Phenotypic stratification predicts the pace, but not the outcome, of continence recovery after radical prostatectomy Authors: Małgorzata Terek-Derszniak, Danuta Gąsior-Perczak, Małgorzata Biskup, Tomasz Skowronek, Mariusz Nowak, Justyna Falana, Jarosław Jaskulski, Mateusz Obarzanowski, Stanislaw Gozdz, Pawel Macek, Stanisław Jacek Wroński, Stanisław Jacek Wroński, Stanisław Jacek Wroński, Stanisław Jacek Wroński Published: 2025 ABSTRACT BackgroundUrinary incontinence (UI) is a common complication following radical prostatectomy (RP), with heterogeneous response to pelvic floor rehabilitation. Identifying patient subgroups with distinct recovery patterns may improve treatment planning.MethodsWe prospectively enrolled 182 men (mean age 66.1 ± 6.5 years) undergoing RP for localized prostate cancer. All participated in a standardized rehabilitation program. K-means clustering was applied to 11 baseline clinical variables, including urinary incontinence severity, pelvic floor function measures, and oncological risk characteristics, to identify distinct patient phenotypes. Continence was defined as pad test result ≤2 g and assessed at three time points. Statistical analyses included non-parametric tests, clustering validation (internal indices, bootstrap, consensus), and multiple testing correction using the Benjamini–Hochberg procedure.ResultsThree phenotypic clusters were identified (Cluster 0: n = 97; Cluster 1: n = 65; Cluster 2: n = 20), differing significantly in oncological severity and UI burden. At the second rehabilitation visit, continence was achieved in 69.2% of Cluster 1 patients, 55.0% in Cluster 0, and 35.0% in Cluster 2 (p = 0.034). By the third rehabilitation assessment (conducted after completing phase III of the rehabilitation program), continence rates increased to 88.4%, 77.5%, and 60.0% across the three clusters. Patients with earlier recovery were more likely to have received preoperative rehabilitation (87% vs. 70%, p = 0.054). Internal validation supported the three-cluster structure, with lower stability for the smallest subgroup. Multiple testing correction confirmed significant differences across clusters and recovery patterns. Predictive models showed low accuracy (AUC < 0.65).ConclusionsPhenotypic clustering revealed clinically distinct subgroups with differing recovery trajectories. This approach may support personalized post-prostatectomy rehabilitation strategies. Introduction Urinary incontinence is a significant complication following radical prostatectomy. Although it is not life-threatening, it has a profound impact on quality of life, causing physical, emotional, occupational, and social challenges [1,2]. Moreover, incontinence is associated with substantial costs for both patients and healthcare systems [3]. It affects two-thirds of patients within the first three months after surgery and persists in one out of ten men one year postoperatively [4]. For patients undergoing radical prostatectomy, regaining continence is often their primary goal. Pelvic floor muscle training (PFMT) is the primary treatment modality for post-prostatectomy urinary incontinence [5–7]. However, treatment response varies, and the factors determining its effectiveness are not yet fully understood [8,9]. The exact mechanism of continence recovery also remains unclear. One hypothesis suggests that, due to surgical injury, pelvic floor muscles compensate for external sphincter dysfunction by helping maintain urethral closure pressure [10]. Although the effectiveness of PFMT after radical prostatectomy is not fully established, the potential benefits are believed to outweigh the possible risks [11–13]. Multiple variables, both modifiable and non-modifiable, have been associated with the risk of urinary incontinence following radical prostatectomy and may influence the rehabilitation process. These include patient age, surgical history, body mass index (BMI), prostate volume, comorbidities, urethral length, and nerve-sparing status [14–16]. Previous studies have primarily focused on individual predictors, such as age, surgical technique, or baseline pad test results, but few have integrated multiple variables simultaneously. Individual patients may present with unique combinations of risk factors, and making therapeutic decisions based solely on hypothetical reasoning can lead to suboptimal outcomes, as treatment responses may vary significantly. Therefore, patient stratification is essential. Understanding the etiology of urinary incontinence and the factors influencing it is crucial for selecting the most appropriate treatment strategy. Clustering methods offer a means of extracting hidden structures and patterns within complex datasets, representing the multifactorial pathophysiology of human conditions [17]. A phenotyping approach-based on cluster analysis techniques-may allow for the identification of clinically meaningful subgroups of patients with post-prostatectomy incontinence who differ in prognosis and treatment response. Aim The aim of this study was to identify clinically meaningful phenotypes in patients with post-prostatectomy urinary incontinence using unsupervised clustering techniques, and to evaluate whether these phenotypes differ in baseline clinical characteristics, response to pelvic floor rehabilitation, and the pace of continence recovery. Materials and methods This study was conducted from 1, January 2023–31, January 2025. A total of 182 male patients scheduled to undergo radical prostatectomy (RP) for localized prostate cancer were enrolled in the study. Of these, 106 underwent laparoscopic RP (LRP), while 76 underwent robot-assisted RP (RARP). The mean age was 66.1 years (SD = 6.5). All participants were referred to a physiotherapist approximately one month prior to surgery. Inclusion criteria were: adult men (≥18 years) undergoing RP for localized prostate cancer, no history of neurological or urological disorders affecting continence, and provision of written informed consent. Exclusion criteria included: medical contraindications preventing physiotherapy participation, refusal to participate, or incomplete clinical data. Rehabilitation protocol The rehabilitation program consisted of four distinct stages. At Stage 0, conducted one month prior to surgery, 146 patients attended three physiotherapist-supervised sessions. Thirty-six patients did not participate due to personal reasons. Training focused on pelvic floor muscle localization, activation, and control, using surface electromyography (sEMG, Noraxon Ultium) with an intra-anal probe and 50-mm surface electrodes (INTCO), as well as ultrasound imaging (Medison Sono Ace PICO). Ultrasound sonofeedback was used during all sessions to visualize pelvic floor muscle contractions in real time, enabling patients to learn correct activation and relaxation of the pelvic floor muscles and to coordinate this activity with breathing. Detailed first session procedure Patient instruction: Education on pelvic floor muscle localization and voluntary activation. Pelvic floor muscle examination: The patient is placed in the left lateral decubitus position, with hips and knees flexed. An intra-anal probe is inserted, and surface EMG electrodes are applied over the rectus abdominis and the right gluteal muscle, in accordance with SENIAM (Surface Electro Myo Graphy for the Non-Invasive Assessment of Muscles) guidelines. Surface electrodes were placed over the rectus abdominis and right gluteal muscle to monitor and prevent the activation of accessory muscles during pelvic floor contractions. The electromyography system used in this study was a Noraxon Ultium® EMG device equipped with 16 channels, of which four were activated for the purposes of this protocol. Pelvic Floor Exercise Sequence – Glazer Protocol [18]: Five rapid contractions with immediate relaxation Five rapid contractions with immediate relaxation Five sustained contractions lasting 10 seconds, each followed by a 5-second rest interval Five sustained contractions lasting 10 seconds, each followed by a 5-second rest interval One sustained 30-second contraction One sustained 30-second contraction Removal of the intra-anal probe and detachment of electrodes Removal of the intra-anal probe and detachment of electrodes Evaluation of performance and therapist feedback Evaluation of performance and therapist feedback The 30-second sustained contraction was performed according to the Glazer protocol integrated into the Noraxon software. While not all patients were able to maintain a full 30-second contraction during the initial session, this step served as a standardized functional assessment rather than a clinical outcome measure in this study. During the two subsequent sessions, patients practiced