Circulating resistin and follistatin levels in obese and non-obese women with polycystic ovary syndrome: A systematic review and meta-analysis

This meta-analysis was performed to resolve the inconsistencies regarding resistin and follistatin levels in women with polycystic ovary syndrome (PCOS) by pooling the available evidence. A systematic literature search using PubMed and Scopus was carried out through November 2020 to obtain all pertinent studies. Weighted mean differences (WMDs) with corresponding 95% confidence intervals (CIs) were calculated to evaluate the strength of the association between the levels of resistin and follistatin with PCOS in the overall and stratified analysis by obesity status. A total of 47 publications, 38 for resistin (2424 cases; 1906 controls) and 9 studies for follistatin (815 cases; 328 controls), were included in the meta-analysis. Resistin levels were significantly higher in PCOS women compared with non-PCOS controls (WMD = 1.96 ng/ml; 95%CI = 1.25–2.67, P≤0.001) as well as in obese PCOS women vs. obese controls, and in non-obese PCOS women compared with non-obese controls, but not in obese PCOS vs. non-obese PCOS patients,. A significantly increased circulating follistatin was found in PCOS patients compared with the controls (WMD = 0.44 ng/ml; 95%CI = 0.30–0.58, P≤0.001) and in non-obese PCOS women compared with non-obese controls and in obese PCOS women vs. obese controls, but, no significant difference in follistatin level was observed in obese PCOS compared with non-obese PCOS women. Significant heterogeneity and publication bias was evident for some analyses. Circulating levels of resistin and follistatin, independent of obesity status, are higher in women with PCOS compared with controls, showing that these adipokines may contribute to the pathology of PCOS.


Search strategy
A comprehensive literature search was performed in PubMed and Scopus to obtain all pertinent human studies published up to November 2020. Following medical subject heading (MeSh) terms and text words were used: (resistin or adipose tissue-specific secretory factor or C/EBP-epsilonregulated myeloid-specific secreted cysteine-rich protein or ADSF or XCP1 or follistatin or activin-binding protein) and (PCOS or polycystic ovary syndrome). The search strategy was limited to articles published in English. Moreover, the reference lists of included studies and review articles were manually searched to obtain other possible relevant studies that may have been missed in the initial search. Two investigators (SHA and LR) independently searched the electronic databases and screened the titles/abstracts and full-text studies after excluding duplicated publications. Any disagreement in the screening process was resolved by involving a third reviewer.

Inclusion and exclusion criteria
Studies were eligible for the current meta-analysis if they met the following criteria: a) studies should be published in English; b) reported circulating levels of resistin or follistatin in women with PCOS compared with healthy female controls; c) reported body mass index (BMI) or weight status of the participants; d) reported the total means of resistin or follistatin and standard deviations (SDs) or sufficient information to calculate them; e) included an acceptable diagnosis of PCOS based on the Rotterdam criteria, the Androgen Excess and PCOS Society criteria, or National Institute of Health (NIH) criteria. Only the most informative or the newest study was included when multiple studies were published based on the same population. Studies were excluded if they enrolled participants with diseases other than PCOS or were reviews, editorials, case reports, conference abstracts, cell line studies, animal studies, letters to editors, and studies without controls. EndNote program was used to facilitate the screening process. Two independent authors (SHA and LR) screened the titles/abstracts of studies to find whether studies are eligible for inclusion. If the abstracts of articles seemed pertinent, then the full-text assessment was done and the screening forms were filled out to select the eligible studies based on the inclusion criteria. Any disagreements were discussed and resolved by consensus with a third reviewer.

Data extraction and quality assessment
Two investigators conducted data extraction and quality assessment of the included publications independently and dissimilarities were resolved by referring back to the original citations. The following information was extracted: first author's last name, country, year of publication, study design, age, numbers of cases and controls, ethnicity, BMI or weight status, and circulating resistin and follistatin levels (means and standard deviations). Because of the variety of definitions for obesity among the studies, it was defined as BMI above 25 or 30 kg/m 2. For articles providing the standard errors of means, the standard deviations (SDs) were calculated by multiplying the standard errors by the square roots of the sample size [27]. For publications reporting the medians and the corresponding interquartile range, medians were considered as means while the SDs were estimated by dividing the widths of the interquartile ranges by 1.35 [2]. The quality of the included studies was evaluated with the Newcastle-Ottawa Scale [28], which covers nine items, and scores range from 0 to 9. Studies with scores of �4, 5 to 6, and 7 or above are categorized as low-quality, medium-quality, and high-quality studies, respectively.

