Multimodal Pelvic Floor Ultrasound Parameters Associated with Postpartum Pelvic Floor Dysfunction: A Retrospective Study

Summary

This retrospective study of 160 patients developed a five-factor postpartum pelvic floor dysfunction (PFD) association model using pelvic floor ultrasound across resting, maximal contraction, and Valsalva states. The findings show strong associations in this single-center dataset but remain exploratory and require external validation before clinical application.

Abstract

Research on multimodal ultrasound parameters associated with postpartum pelvic floor dysfunction (PFD) across resting, maximal contraction, and Valsalva states remains limited. This retrospective study assessed these parameters in 172 patients examined between January 2022 and June 2025. After stratification into PFD and non-PFD groups, 1:1 propensity score matching (PSM) yielded 160 patients (80 per group) for analysis. Ultrasound parameters were evaluated across the three functional states. Logistic regression identified independent factors associated with PFD, and receiver operating characteristic (ROC) curve analysis assessed model performance. After PSM, baseline characteristics were comparable between groups (all p > 0.05). Compared with the non-PFD group, the PFD group had a larger resting levator ani hiatus area, a more inferior bladder neck position, and lower levator ani muscle thickness (all p < 0.001). During maximal contraction, the hiatus area reduction percentage and bladder neck elevation were lower. During the Valsalva maneuver, the hiatus area, bladder neck mobility, and most distal prolapse position were greater, and levator ani avulsion was more frequent (all p < 0.001). A larger resting levator ani hiatus area, lower resting muscle thickness, a lower hiatus area reduction percentage, greater bladder neck mobility during the Valsalva maneuver, and levator ani avulsion were independently associated with postpartum PFD (all p < 0.001). The five-parameter model had an area under the curve (AUC) of 0.870 (95% confidence interval, 0.815–0.926), with 91.3% sensitivity and 88.8% specificity. The model showed favorable internal fitting performance in the development dataset. Because independent validation was not performed, the findings should be considered hypothesis-generating; applications in clinical screening and individualized intervention remain unproven.

Introduction

Postpartum pelvic floor dysfunction (PFD) encompasses a range of pelvic floor symptoms after childbirth, including pelvic organ descent, stress or urgency urinary incontinence, fecal incontinence, chronic pelvic pain, and sexual dysfunction. These conditions may result from pregnancy- and delivery-related injury to the supportive soft tissues of the pelvic floor¹. In this study, PFD was treated as a composite outcome rather than a single disorder; however, group assignment was based only on pelvic organ prolapse, stress or urgency urinary incontinence, and pelvic floor muscle strength, whereas fecal incontinence, chronic pelvic pain, and sexual dysfunction were not used as grouping criteria.

The pelvic floor support system consists of the levator ani muscle, including the puborectalis muscle, pelvic fascia, ligaments, and connective tissues surrounding the pelvic organs. Its structural integrity and functional coordination help maintain the normal position of the pelvic organs and support urinary and fecal function². Childbirth is a major risk factor for PFD. Excessive stretching and laceration of the pelvic floor muscles and fascia during vaginal delivery may contribute to PFD³˒⁴. Because postpartum PFD may have an insidious onset, some patients do not seek medical attention until symptoms or tissue damage have progressed, which may complicate treatment⁵. Therefore, an accurate, noninvasive method for identifying patients at increased risk and developing an association model may support earlier clinical assessment. However, any clinical application would require appropriate validation.

Current methods for evaluating pelvic floor function include physical examination, imaging, and pelvic floor electromyography. Physical examination is simple and inexpensive but is subjective and may not quantify subtle structural or functional abnormalities⁶˒⁷. Pelvic floor electromyography reflects neuromuscular function but does not directly show morphological changes in pelvic floor anatomy⁸. Imaging provides structural and functional information and is widely used for pelvic floor assessment⁹.

Pelvic floor ultrasonography includes two-dimensional (2D), three-dimensional (3D), and four-dimensional (4D) imaging. Two-dimensional ultrasound provides basic morphological information, whereas 3D ultrasound permits reconstruction of the levator ani hiatus and visualization of axial and oblique planes¹⁰. Four-dimensional ultrasound enables dynamic assessment during maneuvers such as maximal pelvic floor muscle contraction and Valsalva. Translabial ultrasound provides a noninvasive view of pelvic floor anatomy and can be repeated during follow-up. Compared with magnetic resonance imaging, ultrasound is generally more accessible and less costly; however, measurements remain operator dependent¹¹.

Pelvic floor morphology and organ position change across functional states. Assessment at rest provides a baseline view of anatomy and structural integrity. Maximal pelvic floor muscle contraction permits evaluation of muscle shortening, coordination, and functional reserve¹²˒¹³. The Valsalva maneuver increases intra-abdominal pressure and permits assessment of pelvic organ descent and pelvic floor support under strain. Together, these states provide complementary information about pelvic floor structure and function¹⁴. Assessment in only one state may therefore provide an incomplete representation of pelvic floor function.

Most previous studies have evaluated a single ultrasound mode, a single functional state, or a specific PFD-related condition¹⁵. Machine learning-based models have also been developed for selected PFD-related outcomes, but models addressing PFD as a composite outcome remain limited¹⁶. In this study, “combined association” referred to integrating multiple main-effect ultrasound parameters obtained across resting, maximal contraction, and Valsalva states into a weighted model; it did not refer to statistical interaction or synergy among predictors. PFD was defined as a composite outcome in accordance with the Expert Consensus on Diagnosis and Rehabilitation Treatment of Pelvic Floor Dysfunctional Diseases (2024 Edition)¹⁷.

