Stress urinary incontinence is a highly prevalent condition affecting 49.8% of women in the United States and is defined as involuntary urinary leakage provoked by physical exertion, sneezing, or coughing.1,2 Females with stress incontinence report significantly diminished quality of life, social isolation, and sexual dysfunction.3–5 Despite the effect of stress urinary incontinence, its precise etiology remains unclear. The relative contributions of genetics and environment in determining stress urinary incontinence have proven to be difficult to ascertain using conventional clinical research methods, largely because observational samples from a general population cannot account for genetic variation between unrelated subjects.6,7 Twins offer a unique study population for investigating genetic causation, allowing for comparison between genetically identical compared with nonidentical twins. “Classic” twin studies allow for a “nature compared with nurture” comparison of identical and nonidentical twin pairs, statistical measurement of differences in intrapair correlation, and estimates of heritability.8
Our study aims to estimate the magnitude of genetic effects on stress urinary incontinence in the United States. We used a classic twin research model, in a large sample of twin sisters, to investigate to what extent female stress urinary incontinence is determined by genetics, environment, or a combination thereof.
MATERIALS AND METHODS
This population cohort study included 882 twin sister pairs who attended the annual “Twins Days” festival in Twinsburg, OH, from 2003 to 2008. This is an annual festival open to participants from throughout the nation. The questionnaire included general demographics and validated condition-specific quality-of-life questionnaires. Included were pairs aged 18 years or older who visited the research “tent” and volunteered to participate in a women's health survey. This was the only information participants were told regarding the study. Excluded from the analysis were subjects with incomplete data. Subjects completed the questionnaire anonymously and independently and received $5 for their participation.
Stress urinary incontinence was identified by an affirmative answer to, “Do you leak urine with coughing, straining, laughing, physical activity, or exercise?,” which is similar to the International Continence Society definition of stress urinary incontinence defined as involuntary urinary leakage provoked by physical exertion, sneezing, or coughing.2 Responses consisted of: “none (0), slight (1), moderate (2), or greatly (3).” The data were dichotomized, with responses of “slight,” “moderate,” or “greatly” considered as positive for stress urinary incontinence. Mild stress urinary incontinence was regarded as clinically relevant to include within our “case” definition, because women experiencing mild symptoms often report significant distress and seek care.9 Subjects with any report of stress urinary incontinence were analyzed as well as a subgroup of subjects with “pure stress urinary incontinence,” which were defined as subjects who reported stress urinary incontinence in the absence of any urge urinary incontinence. In addition, an analysis was made excluding the cohort of women who had previous prolapse or antiincontinence surgery to account for possible confounding of subjects who had their incontinence corrected surgically. The environmental factors analyzed in this study are listed in Table 1. The study was approved by the Ethics Committee and Investigational Review Board of Evanston Hospital, NorthShore University HealthSystem.
Demographic and symptom data were compared between identical and nonidentical twins using a mixed-effects model for continuous covariates and generalized estimating equations for categorical covariates, both of which adjust for standard errors by taking account the correlation within twin pairs.10 The significance level was set at two-sided P<.05 for all hypotheses testing. SAS 9.1 was used for statistical analyses.
In evaluating for potential genetic influences, concordance rates for both identical and nonidentical pairs were calculated. Differences in pairwise concordance rates were compared by likelihood ratio chi square tests.
The second step of the genetic analysis involved the estimation of heritability. Structural equation modeling was applied to quantify the observed phenotypic variance between twins and characterize its genetic and environmental components.11 In the classic twin model, individual differences in the disease trait (phenotypic variance) among identical and nonidentical twins are attributed to three basic factors: A) additive genetic influences; C) common environmental influences referring to shared environmental factors such as family, household, and neighborhood environments; and lastly E) individual-specific or unique environmental factors including all environmental factors not shared by the twins that occurs as one of the twin pairs moves or is exposed to different environments, including random measurement error.11 A genetic model for the covariance of phenotypes is formed from the sum of these variance components: VP=VA+VC+VE.12 The hypothesis underlying the classic twin model assumes that identical twins share the same additive genetic variations, whereas nonidentical twins share just half of the additive genetic variations, and both sets of twins share the same common environmental variations.11
Analyses were performed with the statistical program Mx, which uses maximum likelihood estimation to fit the variance components models.11 Goodness-of-fit statistics were presented for three models: the full model (ACE), which includes genetics and all environmental factors; the model (AE), which explains variance according to genetics and unique environmental factors only; the model (CE), which considers variance explained by a combination of common and unique environmental factors; and lastly, the model (E) that considers the contribution of unique environment and random error only. Reduced models were constructed by removing each parameter sequentially, and we compared the goodness of fit of each reduced model with the full model using a likelihood ratio test. The significance of A, C, and E was sequentially tested using hierarchic chi square identifying the model with the fewest parameters possible that was not significantly different from the full model to explain the pattern of variances and covariances as is consistent with the principle of parsimony.13 The analysis was performed in collaboration with the Center for Clinical Research Informatics, (NorthShore University HealthSystem).
