Are Live-Birth Rates Reduced in Euthyroid Women with Thyroid Autoimmunity Treated with an Assisted Reproductive Technology? The Janus Face of a Meta-Analysis

The association between thyroid autoimmunity (TAI) and miscarriage was first described more than three decades ago, ¹ and thereafter confirmed in many original studies and meta-analyses performed in spontaneous and assisted pregnancies (cf recent review by Bucci et al). ² In the meta-analysis by Busnelli et al, published in this issue of Thyroid, the authors report an increased miscarriage rate (MR) and a reduced live birth rate (LBR) in euthyroid women with TAI who were treated with assisted reproductive technology (ART). ³ These observations confirm the results of their previous meta-analysis on the same topic. ⁴ The other main results in this meta-analysis are the decreased implantation rate (IR), and the absence of an association with ovarian reserve markers (anti-Mullerian hormone, follicle stimulating hormone levels, and the antral follicle count) in women with TAI.

In their previous meta-analysis on this topic in 2016, the authors reported that the odds ratio (OR) for LBR with TAI was 0.73 [95% confidence interval (CI): 0.54–0.99]; p = 0.04, values that are quite similar in this study, although the results now show a smaller CI, and thus resulting in a lower p-value; OR 0.73 [CI: 0.56–0.94]; p = 0.02. However, an important difference between both analyses for the LBR results is that the I² index increased from moderate (41%) to important (71%). The I² index is the percentage of variation across studies due to heterogeneity rather than chance and is one of the other faces of a meta-analysis when more studies are included.

It can be due to statistical or clinical issues. ⁵ Concerning the latter, in the field of ART and thyroid, this can be the type of ovarian stimulation (OS) and subfertility, women's age and race/ethnicity, the definition of TAI and euthyroidism, and finally the type of ART used. The cause of subfertility can be a source of heterogeneity, since ovarian and idiopathic causes have been associated with a higher prevalence of TAI compared with other female and male causes. ⁶ The diagnosis of TAI is also inherent to differences among studies. Many variables have been associated with high(er) thyrotropin (TSH) levels beyond positive thyroid peroxidase antibodies (TPOAb), such as age, body mass index (BMI), and the ethnic background of the women. ⁷

Moreover, the prevalence of TAI has also been shown to be different according to the women's background, a variable that is not always considered in studies. ⁸ The definition of TAI is based mainly on the presence of increased TPOAb levels alone. However, increased thyroglobulin antibodies (TgAb) levels might be important as well, since in one study focusing on subfertile women, it was reported that 5% of them had only positive TgAb compared with 4% with isolated TPOAb. ⁹ This means that in some studies women may have been misclassified. Euthyroidism has been defined differently in all studies included in the meta-analysis; some gave a mean ± standard deviation levels, a median (interquartile range), the reference range of normality of the assay used, and at least two studies did not even describe euthyroidism.

ART is the common term used for three different techniques, a less demanding one (intrauterine insemination with or without a slight OS), the classical in vitro fertilization (IVF) and finally, the intracytoplasmic sperm injection (ICSI), which was mainly used in case of male subfertility, but during the past decade also for other types of subfertility, depending on the experience of the biologist in the concerning center. The authors explored for the source of heterogeneity by performing subgroup analyses for women with TPOAb measured alone and according to the type of ART (respectively) as well as performing and a meta-regression analysis investigating the impact of age and serum TSH on the outcomes.

The results of the secondary analyses for age, TSH, and the measurement of TPOAb alone did not explain the observed heterogeneity. However, concerning the use of ICSI as ART type, TAI was not associated anymore with a lower IR, higher MR, or lower LBR. These results show again another face of the meta-analysis and are as such comparable with those of a meta-analysis by Poppe from a few years ago, in which studies were included of women treated with ICSI only. ¹⁰ Therefore, a question that arises is whether the results of this secondary analysis add evidence to propose ICSI to all women of subfertile couples with TAI? No, it does not.

For the time being, I believe it remains prudent to follow the suggestion given in the recent European Thyroid Association (ETA) guideline on the management of thyroid disorders in women planning ART, proposing to offer ICSI in certain women with TAI. ¹¹ Reasons why the current meta-analysis does not change clinical management are, on one hand, because in the ICSI subgroup analysis only one or two studies were included, and on the other hand, as discussed by the authors and in the meta-analysis by Poppe, ¹⁰ another type of study design is requested. In there is a need for prospective head-to-head trials comparing ICSI with IVF, performed in women with TAI and idiopathic or male subfertility that are matched for age, BMI, TSH, and other relevant covariates. These kinds of studies need to be performed in an international collaboration to overcome statistical power issues, and with respect to ethical issues involved in this research topic.

Furthermore, and to come to a better understanding of the results of studies on ART pregnancies, the embryo quality, endometrial receptivity, and the male factor in the whole process of reproduction should be investigated more in detail. ¹²

However, the negative results observed in the subgroup analysis make sense considering two negative interventional studies with thyroid hormone in subfertile euthyroid women with TAI. ^13,14 Finally, what is now the moral of the Janus-faced meta-analysis story? Reading an abstract is easy, but going into the details of the (supplementary) results and having some knowledge of statistics and clinical context are even better.