Accordingly, this retrospective observational study evaluated multimodal pelvic floor ultrasound parameters obtained at rest, during maximal pelvic floor muscle contraction, and during the Valsalva maneuver in postpartum patients. The study aimed to identify factors independently associated with postpartum PFD and to assess the discrimination and calibration of a combined association model. Because the study was conducted retrospectively at a single center and did not include external validation, the findings should be considered exploratory and hypothesis-generating.

Protocol

The Medical Ethics Committee approved the retrospective use and analysis of data collected between January 2022 and June 2025 under approval number 2026-023 and waived the requirement for informed consent. The research tools used in this protocol are listed in the Table of Materials.

1. Study design

Medical records were retrieved for 175 patients who underwent pelvic floor ultrasound evaluation at the hospital between January 2022 and June 2025. After screening according to the inclusion and exclusion criteria, 172 eligible patients were enrolled. All participants were assessed according to the diagnostic criteria specified in the Expert Consensus on Diagnosis and Rehabilitation Treatment of Pelvic Floor Dysfunctional Diseases (2024 Edition)¹⁷. Patients who met at least one diagnostic criterion were assigned to the PFD group, whereas those without identifiable pelvic floor-related manifestations on clinical assessment were assigned to the non-PFD group.

Before propensity score matching (PSM), the PFD group comprised 87 patients and the non-PFD group 85. One-to-one PSM was then performed to balance baseline confounders, resulting in 80 patients in each group. All matched patients were included in the subsequent analyses, and the study workflow is shown in Figure 1.

Figure 1: Research flowchart. A total of 175 patients were initially screened. After application of the eligibility criteria, 172 patients were included. Following propensity score matching, 160 patients were retained for analysis, with 80 patients in each group. Please click here to view a larger version of this figure.

2. Inclusion criteria

Patients were eligible if they were 20–45 years of age, had a singleton pregnancy, underwent multimodal pelvic floor ultrasound at 42 days ± 2 weeks after delivery, completed standardized multimodal pelvic floor ultrasound in the hospital’s Ultrasound Department in accordance with the Expert Consensus on Clinical Practical Standardized Examination of Pelvic Floor Ultrasound (2022)¹⁸, had complete clinical data, and completed a standardized clinical evaluation of postpartum pelvic floor function in accordance with the Expert Consensus on Diagnosis and Rehabilitation Treatment of Pelvic Floor Dysfunctional Diseases (2024 Edition)¹⁷.

Under this consensus, PFD was defined as a nonorganic disorder associated with weakness, defects, or dysfunction of the pelvic floor. Diagnosis was based on clinical symptoms, physical examination, and ancillary testing. Clinical manifestations included pelvic organ prolapse, stress urinary incontinence, urgency urinary incontinence, fecal incontinence, chronic pelvic pain, and sexual dysfunction. Pelvic organ prolapse was staged using the Pelvic Organ Prolapse Quantification (POP-Q) system. Ancillary assessments included pelvic floor ultrasound, urodynamic testing, and pelvic floor electromyography.

3. Exclusion criteria

Patients were excluded if they had a confirmed diagnosis of PFD before pregnancy, congenital pelvic floor abnormalities, or chronic pelvic pain syndrome; neurological disease, connective tissue disease, severe hepatic or renal dysfunction, diabetic peripheral neuropathy, or another systemic disorder; a history of pelvic surgery or poor healing after repair of a severe perineal laceration; more than 2 weeks of postpartum pelvic floor rehabilitation; a multiple pregnancy, fetal malformation, or severe pregnancy complication; or mental illness or cognitive impairment¹⁹.

Pre-existing PFD or congenital pelvic floor abnormalities could have confounded the assessment of postpartum pelvic floor changes. Neurological, connective tissue, and metabolic disorders could have independently affected pelvic floor neuromuscular function or tissue elasticity. Previous pelvic surgery or failed perineal repair could have altered pelvic floor anatomy and biomechanics. Postpartum pelvic floor rehabilitation lasting more than 2 weeks could have modified muscle morphology and contractile function. Multiple pregnancy, fetal malformation, and severe pregnancy complications could have increased baseline heterogeneity or altered the course of delivery. Mental illness or cognitive impairment could have affected cooperation with standardized ultrasound maneuvers and symptom assessment.

4. Data collection

Two investigators completed standardized training before data collection and extracted the data independently. Both sonographers had more than 5 years of experience in pelvic floor ultrasound. Training included a 2 h didactic session covering the study protocol, standardized image-acquisition planes, and predefined measurement landmarks; practical sessions using five representative cases; consensus discussions regarding measurement techniques; use of a structured measurement template; and a pilot interobserver agreement assessment.

Before formal data extraction, the sonographers independently assessed 20 non-enrolled cases. Intraclass correlation coefficients for key parameters, including levator ani hiatus area, bladder neck position, and levator ani muscle thickness, exceeded 0.85. Formal extraction then proceeded. During data extraction, one repeat measurement was inserted after every 10 cases as an internal calibration check. Because of the retrospective design, formal interobserver and intraobserver reliability statistics were not calculated for the full cohort. Extracted data included baseline clinical characteristics, multimodal ultrasound parameters, and outcome measures.

