponding to generally accepted TSH limits outside pregnancy and the recommended upper cutoff for TSH in the first trimester of pregnancy (13).

A bias may exist because the patients were attended/have been referred to a tertiary care center. No action was taken to limit it.

The study was planned as an exploratory one and no sample size calculation was performed. The normal distribution of quantitative variables was examined using the Kolmogorov–Smirnov test. Because all quantitative variables were non-normally distributed, they were expressed as median (range) and Mann–Whitney U-test was used to compare groups. A chi-square test was used to compare the observed sex distribution versus the expected and to analyze the rate of qualitative variables according to sex. As the expected sex ratio, we used 0.514, which is the sex ratio observed in the city of Barcelona in the study period (14).

A multiple logistic regression analysis (backward method) was used to predict newborn sex using the above-mentioned variables as potential predictors. Sensitivity analyses were performed grouping gravidity and parity as qualitative variables and using mean first-trimester TSH as a quantitative and qualitative variable (five abovementioned categories).

Results

Patient characteristics are shown in Table 1. Primary hypothyroidism caused by autoimmune or non-autoimmune etiologies was the most prevalent thyroid disease and was present in 44.9% of mothers. It was followed by hypothyroidism after treatment for Graves' disease (33.5%), differentiated thyroid carcinoma (18.6%), and surgical treatment of benign conditions (3.0%). Autoimmunity was positive in 75.4% of mothers and mean first-trimester TSH was 2.9 (0.02–24.96) mUI/L. Information for every variable was available in more than 98% of the patients. Maternal characteristics of male and female newborns were comparable with the exception of mean first-trimester TSH, which was higher in mothers of female newborns (3.27 vs. 2.52 mUI/L; p=0.025). The distribution of TSH categories according to newborn sex was shifted toward a higher rate of upper TSH categories in mothers of female fetuses (p<0.021).

When these rates of male and female newborns are expressed in terms of sex ratio at birth, the resulting figure in the 167 pregnancies was 0.485 [95% confidence interval (CI) 0.408–0.561], not statistically different from the expected 0.514. However, the sex ratio differed according to TSH category (Table 1). In mothers with first-trimester TSH ≥5 and <10 mUI/L, the sex ratio was 0.310 [CI 0.142–0.479] and in those with TSH ≥10 mUI/L it was 0.200 [CI 0.000–0.402], both significantly different from the expected (p<0.026 and 0.010, respectively).

The multiple logistic regression analysis confirmed first-trimester maternal TSH as the single predictive factor of newborn male sex (odds ratio 0.900 [CI 0.823–0.984], p<0.020). The results of the sensitivity analyses were concordant (data not shown).

Discussion

To our knowledge, this is the first report describing an association between maternal first-trimester TSH and sex ratio at birth; we report a decrease in mothers with mean first-trimester TSH ≥5 mUI/L.

A trend toward a decreased sex ratio was already present in the hypothyroid group of the seminal article of Haddow et al. (12). We consider that the data of Haddow et al. (although using TSH measurements in the second trimester) support the results presented here. Could the association be causal? We cannot demonstrate that an insufficient compensation of maternal hypothyroidism reduces sex ratio, but that a male fetus shifts the maternal TSH to higher values than a female one appears more unlikely. Specifically, in pregnancies of male fetuses, maternal hormones are shifted to lower concentrations of estradiol (15), which should produce an effect on maternal TSH contrary to the one observed (16).

If we contemplate the possibility that maternal hypothyroidism has a causal influence on sex ratio at birth, we can consider two mechanisms. The first one would be that maternal hypothyroidism induces hormonal changes favoring the conception of a female fetus. Several reports support that the hormonal environment around the time of conception could influence fetal sex, arguing that high concentrations of estrogens and testosterone would increase the probability of a male fetus, while high concentrations of gonadotropins would increase that of a female (17). A second mechanism would be a direct action of low thyroxine and/or high TSH on male fetuses, reported to be more vulnerable to unfavorable conditions (18). We would favor the second mechanism: the association of first-trimester TSH with sex ratio would be the result of thyroid function itself, since the known influences of hypothyroidism on sex steroids and gonadotropins are not those favoring the conception of a male fetus (19). On the other hand, Van Vliet et al. reported in 2003 (20) that male newborns with congenital hypothyroidism had higher rates of absent epiphyses at birth and suggested that this could be due to sexually dimorphic expression of deiodinase enzymes responsible for the final effects of thyroid hormones within tissues. Their findings indicate that male fetuses are more susceptible to the harmful effects of hypothyroidism in the last part of pregnancy; we speculate that this could also be true for fetal losses, most of them taking part in early pregnancy. We cannot elucidate if this would be the result of decreased peripheral thyroid hormones or increased TSH, but the first possibility would seem more likely if we consider the importance of thyroid hormones in fetal life (21).

Although our departing hypothesis was that abnormal TSH would be related to a lower sex ratio, we were surprised by the magnitude of the association in women with high mean TSH levels in the first trimester. Nevertheless, this would be in line with the reported effects in other endocrinologic conditions: in one article addressing the effect of maternal diabetes, the sex ratio at birth was 0.311 (22), whereas in congenital adrenal hyperplasia the sex ratio at birth has been reported to be 0.240 (7). The fact that the sex ratio is affected only in the upper TSH categories explains why the overall sex ratio is not affected.

The main strength of this report is the robustness of the association. Its main weakness is that the information comes from a clinical practice database where TSH measurements were not scheduled at specific weeks and different laboratory methods were used to measure TSH and thyroid antibodies. However, a low precision in measurements should decrease the statistical power of the analysis; the strong negative association of the sex ratio with high maternal TSH levels thus favors a strong association in real life. Another weakness of this study is that we did not have information on the use of assisted reproduction techniques; we have attempted to limit the influence of this factor excluding multiple pregnancies.

In conclusion, although specific mechanisms cannot be elucidated, we describe that in women receiving levothyroxine treatment already before pregnancy, there is a negative association between maternal TSH in the first trimester and the sex ratio at birth. An effect of the maternal thyroid status on the sex ratio appears more likely than the reverse. We cannot foresee how these findings could be used to improve clinical outcomes.