Work

Innovative electromyographic monitoring: Online nomogram for quantifying pelvic floor muscle strength in a multicenter study

Year: 2025

Type: article

Abstract: To the Editor: Urinary incontinence (UI) is a major global health issue for women, affecting nearly 50% of the population and incurring significant costs.[1] Key risk factors, including childbirth and aging, contribute to UI morbidity rates between 15.9% and 34%. The 2021 National Institute for Health and Care Excellence (NICE) guidelines recommend supervised pelvic floor muscle (PFM) training as the primary conservative treatment for stress and mixed UI, aiming to enhance PFM strength. However, over 30% of patients struggle to voluntarily contract their PFMs during initial consultations, limiting the effectiveness of exercise routines, even when guided by digital therapeutic devices. The Modified Oxford Scale (MOS),[2] the gold standard for assessing PFM strength through palpation, faces challenges in promoting adherence due to a lack of standardized diagnostic tools and limited education for healthcare providers and the public. An ongoing issue is the scarcity of digital devices that match the assessment precision of trained professionals, especially for PFM strength evaluation. In telehealth, an important question is how to accurately diagnose PFM function and develop personalized rehabilitation programs remotely. Electromyography (EMG) has shown potential for home monitoring across rehabilitation settings by measuring electrical signals from muscle fibers, which indicate muscle recruitment and strength; higher EMG signals suggest greater muscle force. Intra-vaginal EMG, validated against MOS scores, correlates strongly with PFM strength. While EMG devices have aided over 6 million women across 2160 health centers nationwide, they face obstacles in home-based rehabilitation for UI due to unclear diagnostic efficacy in PFM strength assessment. This multicenter, prospective cohort study (chictr.org.cn Identifier: ChiCTR1900021717) included 1722 women from Beijing Hospital between January 1, 2019, and February 16, 2023 (Ethical approval No. 2019BJYYEC-014-02) and the First Hospital of Yunnan Province (from January 1, 2019, to December 31, 2021) for model development, and 337 women from Beijing Shunyi Traditional Chinese Medicine Hospital (from September 1, 2022, to February 16, 2023) for external validation. Postpartum women aged 18–50 years, within three months of delivery, and with no history of pelvic rehabilitation therapy were recruited. Exclusion criteria were incomplete information, refusal of physical examination, and cognitive or mental disorders. We collected demographic and obstetric characteristics, assessing PFM function through digital palpation and EMG. Obstetric data included maternal age, gravidity, parity, height, weight, delivery mode, and newborn weights. Body mass index ([BMI], kg/m2) was calculated at three stages: early pregnancy, during labor, and at enrollment. Maximum newborn weight was defined as the heaviest infant for multiparous women. The delivery mode was classified as vaginal, cesarean, or forceps delivery. PFM strength was evaluated using the MOS, which rates muscle contraction from 0 to 5, 0 indicates no contraction[1]; one indicates a minor muscle flicker[2]; two indicates a weak contraction[3]; three indicates a good contraction without resistance[4]; four indicates a strong contraction with less than 5 seconds of resistance; and five, a strong contraction sustained for 5 seconds.[2] Participants were divided into two groups based on MOS scores: Group 1 (MOS ≤2, labeled “Weak”) and Group 2 (MOS ≥3, labeled “Normal”). intra-vaginal EMG signals were collected and analyzed using the MLD B4Plus and MLD V2 Vaginal Electrode (Medlander Medical Technology Inc., Nanjing, China). Abdominal muscle EMG signals were recorded from surface electrodes over the rectus abdominis and external oblique abdominis, ensuring standardized, accurate recordings of both PFM and abdominal muscle activity. Following the standardized Glazer protocol, participants were instructed to rapidly contract, sustain, and relax their PFMs. This protocol involved four key phases[3]: (1) a 60-second pre-baseline rest (PREBR) to measure mean amplitude (µV); (2) five rapid contractions with 10-second rests, measuring the mean peak amplitude of fast contraction (µV) (PFC); (3) five 10-second endurance contractions with 10-second rests, measuring the mean amplitude of endurance contraction (µV) (MEC); and (4) a 60-second post-baseline rest (POSTBR) to capture mean amplitude after contraction. EMG data from PFMs and