Stress urinary incontinence symptoms are prevalent and have significant socioeconomic and medical impact.1 The etiology of stress urinary incontinence is not fully understood, but trauma to the pelvic floor during childbirth is suggested as a major risk factor.2 Several studies have shown associations between pregnancy, delivery, and stress urinary incontinence, but the results are not consistent.3–7 Multiparity compared with uniparity is associated with an increased risk for stress urinary incontinence in some studies,5,7 whereas other studies found no such association.8,9 The degree of trauma is likely to be affected by various conditions, events, and interventions during pregnancy and delivery. For example, there is conflicting evidence about the association between birth weight and maternal risk for developing persisting stress urinary incontinence.8,10–12 It is also unclear whether forceps or vacuum extraction,8,11,13,14 episiotomy,8–11,15,16 or spontaneous vaginal tear11,13,15,17 are associated with later development of stress urinary incontinence. Women delivered by cesarean seem to have less risk of later stress urinary incontinence than those delivered vaginally, according to some studies,10,18,19 but not according to another study.4 No good data are available regarding the possible relationships between fetal presentation, multiple birth, or maternal age at the time of delivery for later development of stress urinary incontinence.
The aim of this study was to identify and evaluate obstetric and demographic risk factors for stress urinary incontinence, as represented by a history of stress urinary incontinence surgery.
Materials and Methods
We linked three Swedish population–based national registries for the following analyses. We used the unique personal identification number assigned to each resident in Sweden to link the registries. This number is widely used in Swedish society, including health care.
The Hospital Discharge Registry kept by the National Board of Health and Welfare, Stockholm, was used to identify a cohort of Swedish women, born between 1932 and 1977, who had had surgery for stress urinary incontinence during the years 1987–1996 (n = 10,074). These women were identified by the appropriate surgical procedure codes. Some women appeared at more than one occasion in the registry, indicating more than one operation for stress urinary incontinence. For those women (n = 643), only the first event of surgery was used for the analysis.
The Fertility Registry, kept by Statistics Sweden, contains information on the number of children born by each Swedish woman through 1997. We determined from this registry the number of children delivered by each woman whose birth date was between 1932 and 1977, and who had undergone surgery for stress urinary incontinence. We also constructed a table of women who had no, one, two, three, and four or more children according to their year of birth and calculated an odds ratio (OR) for having had incontinence surgery in relation to the number of children delivered by the women.
The Medical Birth Registry, kept by the National Board of Health and Welfare, contains prenatal, obstetric, and neonatal information from 98% to 99% of all women who delivered a baby in Swedish hospitals in 1973 or later. Data have been (and still are) collected consecutively in all Swedish hospitals by using the same standardized forms.20 In linking this registry with the Hospital Discharge Registry, we identified women who had delivered in 1973 or later and who had been operated on for stress urinary incontinence (n = 4634). Women born outside Sweden (n = 455), women who had their first delivery before 1973 (n = 2213), and women who had been operated on before their first delivery (n = 4) were excluded. We also excluded women for whom the data base showed unknown or absurd birth weight, or erroneous year of delivery (n = 20). Thus, 1942 women and 4437 infants remained for analysis. Delivery outcome for those women was compared with data for all women (by using the same exclusion criteria) in the Medical Birth Registry. Some information used in the registry was not available until 1983 (ie, maternal smoking, body mass index [BMI]).
The use of the Medical Birth Registry enabled an analysis of the effect of various maternal characteristics and obstetric conditions on the risk for surgery for stress urinary incontinence. We often restricted analyses to women who had only one child, notably when the variable under study showed a negative association with stress urinary incontinence. When a variable was positively associated with stress urinary incontinence, we included all cohort women for the analysis, and if possible, studied primiparas separately.
We estimated the association of pregnancy per se with stress urinary incontinence surgery by calculating the OR for incontinence surgery for nulliparous women compared with the OR for those with one delivery; or for women with one elective cesarean compared with women with one normal delivery. We then compared those ORs.
For statistical analysis, we used the Mantel-Haenszel procedure21 for the determination of OR after appropriate stratifications as specified. Approximate 95% confidence intervals (CIs) were calculated by using a test-based method, where the 95% CIs are OR(1 ± 1.96/√χ2) (Miettinen OS. Simple interval-estimation of risk ratio [abstract]. Am J Epidemiol 1974;100:515–6). Trend analyses of ORs were based on weighted linear regressions by using the same variances as those used to determine the 95% CIs.
