I leak a little when I run hard. Is it my sport?
Sport type did not explain it: in 32 studies (4,649 women athletes, 18-45) 39% leaked; cohorts training more hours a week leaked a little more (+6% odds/h). Pelvic-floor training helps.
Main caveat: Prevalence, not incidence. Almost all primary studies describe leakage at one time.
Domínguez-Pérez and colleagues asked how common urinary incontinence is in adult women athletes and whether the sport classifications used to explain it — discipline, modality (endurance, ball games, aesthetic and so on), impact on the pelvic floor, competitive level — or training volume account for the wide spread between studies.
They first gathered every systematic review since 2010 on the question (14, of which AMSTAR 2 rated 12 critically low and 2 low confidence) and then, because those reviews mixed adolescents with older women, parous with nulliparous women and elite with recreational athletes, re-pooled the primary studies themselves, restricted to women aged 18-45: 32 studies, 4,649 athletes, 93.9% nulliparous on average, most using the ICIQ-UI Short Form over a 4-week recall.
The pooled prevalence of any leakage was 39% (95% CI 33-46), stress incontinence 33% and exercise-related leakage 39%, with very large heterogeneity.
Point estimates ranged from 50% in rugby to 27% in swimming and were higher in professionals (49%) than amateurs (32%), but in meta-regression — the authors' primary analysis — neither sport discipline, sport modality nor impact category explained the spread (p = 0.304, 0.149 and 0.253), and competitive level was a trend (p = 0.075) that disappeared in nulliparous-only cohorts.
Weekly training hours did go with leakage: OR 1.06 per extra hour a week (95% CI 1.01-1.11), 1.38 per 5.24 hours, consistent across impact categories, competitive levels and both sensitivity analyses.
The authors conclude that sport labels "may misdirect screening and prevention strategies" and that cumulative exposure is the more useful lens, while calling the volume finding "descriptive and hypothesis generating".
Our reading: this updates the base's prevalence figure (REF-727, one of the 14 reviews, 36% from 8 studies) with an adult-only pool of 32 studies that lands in the same place, 39%, and it undercuts the base's practice of tying the moment to ask to a move into higher-impact work (REF-728's practical line); it does not re-test REF-727's comparison with sedentary women, and it cannot say whether a given woman's leakage rises with her own training hours.
Protocol
Design: overview of systematic reviews plus a de novo systematic review and meta-analysis of the primary studies, following Cochrane guidance on overviews; part of a larger overview project on pelvic-floor disorders in nulliparous women athletes; PROSPERO CRD420251055357; PRISMA.
Search: MEDLINE (PubMed), Scopus, Web of Science, CINAHL and the Cochrane Library for reviews from January 2010 to 29 September 2025; forward citation tracking in Google Scholar; English, French, Portuguese or Spanish; peer-reviewed only.
Reviews (umbrella): systematic reviews or meta-analyses reporting urinary-incontinence prevalence in adult women athletes; 817 records, 46 full texts, 14 included (8 systematic reviews, 6 meta-analyses; 2017-2025); AMSTAR 2; overlap across 99 unique primary studies, corrected covered area 6.5%.
Primary studies (meta-analysis): women athletes aged 18-45 (para athletes excluded) in a clearly reported sport discipline; any type of urinary incontinence measured by a validated questionnaire, a clinician-administered assessment or a self-report item aligned with International Continence Society definitions with an explicit recall period; cohort, case-control, cross-sectional or intervention studies, no year limit; 1,384 records (804 from the reviews' reference lists, 580 from forward citations), 114 full texts, 32 included.AXIS appraisal: 14 high, 18 moderate quality.
Who: n = 4649 women athletes from 120 cohorts, 93.9% nulliparous on average; mean ages in the studies about 19-30 years; most from ball games and weight-based sports, most in high- or medium-impact sports, low-impact sports under-represented; four studies used single International Continence Society-aligned items.
Classifications: discipline as reported; modality (technical, endurance, aesthetic, weight-dependent, ball games, power, gravity); impact — high (e.g. gymnastics, basketball, volleyball, high jump, trampoline, powerlifting), medium (e.g. tennis, running, karate, soccer), low (e.g. swimming, cycling, walking); competitive level — amateurs, high level (national, or at least 3 years sub-elite) and professionals (highest national or international level, elite or Olympic centres); weekly training hours.