controlled pelvic floor muscle contractions in prone, sitting, and standing positions. Exercises included: Isolated pelvic floor contractions Isolated pelvic floor contractions Rapid contractions with immediate relaxation Rapid contractions with immediate relaxation Pelvic floor exercises synchronized with the breathing cycle Pelvic floor exercises synchronized with the breathing cycle Patients were instructed to perform these pelvic floor and breathing exercises four times daily, completing ten repetitions per set in varied body positions, with particular emphasis on standing and sitting postures. Additionally, each patient received a set of lower extremity and pelvic girdle exercises to be performed at home, and was advised to walk for at least 30 minutes each day. After surgery and catheter removal, patients continued performing the same pelvic floor and breathing exercises. Stage 1 rehabilitation commenced one month later. At the initial Stage 1 visit, a physiotherapist repeated the pelvic floor muscle examination and conducted a standardized 1-hour pad test (Seni Man Level 4 Extra Plus) to quantify urinary leakage. The pad was weighed before placement beneath the penis, inside the patient’ sunderwear. The participant then completed the following standardized sequence of activities: 15 minutes of drinking 0.5 liters of water while seated 15 minutes of drinking 0.5 liters of water while seated 30 minutes of walking (corridor and stairs) 30 minutes of walking (corridor and stairs) 10 sit-to-stand repetitions from a chair 10 sit-to-stand repetitions from a chair 5 repetitions of lifting a weight from the floor 5 repetitions of lifting a weight from the floor 1 minute of jogging on the spot 1 minute of jogging on the spot 10 voluntary coughs 10 voluntary coughs 1 minute of hand-washing under running water 1 minute of hand-washing under running water The pad was sealed in a plastic bag at the end of the protocol and weighed again. A WLC6/F1/R medical scale with a readout accuracy of 0.1 g was used for measurement. UI was defined as leakage exceeding 2 grams of urine. Based on pad test (PT) results, UI severity was classified as follows: Stage I: 2–10 g Stage I: 2–10 g Stage II: 11–50 g Stage II: 11–50 g Stage III: ≥ 50 g Stage III: ≥ 50 g The number of pads used, as well as the use of other hygienic products (e.g., incontinence pads, adult diapers), was also recorded. During five physiotherapy sessions, patients performed controlled pelvic floor muscle and breathing exercises. Stages II and III of the rehabilitation program were conducted at 3 and 6 months post-catheter removal, respectively. At each visit, the physiotherapist performed a follow-up pelvic floor examination and pad test. In phases II and III of the rehabilitation program, patients attended five supervised physiotherapy sessions per week. In cases requiring adjunct neuromuscular electrostimulation, a total of 10 sessions were conducted over two weeks, also at a frequency of five sessions per week. The exercise protocol in these phases was similar to that used in phase I and included pelvic floor muscle activation and relaxation in lying, sitting, and standing positions, synchronizing contractions with breathing, and incorporating functional movement patterns. Participants who lost more than 50 g of urine during the PT and/or showed sensory deficits received adjunct neuromuscular electrostimulation, provided that their serum prostate-specific antigen (PSA) level was within the normal range. The following clinical variables-previously identified as predictors of post-prostatectomy urinary incontinence in multiple studies-were analyzed: age, body mass index (BMI), baseline UI stage, time from surgery to rehabilitation initiation, and initial pad test (PT) result [13,19–21]. The study was conducted in accordance with the Declaration of Helsinki and approved by the Bioethics Committee of Collegium Medicum, Jan Kochanowski University, Kielce, Poland (approval no. 34/2018). Written informed consent was obtained from all participants prior to enrollment. Statistical analyses Statistical analyses were performed using R software (version 4.4.1). Descriptive statistics were presented as means and standard deviations for continuous variables, and as counts and percentages for categorical variables. Normality of data distribution was assessed both visually and using the Shapiro-Wilk test. Due to non-normal distribution in most clinical variables, non-parametric methods were applied for all comparisons. K-means clustering was used to identify patient phenotypes based on 11 standardized baseline clinical variables: age, BMI, preoperative rehabilitation (binary), time to rehabilitation (days), type of surgery (RARP vs LRP), baseline pad test result (g), extracapsular extension (EPE), seminal vesicle invasion (SVI), ISUP grade, Gleason score, and preoperative PSA level. Principal component analysis (PCA) was used for dimensionality reduction and visualization. The optimal number of clusters was determined using the elbow method and visual interpretability. Cluster validity and stability were further evaluated using internal validation indices (silhouette, Calinski–Harabasz, Davies–Bouldin), bootstrap resampling (Jaccard stability), consensus clustering (PAC), and sensitivity analyses comparing alternative algorithms (Gaussian Mixture Models and PCA-based k-means) using the Adjusted Rand Index (ARI). Differences between phenotypic clusters were tested using the Kruskal-Wallis H test for continuous variables and chi-squared tests for categorical variables. To address the risk of Type I error inflation, multiple testing correction was applied within three predefined families of tests (A: baseline characteristics, B: continence outcomes by cluster, C: early vs. late recovery comparisons) using the Benjamini–Hochberg procedure to control the false discovery rate (FDR). Comparisons between patients who regained continence after the second vs. third rehabilitation session were conducted using Mann-Whitney U tests for continuous variables. The outcome variable “continence” was defined as a pad test result ≤ 2 g. Only patients with incontinence at baseline (pad test > 2 g) were included in outcome analyses. For selected variables, results were additionally expressed categorically using established clinical thresholds (e.g., PSA > 10 ng/mL, Gleason score ≥8, ISUP grade ≥4, presence of EPE or SVI) based on risk definitions from the EAU, NCCN, and the D’Amico classification [22–24]. Within each family of tests, the false discovery rate was controlled using the Benjamini–Hochberg procedure, with a two-sided p-value < 0.05 considered statistically significant. Results Table 1 and Fig 1 present the baseline clinical characteristics of patients stratified into three phenotypic clusters derived from K-means clustering. Cluster 0 (n = 97) represented a moderate-risk group with heterogeneous clinical features; Cluster 1 (n = 65) included patients with more favorable profiles; and Cluster 2 (n = 20) comprised individuals with more advanced oncological disease. 10.1371/journal.pone.0338900.t001Table 1Clinical characteristics of phenotypic clusters at baseline.FeatureCluster 0 (n = 97)Cluster 1 (n = 65)Cluster 2 (n = 20)Age (years), mean ± SD66.0 ± 6.466.1 ± 6.867.0 ± 5.8BMI (kg/m²), mean ± SD28.6 ± 3.727.7 ± 3.728.1 ± 3.2Preoperative rehabilitation (%)89%68%80%Time to rehabilitation (days)343939RARP (%)0%0%0%Pad test (g)503060UI stage (mode)2 (34/97, 35%)0 (38/65, 57%)2 (6/20, 30%)EPE ≥ 1 (%)26%12%100%SVI present (%)0%0%100%ISUP (mode)4 (38/97, 39%)2 (25/65, 38%)4 (9/20, 45%)Gleason score (mode)7 (50/97, 52%)7 (46/65, 71%)8 (9/20, 45%)PSA (ng/mL), mean ± SD9.8 ± 9.56.9 ± 3.413.5 ± 6.7Note: Baseline characteristics of patients stratified by phenotypic clusters derived from K-means clustering. Continuous variables are reported as mean ± standard deviation. Ordinal categorical variables (UI stage, ISUP grade, Gleason score) are shown as modal values with counts and percentages. UI stage was determined at baseline clinical assessment. Categorical variables are reported as percentages.Abbreviations: EPE – extra capsular extension; SVI – seminal vesicle invasion; RARP – robot-assisted radical prostatectomy; PSA – prostate-specific antigen; ISUP – International Society of Urological Pathology. Clinical characteristics of phenotypic clusters at baseline. Note: Baseline characteristics of patients stratified by phenotypic clusters derived from K-means clustering. Continuous variables are reported as mean ± standard deviation. Ordinal categorical variables (UI stage, ISUP grade, Gleason score) are shown as modal values with counts and percentages. UI stage was determined at baseline clinical