Birthweight and urinary incontinence after childbirth: a systematic review and meta-analysis

Highlights • High birthweight significantly increases risk of urinary incontinence after delivery.• Increasing birthweight (>3500 and >4000 g) leads to increasing risk of incontinence.• The association is also found in separate analyses on vaginal delivery, primiparous, and stress UI.


Introduction
Urinary incontinence (UI) is a common problem after childbirth. Prevalence estimates vary from 14 -45% [1]. A systematic review reported pooled prevalence of any UI to be 32-36% three months postpartum [2]. Reviews on epidural [3], episiotomy [4], cesarean section [2] and instrumental childbirth [5] have clarified their association with UI postpartum. There is inconsistency in the literature regarding remaining birth parameters. By identifying significant risk factors for UI among women after childbirth, future research can identify and validate preventive measures that can be targeted towards these women. Many cohort and cross-sectional studies have reported data on birthweight, but results have not been pooled. In studies using electromyography heavier babies have been associated with evidence of pudendal nerve damage in the pelvic floor after vaginal birth [6], with uncertain clinical significance. The objective of this study was to review the literature to identify studies reporting on the association of high birthweight on urinary incontinence after all modes of childbirth, and to perform meta-analyses on the association of high birthweight on UI after childbirth. If birthweight can be isolated as a risk factor for postpartum urinary incontinence, patients at particular risk can be identified.

Material and methods
Literature searches were done in Medline, Embase, Svemed+, and Cinahl. Additional search was done in ClinicalTrials.gov and Cochrane Database of Systematic Reviews. A librarian from the University in Bergen assisted in the search in May 09. 2014 and August 24. 2016. Additional search was done May 03. -14.2017. The search included the following MESH terms and free text; urinary incontinence, leak, urine, bladder, delivery, obstetric, postpartum, postpartum period, puerperium, birthweight, infant, new-born, large, small, SGA, LGA. Abstracts and articles in Norwegian, Danish, Swedish, or English were considered. Both conference abstracts and full publications were included. Additional literature was added based on authors' knowledge and after reading references in identified literature. Grey literature was not identified.
In the four-part PICO question for this systematic review, we compared women who gave birth with birthweight >4,000 g or >3,500 g, to women giving birth with birthweight <4,000 g or <3,500 g. The outcome was any UI, and stress UI after childbirth.
Search was done in headings and abstracts. Birthweight was sometimes one of several risk factors included in sub-analysis in papers, often not presented in the abstract, and thereby not found by the search strategy. Additional articles were added based on the authors' knowledge of relevant literature, and after reading references. Identified literature was reviewed separately by both authors. Articles evaluating obstetric risk factors for maternal urinary incontinence in title or abstract were reviewed in full by both authors. When discrepancies between the two authors occurred (Seim, Wesnes), the article was discussed with a third researcher (Rortveit, see Acknowledgement). Criteria for inclusion were that the article or conference abstract evaluated birthweight as a possible risk factor for maternal urinary incontinence, with results presented in Norwegian, Danish, Swedish, or English The process for selecting studies is presented in Table 1. Information about origin, study design, response rate, number of participants, method of data collection, adjusted results, time of UI, mode of childbirth, BMI, weight of new-born, age, parity, and main findings for all included studies were extracted.
No reviews of this topic were identified. For obvious reasons, no randomized controlled studies (RCT) on birthweight and UI have been conducted. A considerable number of cohort studies and cross-sectional studies of high quality have been performed. Even though RCT's provide the highest level of evidence, a summary of results from cohort studies and cross-sectional studies will be essential in order to evaluate a possible causal association between birthweight and UI.