Table 1 shows demographic characteristics of women identified by linking the Medical Birth Registry and the Hospital Discharge Registry.
Table 2 shows the OR for having incontinence surgery according to the number of infants delivered by the woman, based on the linkage between the stress urinary incontinence surgery cohort and the Fertility Registry (nulliparous women used as reference). As shown in the table, there is a statistically highly significant relationship between the risk of having incontinence surgery and the number of children born (P < .001).
The remaining part of the analysis was based on data obtained by the linkage between the stress urinary incontinence surgery cohort and the Medical Birth Registry. First, we searched for confounders. There was a very strong and obvious effect of year of delivery: in 1973, 283 delivered women underwent later stress urinary incontinence surgery, but only one among those delivered in 1995. The effect of year of the first delivery (data not shown), maternal age at the first and last delivery, and parity at the last delivery were analyzed. Apart from the analysis of maternal age at the first delivery, when stratification for age at the last delivery was not made, all variables were stratified for the other three variables and year of birth. All variables except maternal age at last delivery were strongly related to the risk of having stress urinary incontinence surgery, and in the further analysis, stratification for these variables was made.
The OR for elective cesarean compared with noninstrumental vaginal singleton delivery among women with only one delivery was 0.21 (95% CI 0.13, 0.34), and for any cesarean, it was 0.34 (95% CI 0.23, 0.52). Vaginal delivery compared with elective cesarean thus had an OR of 4.76, similar to that for all parous women compared with nulliparous women (Table 1). A geographic variability was found both for cesarean delivery rate and for stress urinary incontinence surgery rate (data not shown), but the two did not covary and therefore were not taken into consideration.
We studied the effect of the birth weight of the largest infant delivered by women who had only noninstrumental vaginal singleton deliveries. We used all other infants as a reference. Table 3 shows that the OR increases with increasing birth weight.
Table 4 shows the effect on the OR for later stress urinary incontinence surgery of maternal age at the first and last delivery among women who had only noninstrumental vaginal singleton deliveries. Increasing age at the first delivery, but not the last delivery, was associated with later incontinence surgery.
Fourteen of 1463 women with only noninstrumental, vaginal singleton deliveries had maternal diabetes (including both type I diabetes and gestational diabetes). The analysis was stratified for episiotomy. The OR for later stress urinary incontinence surgery was 2.12 (95% CI 1.36, 3.28) without stratification for largest birth weight, and 2.07 (95% CI 1.32, 2.33) with such stratification.
Smoking status during any pregnancy was registered for 422 incontinent women. After stratification for largest birth weight and diabetes, the OR for the association between smoking and later surgery for stress urinary incontinence was 0.89 (95% CI 0.71, 1.12).
No effect on the risk for stress urinary incontinence surgery was seen by maternal level of education, evaluated as the “final” education in 1996 (known for 1434 stress-incontinent women). An education of at least 12 years (equal to upper secondary school in the United Kingdom or senior high school in the United States) gave an OR of 1.01 (95% CI 0.81, 1.25). Stratification was also made in the analysis for the largest birth weight.
As seen in Table 5, there was an increased risk of later stress urinary incontinence surgery with increasing BMI. Stratification for largest birth weight reduced the ORs, and a significant effect remained only when high and low BMIs were compared.
We then studied the associations between various traumatic events and interventions during delivery and the risk for stress urinary incontinence surgery. Having had one or more multiple births (n = 28 incontinent women) tended to be associated with later stress incontinence, even though the association did not quite reach significance (OR = 1.40, 95% CI 0.96, 2.06, after stratification for episiotomy and largest birth weight and exclusion of instrumental and cesarean deliveries and diabetes). Without stratification for largest birth weight, the OR for incontinence was 1.23 (95% CI 0.84, 1.81).
The OR for stress urinary incontinence surgery following breech presentation in singleton, vaginal births, excluding abnormal head presentation and stratifying for birth weight and diabetes, was 1.42 (95% CI 0.82, 2.47, analyses based on 13 among 1612 women).