Statistical analysis
The summary weighted mean difference (WMD) with 95%CIs was estimated for circulating levels of resistin and follistatin in PCOS cases versus controls (or between obese and non-obese PCOS women). The I 2 statistic and Cochrane Q statistic were applied to assess the statistical heterogeneity across studies. A p value<0.1 for Q-statistic or I 2 larger than 50% was considered significant heterogeneity. When significant evidence for heterogeneity was detected, the random-effects model was used for analyses; otherwise, a fixed-effect model was applied. The sensitivity analysis was conducted by removing one study each time and recalculating WMD with a 95% confidence interval (CI) to find the effect of individual studies on the pooled effect sizes. Publication bias was evaluated using funnel plot asymmetry and Egger's linear regression method [29] and P values �0.05 were considered statistically significant. All analyses were done with the use of Stata software (Version 13.0; StataCorp, College Station, TX, USA).

Search result and study characteristics
The initial search strategy found 870 studies, of which 805 studies were excluded after screening by titles/ abstracts or duplicate publications, and 65 studies were eligible for full-text assessment. Finally, a total of 47 articles, 38 for resistin (2424 cases; 1906 controls) [7,9,11,14, and 9 studies for follistatin (815 cases; 328 controls) [15,17,[64][65][66][67][68][69][70], were included in the present meta-analysis based on the inclusion criteria. The flowchart summarizing the screening process is reported in Fig 1. The included studies were published between 2001 and 2020. The included studies also had subpopulations based on obesity status, which were analyzed separately. For resistin, there were 20 studies comparing circulating resistin in healthy nonobese PCOS women vs. non-obese control, 12 studies on obese PCOS vs. obese healthy women, and 18 studies on obese PCOS vs. non-obese PCOS patients [7, 9, 30, 32-34, 36-39, 44, 65], and 2 studies on obese PCOS vs. non-obese PCOS patients [65,67]. The quality of studies was medium to high, with scores ranging from 4 to 8 (S1 Table).
The characteristics of the included articles are reported in Table 1.
Met-regression, sensitivity analysis, and publication bias. Meta-regression analysis coefficients for serum resistin and follistatin levels in the examined group of studies showed that difference of mean age (Coefficient: 0.24, SE: 0.22, p = 0.30) and sample size (Coefficient: 0.003, SE: 0.003, p = 0.50) did not change the relation of resistin to PCOS. For follistatin, it was found that the relationship between follistatin and PCOS is modified by the difference of mean age between the PCOS patients and healthy controls (Coefficient: -0.06, SE: 0.01, p = 0.007) so that by an increase in the difference of mean age between the PCOS patients and healthy controls, there was a significant reduction in the mean difference of follistatin between the groups (Fig 7). Sensitivity analysis was conducted by deleting the studies one by one in the meta-analysis each time to reflect the impact of the single studies on the pooled WMD. Finally, no single study considerably affected the pooled WMD for resistin (S2 Table) and follistatin (S3 Table),  showing that the results of this meta-analysis were statistically robust. A significant evidence for publication bias was detected by Egger's test for studies on resistin (t = 2.14, p = 0.03) ( Fig  8 and Table 2).