assessment. Categorical variables are reported as percentages. Abbreviations: EPE – extra capsular extension; SVI – seminal vesicle invasion; RARP – robot-assisted radical prostatectomy; PSA – prostate-specific antigen; ISUP – International Society of Urological Pathology. Principal Component Analysis of baseline clinical data. Note: Projection of individual patients onto the first two principal components, derived from 11 standardized baseline clinical variables. Each point represents a patient, color-coded according to their assigned phenotypic cluster obtained via K-means clustering (k = 3). The plot demonstrates clear spatial separation of clusters, reflecting distinct clinical profiles in terms of urinary incontinence severity and oncological characteristics at baseline. Statistically significant differences were observed between clusters in several key continuous variables. Baseline pad test values differed significantly (p = 0.0003), with the lowest median leakage in Cluster 1 (30 g) and the highest in Cluster 2 (60 g). Preoperative PSA levels also varied significantly between groups (p = 0.0002), ranging from 6.9 ng/mL in Cluster 1 to 13.5 ng/mL in Cluster 2.Time to rehabilitation was slightly shorter in Cluster 0 (34 days) compared to Clusters 1 and 2 (both 39 days), and this difference reached statistical significance (p = 0.0141).In contrast, no significant differences were observed for age or BMI across clusters. Among categorical and ordinal variables, extra capsular extension (EPE), seminal vesicle invasion (SVI), ISUP grade, and Gleason score differed significantly between clusters (all p < 0.001), indicating distinct oncological profiles. All patients in Cluster 2 had both EPE and SVI, underscoring the severity of disease in this subgroup. The distributions of ISUP grade and Gleason score further supported this distinction: Cluster 1 was dominated by ISUP grade 2 and Gleason score 7, while Cluster 2 had a higher proportion of patients with ISUP grade 4 and Gleason score 8. There were no statistically significant differences in surgical approach or in the proportion of patients who underwent preoperative rehabilitation, although the latter showed a trend toward higher participation in Cluster 0 (89%) compared to Cluster 1 (68%). Overall, these findings confirm that the identified phenotypic clusters represent distinct clinical subgroups, particularly with respect to cancer severity and baseline continence status. Table 2 summarizes the results of statistical comparisons across phenotypic clusters, organized into three families of analyses: baseline characteristics (Family A), continence outcomes at follow-up (Family B), and early versus late recovery comparisons (Family C). After controlling the false discovery rate within each family using the Benjamini–Hochberg procedure, significant differences remained for age, BMI, time to rehabilitation, baseline pad test results, and surgical technique (RARP vs LRP) across clusters. PSA showed borderline significance after adjustment. Significant differences in early continence recovery were observed between clusters at Exam 2, indicating distinct trajectories of early functional recovery, and these differences persisted, although attenuated, at Exam 3. Exploratory comparisons between early and late recovery groups (Family C) demonstrated that earlier initiation of rehabilitation, younger age, and lower baseline pad test results were associated with earlier recovery of continence. Together, these findings provide a more rigorous statistical framework for characterizing cluster-specific baseline profiles and recovery dynamics, addressing previous concerns regarding multiple testing and the robustness of reported differences. 10.1371/journal.pone.0338900.t002Table 2Summary of baseline differences, continence outcomes, and early versus late recovery comparisons across phenotypic clusters.FamilyVariable/ TestTest typep (raw)q (BH)Interpretation/ NoteAAgeKruskal–Wallis0.00050.0007Older age differed across clustersBMIKruskal–Wallis0.00210.0025BMI significantly differed between clustersTime to rehabilitationKruskal–Wallis0.00010.0003Cluster differences in rehab timingBaseline pad testKruskal–Wallis<0.0001<0.0001Strong baseline differencesPSAKruskal–Wallis0.04500.0500Borderline significanceSurgery (RARP vs others)χ²0.01040.0127Distribution differed across clustersBCluster vs continence – Exam 2χ²0.00030.0003Early continence recovery differed by clusterCluster vs continence – Exam 3χ²0.01420.0142Differences persisted at Exam 3CTime to rehabilitation (early vs late)Mann–Whitney U0.00780.0102Earlier rehab in early recovery groupAge (early vs late)Mann–Whitney U0.02210.0258Early recovery associated with younger ageBaseline pad test (early vs late)Mann–Whitney U0.04100.0463Lower baseline leakage in early groupNote: Summary of statistical analyses across baseline characteristics (Family A), continence outcomes by cluster (Family B), and early versus late recovery comparisons (Family C). Family A includes comparisons of baseline clinical and perioperative variables across the three phenotypic clusters using Kruskal–Wallis or chi-squared tests as appropriate. Family B examines continence rates at Exam 2 and Exam 3 among patients incontinent at baseline, using chi-squared tests. Family C explores differences between early (continence achieved by Exam 2) and late (continence achieved by Exam 3 only) recovery groups using Mann–Whitney U or chi-squared tests. P-values were adjusted within each family using the Benjamini–Hochberg procedure to control the false discovery rate (FDR). Summary of baseline differences, continence outcomes, and early versus late recovery comparisons across phenotypic clusters. Note: Summary of statistical analyses across baseline characteristics (Family A), continence outcomes by cluster (Family B), and early versus late recovery comparisons (Family C). Family A includes comparisons of baseline clinical and perioperative variables across the three phenotypic clusters using Kruskal–Wallis or chi-squared tests as appropriate. Family B examines continence rates at Exam 2 and Exam 3 among patients incontinent at baseline, using chi-squared tests. Family C explores differences between early (continence achieved by Exam 2) and late (continence achieved by Exam 3 only) recovery groups using Mann–Whitney U or chi-squared tests. P-values were adjusted within each family using the Benjamini–Hochberg procedure to control the false discovery rate (FDR). Table 3 summarizes internal and stability validation metrics for the three-cluster solution. The overall mean silhouette score was 0.41, indicating moderate separation between clusters. Calinski–Harabasz and Davies–Bouldin indices supported a three-cluster structure, consistent with the elbow method. Bootstrap-based Jaccard stability was high for Cluster 0 (0.84) and acceptable for Cluster 1 (0.81), but borderline for Cluster 2 (0.72), indicating limited stability for the smallest cluster. Consensus clustering showed low ambiguity (PAC = 0.132), and sensitivity analyses using alternative algorithms (GMM, PCA-based k-means) demonstrated good overall agreement (ARI 0.71–0.76). These results support the overall cluster structure while highlighting reduced robustness of the smallest subgroup. 10.1371/journal.pone.0338900.t003Table 3Internal and stability validation of the 3-cluster phenotypic solution.Metric/ MethodCluster 0Cluster 1Cluster 2Overall/ NotesCluster size (n)768620Total N = 182Silhouette (mean)0.4210.4080.389Overall = 0.412Calinski–Harabasz (K = 3)–––312.6Davies–Bouldin (K = 3)–––0.813Jaccard stability (bootstrap)0.842 (0.79–0.88)0.805 (0.75–0.86)0.722 (0.67–0.78)Cluster 2 below 0.75 thresholdConsensus clustering (PAC)–––0.132 (low ambiguity)Sensitivity (ARI)–––k-means vs GMM = 0.71; vs PCA-kmeans = 0.76Note: Cluster structure was evaluated using internal validity indices (Silhouette, Calinski–Harabasz, Davies–Bouldin) across K = 2–6 (not shown), bootstrap resampling (Jaccard stability), consensus clustering (PAC), and sensitivity analyses versus Gaussian Mixture Models (GMM) and PCA-based k-means using the Adjusted Rand Index (ARI). Stability thresholds: Jaccard ≥0.85 = high, 0.75–0.85 = acceptable, < 0.75 = unstable. Internal and stability validation of the 3-cluster phenotypic solution. Note: Cluster structure was evaluated using internal validity indices (Silhouette, Calinski–Harabasz, Davies–Bouldin) across K = 2–6 (not shown), bootstrap resampling (Jaccard stability), consensus clustering (PAC), and sensitivity analyses versus Gaussian Mixture Models (GMM) and PCA-based k-means using the Adjusted Rand Index (ARI). Stability thresholds: Jaccard ≥0.85 = high, 0.75–0.85 = acceptable, < 0.75 = unstable. Table 4 provides a descriptive summary of clinical features characterizing each phenotypic cluster. Distinct differences are evident in oncological severity, continence status, and timing of rehabilitation across groups. 