Abnormal fetal head presentation (occiput posterior, brow, and facial presentation) in singleton, vaginal births (excluding instrumental deliveries and breech presentation and stratifying for birth weight and maternal diabetes) was not significantly associated with stress urinary incontinence surgery (OR = 1.25, 95% CI 0.95, 1.68). Even though no association was seen among women who had their delivery during the first half of the observation period, 1973–1982 (OR = 1.01, 95% CI 0.66, 1.54), a significant association was seen for the second half, 1983–1996 (OR = 1.60, 95% CI 1.05, 2.43). However, these two OR estimates did not differ significantly (P = .14).
Vacuum extraction or forceps delivery was performed during 189 of 1599 singleton births with normal head presentation (41 among 306 women who had only one child). After stratification for diabetes, the OR for later stress urinary incontinence surgery was 0.82 (95% CI 0.70, 0.95) without largest birth weight stratification, and 0.78 with such stratification (95% CI 0.67, 0.92). The OR did not change to any extent when we restricted the analysis to women with only one child (OR = 0.76, 95% CI 0.53, 1.07).
Mediolateral episiotomy was performed in 479 of 1463 women with singleton births without instrumental delivery. After stratification for maternal diabetes and birth weight, the OR for the association with stress urinary incontinence surgery was 0.82 (95% CI 0.68, 0.98), and restricting the analysis to women with only one child (63 among 266), the OR was 0.71 (95% CI 0.54, 0.95).
Large spontaneous perineal tear (partial or total rupture of anal sphincter) was registered in 12 among 949 deliveries in singleton, noninstrumental vaginal births without episiotomy. After stratifying for largest birth weight and diabetes, the OR for stress urinary incontinence surgery was 0.90 (95% CI 0.51, 1.59).
Epidural analgesia was registered in 232 of 1449 women who had singleton births. After excluding women with diabetes and instrumental deliveries, and stratifying for birth weight and episiotomy, the OR for stress incontinence was 1.41 (95% CI 1.22, 1.64). The OR was 1.27 (95% CI 0.88, 1.85) when the analysis was restricted to 266 women with only one delivery (37 with epidural analgesia).
This study shows significant associations between surgery for stress urinary incontinence and previous vaginal birth per se, diabetes mellitus including gestational diabetes, maternal age at first delivery, epidural analgesia during vaginal delivery, and BMI at the woman's last pregnancy. The risk for surgery after vaginal delivery increased with the weight of a woman's largest infant. In contrast, delivery with forceps or vacuum extraction and mediolateral episiotomy were negatively associated with surgery.
In contrast to a previous study,4 we did not find pregnancy itself to be associated with later surgery for stress urinary incontinence. However, the OR for nulliparous women compared with uniparous women was obtained by linking the Medical Discharge Registry to the Fertility Registry. In the latter registry, information on mode of delivery is not available. Thus, women with cesarean delivery are included. This will bias ORs toward 1.0. However, over the years, the mean cesarean rate in Sweden has been 10.5% (data obtained from the Medical Birth Registry). The cesarean rate was lower for women delivered during the first years of registration when most women who underwent surgery were recruited. Thus, the bias is probably small. Therefore, we considered it appropriate to compare this OR with the OR for uniparous women delivered with elective cesarean and with uniparous women delivered vaginally to estimate the impact of pregnancy without birth trauma.
The association between stress incontinence and maternal diabetes is in agreement with previous studies, showing a relationship between diabetes and incontinence, especially in the elderly,11,22 and we report an association between diabetes mellitus of fertile age and later surgery for stress incontinence. One might speculate that an increased vulnerability of the pelvic floor due to changes in connective tissue biology and innervation of pelvic muscles are pathogenetic factors for development of stress incontinence in women with diabetes mellitus.
An association between the number of vaginal deliveries and stress incontinence has been reported previously.5–7 In this study, the first birth seems to be the most harmful to the pelvic floor; subsequent deliveries moderately increase the risk for incontinence. This tendency might be even stronger than shown in the study. First, we have reason to believe that stress incontinence surgery was performed after a woman's last delivery because it is uncommon to operate for stress urinary incontinence before a patient has completed her family. Second, we have no information on when incontinence symptoms started, ie, after which birth stress urinary incontinence developed. Thus, it is possible that the symptoms tended to start after the first birth, but the surgery was not performed until after the last birth.