5. Outcome classification

The primary outcome was postpartum PFD, assessed at 42 days ± 2 weeks after delivery in accordance with the Expert Consensus on Diagnosis and Rehabilitation Treatment of Pelvic Floor Dysfunctional Diseases (2024 Edition)¹⁷ and International Continence Society definitions.

A patient was assigned to the PFD group if she met at least one of the following criteria: pelvic organ prolapse of POP-Q stage II or higher, with the most distal portion of the prolapse located within 1 cm above or below the hymenal ring; stress or urgency urinary incontinence confirmed by a specialist on the basis of medical history and physical examination, including a cough stress test when applicable, with involuntary urine leakage occurring at least once per week during the previous 3 months; or pelvic floor muscle strength below grade 3 on the Modified Oxford Scale during manual vaginal assessment.

Patients who met none of these criteria were assigned to the non-PFD group. Fecal incontinence, chronic pelvic pain, and sexual dysfunction were not used for group assignment.

6. Observation indicators

All ultrasound measurements were performed according to the standardized protocols in the Expert Consensus on Clinical Practical Standardized Examination of Pelvic Floor Ultrasound (2022)¹⁸. Baseline pelvic floor ultrasound measurements were retrieved from pre-pregnancy health screening records or routine first-trimester ultrasound examinations performed at ≤12 weeks of gestation. Postpartum ultrasound measurements were obtained from standardized examinations conducted at 42 days ± 2 weeks after delivery.

1. Indicators at rest
   The levator ani hiatus area was measured using three-dimensional (3D) ultrasound. The hiatus was bounded by the inferior margin of the pubic symphysis and the medial borders of the bilateral levator ani muscles, and the cross-sectional area was recorded²⁰.
   Bladder neck position was measured using two-dimensional (2D) ultrasound as the vertical distance from the midpoint of the bladder neck to the inferior margin of the pubic symphysis. In this study, a greater distance indicated a more inferior, or caudal, bladder neck position; that is, the bladder neck was located farther from the pubic symphysis in the caudal direction²¹.
   Levator ani muscle thickness was measured using 2D ultrasound at the midpoint of the pubic segment of the muscle, excluding the surrounding fascia²².
2. Indicators during maximal pelvic floor muscle contraction
   Patients performed maximal pelvic floor muscle contraction, similar to holding urine or stool, for 3–5 s. Dynamic images were acquired using four-dimensional (4D) ultrasound, with 3D reconstruction and 2D localization.
   The reduction percentage of the levator ani hiatus area was calculated from measurements obtained at rest and during contraction as follows:
   Reduction percentage = (resting area − contraction area) / resting area × 100%²³˒²⁴
   Bladder neck elevation was measured using 2D ultrasound as the vertical upward displacement of the bladder neck midpoint from rest to maximal contraction²⁵.
   Levator ani avulsion was assessed during maximal contraction using 3D volume reconstruction. Continuity at the origin and insertion of the levator ani muscle was evaluated, and each patient was classified as having no avulsion, unilateral avulsion, or bilateral avulsion. The diagnostic criteria followed the Expert Consensus on Clinical Practical Standardized Examination of Pelvic Floor Ultrasound (2022)¹⁸. Assessment was performed during contraction because contraction improved delineation of the muscle borders and avulsion defects.
3. Indicators during the valsalva maneuver
   Patients received standardized instructions for performing the Valsalva maneuver while avoiding concomitant levator ani contraction. When necessary, biofeedback training was provided to facilitate pelvic floor relaxation²⁶. Dynamic pelvic floor changes were observed with 4D ultrasound, and 2D and 3D ultrasound were used for measurement and assessment.
   The maneuver was performed as forced expiration against a closed glottis, and patients were instructed to push downward as if defecating while consciously relaxing the pelvic floor²⁷. The maneuver lasted at least 6 s and was repeated until an optimal image was obtained²⁷.
   The levator ani hiatus area was measured using 3D ultrasound at maximal dilation²⁸. Bladder neck mobility was measured using 2D ultrasound as the maximum vertical displacement of the bladder neck midpoint from rest to the Valsalva maneuver¹¹. Bladder neck elevation during maximal contraction and bladder neck mobility during the Valsalva maneuver were treated as distinct indicators. The former represented active lifting during muscle contraction, whereas the latter represented passive descent under strain.
   The most distal position of pelvic organ prolapse was measured using 2D ultrasound relative to the inferior margin of the pubic symphysis, which was defined as the zero point. In accordance with the convention used in this study, the caudal, or inferior, direction was recorded as positive, whereas the cephalic, or superior, direction was recorded as negative. Larger positive values, therefore, indicated a more distal prolapse position and greater prolapse severity²⁹.
   Levator ani muscle thickness was measured using 2D ultrasound at the midpoint of the pubic segment during the maneuver³⁰.

7. Statistical analysis

Propensity score matching (PSM) was used to balance confounding factors between the PFD and non-PFD groups. The propensity score model included age, pre-pregnancy body mass index, parity, mode of delivery, duration of labor, fetal weight, degree of perineal laceration, and pregnancy complications. These covariates were selected based on their documented associations with PFD and their availability in the medical records.

Propensity scores were estimated using logistic regression. One-to-one nearest-neighbor matching without replacement was performed using a caliper of 0.02. Only pairs with a propensity score difference of 0.02 or less were matched. Baseline characteristics were compared after matching to assess covariate balance.