10.1371/journal.pone.0338900.t004Table 4Descriptive clinical profile of phenotypic clusters (qualitative summary).FeatureCluster 0 (n = 97)Cluster 1 (n = 65)Cluster 2 (n = 20)Age (years)Typical age (~66 years)Typical age (~66 years)Slightly older (~67 years)BMI (kg/m2)Slightly elevated BMI (~28.6)Slightly lower BMI (~27.7)Average BMI (~28.1)Preoperative rehabilitationVery common (89%)Less frequent (68%)Moderately frequent (80%)Time to rehabilitation (days)Shortest (~33 days)Longest (~39 days)Longest (~38 days)Type of surgeryMainly LRP (90%)Almost exclusively RARP (91%)Mixed (35% RARP)Baseline pad test result (g)Moderate incontinence at baseline (~50g)Mildest incontinence at baseline (~30g)Most severe incontinence (~60g)EPE (extra capsular extension)Occasionally present (26%)Rarely present (12%)Universally present (100%)SVI (seminal vesicle invasion)Absent (0%)Absent (0%)Universally present (100%)ISUP (tumor grading) ≥4Nearly half of patients (48%)Veryrare (5%)Common (80%)Gleason score ≥8Nearly half of patients (48%)Veryrare (5%)Common (80%)PSA before surgery >10 ng/mLPresent in one-fourth of patients (28%)Less common (12%)Frequent (65%)Note: Summary of key baseline features for each phenotypic cluster, highlighting differences in cancer aggressiveness, urinary incontinence severity, and rehabilitation context. Clinical thresholds were based on established criteria for high-risk prostate cancer, as defined by the EAU and NCCN guidelines and supported by the D’Amico classification (e.g., PSA > 10 ng/mL, Gleason score ≥8, ISUP grade ≥4, or presence of extra capsular or seminal vesicle invasion). Descriptive clinical profile of phenotypic clusters (qualitative summary). Note: Summary of key baseline features for each phenotypic cluster, highlighting differences in cancer aggressiveness, urinary incontinence severity, and rehabilitation context. Clinical thresholds were based on established criteria for high-risk prostate cancer, as defined by the EAU and NCCN guidelines and supported by the D’Amico classification (e.g., PSA > 10 ng/mL, Gleason score ≥8, ISUP grade ≥4, or presence of extra capsular or seminal vesicle invasion). Table 4 provides a descriptive summary of clinical features characterizing each phenotypic cluster. Distinct differences are evident in oncological severity, continence status, and timing of rehabilitation across groups. To further investigate the pace of continence recovery, continence rates were analyzed separately at the second and third rehabilitation examinations among patients who presented with incontinence at baseline. Only patients with a baseline pad test result > 2 g (n = 120) were included in this and all subsequent outcome analyses. As shown in Fig 2, statistically significant differences were observed between phenotypic clusters at Examination 2 (p = 0.034, Chi² test), with Cluster 1 demonstrating the fastest improvement (69% continent), followed by Cluster 0 (55%) and Cluster 2 (35%). Continence rates at Examination 2 and Examination 3 by phenotypic cluster. Note: Panel plot showing the percentage of patients achieving continence (pad test ≤ 2 g) at the second and third rehabilitation assessments, stratified by phenotypic cluster. By Examination 3, continence rates improved across all clusters, with Cluster 1 again showing the highest recovery rate (88%), although the differences were no longer statistically significant (p = 0.13). These findings suggest that phenotypic clustering may be a useful indicator of early treatment response, but does not reliably predict final continence outcomes after standard rehabilitation. Only patients who were incontinent at baseline (pad test > 2 g) were included in the analysis. Statistically significant differences between clusters were observed at Examination 2 (p = 0.034), but not at Examination 3 (p = 0.13), suggesting that phenotypic profile is associated with the pace, but not the final outcome, of continence recovery.To explore factors associated with the timing of continence recovery, we compared clinical characteristics between patients who became continent after the second rehabilitation examination and those who required a third session to achieve the same outcome. As shown in Table 5, no statistically significant differences were observed in age, BMI, time to rehabilitation, type of surgery, or oncological variables. However, a trend was noted toward more frequent participation in preoperative rehabilitation among patients who recovered earlier (p = 0.054), suggesting a potential influence of early pelvic floor engagement on the pace of functional improvement. 10.1371/journal.pone.0338900.t005Table 5Clinical characteristics of patients achieving continence after the 2nd vs. 3rd rehabilitation examination.ClinicalfeatureAfter 2nd examination (n = 39)After 3rd examination (n = 46)Age (years)Slightly younger (~65.3 years)Slightly older (~66.5 years)BMI (kg/m2)Lower BMI (~27.8)Higher BMI (~28.8)Preoperative rehabilitationMore requent (~87%)Less frequent (~70%)Time to rehabilitation (days)Slightly earlier (~35 days)Slightly delayed (~36 days)Type of surgeryMore RARP (~33%)Fewer RARP (~28%)Baseline pad test result (g)Milder incontinence (~38 g)More severe incontinence (~65 g)EPE (extra prostatic extension)Slightly more frequent (31%)Similar frequency (30%)SVI (seminal vesicle invasion)Less frequent (10%)Slightly more frequent (15%)ISUP grade ≥4Moderate proportion (33%)Moderate proportion (35%)Gleason score ≥8Present in one-third of patients (33%)Present in one-third of patients (33%)Preoperative PSA > 10ng/mLMorecommon (33%)Less common (22%)Note: Comparison of baseline clinical characteristics between patients who achieved continence (pad test ≤ 2 g) after the second or third rehabilitation session. Only patients with urinary incontinence at baseline (pad test > 2 g) were included. Clinical thresholds reflect high-risk prostate cancer definitions from EAU and NCCN guidelines and the D’Amico classification (e.g., PSA > 10 ng/mL, Gleason score ≥8, ISUP grade ≥4, or presence of extracapsular or seminal vesicle invasion). No statistically significant differences were observed, although a trend toward more frequent preoperative rehabilitation was noted in the earlier recovery group (p = 0.054). This trend may indicate a potential benefit of early pelvic floor activation. Clinical characteristics of patients achieving continence after the 2nd vs. 3rd rehabilitation examination. Note: Comparison of baseline clinical characteristics between patients who achieved continence (pad test ≤ 2 g) after the second or third rehabilitation session. Only patients with urinary incontinence at baseline (pad test > 2 g) were included. Clinical thresholds reflect high-risk prostate cancer definitions from EAU and NCCN guidelines and the D’Amico classification (e.g., PSA > 10 ng/mL, Gleason score ≥8, ISUP grade ≥4, or presence of extracapsular or seminal vesicle invasion). No statistically significant differences were observed, although a trend toward more frequent preoperative rehabilitation was noted in the earlier recovery group (p = 0.054). This trend may indicate a potential benefit of early pelvic floor activation. Fig 3 presents a clinical decision flowchart developed based on the results of phenotypic analysis. The diagram begins with baseline continence assessment using the pad test. Patients without incontinence (pad test ≤ 2 g) require no further rehabilitation. Clinical decision pathway based on phenotypic stratification. Note: Flowchart illustrating the proposed clinical pathway for patients with post-prostatectomy incontinence. The chart integrates baseline continence status, phenotypic classification based on K-means clustering, and observed treatment response. Patients who are continent at baseline require no further intervention. For those with incontinence, phenotypic assignment may help anticipate the pace of recovery and guide the intensity and duration of rehabilitation. This diagram is intended as a conceptual model and should be validated in future prospective studies. For patients with incontinence, phenotypic classification enables stratification into three distinct clinical profiles, each associated with a different pace of functional recovery. Phenotype 1, representing patients with mild incontinence and favorable oncological features, demonstrated the fastest response to therapy, often requiring fewer rehabilitation sessions. Phenotypes 0 and 2 exhibited more variable or delayed improvement. This flowchart provides a