Discussion

This umbrella review and meta-analysis synthesised evidence from 14 systematic reviews and 32 primary studies to examine UI prevalence in adult women athletes and explore sport-related factors contributing to between-cohort variability. The findings indicate that UI is common in this population (39% overall), but characterised by substantial heterogeneity across cohorts. At the population level, commonly used sport-label classifications such as discipline, modality or impact category did not consistently explain differences in prevalence. Although some subgroups showed directional differences (e.g. higher point estimates in rugby and among professionals), these patterns were not robust moderators in meta-regression. In contrast, weekly training volume emerged as the most consistently reported exposure-based correlate of UI odds, although this association should be interpreted cautiously within a predominantly cross-sectional evidence base. Notably, no sport-label classification meaningfully reduced between-cohort heterogeneity.

The interpretation of sport-related patterns in UI among women athletes is fundamentally constrained by the quality of the existing synthesis literature. In this umbrella review, most included systematic reviews and meta-analyses were rated as low or critically low confidence, largely owing to absent protocol registration, inadequate assessment of bias in primary studies and limited consideration of methodological limitations when interpreting pooled estimates. Consequently, prior syntheses have reported widely divergent prevalence estimates and drawn inferences from heterogeneous populations and inconsistent sport classifications. By integrating a structured overview of reviews with a de novo meta-analysis of primary cohorts, the present study helps address these limitations and provides a more coherent athlete-centred synthesis of sport-related correlates of UI.

The absence of consistent associations between sport-label classifications and UI prevalence should not be interpreted as evidence of biomechanical equivalence between sports. Rather, these findings highlight limitations in how sport-related exposure is currently operationalised in the literature. Broad classifications such as “high-impact” or “low-impact” are coarse proxies that may inadequately capture pelvic floor loading, which is likely influenced by multiple interacting factors including movement patterns, intensity, fatigue, technique, recovery and individual neuromuscular characteristics [6, 25]. Previous reviews have similarly cautioned against over-interpreting sport type as a surrogate for pelvic floor stress, noting wide variability in UI prevalence even within the same disciplines [17, 22, 23, 42]. Importantly, this variability persisted even among cohorts examining similar populations or using only validated screening tools, as demonstrated across primary and sensitivity analyses. This pattern suggests that heterogeneity reflects more than methodological noise alone and likely indicates exposure misclassification when broad sport labels are used as proxies for pelvic floor loading. Exposure misclassification is therefore likely and may obscure sport-specific mechanisms when analyses are conducted at the population level.

Training volume showed the strongest association with UI prevalence, with higher weekly training hours linked to increased odds of UI. Similar observations have been reported in both athletic and recreationally active populations, where cumulative exercise exposure has been associated with pelvic floor symptoms independent of sport type [77–79], particularly at a high competitive level [80]. Cumulative exposure time may capture repeated loading, fatigue accumulation or insufficient recovery. However, it should not be interpreted as a direct mechanistic proxy for pelvic floor load. Rather, it reflects that the available data did not allow granular characterisation of true training intensity or pelvic floor-specific loading.

From a load–injury perspective, training volume and mechanical impact represent interrelated components of overall exposure, with injury risk depending on both the magnitude of load (e.g. intensity) and its accumulation over time (e.g. frequency and volume) [26]. In the present evidence base, however, only two studies reported Metabolic Equivalent of Task (MET)-based intensity metrics [56, 74], limiting dose–intensity modelling. Moreover, several cohorts explicitly report regular cross-training [47, 52, 56, 58, 63], suggesting that the total workload may be underestimated when sport-specific exposure is considered in isolation. Together, these limitations restrict the extent to which observed volume effects can be attributed to exposure duration per se, cumulative fatigue, or interactions between training duration and intensity.

These constraints, alongside the limited explanatory value of nominal sport classifications or generic mechanical proxies, support a more exposure-informed conceptual approach to pelvic floor stress and UI risk in women athletes. These findings indicate that pelvic floor exposure is unlikely to be meaningfully represented by sport type alone and instead requires a multidimensional construct integrating training exposure, pelvic floor-specific function and symptom response, where feasible. Specifically, a conceptual “pelvic floor dose” framework should integrate (i) exercise exposure describing the amount and scheduling of sport participation (e.g. weekly training hours, session frequency, competition density) [6, 26]; (ii) load descriptors capturing how that exposure is experienced, including external load (e.g. jump counts, sprint distance, ground contacts) and internal load (e.g. heart rate-based load, perceived exertion or neuromuscular responses such as electromyography) [26, 82, 83]; (iii) pelvic floor-specific structure and function, including muscle capacity, fatigability and displacement of pelvic support structures (e.g. ultrasound, or intravaginal pressure-based measures) [84–87]; and (iv) individual symptom responses during and following sport participation (e.g. leakage during specific tasks, symptom frequency, severity or recovery patterns) [10, 11].

