ORIGINAL ARTICLE

Exploratory modeling of urinary incontinence risk among female athletes the role of smoking and training duration

Authors

Iva Lončarić Kelečić 1, 2 Snježana Schuster 3, 4 Jasna Mesarić 1 ✉

Affiliations

  1. Libertas International University, Zagreb, Croatia
  2. University Clinical Hospital Center Zagreb, Croatia
  3. University of Applied Health Sciences, Zagreb, Croatia
  4. Faculty of Kinesiology, University of Zagreb, Croatia
Phys Act Rev. 2026;14(2):100-112.DOI: 10.16926/par.2026.14.25

Abstract

Background:

Urinary incontinence (UI) is commonly reported among female athletes, yet its determinants remain insufficiently understood. This study aimed to examine whether daily training duration contributes to the likelihood of UI beyond previously observed associations with smoking.

Methods:

A secondary analysis of cross-sectional data from 70 Croatian female athletes was conducted. Multivariable logistic regression models were used to assess associations between UI and age, body mass index (BMI), smoking, and daily training duration as a simplified proxy for external training load, including an exploratory interaction between smoking and training duration.

Results:

Smoking was the only variable significantly associated with UI in the main-effects model (odds ratio [OR] = 4.18, 95% CI 1.05–16.69). Age, BMI, and training duration were not independently associated with UI. In the exploratory interaction model, smoking was associated with higher odds of UI, although with limited precision (OR = 8.26, 95% CI 1.00–68.18), while the interaction between smoking and training duration showed a positive but statistically nonsignificant association (OR = 14.72, 95% CI 0.89–243.95), indicating substantial uncertainty in the estimate.

Conclusions:

These findings suggest that behavioral factors may play a more relevant role in UI among female athletes than training duration alone. Although the exploratory interaction pattern should be interpreted with caution, it may suggest a possible difference in the relationship between training duration and UI according to behavioral factors such as smoking. Further studies with larger samples and multidimensional measures of training exposure are needed.

Keywords

  • Urinary incontinence
  • female athletes
  • smoking
  • training duration
  • logistic regression

Full Text

INTRODUCTION

Urinary incontinence (UI) is defined as any involuntary loss of urine and represents a common health problem among females, often associated with reduced quality of life and psychosocial burden [1,2]. Although UI is frequently discussed in relation to aging, childbirth, or pelvic floor dysfunction, increasing attention has been directed toward its occurrence among physically active women and female athletes.

Female athletes constitute a specific population with distinct physiological demands and health characteristics associated with intensive training and participation in competitive sports [3]. Evidence suggests that participation in sport, particularly high-impact activities that involve repeated increases in intra-abdominal pressure, may contribute to stress on pelvic floor structures and increase the risk of UI [4]. Systematic reviews and meta-analyses indicate that urinary leakage is reported across a wide range of sports, with prevalence estimates suggesting that approximately one-quarter to one-third of female athletes experience UI, depending on sport type and intensity [4–6]. Observational studies further support these findings, showing higher odds of UI among female athletes compared to non-athletic controls and highlighting its potential impact on training and sports participation [7–9].

Beyond sport type and impact-related characteristics, some evidence suggests that overall training exposure may also be relevant for urinary symptoms in female athletes. For example, longer daily training duration has been observed in athletes with UI compared to continent counterparts [10], while greater weekly training volume has been associated with both the presence and severity of lower urinary tract symptoms in female athletes [11]. However, findings across studies remain inconsistent.

In addition to sport-related mechanical factors, lifestyle characteristics may also contribute to the development of UI. Population-based epidemiological research suggests that modifiable behaviors such as smoking may be associated with UI in females, particularly in cases of heavy smoking [12]. However, findings in athletic populations remain inconsistent. For example, a recent study in female weightlifters, despite examining both training-related and lifestyle factors, did not identify significant associations with UI, including for smoking and training characteristics [13]. In contrast, a pilot study of Croatian female athletes reported UI in approximately one-quarter of participants, generally mild to moderate in severity, with smoking as the only factor significantly associated with it, whereas other sport-related characteristics were not statistically significant [14].

Previous findings suggest that, beyond being common, the condition is complex, and commonly assessed sport-related and lifestyle factors may not fully explain UI in female athletes, possibly reflecting limitations in how these factors have been analytically examined. Much of the previous research has relied on bivariate analyses, without examining multiple predictors or their potential interactions. In this context, given that smoking may affect connective tissue properties [15–17] and that mechanical loading during physical activity is characterized by repeated increases in intra-abdominal pressure [18,19], it may be relevant to explore the potential interplay between these factors, including whether daily training duration, as a simplified proxy for external training load, contributes to UI risk independently or in interaction with smoking.

Secondary analysis of existing datasets [20,21] may provide an opportunity further to explore potential determinants of UI in athletic populations. By applying multivariable modeling [22] to previously collected data, it is possible to examine whether behavioral factors, such as smoking, and sport-related characteristics jointly contribute to the occurrence of UI. Accordingly, the present study aimed to conduct a secondary analysis of data collected from Croatian female athletes to investigate potential predictors of UI using multivariable statistical modeling. In particular, this study sought to address the following research question (RQ).