Birth weight
Mean birth weight in Europe and USA [7] is approximately 3500 g. Weight cut-off at 3500 g and 4000 g gives information on the association between UI and birthweight beyond mean, as well as extreme birthweight, respectively. Weight cut-off on 3500 g and 4000 g were most common in identified studies, and were therefore chosen for this review.
Birthweight in one study was originally analysed according to the 50th and 90th percentile for birthweight (3,541 g and 4,180 g, respectively). These data have been re-categorized into 3500 g and 4000 g, and data has thereafter been reanalysed and stratified for mode of childbirth. Data were adjusted for BMI and weight loss after childbirth [8]. Birthweight in one study was categorized according to birthweight quartiles [9]; birthweight >3925 g from this study was included in analyses on the association between birthweight >4000 g and UI. Boyles et al used pounds [10]; birthweight of >8 lb. (3639 g) from this study was included in analyses on the association between birthweight >3500 g and UI.

Urinary incontinence
Information on UI after childbirth was categorized as questionnaires, objective testing, structured interviews, phone interviews, or reviews of existing medical records. No minimum cut-off for frequency, amount or severity of UI was set to be included in this article. Stress UI is more common after childbirth than urgency UI and mixed UI. The prevalence of pure stress UI is reported to be 2 -8 times higher than the prevalence of pure urgency UI in pregnancy [1].
The stress/urgency ratio is reduced postpartum as prevalence of stress UI decline. Several studies focus solely on stress UI [11][12][13][14], but few studies have data on urgency UI. Data on any UI was used in this meta-analysis. Separate analyses on available data on stress UI were also performed.

Assessment of quality and bias
Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Initiative has published recommendations on how to report data in cohort and cross-sectional studies [15]. STROBE was used to assess methodological quality. There was a high threshold to exclude studies. Studies with insufficient methodological information on study design, setting, statistical methods, and study participants were excluded from the systematic review and meta-analysis. Selection bias can affect the meta-analytic estimate. Selection bias was therefore considered in studies used in the meta-analyses, based on setting, study population and response rate. Risk of selection bias was considered as low, medium, high, and unclear. Information on studies included in meta-analyses regarding selection bias, adjustment of effect estimates, and reporting of all data (effect estimate, N in exposed and unexposed group, N with UI and continence) were collected. Funnel plot asymmetry based on standard error (log [OR]) was used to explore possible reporting biases.

Data synthesis
Birthweight was categorized as a dichotomous variable with two categories; <4000 g vs >4000 g, and <3500 g vs >3500 g, respectively. Available original data from included studies on birthweight (!4000 g and !3500 g, respectively) and UI were extracted. Adjusted effect estimates were extracted when available. Unadjusted effect estimates were used when adjusted estimates were not presented; raw data and absolute numbers were then converted to unadjusted OR. Relative risk in one study was treated as OR [16]. To enable comparison across studies, Log [OR] with SE were calculated for each included study. Results were pooled and combined in meta-analyses. Estimates were inserted into Review Manager 5.3 for meta-analyses.
To reduce diversity in study characteristics, separate sub-group meta-analyses were performed according to type of UI, mode of Table 1 Flowchart of included and excluded studies.    delivery, primiparous women, and UI 3 -18 months postpartum.
Mode of delivery was categorized as any vaginal delivery or any CS. Main findings on the association between birthweight and UI are presented as odds ratio, and in Forest plot figures. Both adjusted data and unadjusted data are presented in Forest plot figures. Heterogeneity among studies was assessed by I2. An I2 of 0% to 40% represents minimal heterogeneity, while 75% to 100% represents considerable heterogeneity. Adjusted data and unadjusted data had in general moderate to substantial heterogeneity in effect estimates. Random effect estimates were therefore used.
The review was registered in PROSPERO (73021); NHS' International prospective register of systematic reviews. The review adheres to the PRISMA guidelines and MOOSE guidelines for meta-analyses and systematic reviews of observational studies.

Results
A total of 477 articles were identified. Fifteen external articles were added based on the authors' knowledge of relevant literature, and after reading references. A total of 385 articles remained after removing duplicates (Table 1). Fifty-seven articles (N = 164,600) were included in this systematic review. Descriptive data are presented in Table 2. Twenty-two articles had data that could be included in the meta-analyses. Descriptive data are presented in Table 3.
Selection bias was considered in studies used in the metaanalyses. Risk of selection bias was considered as high in 2/22 studies [13,17], moderate in 2/22 studies [18,19], unclear in 2/22 studies [12,20], and low in 16/22 studies (Table 3). Unadjusted association between birthweight and UI was reported in 9/22 studies. Funnel plot did not reveal publication bias, as it spread evenly on both sides of the average, creating a roughly funnelshaped distribution.
There was a significant positive association between high birthweight and UI after childbirth in 35% (20/57) of the studies. There was a non-significant positive association in 19% (11/57) of the studies. There was no association in 46% (26/57) of the studies. A significant protective association between high birthweight and urgency UI was also found in one of the above studies.