Analysis: random-effects pooling of logit-transformed prevalence with 95% CIs and prediction intervals; subgroups only with three or more cohorts; random-effects meta-regression as the primary test of moderators, pairwise contrasts as odds ratios; training volume as a continuous moderator per hour and per SD; sensitivity analyses restricted to validated instruments and to nulliparous samples; Egger's test where k ≥ 10; R 4.1.2.
Changes from the protocol (stated by the authors): scope narrowed to urinary incontinence; the planned comparison with the general population or sedentary women dropped because parity, age and measurement were incompatible; the de novo meta-analysis and the continuous volume meta-regression added; mixed-parity cohorts kept and handled by nulliparous-only sensitivity analyses.
Funding and interests (as printed): "Funding for open access publishing: Universidad de Granada/CBUA. This research was supported by the Unit of Research Excellence of the University of Granada, Melilla Campus, UECUMel (UCE-PP2024-02).Irene Sevilla-Arrabal was supported by the Research Initiation Grant for Undergraduate Students, University of Granada-Banco Santander (P5A2024-81).The funder did not influence the study design, the collection, analysis and interpretation of the data, the writing of the report and the decision to submit the paper for publication." Conflicts: the five authors "have no conflicts of interest that are directly relevant to the content of this article". No device maker, federation or data platform is named.The work is part of a doctoral thesis (NDP).
Key Results
| Analysis (women athletes 18-45) | Estimate | Does it explain the spread? |
|---|---|---|
| Any urinary incontinence | 39% (95% CI 33-46), I² 94.3% | — |
| Stress / urgency / mixed / exercise-related | 33% (27-41) / 14% (7-27) / 12% (6-24) / 39% (28-50) | I² > 93% for each |
| Nulliparous samples only | any 34%, stress 33%, urgency 16%, exercise-related 28% | heterogeneity persisted (I² 81-96% across restrictions) |
| Validated instruments only | any 42%, stress 32%, urgency 13%, exercise-related 38% | — |
| Sport discipline (11 with ≥3 cohorts) | rugby highest 50% (40-59), swimming lowest 27% (18-37) | No: subgroup p = 0.058; meta-regression QM(10) = 11.71, p = 0.304 (nulliparous only p = 0.773; validated only p = 0.327) |
| Sport modality | ball games 43% (35-52), technical 18% (10-31) | No: subgroup p = 0.018 but meta-regression QM(4) = 6.76, p = 0.149 |
| Impact category | medium 43% (35-52), high 34% (27-42), low 33% (19-50); low vs high OR 0.93 (0.45-1.95), medium vs high 1.47 (0.88-2.46) | No: subgroup p = 0.257; meta-regression QM(2) = 2.75, p = 0.253 |
| Competitive level | professionals 49% (39-59), amateurs 32% (23-42); professional vs amateur OR 2.05 (1.10-3.83), p = 0.024; high level vs amateur 1.51 (0.75-3.02) | Trend only: subgroup p = 0.074; meta-regression QM(2) = 5.19, p = 0.075; nulliparous only p = 0.260; validated only p = 0.050 |
| Weekly training hours (2,362 athletes) | OR 1.06 per extra hour a week (1.01-1.11), p = 0.011; per SD of 5.24 h, OR 1.38 (1.07-1.76) | Yes, modestly: no difference by impact (p = 0.454) or level (p = 0.750); nulliparous only 1.06 (1.01-1.11), n = 1,308; validated only 1.08 (1.02-1.13), p = 0.007, n = 2,188 |
Runners: "athletics or track and field" was the reference discipline in the pairwise table (8 cohorts, 87 of 171 athletes with leakage, a raw count; the pooled estimate is in a figure only).Soccer had lower odds than athletics (OR 0.34, 0.13-0.90, p = 0.030) and swimming a non-significant tendency (0.37, 0.13-1.05, p = 0.063); of the 55 pairwise discipline contrasts, 2 had p below 0.05 (soccer vs athletics, soccer vs rugby), close to the 2.75 that chance alone would give at that threshold (our arithmetic), and the overall test was null.The endurance modality (12 cohorts, 175 of 532) did not differ from aesthetic (OR 1.76, 0.75-4.11) or ball-game sports (0.81, 0.42-1.56).
What the earlier reviews said (umbrella): prevalence 5.7% to 80.0% across reviews; impact-stratified ranges higher in high-impact (12.2-61.2%) than low-impact sports (3.8-37.5%), but pooled estimates overlapped — 36.1% (26.5-46.8) in mixed sports and 44.1% (36.1-52.3) in low-impact sports in Teixeira 2018 (REF-727), 25.9% (23.5-28.3) and 39.9% (34.5-45.2) in high-impact sports.The 80% figures came from adolescent and young elite trampolinists; in adult nulliparous cohorts estimates were about 10-40%. CrossFit 32.1-44.5%, powerlifting 41.0-48.8%, weightlifting 36.6-54.1%, professional volleyball 9-30% during practice. All 14 reviews lacked a list of excluded studies.