Continuous data were assessed using the Kolmogorov–Smirnov test for normality and Levene’s test for homogeneity of variance. Normally distributed continuous data with homogeneous variance were reported as mean ± standard deviation and compared using the independent-samples t-test for between-group comparisons and the paired t-test for within-group comparisons.

Non-normally distributed continuous data were reported as median and interquartile range [M (Q1, Q3)] and compared using the Mann–Whitney U test for between-group comparisons and the Wilcoxon signed-rank test for within-group comparisons. Categorical data were reported as n (%) and compared using the chi-square test.

Because of the limited sample size and collinearity among multiple ultrasound parameters, univariate screening was used as a dimensionality-reduction step. Variables with p < 0.05 in the univariate analysis were entered into the subsequent multivariable logistic regression model. This approach was intended to identify dominant main-effect predictors across the three functional states rather than to test statistical interactions or synergistic effects.

In this study, multimodal combined evaluation involved acquiring ultrasound data across resting, maximal contraction, and Valsalva states and integrating models of significant main-effect variables. A combined association model was constructed using the full matched dataset. A receiver operating characteristic (ROC) curve was generated to assess the model's internal-fit performance in the development cohort, and the area under the curve (AUC), sensitivity, and specificity were calculated. These performance metrics were descriptive measures of model fit in the development cohort and did not represent validated predictive accuracy.

Bootstrap resampling with 1,000 iterations was performed solely to derive empirical standard errors, bootstrap p values, and bias-corrected 95% confidence intervals for the regression coefficients. This procedure was used to assess parameter estimation and coefficient stability. It did not provide optimism-corrected AUC estimates or equivalent adjustments to model-performance metrics and was therefore not considered model validation.

No internal validation, such as a training/test split or k-fold cross-validation, and no external validation in an independent cohort were performed. Consequently, the reported AUC, sensitivity, and specificity reflected discrimination within the development dataset only and remained subject to overfitting. The generalizability and clinical applicability of the model, therefore, remain unproven and require independent validation. All statistical tests were two-sided, and p < 0.05 was considered statistically significant.

8. Sample size calculation

No a priori sample size calculation was performed. The final sample included 80 patients in the PFD group and 80 in the non-PFD group, determined by the number of eligible patients with complete data.

A post hoc power analysis was performed using software. One previous retrospective cohort study included 320 postpartum patients, of whom 114 had PFD and 206 did not, and developed a model based on pelvic floor ultrasound parameters³¹. Another study used multimodal ultrasound to predict postpartum PFD in 200 patients, including 67 patients with PFD and 133 without PFD³².

Based on the reported between-group differences and model discrimination in these studies, Cohen’s d values for key ultrasound parameters ranged from 0.50–0.80. Using a conservative Cohen’s d of 0.65, a two-sided significance level of α = 0.05, and a type II error rate of β = 0.05, the estimated required sample size was approximately 63 patients per group. The actual sample size of 80 patients per group exceeded this estimate. The minimum detectable effect for levator ani muscle thickness was additionally estimated as 0.07, which was smaller than the observed effect size of 0.701.

Results

Baseline data
Before propensity score matching (PSM), age, neonatal birth weight, and total duration of labor differed significantly between the two groups (all p < 0.05). After PSM, no baseline characteristic differed significantly between the groups (all p > 0.05). These findings indicated that PSM balanced the measured baseline confounders between the groups (Table 1).

Indicators	Before PSM				After PSM
	Non-PFD group (n = 85)	PFD group (n = 87)	p	Effect size	Non-PFD group (n = 80)	PFD group (n = 80)	p	Effect size
Age (years, mean ± SD)	28.39 ± 2.87	29.37 ± 3.29	0.039	Cohen’s D = 0.317	28.79 ± 2.45	29.04 ± 2.36	0.512	Cohen’s D = 0.104
Pre-pregnancy BMI (kg / m2,mean ± SD )	21.81 ± 1.08	21.56 ± 1.04	0.129	Cohen’s D = 0.232	21.72 ± 1.06	21.55 ± 1.06	0.29	Cohen’s D = 0.168
Parity (n %)			0.666	Phi = 0.033			0.614	Phi = 0.040
Primiparity	57 (67.1%)	61 (70.1%)			52 (65%)	55 (68.8%)
Multiparity	28 (32.9%)	26 (29.9%)			28 (35%)	25 (31.2%)
Delivery mode (n %)			0.252	Phi = 0.087			0.574	Phi = 0.044
Transvaginal	67 (78.8%)	62 (71.3%)			63 (78.8%)	60 (75%)
CS	18 (21.2%)	25 (28.7%)			17 (21.2%)	20 (25%)
Fetal weight (g, mean±SD)	3279.63 ± 155.31	3227.98 ± 165.26	0.036	Cohen’s D=0.322	3277.71 ± 157.74	3243.48 ± 162.15	0.178	Cohen’s D = 0.214
Duration of labor (h, mean ± SD)	8.78 ± 1.48	8.34 ± 1.42	0.044	Cohen’s D = 0.309	8.75 ± 1.51	8.41 ± 1.46	0.147	Cohen’s D = 0.230
Degree of perineal laceration (n %)			0.646	Phi = 0.035			0.391	Phi = 0.068
No laceration/Grade Ⅰ	67 (78.8%)	71 (81.6%)			65 (81.2%)	69 (86.2%)
Grade Ⅱ	18 (21.2%)	16 (18.4%)			15 (18.8%)	11 (13.8%)
Pregnancy complications (n %)	14 (16.5%)	11(12.6%)	0.476	Phi = 0.054	12 (15%)	11 (13.8%)	0.822	Phi = 0.018

Table 1: Baseline clinical characteristics [mean ± SD or n (%)]. Baseline clinical characteristics of participants before and after propensity score matching. Continuous variables are presented as mean ± SD, and categorical variables are presented as n (%).