Birthweight >4000 g and stress UI
Data from four European [11][12][13][14] and one American [23] study were available for meta-analyses. Time of recording UI varied from 3 days postpartum [11] to several decades after childbirth 14] Weight >4000 g was associated with a significant increased risk of stress UI (OR 1.52, 95% CI 1.03 -2.25) when analysing available data from these five studies with a total of 15,806 women.

Birthweight >4000 g and UI after vaginal birth
Seven studies gave data for meta-analyses on 19,907 women on the association between UI and birth weight >4000 g among women delivering by any vaginal birth; three cohort studies [8,24,30] and four cross-sectional studies [11][12][13][14] were identified. Three large studies with adjusted data were included; Rortveit et al.14] enrolled 11,397 women; Wesnes et al. [8] enrolled 5,219 women, and Thom et al. enrolled 1,521 women. However, mean age, parity and time of UI varied in these studies. Weight >4000 g was associated with a significantly increased risk of UI after vaginal birth (OR 1.41, 95% CI 1.14 -1.75) (Fig. 3).
Only one study had additional data on birthweight >4000 g and birth by CS [8]. OR for UI after birth by any CS of child >4000 g compared to <4000 g was 1.38 (95% CI 0.84 -2.28).

Birthweight >4000 g and UI among primiparous women
Five cohort studies [8,20,23,26,29] and one cross-sectional study [18] had data on 11,643 women for meta-analyses on birthweight >4000 g among primiparous women. All studies had data on UI 2 -18 months postpartum. Three studies included only women who were continent before pregnancy [8,26,29]. Weight >4000 g was associated with a non-significantly increased risk of UI among primiparous women (OR 1.46, 95% CI 0.95 -2.26). Only two studies gave adjusted effect estimates [8,20], leading to an OR of 2.48 (95% CI 0.70 -8.71) for UI among primiparous women delivering babies >4000 g compared to <4000 g. However, due to heterogeneity in effect estimates, I2 was 96%. Unadjusted analyses gave OR of 1.15 (95% CI 0.88 -1.50). Fig. 1. Forest plot of the association between urinary incontinence and birthweight >4000 g vs <4000 g, stratified for adjusted and unadjusted data. Fig. 2. Forest plot of the association between urinary incontinence 3 -18 months postpartum, and birthweight >4000 g vs <4000 g, stratified for adjusted and unadjusted data.

Birthweight >3500 g stress UI
Two studies had unadjusted data on the association between birthweight >3500 g and stress UI [16,31]. Data was collected 6 weeks -1 year after childbirth. None of these studies reached statistical significance on the association between birthweight >3500 g and stress UI. Birthweight >3500 g was associated with a non-significantly increased risk of stress UI in meta-analyses of 2525 women (OR 1.33, 95% CI 0.97 -1.82).

Birthweight >3500 g and UI after vaginal childbirth
Three large studies with 5599 [10], 2390 [16], and 5219 [8] participants had data on the association between birthweight >3500 g and UI after vaginal childbirth. All studies presented adjusted data on UI 3 -12 months postpartum with OR 1.22, 1.30 and 1.25, respectively. In meta-analyses, weight >3500 g was associated with a significantly increased risk of UI after vaginal childbirth (OR 1.26, 95% CI 1.15 -1.37). I2 was 0%.