Among 882 twin pairs (n=1,764) with complete data pertaining to stress urinary incontinence, there were 765 identical and 117 nonidentical twin sister pairs. Demographic and medical characteristics are summarized in Table 1. The mean age for both sets of pairs was 41.3±16.3 years (range, 18–85 years) with a mean body mass index (calculated as weight (kg)/[height (m)]2) of 26.1±6.5 (range, 13.5–55.8) and mean parity of 1.3 (range, 0–7). Race was predominantly white (89.7%). The overall prevalence of stress urinary incontinence was 41.6% and 47.4% (P=.18) in identical and nonidentical twins, respectively. No differences were found between identical and nonidentical twins with respect to demographics, including age, race, parity, body mass index, menopausal status, tobacco use, mode of delivery (vaginal or cesarean, singleton or multiple gestation), and prior pelvic surgery.
Concordance rates for stress urinary incontinence were similar in identical and nonidentical twins, 79.5% and 78.6%, (P=.83), respectively, indicating noncontributory genetic influences. The dominant influence of environment on stress urinary incontinence was confirmed by model fitting. Sequential structural equation modeling of the categorical data demonstrated that the CE model, including shared environmental and unique environmental effects, provided the best fit to the data (Table 2). Shared environmental factors contributed 77.6% (95% confidence interval [CI], 41.4–83.8; P<.001) of the variance among twins compared with unique environmental factors, which contributed 20.9% (95% CI, 15.8–26.7; P<.001). The heritability (genetic contribution of the variance) of stress urinary incontinence in the ACE model was 1.49%, which was not statistically significant (95% CI, 0–38.8; P=.46).
After removing the cohort of women who had prior antiincontinence or prolapse surgery, our analysis still revealed that the CE model including shared environmental and unique environmental effects fit the data best. The sensitivity analysis revealed that shared environmental factors between identical and nonidentical twins accounted for 72.6% (95% CI, 32.2–78.4%) of the variance, whereas unique environmental factors contributed 27.4% (95% CI, 21.3–34.1%), both statistically significant (P<.001). The heritability of stress urinary incontinence computed in the ACE model was 0.00% in both identical and nonidentical twins (95% CI, 0–41.7%; P=.50).
Additional analysis was performed for the subgroup of 458 twin pairs with “pure stress urinary incontinence.” The results indicate a similar pattern with respect to heritability. The “pure stress urinary incontinence” subgroup consisted of 406 identical pairs and 52 nonidentical twin pairs. The overall prevalence of pure stress urinary incontinence was 21.7% and 18.3% in identical and nonidentical twins, respectively (P=.52). The pairwise concordance rates in identical and nonidentical twins was not significantly different, 86.2% and 90.4%, respectively (P=.39), indicating noncontributory genetic effects. Shared environmental factors between identical and nonidentical twins accounted for 83.6% (95% CI, 38.5–89.8%) of the variance, whereas unique environmental factors contributed 16.3% (95% CI, 10.2–24.8%), both statistically significant (P=.003, <.001). The heritability of stress urinary incontinence computed in the ACE model was 0.00%, not statistically significant, in both identical and nonidentical twins (95% CI, 0–42.3%; P=.50).
This classic twin study, involving ACE modeling of more than 1,700 subjects, revealed no significant genetic contribution to stress urinary incontinence in middle-aged women. Identical twins were not found to have higher concordance for stress urinary incontinence than nonidentical twins, and environmental factors were the main contributors for the development of female stress urinary incontinence. Although the classic twin study design does not delineate which precise factors are unique compared with shared, previous twin research performed at our center suggests that obstetric factors such as mode of delivery represents the single most important factor.14 This work reported by Goldberg et al14 found stress urinary incontinence to be significantly more likely among women who had undergone vaginal (odds ratio [OR], 2.3; 95% CI, 1.1–4.5; P=.019) rather than cesarean delivery (OR, .44; 95% CI, 0.22–0.87; P=.019). Thus, stress urinary incontinence within our study population appears to be a disease trait determined by environment and not genetics with obstetric factors playing an important role.