Indicators at rest
As shown in Table 2, levator ani hiatus area, bladder neck position, and levator ani muscle thickness at rest on antepartum baseline did not differ significantly between the groups (all p > 0.05). After delivery, compared with the non-PFD group, the PFD group had a larger levator ani hiatus area (p < 0.001; 95% CI, -6.29–-4.37), a more caudal bladder neck position (p < 0.001; 95% CI, -0.78–-0.72), and lower levator ani muscle thickness (p < 0.001; 95% CI, 0.15–0.38). These findings indicated greater morphological changes in the pelvic floor structures of the PFD group at rest.

Indicators	Time	Non-PFD group (n = 80)	PFD group (n = 80)	p	95% CI of the difference	Effect size (Cohen’s D)
Levator ani hiatus area (cm2)	Antepartum baseline	19.71 ± 2.85	20.27 ± 2.89	0.223	- 1.45, 0.34	0.193
	Postpartum	21.82 ± 2.59*	27.15 ± 3.48*	< 0.001	-6.29, -4.37	1.738
Bladder neck position (cm)	Antepartum baseline	3.15 ± 0.11	3.14 ± 0.11	0.564	- 0.02, 0.04	0.091
	Postpartum	3.80 ± 0.08*	4.55 ± 0.11*	< 0.001	- 0.78, - 0.72	7.599
Levator ani muscle thickness (mm)	Antepartum baseline	4.07 ± 0.46	4.02 ± 0.36	0.446	- 0.08, 0.18	0.121
	Postpartum	3.66 ± 0.35*	3.40 ± 0.40*	< 0.001	0.15, 0.38	0.701

Table 2: Comparison of indicators at rest (mean ± SD). Comparison of pelvic floor indicators at rest between the non-PFD and PFD groups during the antepartum baseline and postpartum periods. Data are presented as mean ± SD. Intergroup comparisons are reported using p values, 95% confidence intervals, and Cohen’s d effect sizes. *p < 0.05 versus antepartum baseline status within the same group using paired samples t-test..

Indicators during maximal pelvic floor muscle contraction
During maximal pelvic floor muscle contraction, the reduction percentage of the levator ani hiatus area was significantly higher in the non-PFD group than in the PFD group (p < 0.001; 95% CI, 9.41–10.28). Bladder neck elevation was also significantly greater in the non-PFD group than in the PFD group (p < 0.001; 95% CI, 0.89–1.01) (Table 3). These findings indicated reduced pelvic floor muscle contraction in the PFD group, as reflected by lower hiatus reduction and reduced bladder neck elevation.

Indicators	Non-PFD group (n = 80)	PFD group (n = 80)	p	95% CI of the difference	Effect size (Cohen’s d)
Reduction percentage of the levator ani hiatus area (%)	22.01 ± 1.08	12.16 ± 1.63	< 0.001	9.41 to 10.28	7.113
Bladder neck elevation (cm)	1.75 ± 0.11	0.80 ± 0.25	< 0.001	0.89 to 1.01	4.958

Table 3: Comparison of pelvic floor indicators during maximal contraction (mean ± SD). Comparison of pelvic floor indicators during maximal contraction between the non-PFD and PFD groups. Data are presented as mean ± SD. Intergroup comparisons are reported using p values, 95% CIs, and Cohen’s d effect sizes.

Indicators during the Valsalva maneuver
On antepartum baseline ultrasound, the levator ani hiatus area, bladder neck mobility, the most distal position of pelvic organ prolapse, and levator ani muscle thickness during the Valsalva maneuver did not differ significantly between the groups (all p > 0.05).

After delivery, compared with the non-PFD group, the PFD group had a larger levator ani hiatus area (p < 0.001; 95% CI, -4.41–-2.56), greater bladder neck mobility (p < 0.001; 95% CI, -1.54–-1.42), and a more distal position of pelvic organ prolapse (p < 0.001; 95% CI, -1.82–-1.47). Levator ani muscle thickness was significantly lower in the PFD group than in the non-PFD group (p < 0.001; 95% CI, 0.86–0.94). These findings indicated greater impairment of pelvic floor structural stability and support under increased abdominal pressure in the PFD group (Table 4).