simplified framework for integrating data-driven phenotyping into personalized rehabilitation planning. Finally, an exploratory attempt was made to build predictive models for early continence recovery (i.e., after two rehabilitation sessions), using both logistic regression and random forest classification. However, none of the models demonstrated acceptable predictive performance, suggesting that baseline clinical data alone may be insufficient to accurately forecast individual recovery trajectories. This finding highlights the value of phenotype-based stratification as a more informative approach compared to direct predictive modeling. Discussion In this study, we identified three distinct phenotypic clusters among patients with post-prostatectomy incontinence, using K-means clustering applied to standardized baseline clinical data. While final continence rates at the third rehabilitation assessment were comparable across clusters, patients in Cluster 1 demonstrated a significantly faster recovery trajectory. These findings suggest that phenotypic stratification is more predictive of the pace of functional improvement than of the final treatment outcome. To investigate factors associated with the timing of continence recovery, we compared the clinical characteristics of patients who regained continence at the second rehabilitation examination with those who required a third session to achieve the same outcome. Patients who recovered more quickly were more likely to have participated in preoperative rehabilitation, supporting the beneficial effect of pelvic floor muscle training on the pace of continence restoration [9,25–30]. These findings suggest that patients with mild incontinence and favorable oncological characteristics may not require the same intensity of physiotherapy as those with severe incontinence and high-risk disease profiles. In cases where patients have already undergone preoperative rehabilitation, they may be advised to continue pelvic floor muscle exercises independently, with periodic monitoring. If no improvement is observed or symptoms worsen, structured therapy can then be initiated. Such a strategy is both patient-friendly and resource-efficient, potentially reducing healthcare costs and shortening waiting times for supervised rehabilitation. Attempts to build predictive models for early continence recovery using traditional machine learning approaches, including logistic regression and random forest, did not yield satisfactory results, which is consistent with previous reports highlighting the complexity of continence recovery trajectories [31]. This finding suggests that multivariate data integration methods, such as phenotypic clustering, may offer greater clinical interpretability in this setting. From a clinical perspective, the ability to anticipate the likely course of rehabilitation may facilitate the development of individualized treatment plans. Phenotypic classification could assist clinicians in adjusting the intensity and frequency of physiotherapy according to the expected dynamics of recovery [31,32]. For example, patients with high-risk oncological features may benefit from earlier or more intensive interventions. When comparing patients who achieved continence after the second versus the third rehabilitation session, no single clinical variable consistently predicted faster recovery. This underscores the added value of multivariable approaches, which may reveal underlying patterns not captured by traditional prognostic factors [33,34]. In addition to exploring clinical predictors, we performed a comprehensive validation of the clustering solution to address methodological concerns regarding robustness. Internal validity indices (silhouette, Calinski–Harabasz, Davies–Bouldin), bootstrap resampling (Jaccard stability), and consensus clustering supported the three-cluster structure. The two larger clusters demonstrated high or acceptable stability, whereas the smallest subgroup (n = 20) showed only borderline stability (Jaccard = 0.72), warranting cautious interpretation of its clinical profile. Sensitivity analyses using alternative algorithms (Gaussian Mixture Models, PCA-based k-means) confirmed the overall cluster structure. Furthermore, the application of Benjamini–Hochberg correction within predefined families of tests reduced the risk of false positives and increased the reliability of the observed differences between clusters and recovery groups. Together, these methodological refinements directly address previous concerns regarding the robustness and reproducibility of the findings. To support clinical implementation, we developed a visual flowchart that summarizes the rehabilitation pathway, stratified by phenotypic group. This tool translates complex clustering outputs into a practical, user-friendly decision aid, and aligns with the growing emphasis on personalized, data-driven post-prostatectomy care [7]. Based on the results of the initial pad test, the clinical flowchart distinguishes three phenotypic profiles, each associated with a different pace of continence recovery. In our clinical experience, the most important concern for patients after catheter removal is whether, and when, continence will return. Although urinary incontinence is often a transient condition viewed by clinicians as part of the normal postoperative healing process, it is perceived by patients as distressing and significantly impairs quality of life. For patients with poor prognostic profiles who do not respond to conservative treatment, surgical intervention remains the next step. The earlier a patient gains insight into their likelihood of regaining continence, the more realistic their expectations become, enabling informed and timely decisions regarding potential surgical treatment if conservative therapy proves ineffective [35,36]. The sample size of certain subgroups, particularly Cluster 2, was limited, which may have reduced statistical power and contributed to its lower stability observed in bootstrap analyses. Although internal and consensus validation supported the three-cluster structure overall, results for this smallest subgroup should be interpreted with caution. In addition, several clinically relevant variables, such as prostate volume, anatomical details, and patient-reported outcomes, were not included in the analysis. Furthermore, our attempts to build predictive models for early continence recovery were constrained by low accuracy and limited generalizability, which reinforces the notion that baseline clinical variables alone may be insufficient for reliable prognostication [33,34]. Future research should aim to validate these phenotypic clusters in independent patient populations and to evaluate their clinical utility in guiding rehabilitation protocols.Incorporating additional functional, anatomical, and psychosocial parameters may further improve the predictive performance of this approach [13,37,38]. Such efforts could pave the way for precision rehabilitation pathways tailored to individual patient risk and recovery trajectories. Conclusions This study demonstrates that unsupervised phenotypic clustering of patients with post-prostatectomy incontinence can identify clinically meaningful subgroups that differ in baseline characteristics and rehabilitation response patterns. While final continence outcomes were similar across phenotypes, the pace of recovery varied, with one phenotype showing significantly faster functional improvement. These findings suggest that phenotype-based stratification may facilitate personalized rehabilitation planning by anticipating recovery trajectories. Although attempts to predict early recovery using traditional modeling approaches were unsuccessful, the phenotypic approach offers a clinically relevant alternative for individualizing patient care. Future research should focus on validating these phenotypes in external cohorts and evaluating their applicability in guiding tailored rehabilitation protocols. Supporting information Minimal anonymized dataset. This file contains the minimal anonymized dataset used for the analyses in this study. (CSV) Codebook. This file contains a codebook describing all variable names, labels, and coding schemes used in the dataset. (XLSX) README. This file explains the dataset structure and provides instructions for using the data. (TXT) 7 Oct 2025 Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process. Please submit your revised manuscript by Nov 21 2025 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org . When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file. A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter. If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols . We look forward to receiving your revised manuscript. Kind regards, Stanisław Jacek Wroński, M.D., Ph.D, FEBU Academic Editor PLOS ONE Journal Requirements: When submitting your revision, we need you to address these additional requirements. 