Introduction

The existence of inherited risk factors for pelvic floor disorders has been recognized for > 150 years [1], and multiple studies have confirmed familial aggregation of pelvic organ prolapse (POP). Three large meta-analyses demonstrated a significant impact of family history on the development of or recurrence of POP with odds ratios ranging between 1.84 to 2.64 [2,3,4] with an affected first-degree relative (mother or sister). Large population database studies have shown similar results. In a Swedish registry including data for 61,323 women with a history of POP surgery, the relative risk of prolapse surgery was found to be 6.58 (95% CI 6.32–6.86) for their sisters and 2.56 (2.41–2.73) for their mothers [5]. These results were further clarified in a population-based study in the USA involving 453,522 total women and 4628 women with a history of POP surgery that found that risk increased with increasing numbers of affected relatives, from RR of 2.36 (95% CI 2.15–2.58) for ≥ 1 affected first-degree relative to RR 6.26 with ≥ 3 first-degree relatives (95% CI 1.29–18.20) [6]. Having ≥ 3 affected third-degree relatives (first cousins) carried a similar risk to having one affected first-degree relative. A relevant family history is also associated with earlier onset disease [7]. Maternal inheritance of POP has been found to be a more significant contributor to the development of POP, but paternal inheritance also contributes to risk [6, 7].

Family studies, particularly those involving nuclear family members, provide limited information on heritability, as they do not control for shared exposure to environmental risk factors. Twin studies have been used to formally quantify the heritability of prolapse. In a sample of 16,886 Swedish twins aged > 50 years, heritability was estimated as 43% for prolapse surgery [8], suggesting prolapse is of similar heritability to other pelvic floor disorders including urinary incontinence.

Given the strong heritability findings, genetic studies are justified to find POP predisposition variants. Early linkage studies identified target regions that have prompted multiple follow-up candidate gene studies. The first linkage analysis investigated a single three-generation Filipino pedigree with six affected women with early-onset POP, and they identified the candidate gene LAMC1 under their 1q31 linkage peak [9]. Two additional linkage studies involving women of European descent identified the chromosome 9q21, 10q24–26 (includes candidate gene LOXL4), and the 17q25 (includes candidate gene TIMP2) regions as showing significant evidence of linkage [10, 11]. A follow-up study involving Russian women with POP identified a significant haplotype association in the 9q21 region with results driven primarily by SNP rs12237333 [12]. These linkage analyses have been followed by multiple candidate gene studies and recently genome-wide association studies (GWAS) that are the main focus for this systematic review.

Objective

Identification of the genetic variants underlying the heritability of POP would provide useful markers for clinical risk, prognosis, and treatment response. In addition, these insights should help explain the pathogenesis of POP, potentially offering new drug targets and preventative strategies. The aim of this systematic review was therefore to assess which polymorphisms and/or genetic loci had been tested for an association with pelvic organ prolapse in women, while assessing the strength, consistency, and potential for bias, among published associations.

Materials and methods

Eligibility criteria

This review updates an earlier review using the same eligibility criteria and including all prolapse studies from that work [13]. The protocol for the earlier work was prospectively registered (PROSPERO 2011:CRD42012001983), and we made no changes to the methods [14]. We pre-specified inclusion of both case-control and cross-sectional designs, with both population-based samples and other sampling methods. We included association studies testing for any genetic polymorphism at the nucleotide level, including SNPs, deletions, duplications, and copy-number variants, but excluded larger microscopic variants at the karyotype level.

There are no gold standard diagnostic methods. For pelvic organ prolapse, validated staging systems, including POP-Q, have been widely used, but again there is no universally accepted criterion for diagnosis. We therefore expected to accept diagnostic criteria for prolapse as specified within each study. In view of heterogeneity in definitions across studies, we tested for heterogeneity between studies with different criteria in different settings. We accepted definitions based on symptom questionnaires, clinical examination, or other validated assessments. We considered the population of interest as women aged ≥ 18 years.

