Thyroid autoimmunity and IVF/ICSI outcomes in euthyroid women: a systematic review and meta-analysis

Thyroid autoimmunity (TAI) is defined as the presence of antithyroid antibodies – anti-thyroid peroxidase (TPO-Ab) and/or anti-thyroglobulin (TG-Ab) antibodies – and is the most common autoimmune condition in women of reproductive age, with a prevalence of 8–14% [1, 2]. Significantly, TAI has an even higher prevalence of up to 20% in infertile women [3]. Even when thyroid function remains normal, TAI has been associated with adverse pregnancy outcomes including spontaneous miscarriage and preterm birth [1, 4, 5]. Ovarian stimulation used during IVF/ICSI places additional demands on thyroid function [6, 7], which may potentially expose women with TAI (TAI+) to additional risks. Consequently, there has been intense interest in whether TAI compromises outcomes during IVF/ICSI treatment.

Multiple systematic reviews have investigated the relationship between TAI and IVF/ICSI outcomes, however there is substantial discrepancy between findings. Toulis et al. (2010) [8], found that sub-fertile TAI+ women undergoing IVF were at significantly greater risk of miscarriage than women without TAI (TAI-) but did not identify differences in either clinical pregnancy rates (CPR) or delivery rates. Thangaratinam et al. (2011) [9] also found a significantly higher miscarriage rate (MR) amongst TAI+ women. TAI+ women were found to have a greater risk of spontaneous miscarriage in addition to lower live birth rates (LBR) with no reductions in either CPR or implantation rates (IR) [6]. On the contrary, He et al. (2016) [10] found no association between TAI positivity and clinical pregnancy, miscarriage rate or delivery rate. Likewise, Leiva et al. (2017) [11] found no association between TAI and increased risk of pregnancy loss.

For a number of reasons, it is important that this topic be revisited. Firstly, earlier meta-analyses combined outcomes that were defined differently in different studies. For instance, some papers considered miscarriage to include both clinical and biochemical pregnancy losses (BPL) (e.g. Muller et al. 1999 [12], Poppe et al. 2003 [13]) whereas others included only clinical losses (e.g. Chai et al. 2014 [14]; Lucaszuk et al. 2015 [15]). In the most recent comprehensive meta-analysis on the effect of TAI on IVF outcomes, all miscarriage types, whether biochemical or clinical were pooled into a single analysis [6]. For uniformity, it is critically important to analyse each miscarriage type separately since biochemical and clinical losses have very different frequencies and aetiologies [16]. Biochemical pregnancy loss in IVF is estimated to have an incidence of up to 22% [17], compared to clinical miscarriage which has an incidence of ~ 15% [18, 19]. A number of mechanisms have been proposed by which, TAI might increase miscarriage risk including detrimental effects on the endometrium, embryo toxicity, general immune disturbance and compromised ability of thyroid function to satisfy increased demands of pregnancy [2]. It is conceivable that very early post-implantation pregnancy stages might have a different vulnerability to these threats than later stages. To explore whether stage-specific vulnerability to TAI might exist, we not only analysed outcomes for all miscarriages, but also differentiated between biochemical and clinical losses whenever possible. For LBR analyses, some papers calculated LBR on a per-cycle basis whereas others calculated on a per-clinical pregnancy (CP) basis; again, these were combined into a single analysis [6]. As with MR outcomes, it is important that like-for-like livebirth outcomes are compared. Notably, similar concerns also apply to the two meta-analyses on this topic performed prior to 2016 [8, 9].

A second critically important consideration is that female age was not reported in 3 of the 12 studies included in the most recent comprehensive meta-analysis [6]. This is pivotal since arguably the most important determinant of pregnancy outcome is female age due to the effects of ageing on oocyte quality, which ultimately impacts aneuploidy rates [20]. Moreover, when age was analysed for the 9 papers in which maternal age was known, Busnelli et al. (2016) [6] found that TAI+ women were significantly older than TAI- women. Based on these analyses, therefore, it is very unclear to what extent the older age of the TAI+ cohort contributed to the adverse reproductive outcomes that were observed.

Thirdly, since the last meta-analysis, at least 4 further studies analysing women of known ages have been published [21‐24]. Busnelli et al. (2016) [6] included a total of 428 TAI+ women of known ages. These 3 newer studies [21‐24] contain an additional 390 TAI+ women.

