Effect of antithyroid antibodies on women with recurrent miscarriage: A meta‐analysis

Abstract Problem The effect of thyroid autoimmunity (TAI) on the prevalence of recurrent miscarriage (RM) is highly debatable. No meta‐analysis has been published in the past decade to investigate the impact of TAI on women with RM. Method of Study Systemic literature search was conducted on PubMed, Embase, Cochrane, and Web of Science databases. English language literatures published between 1993 and 2019 were selected. We assessed the relationship between the prevalence of RM and thyroid peroxidase antibodies (TPO‐Ab) or antithyroid antibodies (ATA) and evaluated the thyroid‐stimulating hormone (TSH) level in TPO‐Ab‐positive women with RM. We also observed the treatment effect with levothyroxine (LT4) for RM. Review Manager 5.3 software was used to obtain the pooled odds ratios (OR). Results Analysis of 22 eligible studies revealed significant association between TPO‐Ab and the prevalence of RM (OR = 1.85; 95% CI, 1.38 to 2.49; P < .001)(n ≥ 3), (OR = 1.82; 95% CI, 1.13 to 2.92; P = .01) (n ≥ 3). Women with ATA + had higher risk of RM (OR = 2.36; 95% CI, 1.71 to 3.25; P < .00001)(n ≥ 3), (OR = 2.34; 95% CI, 1.70 to 3.22; P < .00001)(n ≥ 2). RM women with TPO‐Ab had higher TSH level when compared with those negative for TPO‐Ab (random‐effect SMD = 0.60; 95% CI, 0.31 to 0.90; P < .0001). We also found beneficial effects of LT4 supplementation on the outcome of live birth rate (LBR) among pregnant women with TPO‐Ab (OR = 3.04; 95% CI, 0.69 to 13.36; P = .14). Conclusion The presence of serum antithyroid antibodies does harms to women and can even lead to recurrent miscarriage; LT4 treatment may have beneficial to RM women.


| INTRODUC TI ON
Thyroid disease is one of the most frequent endocrine conditions in women of childbearing age. 1 The most common cause of thyroid dysfunction is thyroid autoimmunity (TAI). 2 TAI is defined as the presence of antithyroid antibodies (ATA), specifically thyroid peroxidase antibodies (TPO-Ab) and/or thyroglobulin antibodies (Tg-Ab). 3 With a prevalence of 5%-20%, TAI is the most common autoimmune condition in women of reproductive age. 4 Thyroid hormones play a role in menstrual cycle and in achieving fertility as they affect the actions of follicle-stimulating hormone and luteinizing hormone on steroid biosynthesis by specific T3 sites on oocytes. 5 Thyroid autoimmunity has been found to be related to subclinical hypothyroidism (SCH), 6 which is defined as high levels of serum thyroid-stimulating hormone (TSH) despite normal levels of serum free thyroxine (FT4). 7 Both thyroid dysfunction and thyroid autoimmunity are known to cause adverse pregnancy outcomes during all trimesters of pregnancy. 8 Nevertheless, recent evidence shows an association between euthyroid women with thyroid autoimmunity and poor obstetric outcomes. The presence of ATA, particularly TPO-Ab, has been associated with miscarriage, preterm birth, and post-partum thyroid disease. [9][10][11] Recurrent miscarriage (RM) has previously been defined as three or more pregnancy losses 12 which affects 1% of couples. However, in recent years, more guidelines have redefined it as two or more pregnancy losses 13,14 which affects < 5% of couples. 15 RM places a severe physical, emotional, and financial burden on many families and our communities. An effective management and treatment is necessary. A higher prevalence of TPO-Ab in women with RM has been found in several studies, varying from 19% to 36%. [16][17][18] A majority of women with TAI have detectable antibodies such as TPO-Ab and sometimes TG-Ab, whereas few have TSH receptor antibodies. 19 The prevalence of TPO-Ab ranges for 8% to 14% in women of reproductive age. 20 The prevalence of positive-TPO-Ab among pregnant women in countries with good iodine supply has been reported between 5.1 21,22 and 12.4%. 23 The effect of thyroid autoimmunity especially TPO-Ab on the clinical outcome of RM is highly debated. In recent years, no new meta-analysis has been published to reveal the relationship between RM and TPO-Ab or ATA. 24 And the expanded definition of RM affects differential amounts of women of childbearing age. In the increasing affected women, the assay of TPO-AB or ATA whether can effectively predict RM, it need more data.
In recent researches, TPO-Ab is associated with unexplained RM and these women may benefit from treatment with levothyroxine (LT4) 25,26 with decreasing TPO-Ab levels after 2-3 months treatment. 26 But in other studies, euthyroid women with TPO-Ab undergo a treatment with levothyroxine, compared with no levothyroxine treatment, did not reduce miscarriage rate or increase live birth rate. 27,28 It is therefore possible that patients with RM and TPO-Ab benefit from the substitution of thyroid hormones in terms of a lower rate of miscarriage, but currently there are no data specifically on patients with RM.
Currently, the effect of thyroid autoimmunity by itself on the prevalence of RM has not been established yet. Moreover, there was only one guideline(reference) 15 with regard to the need of treatment for women with RM, but there is no management for thyroid autoimmune in the guideline, more information is needed for developing a new guideline. Combining data on this controversial issue might reveal useful information for the counseling and management of euthyroid TPO-Ab + or ATA + women with RM. This study aims to conduct a systematic review and meta-analysis to gain insight into the clinical significance of TAI in women with RM.