Indicators	Time	Non-PFD group (n = 80)	PFD group (n = 80)	p	95% CI of the difference	Effect size (Cohen’s D)
Levator ani hiatus area (cm2)	Antepartum baseline	22.01 ± 3.02	22.40 ±2.99	0.413	- 1.33, 0.55	0.13
	Postpartum	25.05 ± 2.92*	28.54 ± 3.00*	< 0.001	- 4.41, - 2.56	1.179
Bladder neck mobility (cm)	Antepartum baseline	1.20 ± 0.14	1.20 ± 0.14	0.956	- 0.05, 0.04	0.009
	Postpartum	2.02 ± 0.22*	3.50 ± 0.14*	< 0.001	- 1.54, - 1.42	8.032
Most distal position of pelvic organ prolapse (cm)	Antepartum baseline	- 0.32 ± 0.57	- 0.20 ± 0.56	0.192	- 0.29, 0.06	0.207
	Postpartum	0.53 ± 0.56*	2.18 ± 0.56*	< 0.001	- 1.82, - 1.47	2.935
Levator ani muscle thickness (mm)	Antepartum baseline	4.24 ± 0.12	4.25 ± 0.11	0.731	- 0.04, 0.03	0.055
	Postpartum	3.75 ± 0.11*	2.85 ± 0.11*	< 0.001	0.86, 0.94	7.999

Table 4: Comparison of indicators during the Valsalva maneuver (mean ± SD). Comparison of pelvic floor indicators during the Valsalva maneuver between the non-PFD and PFD groups at the antepartum baseline and postpartum time points. Data are presented as mean ± SD. Intergroup comparisons are reported using p values, 95% CIs, and Cohen’s d effect sizes. *p < 0.05 versus antepartum baseline measurements within the same group using paired samples t-test.

Levator ani avulsion
Among the 80 patients in the non-PFD group, 76 had no levator ani avulsion (95.0%), 4 had unilateral avulsion (5.0%), and none had bilateral avulsion. Among the 80 patients in the PFD group, all had levator ani avulsion: 32 had unilateral avulsion (40.0%), and 48 had bilateral avulsion (60.0%). The distribution of avulsion status differed significantly between the groups (p < 0.001), indicating an association between levator ani avulsion and postpartum PFD (Table 5).

Group	No avulsion	Unilateral avulsion	Bilateral avulsion
Non-PFD group (n = 80)	76 (95.0%)	4 (5.0%)	0 (0.0%)
PFD group (n = 80)	0 (0.0%)	32 (40.0%)	48 (60.0%)
p	< 0.001
Effect size (Cramér’s V)	0.955

Table 5: Comparison of levator ani avulsion incidence [n (%)]. Comparison of levator ani avulsion incidence between the non-PFD and PFD groups. Categorical data are presented as n (%). Intergroup differences are reported using the p value and Cramér’s V effect size.

Factors independently associated with PFD
Multivariable logistic regression was performed using the ultrasound indicators measured across the functional states.
Levator ani hiatus area at rest (B = 0.514, p < 0.001; OR = 1.672; 95% CI, 1.432–1.952), levator ani muscle thickness at rest (B = 0.336, p < 0.001; OR = 1.400; 95% CI, 1.196–1.637), reduction percentage of the levator ani hiatus area during maximal pelvic floor muscle contraction (B = −1.114, p < 0.001; OR = 0.328; 95% CI, 0.195–0.552), bladder neck mobility during the Valsalva maneuver (B = 0.321, p = 0.002; OR = 1.379; 95% CI, 1.083–2.362), and levator ani avulsion (B = 1.099, p < 0.001; OR = 3.003; 95% CI, 1.565–5.747) were independently associated with postpartum PFD. However, the variable levator ani avulsion exhibited quasi-complete separation between the two groups (100% prevalence in the PFD group versus 5% in the non-PFD group). Consequently, the maximum likelihood estimate for this variable may be unstable, and the adjusted odds ratio (OR = 3.003) and its 95% confidence interval should be interpreted with caution; they should not be regarded as a precise or reliable effect estimate. The strong association between avulsion and PFD is best supported by the descriptive data presented in Table 5 rather than by the regression-derived point estimate.

Bladder neck position at rest, bladder neck elevation during maximal pelvic floor muscle contraction, levator ani hiatus area during the Valsalva maneuver, the most distal position of pelvic organ prolapse, and levator ani muscle thickness during the Valsalva maneuver were not significantly associated with PFD (all p > 0.05) (Table 6).

Bootstrap resampling was performed to assess the stability of the logistic regression estimates. Bootstrap p-values and 95% confidence intervals for the regression coefficients were calculated for all predictors. Variables that were statistically significant in the conventional logistic regression analysis remained significant after bootstrap resampling, with bootstrap p-values < 0.05 and bootstrap 95% confidence intervals that did not include zero. Variables that were nonsignificant in the conventional analysis remained nonsignificant after bootstrap resampling.

Condition	Indicators	B	p	OR	95% CI for OR	Bootstrap p	Bootstrap 95% CI for β
Rest	Levator ani hiatus area	0.514	< 0.001	1.672	1.432–1.952	< 0.001	0.420–0.670
	Bladder neck position	0.732	0.188	2.08	0.700–6.182	0.202	−0.223–2.348
	Levator ani muscle thickness	0.336	< 0.001	1.4	1.196–1.637	< 0.001	0.179–0.493
Maximal pelvic floor muscle contraction	Reduction percentage of the levator ani hiatus area	−1.114	< 0.001	0.328	0.195–0.552	< 0.001	−1.630 to −0.592
	Bladder neck elevation	0.044	0.411	1.045	0.941–1.159	0.214	−0.060–0.150
	Levator ani avulsion	1.099	< 0.001	3.003	1.565–5.747	< 0.001	0.448–1.749
Valsalva maneuver	Levator ani hiatus area	0.063	0.955	1.065	0.118–9.635	0.154	−2.140–2.260
	Bladder neck mobility	0.321	0.002	1.379	1.083–2.362	0.002	0.080–0.860
	Most distal position of pelvic organ prolapse	0.368	0.192	1.444	0.832–2.508	0.085	−0.184–0.919
	Levator ani muscle thickness	0.481	0.729	1.618	0.107–4.583	0.241	−2.230–1.520

Table 6: Analysis of factors independently associated with pelvic floor dysfunction. Multivariable logistic regression analysis of factors independently associated with pelvic floor dysfunction. B represents the regression coefficient. Lowercase italic p values and bootstrap-corrected p values are reported. ORs and corresponding 95% CIs, including raw and bootstrap-corrected estimates, are provided for each predictor.

