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

Background Thyroid autoimmunity (TAI) – the presence of anti-thyroid peroxidase and/or anti-thyroglobulin antibodies – affects 8–14% of reproductively-aged women. It is hotly debated whether TAI adversely affects IVF/ICSI outcomes. This systematic review and meta-analysis evaluated the relationship between thyroid autoimmunity (TAI) and IVF/ICSI outcomes, both overall and amongst euthyroid women of known age using strict criteria for grouping pregnancy outcomes. Methods The review was registered with PROSPERO: CRD42019120947. Searches were undertaken in MEDLINE, EMBASE, Web of Science and Cochrane Database from Inception-March 2020. Primary outcomes were clinical pregnancy rate, clinical miscarriage rate, biochemical pregnancy loss, livebirth rate per-cycle and live birth rate per clinical pregnancy (CP). Results 14 studies were included in the meta-analysis. Compared with women who tested negative for thyroid autoantibodies (TAI-), there was no significant difference in clinical pregnancy rate overall (OR 0.86; 95%CI [0.70, 1.05]; P = 0.14; 11 studies; I2 = 29.0%), or in euthyroid women (OR 0.88; 95%CI [0.69, 1.12]; P = 0.29; 10 studies; I2 = 32.0%). There was also no significant difference in clinical miscarriage rate overall (OR 1.04; 95%CI [0.52, 2.07]; P = 0.908; 8 studies; I2 = 53%), or in euthyroid women (OR 1.18; 95%CI [0.52, 2.64]; P = 0.69; 7 studies; I2 = 54%). There was no significant difference in biochemical pregnancy loss (OR 1.14; 95%CI [0.48, 2.72]; P = 0.769; 4 studies; I2 = 0.0%), live birth rate per cycle (OR 0.84; 95%CI [0.67, 1.06]; P = 0.145; I2 = 1.7%), live birth rate per clinical pregnancy (OR 0.67; 95%CI [0.28, 1.60]; P = 0.369; I2 = 69.2%), both overall and in euthyroid women as all studies included consisted of euthyroid women only. There was also no significant difference in number of embryos transferred, number of oocytes retrieved, mean maternal age or TSH levels overall or in euthyroid women. Conclusion The findings of the present study suggest that thyroid autoimmunity has no effect on pregnancy outcomes in euthyroid women alone, or in euthyroid women and women with subclinical hypothyroidism. Supplementary Information The online version contains supplementary material available at 10.1186/s12958-020-00671-3.


Background
Thyroid autoimmunity (TAI) is defined as the presence of antithyroid antibodiesanti-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 postimplantation 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.
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.

Materials and methods
Protocol and registration PRISMA guidelines were followed and the review was registered with PROSPERO (Registration no. CRD42019120947).

Search strategy
Literature searches were conducted via MEDLINE, EMBASE, The Cochrane Library (Cochrane Database of Systematic Reviews, Cochrane Central Register of Controlled Trials (CENTRAL), Cochrane Methodology Register), and Web of Science (Science and Social Science Citation Index) from inception to April 2019. Searches were rerun in March 2020 prior to publication. The Medline search utilised a combination of text and medical subject headings (MeSH): ("Thyroid Gland"[-Mesh] AND "Autoimmunity"[Mesh]) OR "thyroid autoimmunity" OR "thyroid autoantibody" OR "thyroid autoantibodies" OR antithyroid peroxidase OR "antithyroglobulin antibody", "Fertilization in Vitro"[Mesh] OR IVF OR "invitro fertilization" OR "in vitro fertilization" OR "in-vitro fertilization" OR "invitro fertilization" OR "invitro fertilisation" OR "in vitro fertilisation" OR "invitro fertilisation" OR "invitro fertilisation" OR "Sperm Injections, Intracytoplasmic"[Mesh] OR "intracytoplasmic sperm injection" OR ICSI OR ART [tiab] OR "reproductive techniques, assisted"[MeSH Terms] OR "assisted reproductive technology". Search terms were set by the authors and were adapted for use with other databases. No language restrictions were implemented. Peer reviewed case control and cohort studies were eligible for inclusion.

Study selection
Title and abstract screening were performed independently by two authors (A.V. and W.W.) to eliminate studies deemed irrelevant. The full text versions of relevant studies were retrieved and assessed for eligibility by the same author. Any uncertainty was resolved through discussion with a third author. Studies were excluded if: (i) women had overt thyroid dysfunction, (ii) maternal age was not reported, (iii) women were being treated with levothyroxine, (iv) ART techniques aside from IVF/ICSI were used. The quality of the studies was assessed using the Newcastle Ottawa scale [27].