abdominal muscles were recorded as PFC-PFM, MEC-PFM, PREBR-PFM, POSTBR-PFM for PFM, and PFC-abdominal muscles (PFC-AM), MEC-AM, PREBR-AM, and POSTBR-AM for abdominal muscles. The sample size calculation for the prediction model followed the four-step approach outlined by Richard et al[4] for studies with dichotomous outcomes, aiming to include 20 predictive factors. Based on a 30% prevalence of weak PFM strength (MOS <3), a final sample size of 1079 cases was determined following essential prediction model criteria. Statistical analysis was conducted using R software (version 4.0.0; http://www.R-project.org/). Normally distributed data were expressed as mean ± standard deviation, while non-normally distributed data used median (25th to 75th percentile). Categorical data were presented as counts and percentages. A two-sided P value <0.05 was considered statistically significant. The primary dataset was randomly split (70:30) for model training and testing, with independent external validation. To address potential multicollinearity, LASSO logistic regression was employed with an alpha (α) value of 1. This regularization technique identified relevant indicators for constructing the prediction model. The optimal model was determined through 10-fold cross-validation, selecting the lambda (λ) value that minimized binomial deviation plus standard error, yielding regression coefficients (β). Odds ratios (OR) and 95% confidence intervals (CI) assessed correlation strength. The nomogram’s performance in evaluating PFM strength was measured using the C index, calibration curves, and receiver operating characteristic (ROC) curves to determine predictive accuracy. Internal validation involved constructing a logistic regression model in the testing set, while external validation used the training set model. Decision curve analysis (DCA) quantified the nomogram’s practical application value across different threshold probabilities in the training, testing, and external validation sets. The overall characteristics of participants in cohorts 1 (nomogram development) and 2 (external validation) are summarized as follows [Supplementary Table 1, https://links.lww.com/CM9/C603]: cohort 1 included 1722 parous women (32.3 ± 4.0 years), and cohort 2 included 337 women (33.0 ± 4.1 years). Delivery methods in cohort 1 were 76.3% (1314/1722) vaginal, 20.7% (356/1722) cesarean, and 3.0% (52/1722) forceps, compared to 45.7% (154/337), 42.4% (143/337), and 11.9% (40/337) in cohort 2. Normal PFM strength was seen in 64.9% (1118/1722) of cohort 1 and 67.7% (228/337) of cohort 2, with UI prevalence at 12.49% and 11.88%, respectively. There are 64.9% (1118/1722) and 67.7% (228/337) women exhibiting normal PFM strength in cohort 1 and cohort 2, respectively. A prediction model was developed from cohort 1, randomly splitting into training (70%, n = 1209) and test (30%, n = 513) sets. Among 16 candidate clinical features, five EMG signal indicators with nonzero β were kept in the LASSO regression model derived from the training set. These EMG signals were PREBR-PFM, MEC-PFM, POSTBR-PFM, PFC-AM, and MEC-AM [Supplementary Table 2, https://links.lww.com/CM9/C603]. After visualization, the training set’s logistic regression model was transformed into a nomogram with colored strips [Figure 1A]. Key contributors to assessing PFM strength are depicted in the nomogram. Each EMG parameter was assigned a score on the upper scale, contributing to the total score calculated by adding scores of the five EMG signals. PFM strength could be assessed by locating the total score at the total point line. Two colored strips, green and red, covered the total point and probability lines. The green strip indicated normal PFM strength, while the red one signaled the risk of weak PFM strength. The discriminative threshold for distinguishing between declining and normal strength was determined by ROC analysis in the training set. If the predicted probability of normal PFM strength was below 64.6%, the participant was considered to be at risk of PFM weakness, with a sensitivity of 0.668 and specificity of 0.748. In addition, a dynamic nomogram was developed as a web-based calculator accessible at https://icebear007guifangz.shinyapps.io/PFM_EMG/ [Figure 1B]. The C-indexes of the nomogram were 0.764 for the model training and 0.811 for internal validation. AUCs in the training, testing, and validation sets were 0.764 (95% CI: 0.735–0.792), 0.811 (95% CI: 0.772–0.850) and 0.764 (95% CI: 0.709–0.818), respectively, indicating that the nomogram provided a satisfying ability for the assessment of PFM strength. The calibration curves presented excellent