RQ: How does daily training duration, used as a simplified proxy for external training load, contribute to UI risk among Croatian female athletes beyond the previously observed association with smoking?

MATERIAL AND METHODS

Study design

The present study represents a secondary analysis of a previously published cross-sectional dataset collected among Croatian female athletes [14]. Whereas the initial study focused on assessing the occurrence, severity, and UI-related factors, the current research advances beyond bivariate analysis by employing multivariable modeling to explore possible predictors of UI using the same dataset. Specifically, this analysis aimed to explore whether daily training duration, serving as a simplified proxy for external training load, contributes to the likelihood of UI, in addition to the previously identified link with smoking.

Participants

The original study included a convenience sample of Croatian female athletes.

Eligible participants were female athletes aged 18 or older who actively participated in competitive sports, could understand the questionnaire in Croatian, and were willing to provide informed consent. Athletes who were retired from sports, on extended sick leave, pregnant, or on maternity leave were considered ineligible and, therefore, were not included in the study.

A total of 76 responses were initially obtained; 70 complete questionnaires met the eligibility criteria and were included in the analysis. Six responses were excluded due to ineligibility. Participants had a mean age of 22.64 ± 4.56 years and included athletes from 12 sports disciplines, both team and individual.

Data collection and variables

Data were collected in May 2022 using an anonymous online survey distributed via social media, applying a snowball sampling approach. The survey included items on sociodemographic and anthropometric characteristics, sport participation variables, and selected medical and lifestyle factors. Self-reported body height and weight were used to calculate body mass index (BMI). Sport participation variables included age at entry into competitive sport, years of competitive experience, training duration, weight-training volume, and rest periods between training sessions. Training duration referred to the average number of hours of training per day, as reported at the time of survey completion. Medical and lifestyle information included smoking status, oral contraceptive use, parity, urogenital problems, constipation, and previous epidural or spinal anesthesia.

For the present secondary analysis, age, BMI, smoking status, and average daily training duration were selected as candidate predictors of UI. This selection was guided by their theoretical relevance, prior reporting in the literature, and their availability and completeness in the original dataset. The selection of predictors was further informed by a conceptual framework that distinguishes demographic, behavioral, and exposure-related factors potentially associated with pelvic floor function, enabling the simultaneous consideration of individual characteristics, modifiable lifestyle factors, and sport-related mechanical exposure. Smoking was included as a key behavioral variable, while daily training duration was operationalized as a simplified proxy for external training load. Although external load is a multidimensional construct encompassing factors such as intensity, frequency, and sport-specific mechanical demands, only training duration was available in the dataset and was therefore used as an approximate indicator, consistent with conceptual frameworks distinguishing objective workload from subsequent physiological responses [23–25].

Additional variables (e.g., parity, selected medical conditions, and sport-related characteristics) were examined descriptively. Still, they were not included in the multivariable models due to low prevalence, limited variability, or insufficient distribution across outcome groups. For descriptive purposes, sports were additionally grouped according to their expected mechanical load on the pelvic floor, based on previous literature distinguishing high-impact activities characterized by repetitive increases in intra-abdominal pressure and ground reaction forces from lower-impact disciplines [18,19]. High-impact sports included disciplines characterized by repetitive jumping, landing, and running (e.g., handball, athletics, gymnastics, and combat sports). In contrast, low-impact sports included disciplines with lower vertical loading and reduced ground reaction forces (e.g., swimming, water polo, and racket sports).

Outcome measure

Urinary incontinence symptoms were evaluated using the Croatian version of the International Consultation on Incontinence Questionnaire – Urinary Incontinence Short Form (ICIQ-UI SF). The ICIQ-UI SF assesses the frequency and amount of urinary leakage and the impact of symptoms on everyday life; a total score ranges from 0 to 21, with higher scores indicating greater symptom severity.

The ICIQ-UI SF is a widely used and validated instrument for assessing UI symptoms and their impact on quality of life, and, consistent with previous studies using the instrument, including those on athletes, UI was operationalized as the presence of any reported leakage (ICIQ-UI SF total score > 0) [8,26–28]. This dichotomization reflects the presence of any urinary leakage rather than clinically diagnosed incontinence.

Internal consistency of the ICIQ-UI SF, evaluated using Cronbach’s alpha (α) and McDonald’s omega (ω) [29,30], was high in this sample (McDonald’s ω = 0.933;

Cronbach’s α = 0.896), indicating good coherence of the measure. Because reliability assessment was not central to the present research question, detailed item-level indices are not reported.

Statistical analysis

Statistical analyses were conducted using JASP (Jeffreys’s Amazing Statistics Program), version 0.95.4 (University of Amsterdam, The Netherlands). Descriptive statistics were calculated for all variables. Normality of distribution was assessed using the Shapiro–Wilk test in the overall sample and within subgroups defined by UI status. Most continuous variables, including key variables such as age, body mass index (BMI), and average training hours per day, deviated from normality, particularly in the larger subgroup. Therefore, continuous variables were summarised using both mean (standard deviation, SD) and median (interquartile range, IQR) to provide a comprehensive description of the data. Categorical variables were presented as absolute frequencies (n) and percentages (%). Only complete cases were included in the analysis; no imputation procedures were applied.