Search strategy

We updated the earlier systematic review, using an identical search strategy [13]. We combined searches from PubMed, HuGE Navigator, and an extensive selection of genetic, urological, and urogynaecological conference reports. In this update we searched PubMed from January 1, 2015, to November 1, 2020, using a combination of genetic and phenotype keywords and MeSH terms:

(polymorphism OR SNP OR CNV OR "copy number variation" OR mutation OR genetic OR chromosome OR VNTR OR InDel OR microsatellite) AND (prolapse OR "Pelvic Organ Prolapse"[MeSH]) NOT mitral NOT carcinoma [Title] NOT cancer [Title] NOT (animals[mh] NOT humans[mh])

In this update we searched HuGE Navigator, also from January 1, 2015, to November 1, 2020, using the phenotype indexing term “pelvic organ prolapse.”

In addition, we searched conference abstracts for annual meetings of the American Society of Human Genetics, American Urological Association, American Urogynecologic Society, European Association of Urology, European Society of Human Genetics, International Continence Society, International Urogynecological Association, and Society of Gynecologic Surgeons 2005–2020.

Screening and data extraction

We developed standardized data forms for this study and conducted pilot screening and data extraction training exercises to achieve a high level of consensus between reviewers. All screening and data extraction were then performed independently and in duplicate by methodologically trained reviewers. Reviewers screened study reports by first screening titles and abstracts to select papers for full-text assessment and then screening full-text papers to confirm eligibility of the articles. Screening discrepancies were resolved by adjudication. We hand searched reference lists of all included articles, applying the same standardized screening process. When more than one report was identified for the same association in the same study population, we included the publication with the largest sample size.

We contacted study authors by email, with a reminder after 1 month, for clarifications, additional information about methodology, and additional subgroup analyses where necessary. Data extracted included information on the setting for each study, details of the sampling strategy and sampled populations (age, parity, ethnic/racial composition, and BMI), the overall sample size and proportion genotyped, the outcome assessments used and phenotypic definitions, the genotyping method employed, and the genotyping quality control applied. Where possible we extracted or requested from authors full genotype frequencies among both cases and controls.

Statistical analysis and risk of bias assessments

For polymorphisms assessed in ≥ 2 studies for the same phenotype and evaluated with similar case definitions, we conducted fixed or random effects meta-analyses as appropriate using the Metan package (Stata 12.1). In situations where a proxy SNP had been selected for genotyping in one or more studies, in high linkage disequilibrium (defined as D′ ≥ 0.8) with another SNP of interest, these SNPs were considered as being equivalent for meta-analysis purposes; results are reported based on the original significant SNP identifier. Linkage disequilibrium was assessed between pairs of SNPs using the LDpair tool [15, 16] and an appropriate racially and ethnically matched population (e.g., Utah residents from North and West Europe [CEU] for Caucasian European populations). In all cases we worked from genotype or allele frequencies rather than using precalculated effect sizes. In the absence of a clear rationale supporting any specific model of inheritance, we used the allelic association test and co-dominant models of inheritance for all polymorphisms. We assessed the credibility of pooled associations using the interim Venice criteria [17] (see Table 1). We used the I2 statistic as a measure of between study heterogeneity. We recalculated the power of each study and retested for departure from Hardy-Weinberg equilibrium. We made assessments of risk of bias in phenotype definitions, genotyping, and population stratification. We used the Harbord test of funnel plot asymmetry and the significance chasing bias test [18] to investigate possible reporting biases. Throughout these assessments we used p < 0.05 as the criterion for significance, except in relation to GWAS, where p < 5 × 10−8 is accepted as the criterion for significance. Reporting of this review complies with recommendations of both the HuGE Handbook and the PRISMA statement.

Table 1 Summary of interim Venice guideline ratings of credibility of genetic associations
Full size table

Narrative summaries

For completeness of this review, we additionally provide summaries of the four genome-wide association studies (GWAS) reported to date. Where possible, significant GWAS findings have been included in meta-analyses. However suggestive and non-significant GWAS findings are typically not reported; hence, we are unable to include most null findings from GWAS in the meta-analyses.

Results

Included studies

We screened 504 new abstracts for this review (Fig. 1), eventually including 46 published and 7 unpublished studies, of which 20 had been previously included in the review we updated [13]. A large majority of studies had enrolled either women of European or East Asian descent, with limited representation of other ethnicities.

