Phenotyping clinical disorders: lessons learned from pelvic organ prolapse

One major reason to identify genetic risk factors for a clinical disease is to advance scientific knowledge about the pathophysiology of the disease in order to facilitate targeted diagnostic, therapeutic and preventative interventions. For example, pelvic organ prolapse is thought to have a multifactorial etiology with a genetic component; 12 , 13 , 22 , 23 however, the underlying mechanisms that lead to defects are not yet known. By understanding the genetic susceptibility for a disease, mutation screening may be developed and utilized for risk-stratification to individualize recommendations for prevention or treatment. In studies of common, complex diseases, a critical component is to characterize, or “phenotype,” the disorder accurately to facilitate the identification of genetic effects. The process of developing a phenotype definition involves the use of epidemiologic, biological, molecular and/or bioinformatic methods to systematically select characteristics of a disorder that might result from distinct genetic influences. Defining the phenotype is critical because poorly-phenotyped studies will misclassify participants, which decreases the power to identify significant genetic effects. While it is important to carefully define and phenotype “cases” with a disease, trait, or susceptibility, it is also necessary to define “controls” without disease as genetic epidemiologic studies will often compare cases to controls. To illustrate these concepts in further detail and to discuss the nuances of designing genetic epidemiologic studies, we will use pelvic organ prolapse as an example to highlight critical considerations in phenotyping clinical disorders.

After defining the cases and controls, it is necessary to identify risk factors and potential confounders. The presence of risk factors, in either cases or controls, will be adjusted for in the statistical analysis in order to ultimately estimate the contribution from genetic factors. Family history may be used to further refine the phenotype, if desired. Given that the etiology of prolapse is likely multifactorial with a genetic component as well as environmental factors, which include health behaviors, medical conditions, and exposures, it is important to collect data on other risk factors and to think about potential exclusion criteria. The presence of risk factors, in either cases or controls, will be adjusted for in the statistical analysis in order to ultimately estimate the contribution from genetic factors. Data collection should include detailed information regarding obstetrics history, including parity; mode of deliveries (vaginal, cesarean and assisted vaginal deliveries), age at first delivery, infant birth weight, and perineal trauma; diseases associated with Valsalva maneuvers, such as emphysema/COPD or constipation; prior surgical history, including hysterectomy; prior anti-incontinence and prolapse procedures; medications, especially hormone replacement therapy and steroid use; smoking history; and body mass index. 22 Another consideration is to collect additional markers for connective tissue issues such as joint hypermobility, varicose veins, rectal prolapse, and hernias. 45 However, those with connective tissue disease, such as Ehlers-Danlos or Marfans syndrome, should be excluded as these subjects may have an inherently higher risk for prolapse. Previous studies examining the risk of pelvic organ prolapse in nulliparous and parous sister pairs have observed a concordance rate of prolapse suggesting a familial predisposition. 23 Studies involving extended pedigrees using a population database have shown similar results with significantly increased relative risks in first and third degree relatives (RR= 4.15, p<0.001, RR=1.24, p=0.05, respectively) and increased average pairwise relatedness for all individuals with prolapse (p<0.001). 46 Hence, having close and distant relatives with pelvic organ prolapse are known risk factors for the disease. Study of extended pedigrees with a significant excess of prolapse has multiple benefits including (1) affected relatives are more likely to share the same genetic risk factors than unrelated cases with prolapse, (2) a significant excess number of prolapse cases in a family results in a high probability that the cases are attributable to genetic effect(s), and (3) knowledge of pedigree structure increases power of the study as it can lead to a predictable correlation structure of the genetic risk. Selective sampling of prolapse cases based on family history of prolapse increases the likelihood that genetic effects will be detected in a study. Prolapse often coexists with other pelvic floor disorders, such as stress or urgency urinary incontinence. Currently, we do not know if stress incontinence and prolapse share the same casual pathway or if they are distinct genetic conditions. The incorporation of coexisting pelvic floor disorders becomes an important issue when defining cases and controls because if stress incontinence and prolapse share the same causal pathway, there is the potential to introduce misclassification bias if controls have stress incontinence even if they do not have prolapse. Thus, one recommendation is to collect additional data to define the phenotypes of stress and urge incontinence when conducting a genetic study of prolapse until we have more data regarding the etiologies of these conditions.

Pelvic organ prolapse is a major women’s health issue with a prevalence as high as 40% in postmenopausal women. 24 Risk factors for prolapse include advancing age, Caucasian race, higher body mass index (BMI), increasing parity, smoking, chronic constipation, chronic cough and family history of prolapse. 22 , 23 , 25 – 28 Although childbirth is often considered a major risk factor, a majority of parous women do not develop prolapse while some nulliparous women do develop the condition. Furthermore, the development of prolapse is often temporally distant from birth injury, thereby strongly suggesting that additional factors beyond vaginal parity, such as genetic susceptibility, that contribute to or protect against prolapse.