Small-study effects: funnel-plot asymmetry for sport discipline (Egger p = 0.043), not for competitive level, impact or modality.
Intensity: only two studies reported a MET-based intensity measure; several cohorts reported regular cross-training, so total workload may be underestimated.
Numbers that do not add up: the abstract gives the competitive-level trend as p = 0.08, the results as 0.075; the methods say data "were aggregated to one cohort per study before pooling" while the results report 120 cohorts from 32 studies, and the number of cohorts differs from one analysis to the next; in the impact and competitive-level tables the n and k columns do not line up consistently from row to row (our reading), so the size of each contrast cannot be checked from the table.
Mechanisms
Measured: across cohorts, leakage prevalence did not track sport labels and rose with the cohort's mean weekly training hours; the association held within impact categories and competitive levels and in nulliparous-only and validated-instrument analyses.
Authors' interpretation: sport labels such as high or low impact are coarse proxies for pelvic-floor load, which depends on movement patterns, intensity, fatigue, technique, recovery and the individual, so the null "should not be interpreted as evidence of biomechanical equivalence between sports"; specific tasks (jumping and landing, heavy squats and deadlifts) do provoke leakage.Weekly hours may capture repeated loading, fatigue accumulation or insufficient recovery but are not a mechanistic proxy for pelvic-floor load; the professionals' trend likely reflects greater cumulative exposure; high prevalence in nulliparous athletes shows that exposure-related risk operates independently of childbirth.They propose a "pelvic floor dose" combining exposure, external and internal load, pelvic-floor function and symptom response.
Inference: a between-cohort association can arise because high-volume cohorts differ in other ways (level, age, sport, how leakage was asked about); whether one woman leaks more when her own hours rise is not tested. Nothing here says that cutting training hours reduces leakage.
Practical Application
I leak a little when I run hard. Is it my sport? Sport type did not explain it: in 32 studies (4,649 women athletes, 18-45) 39% leaked; cohorts training more hours a week leaked a little more (+6% odds/h). Pelvic-floor training helps.
Ask every woman, in every sport: the impact category did not explain leakage (low vs high impact OR 0.93, 0.45-1.95), and the authors warn that sport labels "may misdirect screening and prevention strategies".The base's REF-728 practical line ties the moment to ask to a move toward higher-impact work because "Impact level is a major risk modifier in the review's background figures"; by this review the question belongs in any sport.A rise in weekly hours as a moment to ask is our inference from the between-study volume association, not tested here; the base's own practice is to ask it as a normal load question (REF-727).
For runners: running sits in the medium-impact class and in athletics and endurance sports, which contribute few cohorts; there is no evidence here that running leaks more or less than other sports. For cyclists, the low-impact estimate (33%, 19-50) is imprecise and has no cycling-only figure; REF-1269 found cyclists reported no more urinary symptoms than swimmers or runners.
How it updates [REF-727]: Teixeira 2018 (REF-727) is one of the 14 reviews here (AMSTAR 2 critically low; 36.1% from 8 studies); this adult-only pool of 32 studies gives 39% (33-46), the same order. REF-727's "177% higher risk than sedentary women" is neither confirmed nor refuted: the authors dropped the sedentary comparison. REF-728's background range ("~10.9% in low-impact sport such as cycling to ~80% in high-impact sport such as trampoline gymnastics") rests partly on adolescent elite trampolinists, which this review excluded.Updating, not retiring, is the base's call to make.
What it is good for as a source: the current best estimate of how common leakage is in adult, mostly nulliparous women athletes, and the evidence that sport type is a poor screening filter. What it does not show: incidence or who will develop leakage; a dose-response within a woman; any effect of reducing hours; women after birth, in perimenopause or adolescents; any estimate for cycling or running alone.
Limitations
Prevalence, not incidence. Almost all primary studies describe leakage at one time; whether training hours cause leakage, prolong it or make it recur cannot be told (the authors' limitation).
Self-report and heterogeneity. Leakage was mostly self-reported by questionnaire with recall windows from 2 weeks to lifetime; heterogeneity stayed very high after every restriction.
Study-level association. The volume finding compares cohorts' average hours with cohorts' prevalence (ecological), with intensity in only two studies and cross-training unrecorded; it is a hypothesis, as the authors say.