| Literature search strategy
The meta-analysis was conducted according to the Preferred

| Study selection
Only English language literatures were included in our study.
Published cohort or case-control studies (retrospective or prospective) that described at least 10 patients were eligible for inclusion.
Studies were excluded if only TAI-positive women were reported without a control group of TAI-negative women. Studies were excluded if women were known to have overt biochemical hypothyroidism or hyperthyroidism history or were receiving any treatment for thyroid dysfunction.
The quality of the included studies was assessed by the Newcastle-Ottawa scale, a validated modality for assessing observational and non-randomized studies. 29 The scale uses a score system based on three major criteria: selection of participants, comparability of study groups, and completeness of follow-up. A quantitative appraisal of overall quality of each observational study was obtained, and scores ranged from 6 to 8 (Table 1).

| Data extraction
Two authors independently evaluated all articles and abstracted data on standardized forms. Information regarding study characteristics (author name, year of publication, study design, sample size), methodology (definition of RM, thyroid autoantibodies and hormone measurement method, threshold and time of measurement, study quality) and characteristics of study groups, and outcome (prevalence of RM, serum TSH levels and live birth rate (LBR)) were extracted. Live birth rate was defined as the number of deliveries that resulted in at least one live born baby.

| Statistical analysis
The overall effect of TPO-Ab or ATA on a binary variable (RMprevalence, live birth rate) was assessed by a combined odds ratio (OR). The overall impact of TPO-Ab on a continuous variable (TSH) was assessed by the difference in means for the two groups of patients; and by using the summary data published in the eligible studies, the results for outcomes were expressed as odds ratios (OR) with 95% confidence intervals (CI). 30 The inconsistency of studies' results was measured using Cochrane Q and the I 2 statistics. 31 I 2 value of 0% indicates no observed heterogeneity, I 2 < 25% shows insignificant heterogeneity, I 2 values of 25%-50% shows moderate heterogeneity and I 2 > 50% indicates significant heterogeneity. 31 The odds ratios (OR) were combined using a fixed-effects model when heterogeneity observed among studies was absent to moderate; the DerSimonian & Laird method for a random-effects model was employed when heterogeneity was high (I 2 > 50%). 32,33 All statistical analyses were performed using Review Manager 5.3 software (Cochrane Collaboration); P < .05 was considered statistically significant.
Funnel plots, which graph OR on a log scale (effect) against standard error of log-OR (precision), were generated and visually inspected for asymmetry to determine if the included studies were non-representative of the body of possible studies on the subject (as could result from small study effect or other biases, such as publication and poor-quality bias).

| Search results and study characteristics
The process of literature identification and selection of studies is summarized in Figure 1A  The characteristics of the studies included in the quantitative analysis are summarized in Table 1. Six of these were retrospective cohort studies, three were prospective cohort studies, and four were cohort studies and nine were case-control studies (Table 1).

| Prevalence of RM
In the women with RM (n ≥ 3), comparing TPO-Ab-negative ones, TPO-Ab-positive women showed a higher prevalence of RM (random-effects OR = 1.76; 95% CI, 1.00-3.10; P = .05; Figure 2A). The heterogeneity test result was high (I 2 = 70%). However, on excluding the study by Esplin, 34 the degree of heterogeneity declined from high to low (I 2 = 43%, Figure 2B), indicating that this study was heterogeneous with other studies. Therefore, this study was not included in the following subgroup analysis.

| TSH level in RM patients
We examined the difference of mean basal serum TSH between the TPO-Ab-positive and TPO-Ab-negative group. Six studies compared it with and without TPO-Ab. 2,17,41,[47][48][49] From the metaanalysis, it emerged that TPO-Ab-positive women had significantly higher serum TSH levels (random-effect SMD = 0.60; 95% CI, 0.31 to 0.90; P < .0001; Figure 6A). The I 2 value was 80% indicating the present of heterogeneity. Serum TSH concentrations were significantly increased in a study 47 . However, the subgroup pooled result changed when the study by Federico Mecacci 47 was removed from the meta-analysis (Fixed effect SMD = 0.60; 95% CI, 0.34 to 0.88; P < .00001; Figure 6B). When excluding the study by Mecacci, 47 the combined results of n ≥ 2 subgroup showed mild heterogeneity (I 2 = 12%).

