Androgen levels in the fetal cord blood of children born to women with polycystic ovary syndrome: a meta-analysis

A systematic literature search was conducted by two of the authors (Qi and Hanxiao). The PubMed, EMBASE, and Cochrane library databases as well as the Web of Science were searched using the following terms: “Polycystic Ovary Syndrome” and (“infant” or “pregnancy”) and “umbilical cord blood”. No language restrictions were applied. The last search was published in December 2019. Our study protocol has been submitted to PROSPETO (registration number: CRD42020125668).

Human studies that compared cord blood androgen levels, including those of T and ADION, in the fetal cord blood of mothers with and without PCOS were eligible for inclusion, provided that 1) participants in the PCOS group were diagnosed with PCOS based on the NIH or Rotterdam criteria, and the control group mothers did not have PCOS or PCOS-like symptoms.; 2) the study type could be cross-sectional, case-control or cohort studies; 3) the reports were full-text articles in English. For duplicate articles or data, the most comprehensive or the one published earliest were selected.

Studies were excluded when the titles or abstracts indicated that the papers were not relevant, or the papers were reviews, case reports, or other types without raw data. We conducted a quality assessment for the included studies; we used the Newcastle-Ottawa Scale (NOS) for case-control studies and the Agency for Healthcare Research and Quality (AHRQ) list for cross-sectional studies.

Two of the authors (Tianjiao and Yujing) screened and evaluated all articles independently to extract the related information; differences of opinion were settled through negotiation. Furthermore, all the relevant references of the eligible original articles were examined to ensure data integrity. The collected data included the first names of the authors, year of publication, research type, diagnostic criteria, gestational age at delivery, primary study population, type of umbilical cord blood, method of assay, and the units of T and ADION. The scores of quality assessment were also recorded.

Statistical analysis was performed using Review Manager, Version 5.3 (RevMan 5.3) with a two-sided p-value to compare cord blood androgen levels, including those of T and ADION, in the fetal cord blood of mothers with and without PCOS. For the included studies that used the Mann-Whitney U test, the median, the first and third quartiles were transformed to mean and standard deviation using the formula provided by Wan et al. [19]. Considering the variations in androgen assays and units among the studies, the standardized mean difference was selected to analyze the hazard ratios (HRs), 95% confidence interval (CI), and I-square (I²) using the inverse variance method. To prevent ignoring the possible heterogeneity between studies, we first used more conservative random-effects model rather than fixed effects. When I² was not significant (< 50%), the fixed-effects model was applied. Moreover, a sensitivity analysis was conducted by deleting each study individually to evaluate the quality and consistency of the results. Subgroup analyses were performed according to assay types and the sex of the infants. Visual inspection of the funnel plot was done to assess possible publication bias.

The search strategy identified 325 unique studies (among a total of 459) and 7 articles [11‐13, 15‐18] were scrutinized after literature retrieval and selection. A total of 570 samples including 268 female and 222 male (the rest 80 samples had no mention of sex) qualified for review.

The included studies used various assays to measure serum androgen levels in the cord blood. Four studies [11, 12, 15, 17] used mass spectrograph (MS), others used multiple assays like chemiluminescent immunoassay (CLIA), direct electrochemiluminescence methodology (ECL), and radioimmunoassay (RIA). Two of the 7 studies [11, 16] were cross-sectional studies, the remaining 5 [12, 13, 15, 17, 18] were case-control studies. Most studies [12, 13, 15‐17] collected mixed arteriovenous cord blood, Detti et al. obtained arterial and venous cord blood separately [11], while Barry et al. collected only venous cord blood [18]. The selection process is shown in Fig. 1, the characteristics and quality assessment scores of the eligible 7 studies are listed in Table 1.

Figure 2a shows the Forest plot of T levels difference in all types of cord blood of PCOS and control groups, including male and female newborns and mixed arteriovenous, arterial, and venous cord blood. The results demonstrate that there is no significant difference in cord blood T levels between two groups (7 studies; HR = 0.15; 95%CI = [− 0.17,0.47]; I² = 68%; P = 0.36; random-effects model).

The differences of PCOS diagnosis criteria and assays of T between studies may have resulted in high heterogeneity (I² = 68%). We conducted a subgroup analysis by PCOS diagnosis criteria, NIH and Rotterdam criteria. The results showed there is still a high heterogeneity in NIH subgroup (3 studies; HR = 0.25; 95% CI = [− 0.55,1.05]; I² = 85%; P = 0.53; random-effects model; Fig. 2b). However, no heterogeneity was found in Rotterdam subgroup (3 studies; HR = -0.11; 95% CI = [− 0.38,0.17]; I² = 0%; P = 0.46; random-effects model; Fig. 2b). In both subgroups, there is no significant difference of T level between PCOS and control cord blood (Fig. 2b).

As for subgroups divided by MS and other types of assays. The results showed no heterogeneity in the MS group (4 studies; HR = -0.08; 95% CI = [− 0.28,0.13]; I² = 0%; P = 0.47; random-effects model; Fig. 2c). Then, we created a further grouping according to infants’ sex using the fixed-effects model, the result showed no significant difference in both subgroups (Fig. 2d). The funnel plot indicated that there is no visual publication bias (Fig. 2e).

The Forest plot of ADION levels difference in all types of cord blood is shown in Fig. 3a, the difference in cord blood ADION levels between PCOS and control mothers was not significant (4 studies; HR = -0.10; 95%CI = [− 0.63,0.42]; I² = 79%; P = 0.71; random-effects model).

Sensitivity analysis was conducted by deleting each study individually. We found I² = 0% after removing the study by Daan et al. (Fig. 3b). The subgroups were divided according to infants’ sex after deleting Daan’s study. The funnel plot also showed no bias (Fig. 3d). As presented in Fig. 3c, the ADION level in female newborns’ cord blood was significantly lower than that seen in the control group. In the male newborn group, the difference was not obvious.