Birthweight >3500 g and UI among primiparous women
Four large questionnaire-based studies investigated the association of birthweight >3,500 g on UI [8,10,16,32] in women 3 -12 months after childbirth. Three studies reported adjusted results. Two studies found a significant positive association [8,10] between birthweight >3,500 g and UI among primiparous women, two Fig. 3. Forest plot of the association between urinary incontinence after any vaginal delivery, and birthweight >4000 g vs <4000 g, stratified for adjusted and unadjusted data. Fig. 4. Forest plot of the association between urinary incontinence and birthweight >3500 g vs <3500 g, stratified for adjusted and unadjusted data. studies found a non-significant positive association 16,32]. Metaanalyses from these studies gave an OR of 1.23 (95% CI 1.11 -1.35) for UI among primiparous women delivering infants with birthweight >3500 g compared to <3500 g.

Discussion
This is the first systematic review on the association between birthweight and UI after childbirth. UI postpartum is a common condition, affecting 32 -36% of women [2]. Many risk factors have been brought forth in studies, most are not well documented. Reviews on epidural [3], episiotomy [4], cesarean section [2] and instrumental childbirth [5] have made their association with UI postpartum clear. Some birth variables are more commonly extracted from birth registries (birthweight, head circumference, rupture) and often included in analyses. There has been need to summarize knowledge on these potential risk factors.

Main findings
Birthweight >4000 g and >3500 g were associated with significantly increased risk of UI (OR 1.49 and 1.26, respectively). Separate analyses on stress UI, UI 3 -18 months postpartum, UI after vaginal birth and UI among primiparous women also revealed significantly increased risk of UI.

Strengths and limitations
Prevalence of UI after childbirth varies with time of information gathering, type of UI, mode of childbirth, and characteristics of the study population [1]. Diversity in studies included in metaanalyses needs to be addressed. To control for some of these parameters, separate meta-analyses were performed for time of UI, stress UI, vaginal birth, and parity. Subgroup meta-analyses on the reported variables showed significantly increased OR in the range 1.41 -1.54 for UI and birth weight >4000 g. Corresponding analyses on UI and birth weight >3500 g gave significantly increased OR in the range 1.23 -1. 33. This indicates that the overall risk estimates are robust. Studies have shown that selection bias affects prevalence estimates, but data are still valid for risk estimates [33]. High birthweight is also associated with high BMI in the mother, prolonged birth, CS, rupture, episiotomy, birth by forceps and vacuum. We cannot rule out confounding.
A total of 33/57 studies report on the association between birthweight and UI as secondary finding, without authors reporting values, percentages or risks. Information from these articles was applicable for this review, but not for the metaanalysis.
The literature search also needs to be addressed. Birth weight and UI was the main objective in few articles. Literature search retrieved information from headings and abstracts, therefore it did not retrieve articles where relevant information was solely in the main text. 476 articles were identified by literature search. Author's knowledges of literature, and full reference reading identified 15 external articles (Table 1). A total of 27 articles were not evaluated for inclusion due to foreign language or lack of access. We must therefore accept that the sensitivity of the search was high, but not complete. Publication bias may be a problem, as significant data are more likely to be published, and presented in abstracts.
Categorizing of weight groups and reference groups can affect the results. Birthweight >3500 g and > 4000 g led to OR 1.26 and 1.49, respectively. The reference groups were birthweight <3500 g and < 4000 g, respectively. One study with 5,219 women used lower reference value [8]. OR of UI was 1.6 (95% CI 1.2 -2.0) after birthweight >4,180 g (90 percentile) compared to birthweight 3,540 g (50 percentile) [8]. Another study compared birthweight >3,925 to <3,199, finding OR of 1.4 [28]. Risk estimates in studies might have been higher if reference groups consisted of lower birthweights.
Few studies use weight groups >4500 g [16,34,35]. Few study participants in the exposure group makes it difficult to do metaanalyses. One study reported adjusted OR 1.23 (95% CI 0.87 -1.76) for UI among women delivering babies with birthweight >4,500 g compared to < 4,500 g 35].
Vaginal birth is associated with higher risk of UI than CS in most studies [2]. Weight >4000 g was associated with UI after vaginal birth and CS (OR 1.41 95% CI 1.14 -1.75 and OR 1.38, 95% CI 0.84 -2.28, respectively). Weight >3500 g was also associated with UI after vaginal birth and CS (OR 1.26, 95% CI 1.15 -1.37, and OR 1.04 95% CI 0.67 -1.63, respectively). These meta-analyses on birthweight and UI are in line with what is previous shown regarding birth mode and UI [2]; lower risk of UI after CS, than after vaginal birth. Studies including women delivering by vaginal birth and CS indicate that birthweight is a significant risk factor for UI only in association with vaginal birth [8,35]. Mode of birth is a confounder that is likely to affect the association between birthweight and UI in the remaining sub-analysis. There were, however, no data available for further stratified sub-analysis on CS and vaginal birth.
It is unclear how high birthweight lead to UI. Trauma to the pelvic floor when delivering large babies is a plausible contributing factor. Heavier babies have been associated with EMG evidence of pudendal nerve damage in the pelvic floor after vaginal birth [6,36].
The majority of studies report grams, others report pounds [10]. Several studies analysed on mean birthweight [37], some use percentiles [8] or quartiles [38]. Birthweight was often categorized arbitrary without authors reporting reason for cut-off, most studies uses 500-gram groups (3500 g, 4000 g, and 4500 g). Varying reporting and different weight categorizing makes it difficult to summarize published results. Weight cut-off on 3500 g and 4000 g were most common, and were therefore chosen for this review. As 3500 g generally represents mean birth weight, these weight cut-offs gives important information on the association between UI and birthweight beyond mean, as well as extreme birthweight >4000 g.