Fourthly, half of the 12 studies included in Busnelli et al. (2016) [6] were published prior to 2010. In more recent years, guidelines from the ASRM and the American Thyroid Association [25] suggest that a lower thyroid stimulating hormone (TSH) cut-off (2.5 mIU/mL) should be used for initiating levothyroxine treatment in TAI+ women. Newer studies may therefore include women with more intact thyroid function whereas older studies may have included patients with more severe thyroid imbalances, which may have increased the tendency to adverse outcomes in earlier meta-analyses. Indeed, two of the studies in the recent meta-analysis [12, 26] included women with overtly abnormal TSH levels. It is notable in this regard that the 2011 meta-analysis identified an increased odds ratio (OR) for miscarriage of 3.15 amongst TAI+ women [9] whereas in Busnelli et al. (2016) [6], the OR had declined to less than half this at 1.44 [6]. Furthermore, in He et al. (2016) [10] and Leiva et al. (2017) [11], no significant relationship was found between TAI and IVF outcomes. While He et al. (2016) [10] found no association between TAI positivity and clinical pregnancy, miscarriage rate or delivery rate in euthyroid women, they did find that women with subclinical hypothyroidism (SCH) and TAI were at greater risk of miscarriage and had a lower delivery rate than TAI- women, indicating that the extent of associated thyroid dysfunction, rather than TAI per se, could account in some part for adverse outcomes reported in previous studies.

Here we undertook an updated systematic review and meta-analysis of the overall relationship between TAI and IVF/ICSI outcomes as has been done previously, as well as separate analyses of euthyroid women of known female age using strict criteria for defining pregnancy outcomes.

Here we undertake the most comprehensive meta-analysis to-date regarding the effect of TAI on IVF/ICSI outcomes in women of known ages and TSH status. For the first time, we use strict criteria to analyse BPL and CMR as well as LBR on a per-cycle and per-CP basis in euthyroid women as well as in a larger group, which included women with SCH. We found no difference in any pregnancy outcomes in relation to TAI status. Our results contrast with previous meta-analyses showing increased MR and/or decreased LBR in TAI+ women overall and in euthyroid women undergoing IVF/ICSI [6, 8, 9]. On the other hand, our findings are consistent with those of He et al. (2016) [10] and Leiva et al. (2017) [11]. There are important differences between our study and previous meta-analyses that may explain some of these discrepancies. Our systematic review included the most recent studies such as Devi et al. (2019) [22], Chen et al. (2017) [21] and Sakar et al. (2016) [18], which did not find any association between TAI positivity and pregnancy outcome. We also analysed outcomes in studies that were strictly limited to euthyroid women in addition to the wider group seen in previous studies. We were also strict in analysing like-for-like outcomes (e.g. CMR and LBR per cycle) rather than combining variously defined outcomes into a single analysis. Given the powerful influence of female age on pregnancy outcome through effects on oocyte quality [20], a critically important strength of our study was to restrict analyses to papers that reported female age. Importantly, we found that both groups were comparably aged thereby unequivocally ruling out the possibility of an age effect. In contrast, in the most recent meta-analysis, one quarter of included studies had not reported age and analysis of ages amongst those that did, found that TAI+ women were significantly older than TAI- women [6]. A meta-regression analysis performed by the authors suggested that such differences were unlikely to have impacted their outcomes [6]. However, this cannot be concluded with certainty since, as the authors themselves noted, only 9 studies were included in the meta-regression, which is below the minimum of 10 recommended for avoiding Type II errors [29] added to which, ages were completely unknown for a substantial proportion of included studies. Furthermore, by considering maternal TSH levels, the present study accounted for potential confounding effects due to substantial thyroid dysfunction.

The lack of a detrimental effect of TAI on pregnancy outcomes that we find here is consistent with two recent large RCTs. The first RCT, the POSTAL study, tested whether levothyroxine supplementation could reduce MR in 600 TAI+ euthyroid women (mean age 31.6 years) undergoing IVF and found there was no benefit [36]. Since this study did not include a TAI- group, it is not known whether MR was actually increased in their untreated TAI+ cohort. Indeed, the study was powered against a predicted CMR of 30% for TAI+ women but unexpectedly, the actual MR in their untreated TAI+ arm was only 10.6%, which was no higher, and indeed, somewhat lower than the background rate of ~ 15% for women of a similar age [18, 19]. The recently reported TABLET trial [37] studied a broader population of TAI+ women with a history of miscarriage or infertility and again found no benefit of levothyroxine supplementation. Significantly, however, the rate of combined CMR and BPL was ~ 30%; this rate, which includes biochemical losses that occur with high frequency, also seems to be no higher than the predicted rate in the general population [16]. It is therefore likely that the lack of effect observed in both studies was at least in part because there isn’t a markedly increased risk of miscarriage in the first place in TAI+ women, entirely consistent with our present findings.