Discussion
Recently, changes in circulating levels of resistin and follistatin in women with PCOS have been investigated; although, the results of published studies are still inconclusive. Thus, this meta-analysis was performed to comprehensively analyze circulating levels of these adipokines in PCOS women compared with healthy women by considering weight status as an interacting factor. The results revealed that, overall, resistin and follistatin levels were significantly higher in PCOS women compared with healthy controls. The increased resistin and follistatin levels in PCOS patients were independent of obesity status.
In agreement with our findings, Pandis et al. [30] found that the resistin circulating levels were higher in patients with PCOS. They also found that subjects with a BMI�25 kg/m2 compared with non-obese PCOS and non-obese healthy women have higher resistin levels. However, no difference in resistin levels was reported between patients with PCOS and a BMI<25 kg/m2 and healthy lean women [30], proposing that resistin may have a role in PCOS. Resistin has been reported as a potential link between insulin resistance and obesity [71]. The mRNA level of resistin has been revealed to be down-regulated by antidiabetic thiazolidinediones medicines, which bind to peroxisome proliferator-activated receptor-g (PPAR-g) in fat cells [10]. Resistin expression in adipocytes is increased in morbidly obese individuals compared with lean subjects [72]. In addition, in the adipocytes from women with PCOS, the mRNA level of resistin is 2-fold higher than that in healthy women [14]. It has been identified that there is a positive correlation between serum resistin and free testosterone levels [62]. Increased resistin concentration might increase ovarian androgen production in patients with  [34]. Taken together, these data and our findings, show that resistin may contribute to PCOS and its clinical manifestations.
In line with our meta-analysis, previous studies have reported that circulating levels of follistatin are increased in PCOS independent of body mass index [64,65], suggesting that obesity is not the explanatory factor for the increased levels of follistatin among women with PCOS. The source of circulating follistatin in women is unknown; nevertheless, this protein is released from ovarian granulosa cells in response to FSH stimulation [73]. An increase in the expression of follistatin is a possible mechanism in which follicular growth is inhibited [64]. These findings suggest that changes in the secretion of follistatin from the ovary or other organs might be involved in the pathophysiology of PCOS and describe the alteration in circulation levels of this protein. This is in line with this observation that PCOS is most closely related to areas adjacent to the follistatin gene in PCOS siblings [16]; however, later studies have identified the relationship to be much weaker than what previously thought [74]. Interestingly, in some studies, no significant difference in the level of follistatin was observed in the follicular fluid from the polycystic and healthy ovaries [75,76]. The previous reports that failed to find any difference in the level of follistatin in the follicular fluids could be explained by thecal secretin of this adipokine and the rich blood supply in this region of the ovary resulting in rapid clearance of follistatin into the blood [64]. Moreover, these studies indicate the possibility of the extragonadal origin of the high circulating follistatin levels in women with PCOS. Since both resistin and follistatin are involved in clinical manifestations of PCOS, our findings are clinically important as they could be used as possible therapeutic targets in women with PCOS.
This meta-analysis resolved the inconsistency among the results of the previous studies; these findings are clinically important as an increase in resistin and follistatin observed in the present meta-analysis has been related to adverse metabolic consequences in individuals with PCOS. Higher levels of follistatin and resistin are associated with the lack of pre-ovular follicle development in PCOS or insulin resistance [64,77]. Thus, the modulation of these hormones may be important therapeutic targets for new drugs in patients with PCOS. Based on the results of this study, follistatin and resistin could be studied in large cohort of PCOS patients  to better understand the pathophysiology of this disease. Also, future clinical trials implement interventions to improve circulating/follicular concentrations of these hormones.
To the best of our knowledge, this is the first meta-analysis summarizing available evidence to assess whether there is a relationship between circulating resistin and follistatin levels with PCOS. However, some limitations of the present meta-analysis should be acknowledged. First, the leading limitation of the current study is the remarkable heterogeneity across the included studies, which might decrease the reliability of our results. This heterogeneity may reflect clinical heterogeneity related to the differences in the accuracy and sensitivity of various methods applied to assess circulating levels of investigated adipokines, and differences in geographical regions, physical activity, diet, ethnicity of participants, and concomitant subclinical inflammatory diseases. Furthermore, some included studies had small sample sizes and the background of patients was different, which would lead to low statistical power and inconclusive conclusions among the studies. In the meta-regression analysis, we found that difference of  https://doi.org/10.1371/journal.pone.0246200.g005 mean age is a source of observed heterogeneity among the studies. Nevertheless, in each individual study, cases and controls were matched for age. Our meta-regression was an exploratory analysis to assess whether difference of mean age between the PCOS patients and healthy controls across different studies affects pooled effect sizes despite matching cases and controls for age within studies. We revealed that the relationship between follistatin and PCOS is modified by mean difference of age between the PCOS patients and healthy controls, suggesting that results of studies with larger difference of mean age between the PCOS patients and healthy controls should be interpreted cautiously. Second, the language restriction to English publications may increase the odds of publication bias. Third, the results were based on crude estimates without any adjustment for potential confounders, while a more accurate assessment should consider the confounding factors such as alcohol use, smoking status, dietary patterns, and environmental factors. It is noteworthy that cases and controls were matched for most important confounders such as age and BMI, but, similar to other epidemiological studies, residual confounding could not be ignored.
In conclusion, this meta-analysis suggested that resistin and follistatin levels, independent of obesity status, were higher in women with PCOS compared with those in healthy controls. Hence, resistin and follistatin might play important roles in the development of PCOS and may be useful biomarkers for the treatment of this disorder. Further well-designed studies with large sample sizes should be performed to examine the circulating levels of resistin and follistatin and their role in PCOS.