consistency between nomogram predictions and actual observations for the assessment of PFM strength in the training set, testing set, and independent validation set. The DCA curves for the nomogram in the training, testing, and validation set were presented, which showed good clinical applicability of the nomogram in identifying the women at high risk of weak PFM strength.Figure 1: (A) A nomogram of key contributors to assessing PFM strength. Each EMG parameter was assigned a score on the upper scale, contributing to the total score calculated by adding scores of the five EMG signals. PFM strength could be assessed by locating the total score at the total point line. The green strip indicated normal PFM strength, while the red one signaled the risk of weak PFM strength. (B) A dynamic nomogram was developed as a web-based calculator accessible at https://icebear007guifangz.shinyapps.io/PFM_EMG/. EMG: Electromyomraphy; MEC-AM: Abdominal muscles’ mean amplitude of endurance contraction; MEC-PFM: pelvic floor muscles’ mean amplitude of endurance contraction; PFC-AM: Abdominal muscles’ mean peak amplitude of fast contraction; PFM: Pelvic floor muscle; POSTBR-PFM: pelvic floor muscles’ mean amplitude during post-baseline rest; PREBR-PFM: Pelvic floor muscles’ mean amplitude during pre-baseline rest.This study presents an innovative online digital tool for identifying weak PFM activation using intra-vaginal EMG datasets. Our dynamic nomogram addresses the absence of predictive models specifically focused on quantifying PFM strength, demonstrating excellent discrimination and calibration. Unlike existing models that rely on patient demographics or questionnaires, our tool utilizes objective records of PFM activity. A significant finding is the strong prediction of compromised PFM strength based on reduced EMG patterns during endurance contractions, highlighting the importance of improving PFM endurance in training regimens. Pre- and post-resting PFM EMG patterns serve as predictive indicators, indicating that the tension status of the PFM influences muscle contraction strength. This nomogram is pioneering in quantifying the simultaneous effects of abdominal muscle contractions on PFM strength assessment, revealing that unwanted concurrent contractions significantly impact PFM training and EMG results. Our research established a negative correlation between abdominal muscle EMG and PFM strength assessments, reinforcing the nomogram’s independent correlation between EMG readings and PFM strength, unaffected by obstetric factors such as age, delivery mode, parity, or BMI. The established cutoff value enhances the clinical utility of this online diagnostic tool. To expand the model’s applicability, we are conducting clinical validation trials in older populations and optimizing the model through machine learning techniques. In addition, we have secured patent authorization for a home-based electrophysiological monitoring device and are developing a corresponding cloud-based diagnostic platform. Our findings aim to address the challenges associated with hands-on assessments in remote pelvic floor rehabilitation. A primary strength of this study is its large, multi-center sample size and the use of an independent cohort for external validation. The practicality of the model is emphasized by its online dynamic nomogram and clearly defined cutoff values. However, limitations exist: the model may be less applicable to older and nulliparous women since the data predominantly derive from parous women. Although age and delivery modes are not predictors, further validation across diverse age groups is essential. In addition, variations in electrode placement among participants may influence results, although literature suggests minimal impact on EMG evaluations. Lastly, the model does not incorporate alternative PFM strength assessment methods, such as manometry,[2] which may possibly affect the comprehensiveness of the model. In conclusion, this study developed and validated an online dynamic nomogram for quantifying PFM strength using EMG data, demonstrating high accuracy and reliability. The weak amplitude of MEC-PFM was identified as the key predictor, alongside four other EMG parameters. The nomogram shows promise for use in digital therapeutic devices, especially for remote rehabilitation of UI. Funding This study was supported by National Key R&D Program of China (Nos. 2021YFC2701300; 2021YFC2701303) and National High Level Hospital Clinical Research Funding (Nos. BJ‑2018‑204 and BJ‑2023‑112). Conflicts of interest None.

Source: Chinese Medical Journal

DOI: 10.1097/cm9.0000000000003808