For modeling purposes, the outcome variable was operationalized as the presence of UI. Following commonly used approaches in epidemiological studies using the ICIQ-UI SF, UI was coded as a binary variable: a total score greater than zero indicated the presence of urinary leakage (UI = 1), and a score of zero indicated the absence of symptoms (UI = 0). This operationalization allowed the identification of individuals reporting any degree of urinary leakage and enabled the modeling of UI occurrence using logistic regression [31].

To explore potential predictors of UI, multivariable binomial logistic regression analysis was conducted. Age (years), body mass index (BMI; kg/m²), smoking status, and training duration (hours per day) were entered as predictors.

Age, BMI, and training duration were treated as continuous variables, while smoking status was coded as a binary variable (0 = non-smoker, 1 = smoker). Continuous predictors were entered in their original measurement units to allow interpretation of odds ratios (OR) as the change in odds associated with a one-unit increase in the predictor [32].

Because the research question specifically addressed the potential joint contribution of behavioral and sport-related factors, an additional exploratory model was estimated that included an interaction term between smoking and training duration. To reduce potential multicollinearity and to improve the interpretability of the regression coefficients [33], prior to computing the interaction term, training duration was mean-centered. The interaction model included both lower-order terms (smoking status and mean-centered training duration) and their interaction term.

To ensure that the model was not only accurate at a given threshold but also had substantial overall discriminatory power, model performance was evaluated using metrics of the likelihood ratio test (Δχ²), pseudo-coefficient of determination indices (Nagelkerke R² and McFadden R²), classification accuracy, and the area under the receiver operating characteristic curve (AUC) [34]. Multicollinearity among predictors was assessed using variance inflation factors (VIF).

Given the limited number of UI cases in the sample (n = 17), the number of predictors included in the regression models was intentionally restricted to maintain an acceptable events-per-variable (EPV) ratio and reduce the risk of model overfitting [35,36]. No formal sample size calculation was performed, as this study represents a secondary analysis of an existing dataset.

Ethical considerations

This study represents a secondary analysis of data originally collected through an anonymous online survey conducted in accordance with ethical principles and applicable regulations. In the primary study, participants provided informed consent by voluntarily completing the survey, and no identifiable personal data were collected or stored. The present secondary analysis was conducted using a fully anonymized dataset, with no possibility of participant identification, and therefore in accordance with ethical guidelines for the use of health data [37].

RESULTS

Results are derived from a secondary analysis of 70 Croatian female athletes, of whom 53 reported no UI, and 17 reported UI. The sample included athletes from 12 different sports disciplines, representing both team and individual sports, as previously described in the original dataset publication [14].

Sample characteristics

As shown in Table 1, the median age was 21 years (IQR 6) among athletes without UI and 23 years (IQR 6) among those with UI, with a mean age of 22.64 ± 4.56 years in the total sample. The median BMI was 22.0 kg/m² (IQR 2.4) in the no-UI group and 21.7 kg/m² (IQR 2.9) in athletes with UI; the overall mean BMI was 22.22 ± 2.50 kg/m². Athletes reported entering competitive sport at a median age of 9 years (IQR 3.5) in the no-UI group and 10 years (IQR 2) in the UI group. The median duration of competitive experience was similar between groups (7 years), with IQRs of 6 and 5 years, respectively.

Regarding training characteristics, the median daily training duration was 2 hours (IQR 1) in athletes without UI and 3 hours (IQR 2) in those with UI, with an overall mean of 2.87 ± 1.43 hours per day. Median weekly weight-training duration was 2 hours in both groups (IQR 2 in the no-UI group and 0 in the UI group). The median rest interval between training sessions was 15 hours (IQR 18) among athletes without UI and 7 hours (IQR 19) among those with UI. Athletes in both groups reported a median of 4 weeks of rest per year, with slightly different variability (IQR 3 vs. 1 week).

In terms of lifestyle and sport-related characteristics, 14 athletes (20%) reported smoking. The proportion of smokers was higher in the UI group compared to the no-UI group (41.2% vs. 13.2%). Participation in high-impact sports was common in both groups (58.8% in athletes with UI and 66.0% in those without UI).

Parity was rare in the sample (n = 2).

The ICIQ-UI SF total score was 0 among athletes without UI and ranged from 4 to 11 among those with UI, with a mean score of 1.56 ± 3.03 in the overall sample.

Multivariable analysis of factors associated with urinary incontinence

Multivariable logistic regression analysis is presented in Table 2, and the corresponding adjusted odds ratios are illustrated in Figure 1. As shown in the model, smoking was the only variable significantly associated with UI. Athletes who reported smoking had higher odds of UI compared with non-smokers (OR = 4.18, 95% CI 1.05–16.69, p = 0.043). In contrast, age, BMI, and average training duration were not significantly associated with UI. Overall, the model demonstrated modest explanatory power and moderate discrimination, although the likelihood-ratio test did not reach conventional statistical significance.

