Paraoxonase single nucleotide variants show associations with polycystic ovary syndrome: a meta-analysis

We searched MEDLINE using PubMed, Science Direct and Google Scholar for association studies as of August 03, 2019. The terms used were “PON”, “paraoxonase”, “polymorphisms”, “polycystic ovary syndrome” and “PCOS” as medical subject heading and text, unrestricted by language and time span. References cited in the retrieved articles were screened manually for additional eligible studies. Inclusion criteria were: (i) case–control studies evaluating the association between PON variants and PCOS and (ii) genotype frequency data to calculate odds ratios (ORs) and 95% confidence intervals (CIs). Exclusion criteria were: (i) animal studies, (ii) reviews, case report or case series, expert opinion and (iii) unusable genotype data.

Two investigators (AK and NP) independently extracted data and arrived at consensus. The following information was obtained from each publication: indications (yes/no) of whether each article was included in the four previous meta-analyses [18‐21], first author’s name, published year, country of origin, ethnicity, diagnostic criteria, PON SNVs examined and article features needed to tally the Clark-Baudouin score. Table S1 shows the rs numbers (SNVs per study), values under cases and controls that include sample sizes and genotype frequencies as well as minor allele frequencies and P-values for the Hardy-Weinberg Equilibrium (HWE).

We used the G*Power program [22] to evaluate statistical power, where adequacy was set at ≥80% assuming an OR of 1.5 and a genotypic risk level of α = 0.05. Control frequencies from the HWE were calculated from https://​ihg.​gsf.​de/​cgi-bin/​hw/​hwa1.​pl with a two-tailed P < 0.05 indicating deviations. Data distribution was assessed with the Shapiro-Wilk test [23].

Methodological quality of the included articles was assessed with the Clark-Baudouin scale [24], the scores of which range from 0 (worst) to 10 (best) where < 5, 5–7 and > 7 indicate low, moderate and high, respectively.

Figure 1 outlines the study selection process in a flowchart following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (Table S2). Twenty-one citations from the initial search were filtered to yield 12 articles for inclusion [38‐49], all of which examined SNVs in PON1 except Liu et al. [48], which focused on rs7493 SNV in PON2.

Table 1 shows which studies were and were not included in the four previous meta-analyses [18‐21]. Three articles [47‐49] were new additions to the meta-analysis literature and account for the updated associations between PON variants and PCOS. The sole African study [41] was included in the overall LD2 analysis, but not in LD1 (where it was HWE-deviating) nor in subgroup analysis. Asian and Caucasian subjects were in six and five articles of LD1, in five and three articles of LD2, respectively. Geography of the Asians was clearly dichotomous (China and India), which could have been sub-stratified were it not for issues of multiplicity (risk of false-positives) and reduced statistical power [50]. Based on the mean ± standard deviation (7.17 ± 1.19) of the normally distributed (Shapiro-Wilks test: P = 0.18) Clark-Baudouin scores, methodological quality of the component studies was high. Table S1 shows the unduplicated total sample sizes of the included articles (n = 7092), five [38, 39, 46‐48] with adequate statistical power.

This meta-analysis generated 38 comparisons where 11 significant outcomes were confined to LD1 and LD2 (Tables 2 and 3), consigning PON-PCOS associations to five SNVs (rs854560, rs662, rs705379, rs7493 and rs854572).

In the main findings, the two LD1 var outcomes met the criterion of high significance. However, only the Asian outcome met the other criterion of homogeneity. Nevertheless, these two core findings identified the PON rs854560 and rs662 SNVs to be associated with risk of PCOS. The LD groups differed by ethnicity, where a significant codominant risk effect in the LD2 was found in Caucasians (OR 1.32, P^a = 0.02) but not in Asians (OR 1.03, P^a = 0.59). In LD1, on the other hand, significant risk effects were found in all genetic models in Asians (P^a = 0.0001–0.03) but not in Caucasians (P^a = 0.44–0.97). Two notes regarding power outcomes (i) LD1 Asian increased risk effects were also powered and homogeneous in the var comparison; and (ii) LD2 power effects were protective (18–23%) in the recessive and codominant models. Reduced risk effects (15–16%) were also observed in the het outcomes of LD1 (overall and Asian), where the phenotypic difference (increased risk versus reduced risk) between var and het suggests a heterosis phenomenon. Heterosis occurs when subjects heterozygous for a specific genetic polymorphism show a different phenotype from homozygotes [51], conferring a heterozygote advantage (protection). Variations of these effects between the genetic models and ethnic subgroups suggest complex PON-PCOS associations, which is further driven by interactions between genetic and non-genetic risk factors. Gene-gene and gene-environment interactions have been reported to have roles in the associations of PON variants with PCOS. All 12 included articles acknowledged the role of environment but haplotype analysis was addressed in only four (33%) of the component studies [38, 45, 47, 49]. In addition to PON, one article [48] examined SNVs in another gene (superoxide dismutase-2).