Combined association model
A combined association model for postpartum PFD was constructed using the independently associated factors. The logistic regression equation was:

Logit (p) = 0.916 + 0.514X₁ + 0.336X₂ − 1.114X₃ + 0.321X₄ + 1.099X₅

In this equation, X₁ represented levator ani hiatus area at rest (cm²), X₂ represented levator ani muscle thickness at rest (cm), X₃ represented the reduction percentage of the levator ani hiatus area during maximal pelvic floor muscle contraction (%), X₄ represented bladder neck mobility during the Valsalva maneuver (cm), and X₅ represented levator ani avulsion (1 = avulsion; 0 = no avulsion).

Receiver operating characteristic (ROC) curve analysis of the derivation cohort showed an area under the curve (AUC) of 0.870 (95% CI, 0.815–-0.926), with a sensitivity of 91.3% and a specificity of 88.8%, indicating favorable internal discriminative ability within the development dataset. The Hosmer–Lemeshow test yielded a nonsignificant p value of 0.684, suggesting acceptable model calibration in this cohort (Figure 2 and Table 7).

In summary, across resting, maximal contraction, and Valsalva states, multiple ultrasound parameters showed significant between-group differences. The five parameters independently associated with postpartum PFD were integrated into this combined association model. Although the model exhibited good internal fitting performance and calibration in the development cohort, it is important to emphasize that these performance metrics reflect model fit within the derivation sample rather than validated predictive accuracy. As the model was not tested in an independent external cohort, its generalizability remains unproven, and the findings should be interpreted as hypothesis-generating rather than as a basis for clinical screening or individualized intervention.

Figure 2: Receiver operating characteristic curve of the combined association model. The receiver operating characteristic curve shows the discrimination of the five-parameter combined association model for postpartum pelvic floor dysfunction in the matched development cohort. The area under the curve was 0.870 (95% confidence interval, 0.815–0.926). Please click here to view a larger version of this figure.

Indicator	Youden’s index	Sensitivity (%)	Specificity (%)	AUC	95% CI for AUC	Hosmer–Lemeshow p
Combined association model	0.801	91.3	88.8	0.87	0.815–0.926	0.684

Table 7: Performance indicators of the combined association model for postpartum pelvic floor dysfunction. Internal fitting-performance metrics for the combined association model in the development cohort were derived from receiver operating characteristic curve analysis. Youden’s index, sensitivity, specificity, AUC with its 95% CI, and the Hosmer–Lemeshow p value for calibration assessment are presented.

Data Availability:
The de-identified data from 160 matched participants supporting the findings of this study are provided in Supplementary File 1. Direct personal identifiers have been removed.

Supplementary File 1: The de-identified data at the direct-identifier level. Please click here to download this file.

Abbreviations: AUC, area under the curve; BMI, body mass index; CI, confidence interval; CS, cesarean section; OR, odds ratio; PFD, pelvic floor dysfunction; PSM, propensity score matching; ROC, receiver operating characteristic; SD, standard deviation. B, regression coefficient; Cohen’s d, effect size for continuous variables; Cramér’s V, effect size for categorical variables; n, number of participants; p, probability value.

Discussion

This single-center retrospective observational study evaluated pelvic floor ultrasound indicators across three functional states. At rest, the PFD group had a larger levator ani hiatus area, a more caudal bladder neck position, and lower levator ani muscle thickness than the non-PFD group. During maximal pelvic floor muscle contraction, the reduction percentage of the levator ani hiatus area and bladder neck elevation were lower in the PFD group, indicating reduced active pelvic floor muscle contraction. During the Valsalva maneuver, the PFD group had a larger levator ani hiatus area, greater bladder neck mobility, a more distal position of pelvic organ prolapse, and a higher prevalence of levator ani avulsion. Following univariate screening, multivariable logistic regression identified five factors independently associated with postpartum PFD: larger levator ani hiatus area and lower levator ani muscle thickness at rest, a lower reduction percentage of the levator ani hiatus area during maximal contraction, greater bladder neck mobility during the Valsalva maneuver, and levator ani avulsion. The combined association model constructed from these five indicators showed favorable internal fitting performance in the development cohort.

Baseline pelvic floor ultrasound parameters obtained from pre-pregnancy health records or first-trimester ( ≤ 12 weeks) medical examinations did not differ significantly between the groups, whereas marked between-group differences were observed after delivery. During pregnancy, enlargement of the uterus may stretch the pelvic floor muscles, fascia, and ligaments, resulting in collagen remodeling and reduced elasticity of the pelvic floor support structures. Hormonal changes may further reduce pelvic floor tissue tension³³˒³⁴. During vaginal delivery, mechanical compression and stretching by the fetal head may damage pelvic floor muscle fibers and alter nerve conduction³⁵. These changes may contribute to postpartum pelvic floor relaxation and reduced contractile function. The larger levator ani hiatus area observed at rest in the PFD group may reflect impaired support by the levator ani muscle ring and dilation of the hiatus, which may reduce support for the pelvic organs³⁶˒³⁷. A more caudal bladder neck position may indicate relaxation of the periurethral supporting tissues, whereas lower levator ani muscle thickness may reflect atrophic or degenerative changes after muscle-fiber injury. The effects of measurement timing and edema resolution should also be considered³⁸.