Figure 1 illustrates these findings: the CIs for age, BMI, and training hours all cross the null value (OR = 1), indicating no statistically significant association with UI. In contrast, the point estimate for smoking lies to the right of the null line, and its confidence interval does not cross 1, illustrating the increased odds of UI among smokers.

Table 1.

Sample characteristics stratified by urinary incontinence status.

VariableNo UI (n = 53)UI present (n = 17)Total (N = 70)
Age (years)21 (6); 22.30 ± 4.4823 (6); 23.71 ± 4.7722.64 ± 4.56
Body mass index (kg/m²)22 (2.4); 22.07 ± 2.0821.7 (2.9); 22.68 ± 3.5522.22 ± 2.50
Age at entry into competitive sport (years)9 (3.5); 9.41 ± 2.8010 (2); 9.88 ± 3.109.53 ± 2.87
Years of professional experience7 (6); 7.26 ± 4.277 (5); 7.88 ± 4.477.41 ± 4.30
Training duration (hours/day)2 (1); 2.76 ± 1.493 (2); 3.24 ± 1.192.87 ± 1.43
Weight training hours per week2 (2); 2.27 ± 2.542 (0); 2.18 ± 0.932.25 ± 2.22
Rest hours between training sessions15 (18); 14.80 ± 8.467 (19); 13.32 ± 12.2914.44 ± 9.43
Weeks of rest per year4 (3); 4.94 ± 2.784 (1); 4.12 ± 2.714.74 ± 2.75
Smoking, n (%)7 (13.2)7 (41.2)14 (20.0)
High-impact sport participation, n (%)35 (66.0)10 (58.8)45 (64.3)
Parity, n (%)1 (1.9)1 (5.9)2 (2.9)
ICIQ-UI SF total score04–111.56 ± 3.03

Continuous variables are presented as median (IQR) and mean ± SD unless otherwise indicated. Categorical variables are presented as counts (percentages), with only the presence of the characteristic shown. UI: urinary incontinence; IQR: interquartile range; SD: standard deviation; ICIQ-UI SF: International Consultation on Incontinence Questionnaire – Urinary Incontinence Short Form; n: subgroup sample size; N: total sample size.

Table 2.

Multivariable logistic regression model predicting urinary incontinence (UI presence).

Predictorβ (SE)OR95% CIp
Age (years)0.035 (0.067)1.040.91–1.180.596
Body mass index (kg/m²)0.074 (0.119)1.080.85–1.360.531
Smoking (yes vs no)1.429 (0.707)4.181.05–16.690.043
Training duration (hours/day)0.289 (0.197)1.340.91–1.970.142
Model fit statisticsΔχ²(4) = 8.14; p = 0.086; Nagelkerke R² = 0.164; McFadden R² = 0.105; AUC = 0.705; Accuracy = 0.829

UI: urinary incontinence; β: log-odds coefficient; SE: standard error; OR: odds ratio; CI: confidence interval; AUC: area under the receiver operating characteristic curve; ICIQ-UI SF: International Consultation on Incontinence Questionnaire: Urinary Incontinence Short Form. UI presence coded as 1 (ICIQ-UI SF total score > 0). Continuous predictors were entered unstandardized; no evidence of multicollinearity was observed (all VIF < 2).

Figure 1.

Adjusted odds ratios for urinary incontinence among female athletes derived from the multivariable logistic regression model. Points represent adjusted odds ratios (ORs), and horizontal bars indicate 95% confidence intervals. The dashed vertical line represents the null value (OR = 1). Confidence intervals crossing the null line indicate non-significant associations.

Exploratory interaction analysis

An additional logistic regression model, including an interaction term between smoking status and mean-centered training duration, was estimated to examine whether the association between training duration and UI differed by smoking status (Table 3). In this model, neither age nor BMI was significantly associated with UI. Training duration alone was also not significantly associated with UI (OR = 1.12, 95% CI 0.74–1.70, p = 0.601). Smoking was associated with higher odds of UI, although the estimate was imprecise and the p-value was at the conventional threshold of statistical significance (OR = 8.26, 95% CI 1.00–68.18, p = 0.050).

The interaction term between smoking and training duration showed a positive but statistically non-significant association with UI (OR = 14.72, 95% CI 0.89–243.95, p = 0.061). Although the direction of the association may suggest a difference in the relationship between training duration and UI by smoking status, the very wide confidence interval indicates substantial uncertainty and potential instability of the estimate.

The interaction pattern is illustrated in Figure 2. The predicted probability of UI increased with higher values of the smoking–training interaction term. However, because training duration represented only an approximate indicator of exposure and the confidence intervals were wide, this pattern should be interpreted cautiously. Although the likelihood-ratio test for the interaction model was statistically significant, the interaction term itself did not reach conventional statistical significance, and the model did not demonstrate a meaningful improvement in discrimination compared with the main-effects model. Therefore, these findings should be interpreted strictly as exploratory and hypothesis-generating.