Although integration of pelvic floor-specific assessments within broader exposure frameworks currently remains largely confined to research settings, it is essential for moving beyond indirect proxies of load. From a clinical and applied perspective, exposure-aware screening approaches that combine cumulative training demands with pelvic floor-specific considerations and symptom profiling may therefore offer a more informative basis for identifying potentially unsafe exposure patterns than sport type-based risk stratification alone. The high prevalence of UI, together with its association with training volume, supports the incorporation of exposure-aware screening, pelvic health education and timely referral within athlete care pathways. The emerging PFD-SENTINEL screening consensus for female athletes and broader female-athlete health frameworks provide practical entry points for embedding pelvic health in sport medicine workflows [25, 35]. Programmes should address knowledge gaps and stigma that normalise leakage, particularly in elite contexts where symptoms may be under-reported [10, 11, 88]. Given the early onset and high prevalence of UI among nulliparous athletes, interventions should not wait for parity or ageing-related risk factors to emerge. Women’s pelvic health physiotherapists and sport professionals should work closely with sports organisations to embed routine pelvic health screening within standard medical evaluations, ensure access to pelvic health education and actively challenge the cultural normalisation of leakage in sport.

Within this framework, competitive level warrants consideration as a potential modifier of cumulative exposure rather than as an independent risk factor. Prior reviews seldom examined competitive level as a moderator and generally did not perform level-stratified meta-analyses. Cross-sport syntheses focused on sport/impact rather than level [18, 21, 27], with broader reviews mentioning elite athletes without formal quantitative evaluation [16]. In our data, although the competitive level did not reach conventional significance, a subgroup test and a directional pairwise contrast suggested higher odds of UI among professional athletes compared with amateur athletes. Given the substantial within-level heterogeneity and wide prediction intervals, this signal should be interpreted cautiously and considered hypothesis generating. A plausible explanation might be greater cumulative exposure over time at the professional level [10, 11, 88]. However, the present analyses did not identify differential associations between weekly training volume and UI across competitive levels. Prospective level-stratified studies with granular workload metrics are therefore needed to disentangle the contributions of competitive context, cumulative load and recovery to pelvic floor symptom risk. The presence of elevated prevalence estimates among nulliparous professional athletes reinforces that exposure-related risk operates independently of obstetric history.

Several limitations should be acknowledged. Most available data describe prevalence rather than incidence, limiting inference on symptom onset, progression and reversibility in relation to training exposure. Consequently, it remains unclear whether higher training exposure primarily contributes to symptom onset, prolonged persistence or recurrent episodes of UI in women athletes. Pelvic floor-specific exposure metrics were rarely reported, necessitating reliance on indirect proxies such as sport labels and training volume. In this context, sport classification itself represents an additional source of uncertainty; although we applied a classification commonly used in pelvic floor research, alternative systems for categorising sport modalities could produce different estimates. Finally, the restriction to adult women improves aetiological specificity but limits generalisability to adolescent athletes, who represent an important population for future dedicated syntheses. Notwithstanding these limitations, several methodological strengths warrant consideration. Unlike prior syntheses that pooled broad and heterogeneous populations, the present review focused on a biologically and clinically defined population of adult women athletes (aged 18–45 years), improving aetiological specificity. In addition, sport context (discipline and modality), competitive level and continuous training exposure were examined using consistent eligibility criteria, exposure definitions and statistical models. This approach provides a more athlete-specific synthesis of sport-related correlates of UI within the constraints of the available evidence.

Relative to our initial protocol, we introduced some refinements to enhance precision and clinical relevance: (1) scope narrowed to UI only; (2) no planned comparison with the general population/sedentary controls was undertaken because of incompatibility of parity/age and measurement across datasets; (3) we added a de novo meta-analysis of primary studies to the umbrella review because existing reviews overlapped and did not provide the necessary subgroup/exposure granularity; (4) we introduced continuous exposure meta-regressions beyond the protocol’s subgroup framework; and (5) although the protocol targeted nulliparous women aged 18–45 years, some cohorts included mixed parity, addressed via nulliparous-only sensitivity analyses rather than outright exclusion.