During maximal pelvic floor muscle contraction, the lower reduction percentage of the levator ani hiatus area and reduced bladder neck elevation in the PFD group may reflect impaired muscle contractility and reduced active support. During the Valsalva maneuver, greater bladder neck mobility may indicate insufficient support during increased intra-abdominal pressure, whereas a more distal position of pelvic organ prolapse directly reflects reduced pelvic floor support³⁹. A previous cross-sectional study reported that the anteroposterior diameter of the levator ani hiatus during the Valsalva maneuver was greater in postpartum patients with PFD than in those without PFD and was associated with reduced pelvic floor support⁴⁰. Another prospective observational study reported that the levator ani hiatus area and levator ani muscle thickness at rest could be used to evaluate childbirth progression and potential PFD³⁰.

A single-center longitudinal study reported levator ani avulsion in 35 patients (17.9%) at 8 years after their first delivery, which was lower than the rate observed in the PFD group in the present study⁴¹. This difference may reflect variation in the study populations and assessment time points. The previous study included all primiparous patients, whereas the present study focused on patients with PFD, who may have had more severe structural injury. In addition, the previous study evaluated patients 8 years after delivery, whereas the present study evaluated patients at 42 days ± 2 weeks postpartum. Over time, tissue repair and fibrosis may also have affected the detectability of milder avulsion defects.

The principal contribution of this study was the evaluation of ultrasound parameters across resting, maximal contraction, and Valsalva states rather than within a single functional state. The model integrated five independently associated parameters obtained across the three states into a single combined association model for postpartum PFD. Most previous studies evaluated a single functional state, a single ultrasound modality, or a specific PFD subtype. In contrast, the present study evaluated PFD as a composite outcome. The combined model had an area under the curve of 0.870 in the development cohort; however, this value represented apparent performance in the derivation sample. However, this estimate was obtained from the same dataset used to construct the model and was not confirmed in an independent cohort. Therefore, the findings provide preliminary evidence only and do not support immediate clinical screening or intervention.

Study limitations and future directions
The most important limitation was that the combined model was developed and evaluated using the same single-center retrospective dataset. No training/test split, cross-validation, or external validation was performed. Bootstrap resampling was used to assess the stability of regression coefficients but did not provide independent validation of model discrimination. Consequently, the reported area under the curve, sensitivity, and specificity may be optimistic and may not represent performance in new patient populations. Therefore, the model’s generalizability and clinical applicability remain unproven.

Additional limitations included potential selection bias associated with the single-center retrospective design, the absence of long-term follow-up, and operator dependence in ultrasound measurements despite standardized procedures. Electrophysiological indicators, including pelvic floor nerve-conduction studies and electromyography, were not included. Associations between specific PFD subtypes and ultrasound parameters were also not evaluated. In addition, variable selection was based on univariate screening followed by stepwise logistic regression rather than penalized regression methods such as the least absolute shrinkage and selection operator (LASSO). This approach may increase the risk of overfitting and coefficient instability and does not evaluate potential interactions among predictors. Baseline ultrasound measurements were obtained either before pregnancy or during the first trimester, depending on record availability. Although both were treated as early baseline assessments, variation in the timing of baseline measurement may have introduced heterogeneity.

Levator ani avulsion exhibited quasi-complete separation in the dataset (100% prevalence in the PFD group versus 5% in the non-PFD group). Consequently, maximum likelihood estimation for this variable may have produced unstable coefficient estimates. Although Firth penalized likelihood regression is commonly used in such situations, it was not implemented in this analysis. Therefore, the adjusted odds ratio for avulsion should be interpreted with caution, and the strong avulsion–PFD association is best supported by the descriptive data (Table 5) rather than by the regression-derived point estimate. Future studies with larger samples are needed to provide more stable multivariable-adjusted estimates for this variable.

Future studies should use multicenter prospective cohorts with larger samples and independent validation datasets. A longer follow-up could clarify the association between changes in ultrasound indicators and longer-term outcomes. Additional clinical and pelvic floor functional indicators could also be evaluated. Stratified analyses of specific PFD subtypes may identify subtype-specific ultrasound characteristics and clarify whether model performance differs across clinical presentations.

Conclusion
Ultrasound parameters measured at rest, during maximal pelvic floor muscle contraction, and during the Valsalva maneuver were associated with postpartum PFD. The five-parameter combined association model showed favorable internal fitting performance in the development cohort. However, the study was exploratory and did not establish clinical utility. External validation in independent multicenter prospective cohorts is required before the model can be considered for risk stratification, screening, or individualized intervention.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
G*Power	Heinrich Heine University Düsseldorf	Version 3.1.9.7	Sample size calculation
GE RAB4-8-D transperineal 3D/4D volumetric probe	GE Healthcare	RAB4-8-D	Ultrasound measurement
GE Voluson E10 color Doppler ultrasound system	GE Healthcare	GE Voluson E10	Ultrasound measurement
IBM SPSS Statistics	IBM	Version 27.0.1	Statistical analysis