Table 3.

Multivariable logistic regression model including interaction between smoking and training duration.

Predictorβ (SE)OR95% CIp
Age (years)0.021 (0.077)1.020.88–1.190.783
Body mass index (kg/m²)0.094 (0.140)1.100.84–1.450.501
Smoking (yes vs no)2.112 (1.077)8.261.00–68.180.050
Training duration (hours/day, centered)0.112 (0.215)1.120.74–1.700.601
Smoking × Training duration (centered)2.689 (1.433)14.720.89–243.950.061
Model fit statisticsΔχ²(5) = 15.74, p = 0.008; Nagelkerke R² = 0.300; McFadden R² = 0.203; AUC = 0.721; Accuracy = 0.829

UI: urinary incontinence; β: log-odds coefficient; SE: standard error; OR: odds ratio; CI: 95% confidence interval; AUC: area under the receiver operating characteristic curve. UI presence coded as 1 (ICIQ-UI SF total score > 0); training duration was mean-centered prior to computing the interaction term; no evidence of problematic multicollinearity was observed (all VIF < 2). Continuous predictors were entered unstandardized.

Figure 1.

Predicted probability of urinary incontinence according to the interaction term between smoking status and mean-centered training duration. The curve represents predicted probabilities derived from the logistic regression model, including the interaction term (Table 3), with grey shading indicating the 95% confidence interval. The widening of the confidence band reflects increasing uncertainty of the estimates at higher values of the interaction term.

DISCUSSION

The present findings contribute to the growing body of evidence indicating that UI in female athletes cannot be explained solely by sport-related exposure but rather reflects a multifactorial interplay among mechanical, behavioral, and biological factors. UI is recognized as a common condition among women, with complex underlying mechanisms involving pelvic floor dysfunction, neuromuscular control, and tissue-level alterations [1].

Consistent with previous research, the broader literature indicates that UI is highly prevalent among female athletes, particularly in sports characterized by repeated increases in intra-abdominal pressure [4–6]. Comparative studies further demonstrate that elite athletes have significantly higher odds of UI compared with non-athletic controls [7], and observational data suggest that UI may affect up to one-third of athletes and can occur during training and competition [8]. Despite consistent evidence on prevalence, determinants of UI in athletic populations remain insufficiently understood.

In this context, the present study provides a more nuanced perspective by applying multivariable modeling to examine potential predictors simultaneously.

Taken together, these findings suggest that UI in female athletes may reflect the interplay between individual characteristics, modifiable behavioral factors, and sport-related mechanical exposure.

The lack of an independent association between training duration and UI aligns with previous findings suggesting that sport-related characteristics alone may not fully explain the occurrence of UI in athletes [5,6,19]. This supports the notion that simple quantitative indicators of training exposure, such as duration, may not adequately capture the mechanical load on the pelvic floor because they do not account for factors like intensity, movement patterns, or transient increases in intra-abdominal pressure. Previous biomechanical studies suggest that pelvic floor loading during high-impact activities may involve rapid and substantial transient increases in intra-abdominal pressure and dynamic deformation patterns that differ from those under sustained loading [38]. However, given the predominance of high-impact sports in the sample, variability in mechanical exposure may have been limited, potentially reducing the ability to detect associations with training duration. As previously proposed, the effect of physical activity on pelvic floor function likely depends on a balance between adaptive strengthening and mechanical overload, which may vary across individuals and sport types [19]. This may partly explain inconsistencies in findings regarding training-related risk factors.

In contrast, the observed association between smoking and UI is consistent with epidemiological evidence identifying modifiable lifestyle factors as relevant contributors to UI in women [12]. Importantly, this relationship is supported by biological mechanisms. Smoking has been shown to impair connective tissue metabolism, including reduced synthesis of type I and III collagen and increased degradation mediated by matrix metalloproteinases [15–17]. These alterations may weaken the structural integrity of supportive tissues, potentially reducing the pelvic floor's capacity to withstand repeated mechanical loading.

An additional exploratory observation in this study concerns the potential interaction between smoking and training duration, used as a simplified proxy for external training load. Although this interaction did not reach conventional statistical significance, its direction is consistent with a model in which mechanical stress and tissue vulnerability may interact [39]. Repeated increases in intra-abdominal pressure during physical activity have been proposed as one potential mechanism contributing to stress UI [4], and this effect may plausibly be amplified in individuals with compromised connective tissue properties due to smoking. In such a scenario, training exposure alone may not be sufficient to increase UI risk, but may become more relevant in the presence of underlying biological susceptibility. Contemporary models of pelvic floor function, such as the Integral Theory [40] and the Hammock Hypothesis [41], conceptualize urinary continence as a dynamic balance in which the structural capacity of supporting tissues must counteract the mechanical load generated by increases in intra-abdominal pressure [42]. Within this framework, the pattern observed in the present study may reflect a potential mismatch between mechanical demands and tissue resilience; however, this interpretation remains exploratory, and the underlying mechanisms were not directly measured in the present study. Given the wide confidence interval and limited precision of the interaction estimate, this finding should be interpreted as hypothesis-generating rather than indicative of a true interaction pattern.