Data extraction and analysis
Two authors extracted data independently. Study year, setting, design, thyroid antibody assays used and cut-off values, number of TAI+ and TAI-patients, number of IVF/ ICSI cycles performed, type of ART procedure performed and key outcomes of the study were recorded. Extracted data was compared to ensure accuracy, and discrepancies were corrected upon referral back to the original paper.
The primary outcomes measured were CPR, MR, and LBR as; (i) CP: the presence of a gestational sac on ultrasound scan, CPR: the proportion of IVF/ICSI cycles resulting in a CP; (ii) MR: the proportion of pregnancies that miscarried/did not result in a delivery or live birth, CMR: the proportion of clinical pregnancies that miscarried/did not result in a delivery or live birth, BPL: the proportion of biochemical pregnancies that were lost prior to the CP stage (iii) LBR: the proportion of cycles of IVF/ICSI that resulted in the birth of at least one live baby, calculated per cycle and per CP. In cases where outcomes of interest were not reported by the study but could be determined from the raw data, outcomes were calculated manually. Secondary outcomes such as overall MR (including all definitions of MR: BPR + CMR + CMR alone), maternal age, number of oocytes retrieved (NOR), and numbers of embryos transferred (ET) were also recorded.

Statistical method
Data were analysed in STATA 15 [28]. Pooled effect sizes were estimated by applying random effects metaanalysis. The I 2 statistic was used as a measure of heterogeneity between studies. An I 2 value of 0% indicates no heterogeneity whereas values of 25, 50 and 75% reflect low, moderate and high heterogeneity, respectively [29]. Egger's weighted regression test was used to assess publication bias where more than 10 studies were included in the analysis. Sensitivity analyses were performed where moderate levels of heterogeneity were detected.

Results of literature search and description of included studies
Systematic searches of the PubMed, Embase, Web of Science and Cochrane Library databases identified 491 citations, with 332 of these being non-duplicates ( Fig. 1). 274 articles were excluded based on title and abstract screening. The full texts of 58 articles were retrieved. An additional 45 articles were excluded ( Fig. 1) on the following basis: 5 articles did not report maternal age, 7 articles did not include either a TAI+ group or a TAIcontrol group, 3 studies included patients on levothyroxine, 3 were review articles, 3 did not report patient outcomes, 19 were conference abstracts, 1 did not use IVF/ ICSI, 2 were excluded due to patient overlap and 2 studies were excluded during data analysis due to discrepancies in results (additional file 1). One more study was added in March 2020 when searches were rerun. Details of the 14 studies that were included in the final analyses are presented in Table 1.

Age
Age was reported by all 14 studies. One study reported age as the median (25th-75th) [15] and could not therefore be included in our analysis of mean ages. However, since patients in this study had a median age below 37 without any significant difference between TAI+ and TAI-groups [15], data from this study was included in other analyses. Analysis of the 13 remaining studies revealed no significant differences in patient age between all TAI+ and TAI-women (SMD 0.12; 95%CI [− 0.04, 0.28]; P = 0.15; 13 studies; I 2 = 76%) (Fig. 10a). There were also no significant age differences between the sub-group of TAI+ and TAI-women who were euthyroid (SMD 0.13; 95%CI [− 0.10, 0.36]; P = 0.25; 10 studies; I 2 = 82%) (Fig. 10b).

Heterogeneity and sensitivity analysis
Analyses displaying moderate/high heterogeneity (I 2 ≥ 50%) were adjusted following sensitivity analysis ( Table 2). Removal of 1-2 studies was sufficient to reduce heterogeneity in all cases. Following sensitivity

Publication Bias
Results of Egger's weighted regression test indicated no publication bias in mean maternal age, CPR, overall MR or OPR.

Quality analysis
Of the 14 studies included, the quality ranged from 7 to 8 (Table 1). No studies were excluded from the analysis based on poor quality.

Discussion
Here we undertake the most comprehensive metaanalysis 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].  [11]. There are important differences between our study and previous meta-analyses that may explain some of  [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 metaregression 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 Fig. 10 a Overall association between thyroid autoimmunity and mean maternal age. Results showed no significant difference in mean maternal age between TAI+ and TAI-women overall (SMD 0.12; 95%CI [− 0.04, 0.28]; P = 0.15; 13 studies; I 2 = 76%). b Association between thyroid autoimmunity and mean maternal age in euthyroid women. Results showed no significant difference in mean maternal age between TAI+ and TAI-euthyroid women (SMD 0.13; 95%CI [− 0.10, 0.36]; P = 0.25; 10 studies; I 2 = 82%). TAI+, thyroid autoimmune; TAI-, not thyroid autoimmune; SMD, standard mean difference; CI, confidence interval 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 Fig. 11 a Overall association between thyroid autoimmunity and TSH levels. Results showed no significant difference in TSH levels between TAI+ and TAI-women overall (SMD 0.28; 95%CI [− 0.03, 0.59]; P = 0.08; 6 studies; I 2 = 79%). b Association between thyroid autoimmunity and TSH levels in euthyroid women. Results showed no significant difference in TSH levels between TAI+ and TAI-euthyroid women (SMD 0.14; 95%CI [− 0.07, 0.34]; P = 0.19; 5 studies; I 2 = 46%). TAI+, thyroid autoimmune; TAI-, not thyroid autoimmune; SMD, standard mean difference; CI, confidence interval some important data were omitted that may have influenced the results.

Conclusions
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.