In developing a study to evaluate the genetic epidemiology of prolapse, a common approach is to recruit cases with prolapse versus controls without prolapse; however, a critical scientific question is how will pelvic organ prolapse be defined. Prior to discussing different approaches to defining prolapse, it is important to discuss key principles that are unique to genetic epidemiologic studies. It is important to consider the age of onset and natural history of the disease of interest as this will directly impact who to recruit for cases and controls. For example, a 45 year old woman may not show any evidence of prolapse, and thus could be considered a control subject. However, it is possible that she may not develop prolapse until 50 or 60 years of age. In many genetic epidemiologic schemes, these younger “unaffected” subjects are considered “unknowns.” Thus, older women without evidence of prolapse are more ideal control subjects. The same rationale applies to studying other phenotypes, which are characterized by older age at onset, such as coronary artery disease and breast cancer. It is critical to carefully consider race and ethnicity in genetic epidemiologic studies, as different racial/ethnic groups may have different frequencies of genetic variants, which is a concept known as population stratification. These systematic differences in allele frequencies exist because of differences in ancestry rather than the association between genes and diseases. 29 , 30 Confounding by race results when race is associated with both differences in genetic variant frequencies as well as differences in disease prevalence, but there is no association between the genetic variants and disease Thus, population stratification in genetic studies can lead to excess false positive results and failures to detect true associations. 31 Because of population stratification, genetic studies will often focus on one specific race/ethnicity or utilize advanced analytic techniques to adjust for racial genetic differences. 32 – 34 Heterogeneities in complex disease, such as prolapse, arise due to different clinical manifestations, differences in biological pathways leading to disease, and differences in symptom severity. Often phenotypic heterogeneity (inconsistency in the definition of phenotype) across studies makes it difficult to generalize study findings and to replicate genetic associations. This inconsistency in trait definition may be due to lack of consensus or a general disagreement among researchers in the field as to the correct phenotypic definition. Establishing a consistent operational definition and description of the phenotype to be used across studies may help avoid this dilemma. Below, we discuss different methods of phenotyping prolapse and make a recommendation regarding how to phenotype prolapse in order to move towards developing a consistent phenotype definition for future genetic epidemiologic studies.

There are several different clinical definitions for prolapse, and none are universally accepted. Possible definitions include 1) bothersome symptoms, which focuses on subjective factors, 2) surgical treatment for prolapse, and 3) stage of prolapse, which is an objective measure of severity. It has been shown that bothersome symptoms are a critical component of evaluating prolapse 35 and validated questionnaires 36 , 37 include a question regarding bothersome prolapse symptoms: “Do you usually have a bulge of something falling out that you can see or feel in the vaginal area? If yes, how much does it bother you?” 38 While this is key to evaluating treatment outcomes, it is less suited to designing early genetic association studies. Another option for phenotyping prolapse is to define prolapse based on those who have chosen to undergo surgery. Presumably only women with severe enough bothersome symptoms warrant surgical management. 12 One important consideration is that it is necessary to confirm the procedures performed and the indication for surgery. Obtaining prior operative reports can be somewhat challenging, especially if the surgery was in the distant past. Furthermore, this phenotype definition fails to capture women who may have very advanced prolapse but pursue non-surgical treatment options. A more objective method of phenotyping is based on examination findings, and for prolapse, the two primary examination techniques are the pelvic organ prolapse quantification (POP-Q) 39 or Baden-Walker system. 40 A common strategy is to compare cases with POP-Q stage III/IV prolapse versus controls with stage 0/I prolapse. 14 , 41 The rationale is that there may be some degree of prolapse that occurs as a “normal” part of aging, as the proportion of postmenopausal women with a mild degree of prolapse is relatively high. 24 By excluding stage II, the more extreme phenotypes are captured. Selective sampling of more extreme phenotypes is a common approach in genetic epidemiologic studies, as recruitment of more extreme phenotypes increases the probability of observing a genetic effect. For example, a woman with a high genetic risk might be a young, premenopausal, nulliparous woman with stage IV prolapse and a family history of prolapse. In contrast, a low genetic risk woman might be an 80 year old woman who has had 5 vaginal deliveries yet has stage 0 prolapse. While it may be difficult to identify and recruit these ideal extreme phenotypes, it is important to select cases and controls which are clinically disparate. Thus far, a majority of case control association studies have defined cases and controls based on physical examination findings. 14 , 42 17 , 43 , 44 Currently, we recommend using physical examination findings to phenotype prolapse, given that it is an objective measure that is likely to be more reproducible. Also, for initial genetic epidemiologic studies, the definition of cases as stage III to IV and controls as stage 0 to I may increase the power to identify a genetic effect. Although we have focused on the definition of cases for this phenotype, it is also critically important to define control subjects carefully, in order to decrease the risk of misclassification bias. It is necessary to ensure that controls have been evaluated fully for the disease/phenotype of interest and have had equal opportunity for detection of the condition. For example, if we defined prolapse solely based on symptoms, it is possible that an asymptomatic woman who actually has prolapse on examination could be selected as a control. It is important to exclude those who have had prior surgery for prolapse; thus, a careful medical and surgical history and physical examination are needed.