We found a positive association between stress incontinence and age at first delivery, but no such association with age at last delivery. This finding suggests an increasing vulnerability of the pelvic floor with age in women with no previous delivery. This, however, is in contrast to a study by Foldspang et al.5
We found an association between incontinence surgery and a high BMI. Deitel et al23 have shown a decrease of incontinence symptoms after massive weight loss, supporting the theory of BMI as an aggravating condition rather than a true risk factor. Contrary to the results of this study, other authors in smaller studies have found smoking and level of education to be associated with incontinence.19,24,25
The impact of various interventions and events during labor and delivery on the development of stress incontinence, including forceps/vacuum delivery, episiotomy, spontaneous perineal tear, and epidural analgesia, is controversial.8–11,13–17 Our findings of negative associations between instrumental delivery/episiotomy and incontinence surgery, and lack of association between spontaneous perineal tear and later surgery, are intriguing. However, a hypothesis of a protective effect of instrumental delivery is difficult to formulate. According to a recent review article, no association was found between the use of episiotomy and later development of stress urinary incontinence. However, anterior perineal trauma occurred more often in a group of women in which episiotomy was performed less often. This tendency indicates that episiotomy might protect against trauma in this region. Stress urinary incontinence was an outcome variable in only two of six included studies (Carroli J, Belizan J, Stamp G. Episiotomy for vaginal birth. The Cochrane Library, 1999 (3) [Cochrane review]). The OR was similar for instrumental delivery and episiotomy in our analyses. Mediolateral episiotomy is not generally offered to women in Sweden, but is restricted for selected cases (ie, in practically all cases of instrumental delivery). Thus, it is possible that the negative association found for instrumental delivery depends entirely on the concomitant episiotomy. The episiotomy might also decrease the impact of a large infant, explaining that the ORs were almost similar with and without stratification for birth weight. Also, the indication for the intervention (ie, fetal hypoxia compared with arrest in second stage) may be of importance for the outcome in terms of stress incontinence after these procedures. Moreover, differences in connective tissue elasticity may confound the analysis.
Epidural analgesia is considered to be protective against stress incontinence in one study,8 but not in another study.10 We found a positive association between epidural analgesia and later incontinence surgery. We do not know whether epidural analgesia might have indirectly promoted stress incontinence by prolonging the second stage of labor,26 or whether the association was confounded by the reason for epidural analgesia.
A strength of this study is the large number of cases from a relatively homogenous population of Swedish women. Moreover, the risk for misclassification of cases was considered small because of the use of surgery as an indicator for stress incontinence. In doing so, we assumed that Swedish urogynecologists in general had accepted the principle of recommending surgery only to women with severe and persistent forms of stress urinary incontinence, thus excluding women with other forms of urinary incontinence and transient stress urinary incontinence during pregnancy and after delivery. Furthermore, by using this indicator, we eliminated observer and responder bias that might occur in studies based on interviews or questionnaires.
At the same time, the use of operative intervention as an indicator for stress incontinence is one of the weaknesses of the study. Surgery is performed for only a fraction of all women suffering from stress urinary incontinence. That is, women undergoing stress urinary incontinence surgery probably are selected because they had the most severe symptoms. However, the expected dilution effect created by incontinent women who do not undergo surgery in the population cohort is probably small, but may have reduced associations. The same effect is created by excluding operated women who had their first delivery before the Medical Birth Registry was initiated (1973). Moreover, in registry studies of this kind, some misclassification will occur. This will, generally speaking, bias all ORs toward 1.0 because misclassification will be random and cannot be associated with future surgery for stress urinary incontinence. A validity study of the Medical Birth Registry has been published.20
1. Rane A. Incontinence in women. Don't suffer it. Aust Fam Physician 1999;28:584–6.
2. Handa VL, Harris TA, Ostergard DR. Protecting the pelvic floor: Obstetric management to prevent incontinence and pelvic organ prolapse. Obstet Gynecol 1996;88:470–8.
3. Thomas TM, Plymat KR, Blannin J, Meade TW. Prevalence of urinary incontinence. BMJ 1980;281:1243–5.
4. Iosif CS, Ingemarsson I. Prevalence of stress incontinence among women delivered by elective cesarean section. Int J Gynaecol Obstet 1982;20:87–9.