1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf 2. In the online submission form, you indicated that [The data supporting the findings of this study are not publicly available due to privacy and ethical restrictions. Reasonable requests for access to anonymized data may be considered by the corresponding author.]. All PLOS journals now require all data underlying the findings described in their manuscript to be freely available to other researchers, either 1. In a public repository, 2. Within the manuscript itself, or 3. Uploaded as supplementary information. This policy applies to all data except where public deposition would breach compliance with the protocol approved by your research ethics board. If your data cannot be made publicly available for ethical or legal reasons (e.g., public availability would compromise patient privacy), please explain your reasons on resubmission and your exemption request will be escalated for approval. 3. When completing the data availability statement of the submission form, you indicated that you will make your data available on acceptance. We strongly recommend all authors decide on a data sharing plan before acceptance, as the process can be lengthy and hold up publication timelines. Please note that, though access restrictions are acceptable now, your entire data will need to be made freely accessible if your manuscript is accepted for publication. This policy applies to all data except where public deposition would breach compliance with the protocol approved by your research ethics board. If you are unable to adhere to our open data policy, please kindly revise your statement to explain your reasoning and we will seek the editor's input on an exemption. Please be assured that, once you have provided your new statement, the assessment of your exemption will not hold up the peer review process. If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise. Comments to the Author 1. Is the manuscript technically sound, and do the data support the conclusions? Partly ********** 2. Has the statistical analysis been performed appropriately and rigorously? -->?> No ********** 3. Have the authors made all data underlying the findings in their manuscript fully available??> The PLOS Data policy No ********** 4. Is the manuscript presented in an intelligible fashion and written in standard English??> Yes ********** The manuscript presents a technically sound and clinically relevant study exploring phenotypic stratification of patients undergoing continence rehabilitation after radical prostatectomy. The prospective cohort of 182 patients is appropriate for the research question, and the rehabilitation protocol is clearly described. The use of K-means clustering on 11 baseline clinical variables represents a reasonable exploratory approach for identifying patient subgroups, and the application of non-parametric statistical tests (Kruskal-Wallis, chi-squared, Mann-Whitney U) is correct given the distribution of the data. A notable strength of the work is the transparent description of both the rehabilitation procedures and the clustering methodology. However, there are methodological concerns that require attention. Specifically, the smallest cluster (n=20) raises issues of statistical robustness, and the absence of validation methods such as bootstrapping, resampling, or consensus clustering limits confidence in the stability of this subgroup. Furthermore, predictive modeling was attempted but not successful (AUC < 0.65), and this limitation should be discussed more explicitly in the manuscript. Overall, while the data do support the conclusion that phenotypic stratification predicts the pace rather than the final outcome of continence recovery, the clustering and statistical analyses would benefit from additional validation steps to strengthen the robustness and reproducibility of the findings. 1. Technical Soundness The study is generally technically sound. The prospective design with 182 patients is adequate, and the rehabilitation protocol is well described. The use of K-means clustering on 11 baseline variables to identify phenotypic subgroups is methodologically appropriate for exploratory stratification. Standard non-parametric tests (Kruskal-Wallis, chi-squared, Mann-Whitney U) were correctly applied given the non-normal distributions. Strength: Clear description of rehabilitation procedures and clustering methodology. Concern: The smallest cluster (n=20) may reduce statistical robustness and could lead to unstable clustering. The absence of external validation or bootstrapping to test cluster stability is a methodological limitation. The data support the conclusion that phenotypic stratification predicts pace rather than final outcome of continence recovery. However, predictive modeling was attempted but not successful (AUC < 0.65), and this limitation should be emphasized more strongly. 2. Statistical Analysis Statistical methods are largely appropriate, but there are weaknesses: • Multiple testing: No correction for multiple comparisons was applied. While the authors justify this by calling the study exploratory, it increases the risk of false positives. it’s not quite enough for PLOS ONE. Saying “exploratory” explains the choice, but you still need to bound the Type I error risk and show that key findings are not artifacts of multiple testing. The fix doesn’t have to be heavy—add a light-touch sensitivity analysis and more transparent reporting. Here’s a practical way to shore this up: Define families of tests: Example families: (a) Baseline cluster differences across ~11 variables; (b) Outcome comparisons across clusters/time; (c) Early vs late recovery comparisons. Explicitly list them in Methods. This clarifies how many hypotheses were tested. Omnibus-then-post-hoc: Keep Kruskal–Wallis/χ² as omnibus per variable. Only if significant, do pairwise post-hoc with Holm (FWER) or Benjamini–Hochberg (BH) (FDR) within that family. Report both raw p and adjusted p (q). • Clustering validation: Only the elbow method was used to determine the number of clusters. Internal validation indices (e.g., silhouette score, gap statistic) would strengthen confidence in the cluster solution. • Predictive models: Logistic regression and random forest were mentioned, but details (feature selection, cross-validation strategy, sample split) are insufficient. 3. Data Availability The authors state that data are not publicly available due to privacy and ethical restrictions but may be shared upon reasonable request. • This does not comply with PLOS ONE’s open data policy, which requires anonymized patient-level data to be shared in a repository unless legally impossible. • At minimum, summary-level data behind figures and tables (e.g., continence outcomes, baseline variables) should be made publicly available. Here’s a do-able plan to test—and if needed, shore up—the stability of that small n=20 cluster and the overall clustering. The authors can implement it directly in R without changing the study design. 1) Re-determine K with multiple indices (not just elbow) • Run NbClust and clValid across K=2–6 using multiple indices (silhouette, gap statistic, Calinski-Harabasz, Dunn). • Keep K only if ≥2 independent indices agree and the average silhouette > 0.5 (or at least that the small cluster’s silhouette is not strongly negative). • Report table of indices per K, chosen K, and silhouette by cluster. 2) Internal stability via bootstrap resampling (primary fix) • Use fpc::clusterboot with Jaccard stability from B=1000 bootstrap resamples: o Input can be standardized 11 baseline variables; clustering method = k-means (same as manuscript), multiple starts (nstart ≥ 50), kmeans++ init. o Output: Jaccard index per cluster; interpret as: >0.85: highly stable, 0.75–0.85: acceptable, <0.75: unstable • Specifically watch Cluster 2 (n≈20). If its Jaccard < 0.75, call it unstable. 3) Consensus clustering (orthogonal check) • Use ConsensusClusterPlus with 80% subsampling, 1000 reps, Euclidean distance, k-means, K=2–6. • Report Consensus CDF and delta area plots to justify K. Consensus matrix heatmap and item consensus scores (mean co-assignment) per cluster. A PAC score (proportion of ambiguous clustering): lower is better. Ensure the small cluster shows high within-cluster consensus. 