Fig. 1
figure 1

Flowchart outlining the literature search and article evaluation process. a ASHG, ESHG, ICS, IUGA, AUA, SGS, AUGS, and EAU abstracts 2005–2020 using search interfaces at http://www.ics.org/publications/abstracts, http://www.sciencedirect.com/science/journal/15699056, http://www.jurology.com/supplements, http://www.ashg.org/meetings/meetings_abstract_search.shtml, and/or full text search of abstract book PDFs. b Includes reviews (n = 2), inapplicable phenotypes (n = 3), and other study designs including pharmacogenetic studies, gene expression studies, or methylation studies (n = 33)

Full size image

Meta-analyses

We conducted 24 separate meta-analyses for variants in or near 16 different genes or genetic loci. Four of these 12 genes had significant findings in pooled analyses: rs2228480 in the ESR1 gene, rs12589592 in the FBLN5 gene, rs484389 in the PGR gene, and rs1800012 in the COL1A1 gene (Figs. 2, 3, 4, and 5).

Fig. 2
figure 2

Forest plot of meta-analysis of studies of the rs2228480 SNP in the gene ESR1

Full size image
Fig. 3
figure 3

Forest plot of meta-analysis of studies of the rs12589592 SNP in the gene FBLN5

Full size image
Fig. 4
figure 4

Forest plot of meta-analysis of studies of the rs484389 SNP in the gene PGR

Full size image
Fig. 5
figure 5

Forest plot of meta-analysis of studies of the rs1800012 SNP in the gene COL1A1

Full size image

ESR1 gene

ESR1 is an estrogen receptor gene, which was identified as relevant in candidate gene studies because of the epidemiological association between estrogen status and prolapse. Two studies from Taiwan and China assessed the same three variants (rs17847075, rs2228480, and rs2234693) and could be included in meta-analyses [19, 20]. In pooled analyses, rs2228480 showed a large protective effect with low heterogeneity (OR = 0.67, 95% CI: 0.46–0.98, I2 = 0.0%, Venice rating BAB). The risk variant is common in the populations assessed, and so despite the low total sample size (n = 339), this confers moderate epidemiological credibility.

FBLN5 gene

FBLN5 has been investigated as a candidate gene for prolapse as fibulins play a critical role in the assembly of elastic fibers, believed to provide strength and flexibility in the pelvic floor. Three studies from Brazil, Russia, and China assessed the same two variants (rs2018736 and rs12589592) of which two studies could be included in meta-analyses [19, 21, 22]. No significant pooled effect was observed for rs2018736, but a large effect was seen at rs12589592 with moderate heterogeneity (OR 1.43 95% CI 1.11–1.82, I2 = 36.3%, Venice rating BBB). The risk variant is common in the populations assessed, and so despite the low total sample size (n = 568), this confers moderate epidemiological credibility.

PGR gene

PGR has been investigated as a candidate gene for prolapse, as it codes for the progesterone receptor, and changes in serum progesterone cyclically, during pregnancy, and after menopause are all observed to have an influence on prolapse. Two studies from China each assessed the same two common polymorphisms and could be included in meta-analyses [19, 23]. No significant pooled effect was observed for rs500760, but a large effect was seen at rs484389 with moderate heterogeneity (OR = 0.61, 95% CI: 0.39–0.96, I2 = 32.4%, Venice rating CBB). The risk variant is common in the populations assessed, but the low total sample size (n = 336) confers weak epidemiological credibility.

COL1A1 gene

COL1A1 has been investigated as a candidate gene for prolapse as it forms type 1 collagen, the most abundant human collagen. The rs1800012 was identified as a replicated locus in our earlier review, but we could now include six studies with a moderate protective effect with no heterogeneity (OR = 0.80, 95% CI: 0.66–0.96, I2 = 0.0%, Venice rating BAB) [24,25,26,27,28]. The risk variant is common in the populations assessed, and with a moderate sample size (n = 1264), this confers moderate epidemiological credibility.