Overall, these findings highlight the importance of considering measurement limitations and biological susceptibility when interpreting the role of training-related factors in UI risk.

From a practical perspective, these findings suggest that modifiable lifestyle factors, particularly smoking, should be considered in the assessment of UI risk among female athletes. In addition to sport-specific evaluation, routine screening for UI symptoms and relevant lifestyle behaviors may be warranted in athletic settings.

Integrating pelvic floor health education and smoking cessation support into athlete care programs may therefore represent feasible preventive strategies. At the same time, the findings suggest that training-related factors alone may not fully explain UI risk, underscoring the need for a more comprehensive approach that considers both mechanical exposure and individual susceptibility.

Several limitations should be considered when interpreting the findings of this study. The limited number of UI cases led to wide confidence intervals, particularly for the interaction model, indicating reduced statistical precision and potential instability of estimates. In addition, no formal statistical power analysis was conducted, and the study may have been underpowered to detect smaller or interaction effects. The cross-sectional design precludes causal inference, and training duration was used as a simplified proxy for external training load, without accounting for intensity, sport type, cumulative exposure, or sport-specific biomechanical demands. In addition, although some potentially relevant variables, such as parity and selected medical or lifestyle characteristics, were examined descriptively, they could not be meaningfully incorporated into the regression models due to their low prevalence and limited variability within the sample.

Additionally, data were self-reported and may therefore be subject to reporting bias. The use of a non-probability sampling online survey may limit the generalizability of the findings due to potential selection bias and limited representativeness. Furthermore, the analysis was restricted to complete cases without imputation, which may have introduced bias if missing data were not completely at random and may have affected the stability of the results. Given these considerations, the findings should be interpreted with caution, particularly when generalizing beyond young, competitively active female athletes. However, they provide context-specific insights that may be relevant to comparable athletic populations and may serve as a basis for future hypothesis-driven research.

Future studies should aim to incorporate more detailed, multidimensional measures of training load, as well as behavioral and biological factors, to better capture the complexity of UI risk in athletes. Longitudinal designs would be particularly valuable for examining potential interaction effects over time and for clarifying causal pathways.

CONCLUSION

This secondary analysis suggests that behavioral factors may play a relevant role in the occurrence of UI among young female athletes. In this sample, smoking was the only variable significantly associated with UI. In contrast, age, BMI, and daily training duration, used as a simplified proxy for external training load, were not independently associated with symptom presence. Although training duration alone was not associated with UI, the exploratory interaction analysis identified a potential pattern that should be interpreted as hypothesis-generating rather than confirmatory. This pattern may reflect a difference in the relationship between training duration and UI across behavioral factors such as smoking. Given the limited sample size and the small number of UI cases, these findings should be interpreted with caution and confirmed in larger, adequately powered studies in athletic populations.

Ethical Approval and Consent to Participate: The original study was conducted in accordance with the principles of the Declaration of Helsinki. Participation was voluntary and anonymous, and informed consent was obtained from all participants prior to survey completion. As no identifiable personal data were collected, formal ethical approval was not required under applicable regulations. The present study represents a secondary analysis of a fully anonymized dataset.

Funding Statement: This research received no external funding.

Consent for Publication: Not applicable.

Availability of Data and Materials: The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request. Data will be shared in an anonymous form and with appropriate acknowledgment of the original study.

Conflicts of Interest: The authors declare no conflict of interest.