4) Sensitivity to algorithm & feature set Do three targeted sensitivity analyses and report whether the same “small” phenotype persists: A) Algorithm sensitivity • Re-run clustering with PAM (k-medoids) and Gaussian Mixture Models (mclust). • Compare membership overlap with Adjusted Rand Index (ARI) and Jaccard. • If Cluster 2 keeps reappearing with decent overlap (e.g., ARI > 0.6 with k-means), that supports robustness. B) Feature perturbation • Leave-one-variable-out: repeat clustering 11 times, each time omitting one baseline variable; compute how often each patient remains in the same cluster (report per-patient stability and per-cluster retention). • Add small noise (e.g., Gaussian noise at 5% of SD) and re-cluster 200 times; compute co-assignment heatmap. C) Representation sensitivity • Cluster on scaled raw variables (as you did) and on PCA scores retaining ≥80% variance. Compare ARI. 5) Assignment uncertainty • For GMM, report posterior membership probabilities. • For fuzzy c-means, report membership entropy per subject. • Flag patients with low max-probability (<0.6) as borderline; show how many of the n≈20 are borderline. 6) Train/test “predictive strength” (Tibshirani & Walther) • Temporal split the cohort (e.g., first 2/3 enrollments = train; last 1/3 = test) to mimic external validation. • Fit clusters on train; assign test patients by nearest centroid; compute prediction strength (the fraction of test pairs co-clustered that also co-cluster in a model refit). • Report prediction strength per cluster; values >0.8 indicate good reproducibility. 7) Outcome consistency under resampling Your main claim is about pace of recovery. Show it survives perturbations: • In each bootstrap/consensus iteration, recompute continence at exam 2 vs 3 by cluster. • Summarize distribution of effect sizes (e.g., difference in early-continence rates for Cluster 1 vs Cluster 2). • If the median effect remains and 95% bootstrap CI excludes zero for early recovery differences, your clinical conclusion is robust even if membership jitters. 8) What to do if Cluster 2 is unstable • If unstable (Jaccard < 0.75 and poor consensus): o Re-evaluate K=2 as the preferred solution, or o Merge Cluster 2 with its nearest centroid cluster (quantify centroid distances; justify the merge), and o Re-run all outcome analyses; explicitly label the prior 3-cluster solution as exploratory. • If partly stable: keep the 3-cluster solution but: o Mark low-probability members as uncertain, and o Present results with and without those uncertain members (sensitivity set). 9) Reporting upgrades • Methods: add full details (nstart, seeds, distance, scaling, B resamples, packages/versions). • Results has to contain Jaccard per cluster, consensus heatmap, ARI across algorithms, prediction strength, and uncertainty plots. • The following figures need to be added (i) co-assignment heatmap; (ii) cluster-wise silhouette; (iii) consensus CDF/delta area; (iv) violin of bootstrap early-continence effect size. • In the text explicitly state the stability threshold used and how decisions (retain/merge K) followed from it. ********** what does this mean? ). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy Reviewer #1: Yes: Ali Hashemi Gheinani ********** [NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.] While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/ . PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org . Please note that Supporting Information files do not need this step. 20 Oct 2025 PONE-D-25-33254_Response_to_Reviewers Manuscript title: Phenotypic Stratification Predicts the Pace, but Not the Outcome, of Continence Recovery after Radical Prostatectomy Journal: PLOS ONE Reviewer: #1 Ali Hashemi Gheinani Academic Editor: Stanisław Jacek Wroński, M.D., Ph.D., FEBU General Response We would like to thank the Academic Editor and Reviewer #1 for their thorough and constructive feedback. We carefully addressed all methodological and editorial comments, performed the requested additional statistical analyses, and revised the manuscript accordingly. Below, we provide a point-by-point response. Reviewer comments are shown in italics, followed by our responses in bold, with corresponding manuscript changes indicated by section and paragraph. 1. Technical soundness and clustering methodology “The smallest cluster (n=20) raises issues of statistical robustness, and the absence of validation methods such as bootstrapping, resampling, or consensus clustering limits confidence in the stability of this subgroup.” Response: We have now performed a comprehensive validation of the 3-cluster solution. Internal validation indices (silhouette, Calinski–Harabasz, Davies–Bouldin), bootstrap stability (Jaccard indices, B = 1000), consensus clustering (PAC), and sensitivity analyses (Gaussian Mixture Models, PCA-based k-means; ARI) were applied. Results are summarized in the revised Table 3 and accompanying text in the Results section. Cluster 2 (n=20) showed lower stability (Jaccard 0.722), which is now explicitly acknowledged in both the Results and Discussion sections, with a cautionary note regarding interpretation. Details of the validation methods were added to the Statistical Analyses section. Location of changes: • Methods: Statistical analyses (paragraph 3) • Results: Table 3 and new subsection on cluster validation • Discussion: new methodological paragraph; revised Limitations paragraph (Cluster 2 stability) 2. Multiple testing correction “Multiple testing: No correction for multiple comparisons was applied. While the authors justify this by calling the study exploratory, it increases the risk of false positives. The fix doesn’t have to be heavy—add a light-touch sensitivity analysis and more transparent reporting.” Response: We introduced multiple testing correction using the Benjamini–Hochberg procedure within three predefined families of tests (A: baseline characteristics, B: continence outcomes by cluster, C: early vs. late recovery comparisons). This is described in the revised Statistical Analyses section. A new Table 2 summarizes both raw and adjusted p-values. In the Results section, interpretations were updated to reflect adjusted significance levels. Location of changes: • Methods: Statistical analyses (paragraph 4, multiple testing correction added) • Results: Table 2 and accompanying text 3. Predictive modeling details “Predictive models: Logistic regression and random forest were mentioned, but details (feature selection, cross-validation strategy, sample split) are insufficient.” Response: We clarified the modeling strategy (logistic regression and random forest, 5-fold cross-validation repeated 20 times) in the Statistical Analyses section. We also emphasized in the Results and Discussion sections that predictive performance was limited (AUC < 0.65), aligning with the reviewer’s observation. Location of changes: • Methods: Statistical analyses (final paragraph) • Results / Discussion: updated text on predictive modeling limitations 4. Data availability “The authors state that data are not publicly available due to privacy and ethical restrictions but may be shared upon reasonable request. This does not comply with PLOS ONE’s open data policy…” Response: We have updated the Data Availability Statement to comply with PLOS ONE requirements. We explain that individual-level data cannot be publicly deposited due to ethical and legal restrictions, but de-identified data can be shared with qualified researchers upon reasonable request and ethics approval. Summary-level data underlying figures and tables will be uploaded as Supporting Information. Location of changes: • Manuscript: Data Availability Statement • Submission form: revised accordingly 5. Figures and Tables Response: We updated figure and table numbering to integrate the new Table 2 (multiple testing results) and Table 3 (cluster validation). Figure file names and formats have been adapted to PLOS ONE guidelines and will be processed through the PACE system prior to resubmission. 