After carefully phenotyping the disease of interest and identifying risk factors, the next step is to determine what clinical and genetic data resources are available. The two broad categories are those that (1) exist already or (2) need to be collected prospectively. There are a number of existing resources to investigate ( Table 1 ). A valuable source is the National Center for Biotechnology Information’s (NCBI) database of Genotypes and Phenotypes (dbGaP) which is a public repository that archives and distributes the results of studies that have investigated numerous phenotypes with a wide variety of genomic technologies ( http://www.ncbi.nlm.nih.gov/gap ). 47 An investigator can search dbGaP to browse what genetic studies have been conducted, what variables were included in these studies, and what type of analysis was performed. These datasets from dbGaP can be downloaded at no expense to an investigator and are particularly useful if additional phenotype data are available from an existing genome-wide association study (GWAS) dataset on dbGaP but was not analyzed as part of the primary analyses. For example, if a GWAS was performed to primarily evaluate diabetes but other variables of interest were assessed, such as urinary incontinence, an investigator interested in incontinence can request access to the GWAS data and then evaluate the genetic epidemiology of incontinence. Another useful resource is DNA banks or biorepositories that are linked to electronic medical records (EMR). One example is the BioVU DNA Repository at Vanderbilt University, which is designed to support large-scale genomic research. 48 , 49 De-identified DNA samples are linked to electronic medical records (EMRs) of patients from Vanderbilt University Medical Center clinics, and detailed human subjects protection is applied to BioVU. 50 Sample accrual started in 2007 and is ongoing with an average of 600 new samples per week. As of June 2012, there are over 128,480 adult DNA samples. BioVU is part of the eMERGE Network (electronic MEdical Records and GEnomics) which is an NHGRI-funded consortium of DNA biobanks linked to electronic medical records (EMR) with the goal of performing large-scale genomic analyses. 51 To-date eMERGE ( http://www.gwas.org ) has performed genome-wide association studies of peripheral arterial disease, 52 red blood cell traits, 53 HDL cholesterol, 54 and electrocardiographic PR interval. 55 Another potential resource is the UK Biobank ( http://www.ukbiobank.ac.uk/ ). 56 Another option is to establish a biorepository by collecting samples prospectively. This approach ensures the collection of detailed, uniform clinical information as well as phenotype data; however, it is critically important to have the necessary infrastructure to establish a biorepository, and the interdisciplinary collaborations needed to conduct genetic epidemiologic studies. When designing a new genetic epidemiologic study, a useful resource is the PhenX toolkit ( www.phenx.org/ ). 57 PhenX provides 295 well-established, high-quality measures within 21 different domains that can be used by the scientific community to facilitate cross-study comparisons and analyses. 57 Although specific measures for a disease of interest may not be listed (i.e. there is no information regarding pelvic organ prolapse or urinary incontinence), suggestions on how to collect data on possible covariates (i.e. what is the ideal question to use to assess tobacco use?) will help to ensure that these data could be useful for studies of different traits. Lastly, another unique aspect of genetic epidemiologic studies is that after conducting an initial study, it is important to validate your findings in an independent cohort, often called a validation cohort. This is in contrast to randomized controlled clinical trials in which the results do not need to be validated initially to be considered a high quality study. Thus, collaboration is a critical component of conducting genetic epidemiologic studies, as collaborators may be able to provide a validation, or replication, cohort. The strong collaborative nature of this field is evidenced by the numerous consortia that have developed to study different diseases and traits.

In this article, we have outlined important considerations in genetic epidemiologic studies and key issues to consider when phenotyping a clinical disorder. Phenotyping is a critical step in advancing our understanding of the genetic etiology of disorders within obstetrics and gynecology. There are clearly unique considerations in how to operationally define, or phenotype, cases and controls in genetic epidemiologic studies in order to optimize the ability to detect genetic risk factors. Hopefully, in the future there will be dedicated and funded efforts to help uncover the genetic susceptibilities to the multitude of women’s health issues.