Uneven evidence. Clustering in ball games and weight-based sports, few low-impact or endurance cohorts; funnel-plot asymmetry for sport discipline; the sport-classification system is one of several possible (the authors' limitation).
Who. Adults 18-45, mostly nulliparous; adolescents, perimenopausal women and para athletes excluded; languages limited to English, French, Portuguese and Spanish.
The umbrella layer is weak by design of its sources: 12 of 14 reviews critically low and 2 low confidence; the pooled numbers come from the new meta-analysis, not from those reviews.
Reporting inconsistencies in cohort counts and table columns (see Key Results).
Funding and interests: University of Granada research funding and a University of Granada-Banco Santander student grant; no relevant conflicts declared (as printed above). No device maker, federation or data platform is named.
How this entry was checked and changed
Why this level
the rubric's A+ row names umbrella reviews, but this one is held at A- because the 14 reviews it covers are of critically low (12) or low (2) confidence and the pooled numbers come from the authors' own meta-analysis of observational prevalence studies, mostly cross-sectional with self-reported leakage, whose heterogeneity (I² 94.3%) no moderator reduced; the training-volume association is a study-level meta-regression (cohort mean hours against cohort prevalence, 2,362 athletes), not a within-person or prospective dose-response; and the sport-label nulls carry wide confidence intervals. The same level as REF-728 (A-), above REF-727 (B+), which this review includes
Full text (checked 2026-10-28)
Fourteen systematic reviews were included, most rated as low or critically low confidence; AMSTAR 2 classified 12 of 14 included reviews as critically low confidence, and two were rated low confidence (full text)
Thirty-two primary studies (120 cohorts; n = 4649) were meta-analysed; predominantly nulliparous (93.9% in average) (full text)
Analyses were restricted to adult women athletes (aged 18–45 years) to reduce developmental and hormonal heterogeneity, as adolescence and perimenopause represent distinct life stages (full text)
Overall pooled UI prevalence in women athletes was 39% (95% CI 33–46), with substantial heterogeneity (I^2 = 94.3%) (full text)
Pooled prevalences by UI subtype were 33% for SUI (95% CI 27–41), 14% for UUI (95% CI 7–27), 12% for MUI (95% CI 6–24) and 39% for EXUI (95% CI 28–50) (full text)
Sensitivity analyses: only nulliparous samples (UI: 34%, SUI: 33%, UUI: 16%, EXUI: 28%) and only validated screening tools (UI: 42%, SUI: 32%, UUI: 13%, EXUI: 38%) (full text)
Sport discipline: rugby 50%, 95% CI 40–59; swimming 27%, 95% CI 18–37; inconclusive in the meta-analysis (p = 0.058, I^2 = 78%) and not significant in the meta-regression (QM(10) = 11.71; p = 0.304) (full text)
Sport modality: ball games 43%, 95% CI 35–52; technical sports 18%, 95% CI 10–31; meta-regression QM(4) = 6.76; p = 0.149 (full text)
Sport impact: medium-impact 43% (95% CI 35–52), high-impact 34% (95% CI 27–42), low-impact 33% (95% CI 19–50); meta-regression (QM(2) = 2.75; p = 0.253); Low impact vs High impact 0.93 (0.45–1.95) (full text)
Professional athletes 49% (95% CI 39–59), amateurs 32% (95% CI 23–42); meta-regression (QM(2) = 5.19, p = 0.075); Professional vs Amateur 2.05 (1.10–3.83), p = 0.024 (full text)
Weekly training hours: per + 1 h/week OR = 1.06, 95% CI 1.01–1.11, p = 0.011, n = 2362 athletes; increments of 5.24 h/week correspond to OR = 1.38 (95% CI 1.07–1.76); no evidence that the association differed by sports impact (p = 0.454) or competitive level (p = 0.750) (full text)
Soccer vs Athletics or track and field 0.34 (0.13–0.90), p = 0.030; Swimming vs Athletics or track and field 0.37 (0.13–1.05), p = 0.063 (full text)
Commonly used sport-label classifications showed limited explanatory value and "may misdirect screening and prevention strategies"; the volume finding "should be interpreted as descriptive and hypothesis generating" (full text)
No planned comparison with the general population/sedentary controls was undertaken because of incompatibility of parity/age and measurement across datasets (full text)
Funding: Unit of Research Excellence of the University of Granada, Melilla Campus, UECUMel (UCE-PP2024-02); Research Initiation Grant, University of Granada-Banco Santander (P5A2024-81) (full text)