References

  1. Aoki Y, Brown HW, Brubaker L, Cornu JN, Daly JO, Cartwright R. Urinary incontinence in women. Nat Rev Dis Primers. 2017; 3: 17042. doi: 10.1038/NRDP.2017.42
  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. Int Urogynecol J. 2010; 21(1): 5–26. doi: 10.1007/S00192-009-0976-9
  3. Castanier C, Bougault V, Teulier C, Jaffré C, Schiano-Lomoriello S, Vibarel-Rebot N, et al. The Specificities of Elite Female Athletes: A Multidisciplinary Approach. Life (Basel). 2021; 11(7): 622. doi: 10.3390/life11070622
  4. Teixeira RV, Colla C, Sbruzzi G, Mallmann A, Paiva LL. Prevalence of urinary incontinence in female athletes: a systematic review with meta-analysis. Int Urogynecol J. 2018; 29(12): 1717–1725. doi: 10.1007/S00192-018-3651-1
  5. de Mattos Lourenço TR, Matsuoka PK, Baracat EC, Haddad JM. Urinary incontinence in female athletes: a systematic review. Int Urogynecol J. 2018; 29(12): 1757–1763. doi: 10.1007/s00192 -018-3629-z
  6. Pires T, Pires P, Moreira H, Viana R. Prevalence of Urinary Incontinence in High-Impact Sport Athletes: A Systematic Review and Meta-Analysis. J Hum Kinet. 2020; 73(1): 279–288. doi: 10.2478/hukin-2020-0008
  7. Carvalhais A, Natal Jorge R, Bø K. Performing high-level sport is strongly associated with urinary incontinence in elite athletes: a comparative study of 372 elite female athletes and 372 controls. Br J Sports Med. 2018; 52(24): 1586–1590. doi: 10.1136/bjsports-2017-097587
  8. Rodríguez-López ES, Calvo-Moreno SO, Basas-García Á, Gutierrez-Ortega F, Guodemar-Pérez J, Acevedo-Gómez MB. Prevalence of urinary incontinence among elite athletes of both sexes. J Sci Med Sport. 2021; 24(4): 338–344. doi: 10.1016/j.jsams.2020.09.017
  9. Velázquez-Saornil J, Méndez-Sánchez E, Gómez-Sánchez S, Sánchez-Milá Z, Cortés-Llorente E, Martín-Jiménez A, et al. Observational Study on the Prevalence of Urinary Incontinence in Female Athletes. Int J Environ Res Public Health. 2021; 18(11). doi: 10.3390/ijerph18115591
  10. Wong AMK, Pei YC, Huang PP, Cheng HW, Chou SW, Li CY. Urinary incontinence in young female athletes of Taiwan. J Rehabil Med Assoc ROC. 2002; 30(4): 201–206. doi: 10.6315/3005-3846.2172
  11. Saulicz M, Saulicz A, Saulicz E. Self-assessment of lower urinary tract condition in female competitive cyclists. Healthcare (Basel). 2024; 12(12): 1163. doi: 10.3390/healthcare1212116 3
  12. Hannestad YS, Rortveit G, Daltveit AK, Hunskaar S. Are smoking and other lifestyle factors associated with female urinary incontinence? The Norwegian EPINCONT Study. BJOG. 2003; 110(3): 247–254.
  13. Lopes S, Becam M, Pierrot C, Réard J, Carvalhais A, Vieira Á, et al. Factors associated with urinary incontinence in female weightlifters. Healthcare (Basel). 2026; 14(3): 381. doi: 10.3390/healthcare14030381
  14. Lončarić Kelečić I, Schuster S, Uršej E. Urinary incontinence and risk factors among Croatian female athletes-a pilot cross-sectional observational study. Phys Act Rev. 2023; 11(2): 9–20. doi: 10.16926/par.2023.11.17
  15. Knuutinen A, Kokkonen N, Risteli J, Vähäkangas K, Kallioinen M, Salo T, et al. Smoking affects collagen synthesis and extracellular matrix turnover in human skin. Br J Dermatol. 2002; 146(4): 588–594. doi: 10.1046/j.1365-2133.2002.04694.x
  16. Sørensen LT. Wound healing and infection in surgery. The clinical impact of smoking and smoking cessation: a systematic review and meta-analysis. Arch Surg. 2012; 147(4): 373–383. doi: 10.1001/archsurg.2012.5
  17. Sørensen LT, Toft BG, Rygaard J, Ladelund S, Paddon M, James T, et al. Effect of smoking, smoking cessation, and nicotine patch on wound dimension, vitamin C, and systemic markers of collagen metabolism. Surgery. 2010; 148(5): 982–990. doi: 10.1016/j.surg.2010.02.005
  18. Bø K. Urinary incontinence, pelvic floor dysfunction, exercise and sport. Sports Med. 2004; 34(7): 451–464. doi: 10.2165/00007256-200434070-00004
  19. Bø K, Nygaard IE. Is Physical Activity Good or Bad for the Female Pelvic Floor? A Narrative Review. Sports Med. 2020; 50(3): 471–484. doi: 10.1007/s40279-019-01243-1
  20. Cheng HG, Phillips MR. Secondary analysis of existing data: opportunities and implementation. Shanghai Arch Psychiatry. 2014; 26(6): 371. doi: 10.11919/j.issn.1002-0829.214171
  21. Wickham J. Secondary Analysis Research. J Adv Pract Oncol. 2019; 10(4): 395. doi: 10.6004/jadpro.2019.10.4.7
  22. Wang Z, Shi Y, Xu Z, Xi J, Zhang Y. A multidimensional analysis-based risk prediction model for stress urinary incontinence in middle-aged and elderly women. BMC Womens Health. 2025; 25(1). doi: 10.1186/s12905-025-03975-x
  23. Impellizzeri FM, Marcora SM, Coutts AJ. Internal and External Training Load: 15 Years On. Int J Sports Physiol Perform. 2019; 14(2): 270–273. doi: 10.1123/ijspp.2018-0935