The role of PON1 variants with PCOS have been addressed in four meta-analyses [18‐21], which we compared with the present study in terms of general features and methodology (Table S3). The previous meta-analyses [18‐21] examined PON1 only but our study included PON2 (rs7493) on account of its full LD with rs705379 and rs854572 of PON1. To our knowledge, this is the fifth meta-analysis to address the PON-PCOS associations, but the first to accomplish the following: (i) perform an umbrella review (Table S3); (ii) apply outlier treatment with the most number (n = 12) of included articles; and (iii) operate within LD parameters. These features render our study as most comprehensive, managing to accomplish two things: (i) fill the gaps and update the meta-analysis knowledge of the PON SNV-PCOS associations and (ii) minimize the methodological problems that beset primary studies including limited statistical power, unrecognized confounding factors, misleading definition of phenotypes and stratification of populations [24].

PCOS is a reproductive endocrinopathy [52] that is associated with dyslipidemia, obesity and insulin resistance [53]. These metabolic disorders lead to disease conditions such as hypertension, cardiovascular disease and diabetes mellitus [54]. These disease comorbidities were shown to be related with increased oxidative stress, exhibiting altered physiological conditions such as increased plasma glucose and low antioxidant reserves [55, 56]. Low antioxidant levels in PCOS patients suggest that elevated oxidative status contribute to the battery of cardiometabolic derangements [57]. These perturbations have been attributed to reduced serum PON1 activity in PCOS patients [58], its genetic underpinnings [59] partly explained with the significant findings of our study observed in LD1, involving rs854560 (L55M) and rs662 (Q192R). These two SNVs account for the main PCOS risk associations in this meta-analysis. The rs854560 variant, not rs662, has been reported to affect enzyme concentration [15, 60]. Compared to 55LL genotype in rs854560, 55MM carriers have lower enzyme activity, which may be attributed to the correlation between the 55 M allele and reduced mRNA and protein levels [61]. Moreover, strong LD of this variant with rs662 may partly explain the variation in PON1 catalytic activity [62]. Depending on the assay used, the R allele of rs662 could be associated with increased or decreased PON1 activity [63]. Reduced enzyme activity leads to elevated levels of oxidative stress altering the metabolism in PCOS patients [64]. Oxidative stress profiles in these patients revealed the R allele carriers had impaired physiologic responses involving increased oxidization of low-density lipoprotein [46], high triglyceride levels [38] and elevated insulin resistance [44]. Even in the absence of insulin resistance, oxidative stress levels remain high in PCOS women [65]. This cascade of impaired physiological events, as well as inflammatory responses to cellular injuries caused by oxidative stress, primes the pathophysiology of PCOS as inflammatory mediators that have been known to regulate PON expression [66] and contribute to PCOS pathogenesis [67]. Moreover, genetic combination analysis demonstrated that haplotypes containing the 192R allele was significantly associated with PCOS risk ranging from 1.6 to 8-fold [38, 49]. In sum, the R allele of rs662 appears to be a genetic determinant of PCOS susceptibility in the female population [49]. These genetically susceptible women are the likely candidates who could benefit from the clinical application of our results. PCOS risks along with their comorbidities could be clinically reduced or delayed with modifications of environmental influences meant to reduce levels of oxidative stress.

Limitations of our study include: (i) six (50%) of the included primary 12 articles were underpowered. However, statistical power at the aggregate level was more than adequate and (ii) credible subgrouping was suggested at no more than two levels [50], which left other possible subgroups (e.g. diagnostic criteria) unexamined. On the other hand, strengths of this meta-analysis are: (i) the combination of more studies, larger sample sizes and multiple meta-analysis treatments raised the level of evidence presented in this study; (ii) restricting our analysis to HW-compliant studies minimized the risk of representation and methodological bias [28]; (iii) potency of outlier treatment is evidenced from the LD1 and LD2 analyses, where six (50%) and eight (67%) of the 12 combined post-outlier comparisons were acquired significance and eliminated heterogeneity (Tables 2 and 3); (iv) the overall methodological quality (determined by the Clark-Baudouin Scale) of the included articles was high; (iv) all genotyping used polymerase chain reaction followed by either restriction fragment length polymorphism (11/12 articles) or direct sequencing (1/12 articles) techniques, indicating low-level heterogeneity; (v) all tissue samples were from blood, indicating source homogeneity and (vi) umbrella review of previous meta-analyses enabled comparisons of methodological treatments and findings (Table S3). This provided insight into the evolving consolidation of knowledge into the association genetics of PCOS involving the PON SNVs.