Two previous European studies have examined the heritability of urinary incontinence using a classic twin model.15,16 Altman et al15 in an analysis of female twin pairs identified from crosslinkage of the Swedish Twin Registry and Swedish Inpatient Registry concluded that genetic and nonshared environmental factors contributed approximately 40% and shared environmental factors approximately 20% of the total variance associated with a history of antiincontinence surgery. Rohr et al16 in a cohort of 1,168 female twin pairs from the Danish Twin Registry reported that genetic factors contributed more toward the development of stress urinary incontinence in elderly women (70–94 years), whereas environmental factors played a more dominant role in middle-aged women (46–68 years). This study, along with ours, suggests that although “nurture” appears to play the major role before menopause and closer to the inciting obstetric event, it appears possible that “nature” may tread in as aging ensues. The results of our study strongly underscore the finding that environmental factors play a dominant role over genetics in middle-aged women; however, we were unable to comment on genetic, environmental, or both influences among the elderly or women with previous antiincontinence surgery applying the classic twin model as a result of the paucity of these subjects in our cohort. Of note, however, previous work by our group has found that the relative effect of obstetric factors such as mode of delivery diminishes with aging suggesting that genetic influences may take precedence.14
A post hoc power calculation revealed that our sample of 765 identical and 117 nonidentical twins had more than 80% power to detect 77.6% of variation contributed from shared environmental factors at a .05 significance level. Our data revealed only 1.49% of variation was contributed to genetic factors. Genetic influence to stress urinary incontinence was significantly ruled out (P=.003). Limitations to the current study include recall bias resulting from self-reported data and selection bias by the women who volunteered to participate. Women with more severe symptoms may have not attended the festival or may have been too embarrassed to participate. Also, self-reported zygosity through our questionnaire has not been previously validated and is subject to misclassification. Additionally, our study population was predominantly white and not a random sample. Thus, these findings may not apply to other ethnic groups or the general population. Furthermore, although female stress urinary incontinence is related to other pelvic organ prolapse disorders, the results of this study cannot be generalized to include pelvic organ prolapse, because this was not tested in this study.
This classic twin study suggests that female stress urinary incontinence in premenopausal childbearing women is determined by environmental factors rather than genetics. These findings should underscore the need to focus efforts on preventable environmental risk factors leading to stress urinary incontinence such as the potential effect of pelvic floor injury during pregnancy and childbirth.
1. Dooley Y, Kenton K, Cao G, Luke A, Durazo-Arvizu R, Kramer H, et al. Urinary incontinence prevalence: results from the National Health and Nutrition Examination Survey. J Urol 2008;179:656–61.
2. Haylen BT, De Ridder D, Freeman RM, Swift SE, Berghmans B, Lee J, et al. An International Urogynecological Association (IUGA)/International Continence Society (ICS) joint report on the terminology for female pelvic floor dysfunction. Neurourol Urodyn 2010;29:4–20.
3. Hägglund D, Walker-Engström M, Larsson G, Leppert J. Quality of life and seeking help in women with urinary incontinence. Acta Obstet Gynecol Scand 2001;80:1051–5.
4. Margalith I, Gillon G, Gordon D. Urinary incontinence in women under 65: quality of life, stress related to incontinence and patterns of seeking health care. Qual Life Res 2004;13:1381–90.
5. Norton PA, MacDonald LD, Sedgwick PM, Stanton SL. Distress and delay associated with urinary incontinence, frequency, and urgency in women. BMJ 1988;297:1187–9.
6. Mushkat Y, Bukovsky I, Langer R. Female urinary stress incontinence—does it have familial prevalence? Am J Obstet Gynecol 1996;174:617–9.
7. Hannestad YS, Lie RT, Rortveit G, Hunskaar S. Familial risk of urinary incontinence in women: population based cross sectional study. BMJ 2004;329:889–91.
8. Boomsma D, Busjahn A, Peltonen L. Classical twin studies and beyond. Nat Rev Genet 2002;3:872–82.
9. Shaw C, Das Gupta R, Williams KS, Assassa RP, McGrother C. A survey of help-seeking and treatment provision in women with stress urinary incontinence. BJU Int 2006;97:752–7.
10. Hedekar D, Gibbson R. Longitudinal data analysis. Hoboken (NJ): John Wiley & Sons Inc; 2006.
11. Neale MC, Boker SM, Xie G, Maes HH. Mx: statistical modeling. 6th ed. Richmond (VA): Department of Psychiatry, Medical College of Virginia; 2002.
12. Neale MC, Maes HM. Methodology for genetics studies of twins and families. Dordrecht (The Netherlands): Kluwer Academic; 2004.
13. Akaike H. Factor analysis and A1C. Psychometrika 1987;52:317–32.
14. Goldberg RP, Abramov Y, Botros S, Miller JJ, Gandhi S, Nickolov A, et al. Delivery mode is a major environmental determinant of stress urinary incontinence: results of the Evanston-Northwestern Twin Sisters Study. Am J Obstet Gynecol 2005;193:2149–53.
15. Altman D, Forsman M, Falconer C, Lichtenstein P. Genetic influence on stress urinary incontinence and pelvic organ prolapse. Eur Urol 2008;54:918–22.
16. Rohr G, Kragstrup J, Gaist D, Christensen K. Genetic and environmental influences on urinary incontinence: a Danish population-based twin study of middle-aged and elderly women. Acta Obstet Gynecol Scand 2004;83:978–82.