There has been another recent meta-analysis that did not consider the full breadth of IVF patients as we did here, but instead focused only on studies in which ICSI was performed for fertilisation [38]. Consequently, it only contained 4 studies, 3 of which had been included in the earlier meta-analysis by Busnelli et al. (2016) [6], along with a more recent paper [23] which we included in our study. Significantly, Poppe et al. (2018) [38] found no difference in MR when analyses were strictly restricted to CPs, entirely in keeping with our findings. There have been a number of alternative theories regarding ICSI’s ability to overcome the adverse effects of TAI on pregnancy outcomes. It has been suggested that anti thyroid antibodies may bind antigens in the zona pellucida, thereby disrupting its function [39], and that ICSI may be able to overcome this effect. However, in light of the present study’s findings, it appears more likely that the finding of Poppe et al. (2018) [38] are due to the inclusion of more recent studies and stricter inclusion criteria.

A recent meta-analysis investigating the effect of levothyroxine treatment found that MR in TAI+ women who conceived with ART were not reduced by levothyroxine but that levothyroxine decreased MR among TAI+ women who also had SCH [40] suggesting that SCH rather than TAI per se may confer the detrimental effect. Considering the recent change in guidelines, which now propose that TAI+ women with TSH > 2.5 mIU/mL should be supplemented with thyroxine [25, 41], most women with SCH and TAI were more likely to have been excluded from more recent studies. It is therefore possible that previous negative effects on pregnancy outcomes in TAI+ women may have been due to SCH, rather than TAI itself. It is worth noting that a number of studies included in previous analyses that were not included in the present study did not report serum TSH levels, making it possible that the relationship between TAI and adverse pregnancy outcomes may have been due to thyroid dysfunction. We note, however, that the results of the present study do not support this notion since we found no association between TAI and pregnancy outcome either in euthyroid women or the overall group. Studies such as Chai et al. (2014) [14] and Zhuang et al. (2017) [24] separated their investigation on TAI and IVF based on maternal serum TSH levels < 2.5 and > 2.5 and found no difference in IVF outcomes based on TSH status. Furthermore, A recent retrospective study and meta-analysis found no increased risk of adverse pregnancy outcomes in women with TSH > 2.5 [42]. Thus, whilst this idea may hold some merit, further investigation is required.

We acknowledge that a variety of factors may influence the strength of our findings. Individual studies assessed different antibodies (TPO-Ab and/or TG-Ab) and used different assays as well as different cut-off values for defining TAI positivity (Table 1), which could potentially affect outcomes. Cause of infertility maybe be a potential confounding factor. A number of the included studies reported no significant difference in cause of infertility between the TAI+ and TAI- groups. However, each had varying inclusion criteria, and in recent a recent meta-analysis, TAI has been shown to be more prevalent in conditions associated with poor IVF outcomes such as PCOS [43]. It has been shown that the presence of TPO-Ab in particular is associated with an increased risk of PCOS and endometriosis, and varying aetiologies based on the inclusion criteria and antibodies present may have influenced results [36]. There were also limited details on individual patients such as stimulation protocols used, which could theoretically affect results given that antagonist and agonist protocols could have different effects on thyroid function [44]. The numbers of embryos transferred could also impact pregnancy outcomes. Although we considered this potential confounder, and analysis of the five studies that reported embryo numbers transferred showed no difference in relation to TAI status (Fig. 8), we did not have data on embryo numbers for 9 studies. We note as well that although we adhered to uniform criteria for outcomes, the included studies defined and reported clinical outcomes differently and some studies were excluded based on discrepancies in the reported data, a lack of clarity regarding definitions, or insufficient available data regarding clinical outcomes. It therefore remains possible that some important data were omitted that may have influenced the results.

The results of the present study suggest that TAI has no effect on pregnancy outcomes in women undergoing IVF either overall, or in euthyroid women. The findings of this study may be useful when deciding on treatment plans for TAI+ women undergoing IVF.