Other genes

We conducted further meta-analyses for variants in COL3A1 type 3 collagen (8 studies), COL18A1 collagen type 18 (3 studies), LAMC1 Laminin, gamma 1 (6 studies), ZFAT (3 studies), MMP1 matrix metalloproteinase 1 (3 studies), MMP3 matrix metalloproteinase 3 (4 studies), MMP9 matrix metalloproteinase 9 (4 studies), MMP10 matrix metalloproteinase 10 (2 studies), and four other variants identified from GWAS (rs1455311, rs430794, rs8027714, and rs1810636). None of these meta-analyses showed significant pooled effects. Results are summarized in Table 3. Many genes had been assessed in a single study only and as such require replication for credibility (Table 2).

Table 2 Included studies
Full size table

Narrative summary of GWASes

The first GWAS for POP involved 115 surgically treated, related POP cases who were part of high-risk POP pedigrees and 2976 population-based controls [29]. They identified six variants at chromosomal regions 4q21 (rs1455311), 8q24 (rs1036819), 9q22 (rs430794), 15q11 (rs8027714), 20p13 (rs1810636), and 21q22 (rs2236479). Five of these six SNPs have subsequently been identified as at risk of genotyping error on one or more Illumina arrays, which may have led to spurious association signals [30]. The original study observed nominally or trending towards significance for some variants in a Dutch validation cohort of 76 POP cases. Subsequent independent replication studies [31,32,33, 12, 34, 19, 35] have tested for association at some or all of these six SNPs, with rs1036819 close to ZFAT replicating in one study [19], rs8027714 on chromosome 15q11 replicating in another study [35], and rs1810636 on chromosome 20p13, demonstrating replication in another study [31], but with no overall significant replication for any SNP observed in our meta-analyses (see Table 3).

Table 3 Summary of meta-analyses
Full size table

A further GWAS using African American and Hispanic women from the Women’s Health Initiative Hormone Therapy study [36] included 1427 cases with any diagnosis of POP (grades 1–3) and 317 cases diagnosed with moderate/severe POP (grades 2–3) and 1274 controls without POP (grade 0). Although they did not identify any variants meeting genome-wide significance, they did identify a number of variants that met p < 10−6.

The largest POP meta-analysis of two GWA studies involved 3409 cases from Iceland and 131,444 controls and 11,601 cases and 209,288 controls from UK Biobank, all of which were of European ancestry [34]. POP cases were identified based on ICD 9/10 coding therefore representing women who had presented for care. They identified eight variants at seven loci meeting the genome-wide significance criterion in the meta-analysis with results driven mainly by UK Biobank data. The significant SNPs include rs3820282, rs9306894, rs3791675, rs7682992, rs1247943, rs12325192, rs72624976, and rs1430191. None of the lead POP variants were coding or in high linkage disequilibrium (LD) with coding variants. We can consider them each as having moderate credibility (Venice rating ABB). This study did not replicate any variants identified by earlier GWASes [29, 36] Table 4.

Table 4 Interim Venice ratings of the credibility of replicated associations
Full size table

Finally, a recently reported GWAS utilizing 1329 women with diagnosed and/or surgically treated prolapse and 16,383 hospital controls did not identify any variants meeting genome-wide significance [35]. However, testing associations from previous GWASes showed nominal replication for rs8027714 [29] and for rs12325192, and rs9306894 [34].

Conclusions

Given current evidence supporting a genetic predisposition for pelvic organ prolapse, we have identified four variants through meta-analysis of candidate gene studies significantly associated with POP (rs2228480 in the ESR1 gene, rs12589592 in the FBLN5 gene, rs484389 in the PGR gene, and rs1800012 in the COL1A1 gene). In each meta-analysis we have at most moderate evidence in support of an association with POP. A much larger, recent prospective meta-analysis of two genome-wide association studies has identified eight variants significantly associated with POP [34], with recent evidence of replication for two of these variants in an independent population [35]. As the sizes of GWAS meta-analyses grow, further novel variants are likely to be identified providing novel insights into pathogenesis. Given the impact of pelvic floor disorders on women’s health, additional work needs to be done to provide further validation of POP predisposition variants in a variety of different populations to establish the role of these genes in the pathogenesis of prolapse and to establish a possible role for genetic testing in clinical practice that could improve patients’ outcomes and address the best treatment options.