Interpretation
Birthweight appears to be a risk factor for UI after childbirth. Data in this meta-analysis are gathered from epidemiological cohort and cross-sectional studies. When planning future epidemiological studies, data on birthweight ought to be gathered when possible. Birthweight is likely to effect risk estimates on UI after deliver, and adjustment should be considered in future research.
The Hill Criteria are useful when considering a possible causal relationship between exposure and outcome. Birthweight satisfies several of the Hill Criteria for causation; it is biologically plausible, exposure precedes outcome, and there is consistency in the majority of studies. There is a dose-response association, as birthweight >4000 g gave higher OR for UI than >3500 g.
To predict future risk of UI after childbirth, the UR-CHOICE risk calculator intend to include eight variables, of which birthweight is the only fetal factor [39]. Birthweight is included in the risk calculator due to the high likelihood of causation.
Even though results appear biological plausible, results in the meta-analysis are based on epidemiological research and can thereby not automatically be applied into a clinical setting. As high birthweight is associated with UI postpartum, clinical preventive strategies ought to be identified. Strategies might be targeted on identifying mothers at risk of having babies with high birthweight (for instance by identifying high maternal BMI, identifying previous deliveries of babies with high birthweight), avoiding high birthweight (for instance by avoiding high maternal weight gain during pregnancy, detecting gestational diabetes), detecting high birthweight (for instance by growth charts, ultrasound, symphysis-fundus height measurements), reducing risk of incident UI postpartum (for instance by aiming at normal weight before pregnancy, and at regaining pre-pregnancy weight postpartum, considering CS, performing PFMT), or by treating UI postpartum, for instance by PFMT. Future research will need to find ways to identify which women are likely to give birth to babies >3500 g, and look into the best preventive strategies.
Pelvic floor muscle training is generally recommended in pregnancy and postpartum. It has been unclear which women benefit the most from this training. Women delivering babies with birthweight > 3500 g are women at risk of developing UI. No preventive strategy is validated in this study, but pelvic floor muscle training has documented effect on preventing UI among women delivering heavy babies [40].

Conclusion
We conclude that birthweight appears to be a risk factor for UI after childbirth. A causal relationship between birthweight and UI is biologically plausible. Strategies towards preventing UI postpartum should be targeted towards women at higher risk, like women giving birth to babies with high birthweight.
Author contribution to the manuscript Wesnes: Project development, data collection, manuscript writing, statistical analysis.

Funding, financial disclaimers
None.

Declaration of Competing Interest
The authors report no declarations of interest.