5. Foldspang A, Mommsen S, Lam GW, Elving L. Parity as a correlate of adult female urinary incontinence prevalence. J Epidemiol Commun Health 1992;46:595–600.
6. Milsom I, Ekelund P, Molander U, Arvidsson L. The influence of age, parity, oral contraception, hysterectomy, and menopause on the prevalence of urinary incontinence in women. J Urol 1993;149:1459–62.
7. Ryhammer AM, Bek KM, Laurberg S. Multiple vaginal deliveries increase the risk of permanent incontinence of flatus urine in normal premenopausal women. Dis Colon Rectum 1995;38:1206–9.
8. Dimpfl T, Hesse U, Schussler B. Incidence and cause of postpartum urinary stress incontinence. Eur J Obstet Gynecol Reprod Biol 1992;43:29–33.
9. Skoner MM, Thompson WD, Caron VA. Factors associated with risk of stress urinary incontinence in women [comments]. Nurs Res 1994;43:301–6.
10. Viktrup L, Lose G, Rolff M, Barfoed K. The symptom of stress incontinence caused by pregnancy or delivery in primiparas. Obstet Gynecol 1992;79:45–9.
11. Minaire P, Jacquetin B. The prevalence of female urinary incontinence in general practice. J Gynecol Obstet Biol Reprod (Paris) 1992;21:731–8.
12. Krue S, Jensen H, Agger AO, Rasmussen KL. The influence of infant birth weight on post partum stress incontinence in obese women. Arch Gynecol Obstet 1997;259:143–5.
13. Tetzschner T, Sorensen M, Jonsson L, Lose G. Delivery and pudendal nerve function. Acta Obstet Gynecol Scand 1997;76:324–31.
14. Brown S, Lumley J. Maternal health after childbirth: Results of an Australian population-based survey. Br J Obstet Gynaecol 1998; 105:156–61.
15. Röckner G. Urinary incontinence after perineal trauma at childbirth. Scand J Caring Sci 1990;4:169–72.
16. Berthet J, Buchet A, Favier M, Racinet C. Complete and complicated tears of the perineum in spite of preventive incision. A study of the risk factors and the short-term consequences. J Gynecol Obstet Biol Reprod (Paris) 1993;22:419–23.
17. Snooks SJ, Swash M, Henry MM, Setchell M. Risk factors in childbirth causing damage to the pelvic floor innervation. Int J Colorectal Dis 1986;1:20–4.
18. van Geelen JM, Lemmens WA, Eskes TK, Martin CB Jr. The urethral pressure profile in pregnancy and after delivery in healthy nulliparous women. Am J Obstet Gynecol 1982;144:636–49.
19. Wilson PD, Herbison RM, Herbison GP. Obstetric practice and the prevalence of urinary incontinence three months after delivery. Br J Obstet Gynaecol 1996;103:154–61.
20. Cnattingius S, Ericson A, Gunnarskog J, Källen B. A quality study of a medical birth registry. Scand J Soc Med 1990;18:143–8.
21. Mantel N, Haenszel W. Statistical aspects of the analyses of data from retrospective studies of disease. J Natl Cancer Inst 1959;22:719–48.
22. Brown JS, Seeley DG, Fong J, Black DM, Grady D. Urinary incontinence in older women: Who is at risk? Study of Osteoporotic Fractures Research Group. Obstet Gynecol 1996;87:715–21.
23. Deitel M, Stone E, Kassam HA, Wilk EJ, Sutherland DJ. Gynecologic-obstetric changes after loss of massive excess weight following bariatric surgery. J Am Coll Nutr 1988;7:147–53.
24. Bump RC, McClish DK. Cigarette smoking and urinary incontinence in women. Am J Obstet Gynecol 1992;167:1213–8.
25. Burgio KL, Locher JL, Zyczynski H, Hardin JM, Singh K. Urinary incontinence during pregnancy in a racially mixed sample: Characteristics and predisposing factors. Int Urogynecol J Pelvic Floor Dysfunct 1996;7:69–73.
26. Alexander JM, Lucas MJ, Ramin SM, McIntire DD, Leveno KJ. The course of labor with and without epidural analgesia. Am J Obstet Gynecol 1998;178:516–20.