| Live birth rate
Five studies reported data on the association between LT4 supplementa-   There has been found that women with TA has higher prevalence of recurrent miscarriage. 41 Our contrasting results on prevalence rate in women with thyroid autoimmunity can be explained by several reasons. Among the 22 studies included in our meta-analysis, one showed lower prevalence rate in TPO-Ab positive women. 40 However, the sample size of this study was relatively small.

| D ISCUSS I ON
Additionally, the number of TPO-Ab or ATA-positive women enrolled in the studies was comparatively less than TPO-Ab or ATA-negative women in general.
Therefore, it has been suggested that thyroid auto antibodies may be employed as a marker for at-risk pregnancies. 53 Although the mechanism is not completely understood, it is postulated that TAI results in early pregnancy loss because of activation of immune F I G U R E 1 Eligibility of Studies for Inclusion in Meta-analysis system 54 or act as an infertility factor and results in delayed conception. 55 In this meta-analysis, subjects with thyroid dysfunction or undergoing treatment were excluded in all the studies included.
Although the TSH values in TPO-Ab positive women were significantly higher than that in TPO-Ab negative women. In particular, mean TSH was significantly higher by 0. with >1000 participants, which found that the likelihood of detecting an abnormality after two losses was similar to that after three or four or more losses. 57 In our data, no matter how the definition changes, women with TPO-Ab or ATA have a higher risk of RM than that in TPO-Ab or ATA-negative women. found that treating TAI + women with a combination of levothyroxine, acetylsalicylic acid and prednisolone resulted in higher ovarian responsiveness to gonadotropins and higher pregnancy rates but did not decrease the miscarriage rate. 4 As TAI is hypothesized to be a marker of an underlying generalized autoimmune imbalance, Litwicka et al tested the efficacy of glucocorticoids administered alone and observed significantly higher clinical pregnancy and live birth rates in the treated group. 69 But their sample size was small and all patients were treated in the same center. 69   Antithyroid antibodies are known to occur in normal, healthy populations, and these autoantibodies are five times more common in women than in men. 73 The clear association between TAI and RM suggests that women with RM need to be aggressively tested or treated for antithyroid antibodies. The tests and treatments are not only expensive but also involve potential health risks for the mother as well as the offspring. A recent study evaluated the association of maternal thyroid function during early pregnancy with offspring intelligence quotient and brain morphology in childhood and raised some concerns on possible overtreatment of patients with thyroxine, because offspring of patients with suppressed TSH may have worse neuropsychological outcomes. 44 So we must concern proper screening crowd.
The statement from the American Society for Reproductive Medicine mentioned that the available data support the routine measurement of TSH in infertile women attempting pregnancy, but not that of TPO-abs, unless TSH levels are ≥2.5 mIU/L. 74  According to their guidelines, LT4 administration is recommended in women with TAI and TSH concentrations higher than the pregnancy-specific reference range, whereas it may be considered in women without TAI and TSH concentrations higher than the pregnancyspecific reference range, but below 10 mIU/L. 78 We propose that TSH levels of TAI women are monitored stringently for women with RM before pregnancy and further large scale prospective, randomized, placebo controlled trials are carried out to evaluate the effect of treatment for antithyroid antibodies in euthyroid women with RM.
It should be noted that our meta-analysis had certain limitations. Due to the small number of studies included to analyze certain outcomes, we cannot rule out the existence of publication bias in our analysis. 79 Furthermore, the included studies used different  Women with TAI are prone to develop subclinical hypothyroidism (SCH) during pregnancy, even though they can be euthyroid during the first trimester of pregnancy. 80 Consequently, in order to decipher the effect of TSH on pregnancy outcome in women with RM, it is imperative to conduct further studies.
Our analysis opens the way for more fundamental studies in order to gain deeper insights into the pathophysiological mechanisms of thyroid autoimmunity. The supposed association between TAI and endometrium deserves attention in particular. The process of fertilization in TAI-positive women also needs to be studied in details. Finally, we require further evidence regarding the factors involved in implantation of the embryo, including studies on endometrial receptivity, embryo quality and immunological factors.

| CON CLUS ION
This study reveals the association between thyroid autoimmunity per se and the prevalence of RM. Euthyroid women positive for antithyroid antibodies have higher risk of RM. TSH level in positive-TPO-Ab women with RM is higher than negative-TPO-Ab women.
And LT4 supplementation may effectively increase live birth rate.
More trials involving endometrial receptivity in TAI women should be carried out to decode the effect of TPO-Ab or ATA. More RCTs are necessary to clarify the efficacy of LT4 in the treatment of RM.
Advanced research is also necessary to elucidate the pathophysiological mechanisms of thyroid autoimmunity and its involvement in RM and subsequent pregnancy.

ACK N OWLED G M ENT
We would like to thank Dr Dan-Qing Yu for her expert guidance and continuous encouragement throughout this research project. This project was supported by the National Natural Science Foundation of China (NSFC) (81671487).

CO N FLI C T O F I NTE R E S T
The authors declare that there is no conflict of interest regarding the publication of this article.