  24. Passfield L, Murias JM, Sacchetti M, Nicolo A. Validity of the Training-Load Concept. Int J Sports Physiol Perform. 2022; 17(4): 507–514. doi: 10.1123/ijspp.2022-0247
  25. Impellizzeri FM, Menaspà P, Coutts AJ, Kalkhoven J, Menaspà MJ. Training Load and Its Role in Injury Prevention, Part I: Back to the Future. J Athl Train. 2020; 55(9): 885. doi: 10.4085/1062-6050-500-19
  26. Mikuš M, Ćorić M, Matak L, Škegro B, Vujić G, Banović V. Validation of the UDI-6 and the ICIQ-UI SF - Croatian version. Int Urogynecol J. 2020; 31(12): 2625–2630. doi: 10.1007/S00192-020-04500-4
  27. Jensen LCG, Boie S, Axelsen S. International consultation on incontinence questionnaire -Urinary incontinence short form ICIQ-UI SF: Validation of its use in a Danish speaking population of municipal employees. PLoS One. 2022; 17(4). doi: 10.1371/journal.pone.0266479
  28. Avery K, Donovan J, Peters TJ, Shaw C, Gotoh M, Abrams P. ICIQ: a brief and robust measure for evaluating the symptoms and impact of urinary incontinence. Neurourol Urodyn. 2004; 23(4): 322–330. doi: 10.1002/nau.20041
  29. Flora DB. Your Coefficient Alpha Is Probably Wrong, but Which Coefficient Omega Is Right? A Tutorial on Using R to Obtain Better Reliability Estimates. Adv Methods Pract Psychol Sci. 2020; 3(4): 484–501. doi: 10.1177/2515245920951747
  30. Dunn TJ, Baguley T, Brunsden V. From alpha to omega: A practical solution to the pervasive problem of internal consistency estimation. British Journal of Psychology. 2014; 105(3): 399–412. doi: 10.1111/bjop.12046
  31. Peng CYJ, Lee KL, Ingersoll GM. An Introduction to Logistic Regression Analysis and Reporting. J Educ Res. 2002; 96(1): 3–14. doi: 10.1080/00220670209598786
  32. Szumilas M. Explaining Odds Ratios. Journal of the Canadian Academy of Child and Adolescent Psychiatry. 2010; 19(3): 227.
  33. Osborne JW, Waters E. Four assumptions of multiple regression that researchers should always test. Practical Assessment, Research, and Evaluation. 2002; 8(1): 2. doi: 10.7275/r222-hv23
  34. Steyerberg EW, Vickers AJ, Cook NR, Gerds T, Gonen M, Obuchowski N, et al. Assessing the performance of prediction models: a framework for some traditional and novel measures. Epidemiology. 2010; 21(1): 128. doi: 10.1097/EDE.0b013e3181c30fb2
  35. Peduzzi P, Concato J, Kemper E, Holford TR, Feinstem AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol. 1996; 49(12): 1373–1379. doi: 10.1016/s0895-4356(96)00236-3
  36. Babyak MA. What you see may not be what you get: a brief, nontechnical introduction to overfitting in regression-type models. Psychosom Med. 2004; 66(3): 411–421. doi: 10.1097/01.psy.0000127692.23278.a9
  37. International Ethical Guidelines for Health-related Research Involving Humans: Prepared by the Council for International Organizations of Medical Sciences (CIOMS) in collaboration with the World Health Organization (WHO) [Internet]. 4th ed. Geneva: Council for International Organizations of Medical Sciences; 2016. Available from: https: //www.ncbi.nlm.nih.gov /books/NBK614410/. doi: 10.56759/rgxl7405
  38. Dias N, Peng Y, Khavari R, Nakib NA, Sweet RM, Timm GW, et al. Pelvic floor dynamics during high-impact athletic activities: a computational modeling study. Clin Biomech (Bristol, Avon). 2017; 41: 20–27. doi: 10.1016/j.clinbiomech.2016.11.003
  39. Merino-Muñoz P, Pérez-Contreras J, Inostroza-Ríos F, Vidal-Maturana F, Campbell B, Villaseca-Vicuña R. External and Internal Load of International Matches According to Age Categories and Positions: A Case of Women’s National Team. Phys Act Rev 2025; 13(2): 70–78. doi: 10.16926/par.2025.13.21
  40. Petros PE, Ulmsten UI. An integral theory of female urinary incontinence. Experimental and clinical considerations. Acta Obstet Gynecol Scand Suppl. 1990; 153: 7–31. doi: 10.1111/j.1600-0412.1990.tb08027.x
  41. DeLancey JO. Structural support of the urethra as it relates to stress urinary incontinence: the hammock hypothesis. Am J Obstet Gynecol. 1994; 170(6): 1713–1723. doi: 10.1016/S0002-9378(94)70346-9
  42. Ashton-Miller JA, DeLancey JO. Functional anatomy of the female pelvic floor. Ann N Y Acad Sci. 2007; 1101: 266–296. doi: 10.1196/annals.1389.034

Article information

Published

2026-05-18

Issue

Vol. 14 No. 2 (2026)

Article type

ORIGINAL ARTICLE

License

Copyright (c) 2026 The authors

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

Phys Act Rev. 2026;14(2):100-112.DOI: 10.16926/par.2026.14.25

How to Cite

Lončarić Kelečić, I., Schuster, S., & Mesarić, J. (2026). Exploratory modeling of urinary incontinence risk among female athletes the role of smoking and training duration. Physical Activity Review, 14(2), 100-112. https://doi.org/10.16926/par.2026.14.25