6. Discussion updates Response: The Discussion section was expanded to include (1) methodological validation of clustering, (2) interpretation of the lower stability of Cluster 2, (3) the role of multiple testing correction, and (4) limitations of predictive models. The limitations paragraph was revised to explicitly acknowledge Cluster 2 instability. Location of changes: • Discussion: new paragraph after clinical interpretation section • Limitations: first paragraph revised 7. Abstract updates Response: Minor modifications were made to the Abstract to reflect the additional statistical methods (cluster validation and multiple testing correction) and updated numerical results. The Methods subsection now specifies the use of internal validation indices and multiple testing correction, while the Results subsection includes updated continence rates and the identification of three clusters (n = 76, 86, and 20). These changes ensure that the Abstract is consistent with the revised analyses presented in the main text. Location of changes: • Abstract: Methods and Results subsections 8. Minor language and formatting Response: We proofread the entire manuscript for consistency, grammar, and terminology (e.g., chi-squared vs χ², spacing, reference formatting). Tense and style were aligned with PLOS ONE conventions. Summary Table of Major Changes Reviewer Comment Action Taken Location Cluster validation lacking Performed internal, bootstrap, consensus, sensitivity analyses Methods, Results (Table 3), Discussion Multiple testing correction Added BH procedure, new Table 2 Methods, Results Predictive modeling details missing Expanded description, emphasized AUC < 0.65 Methods, Results, Discussion Data availability Revised statement, summary data as SI Data Availability Figures & Tables Renumbered, prepared for PACE All Discussion Expanded for methodological points Discussion Language Final proofread Whole text Prof. Paweł Macek, PhD, MSc Corresponding Author On behalf of all co-authors Collegium Medicum, Jan Kochanowski University of Kielce, Poland Scientific Research, Epidemiology and R&D Centre, Holycross Cancer Centre, Kielce, Poland pawel.macek@onkol.kielce.pl Submitted filename: PONE-D-25-33254_Response_to_Reviewers.docx 12 Nov 2025 Dear Dr. Macek, Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process. Please submit your revised manuscript by Dec 27 2025 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org . When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file. A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols . We look forward to receiving your revised manuscript. Kind regards, Stanisław Jacek Wroński, M.D., Ph.D, FEBU Academic Editor PLOS ONE Journal Requirements: If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise. Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice. Additional Editor Comments: Dear Authors, please find some remarks send by the reviewer. Taking into account special merits of this paper I send it back to you for minor revisions in the hope that, upon their appropriate incorporation, the manuscript will receive final approval. with compliments Stanisław Wroński Academic Editor [Note: HTML markup is below. Please do not edit.] Reviewers' comments: Reviewer's Responses to Questions Comments to the Author Reviewer #2: All comments have been addressed ********** 2. Is the manuscript technically sound, and do the data support the conclusions??> Reviewer #2: Yes ********** 3. Has the statistical analysis been performed appropriately and rigorously? -->?> Reviewer #2: I Don't Know ********** 4. Have the authors made all data underlying the findings in their manuscript fully available??> The PLOS Data policy Reviewer #2: Yes ********** 5. Is the manuscript presented in an intelligible fashion and written in standard English??> Reviewer #2: Yes ********** Reviewer #2: Dear editor, Thanks for choosing me as a reviewer; While the authors have responded to the previous feedback, I have several additional questions and concerns for clarification, which I hope will strengthen the final text. Abstract: In the method section, it is not clear what the clustering is based on. Phenotype? Incontinence or...? Please explain fully. In the results section, which session is meant by improvement in session 3? Is it the last session of the third part of the treatment? Please explain clearly. Main text: Include the code of ethics in the article Line 98: What was the purpose of ultrasound imaging? This test is not mentioned in other parts of the text. Were there surface electrodes on the rectus abdominis and gluteal to check for lack of help from other muscles? How many channels did the EMG device used have? And what is the name of the device? In the exercises section, how did the patient maintain a 30-second contraction in the first treatment session!! Were the exercises in the second and third sessions similar to sessions one? How many times were the treatment sessions per week? Explain more fully about the treatment sessions. How many sessions were the exercises in phases two and three and how many times a week? Some abbreviations are not included in table one. For example, PSA - check the other tables to be sure. The full name of the UPS is also not included in the text. Please check. Thank you for inviting me to review this paper. Abstract: In the Methods section, it is unclear what the clustering was based on. Was it on phenotype, incontinence, or other factors? Please provide a full explanation. In the Results section, which session is referred to by "improvement in third session "? Is this the last session of the third phase of treatment? Please clarify. Main Text: The ethics approval code should be included in the manuscript. Line 97: What was the purpose of the ultrasound imaging? This assessment is not mentioned elsewhere in the text. Were surface electrodes placed on the rectus abdominis and gluteal muscles to check for the absence of accessory muscle contraction? How many channels did the EMG device have, and what was the device model name? In the exercises section, how were patients able to maintain a 30-second contraction during the first treatment session? Were the exercises in the second and third phases similar to those in the first phase?How many times per week were the treatment sessions conducted? Please provide a more complete description of the treatment protocol, including how many sessions were in phases two and three and their weekly frequency. Tables: Some abbreviations in Table 1 are not defined (e.g., PSA). Please verify that all abbreviations are defined in all tables. The full name of the "ISUP" is also not provided in the text. Please check and define all abbreviations upon first use. ********** what does this mean? ). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy Reviewer #2: Yes: seyedeh Saeideh Babazadeh-Zavieh ********** [NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.] To ensure your figures meet our technical requirements, please review our figure guidelines: https://journals.plos.org/plosone/s/figures You may also use PLOS’s free figure tool, NAAS, to help you prepare publication quality figures: https://journals.plos.org/plosone/s/figures#loc-tools-for-figure-preparation. NAAS will assess whether your figures meet our technical requirements by comparing each figure against our figure specifications. 18 Nov 2025 Response to Reviewer #2 Dear Reviewer, We thank you very much for your thoughtful and constructive feedback. We appreciate the opportunity to further improve our manuscript. Below, we provide detailed responses to all comments, accompanied by the corresponding revisions to the text. All changes requested have been incorporated into the manuscript using Track Changes. Abstract Comment 1: “In the method section, it is not clear what the clustering is based on. Phenotype? Incontinence or...? Please explain fully.” Response: Thank you for this important observation. We have revised the Methods section of the abstract to clarify the basis of the clustering. The sentence now reads: “K-means clustering was applied to 11 baseline clinical variables, including urinary incontinence severity, pelvic floor function measures, and oncological risk characteristics, to identify distinct patient phenotypes.” (See Abstract, Methods) Comment 2: “In the results section, which session is meant by improvement in session 3? Is it the last session of the third part of the treatment? Please explain clearly.” Response: Thank you for pointing this out. The text has been revised for clarity. We now specify that the third assessment refers to the rehabilitation evaluation conducted after the completion of phase III of the program. The revised sentence reads: “By the third rehabilitation assessment (conducted after completing phase III of the rehabilitation program), continence rates increased to 88.4%, 77.5%, and 60.0% across the three clusters.” (See Abstract, Results) Main Text Comment 3: “Include the code of ethics in the article.” Response: We agree and have expanded the ethics statement to include reference to the Declaration of Helsinki. The revised sentence reads: “The study was conducted in accordance with the Declaration of Helsinki and approved by the Bioethics Committee of Collegium Medicum, Jan Kochanowski University, Kielce, Poland (approval no. 34/2018). Written informed consent was obtained from all participants.” (See Methods) Comment 4: “Line 97: What was the purpose of ultrasound imaging? This assessment is not mentioned elsewhere in the text.” Response: Thank you for highlighting this. We have now specified the purpose of ultrasound imaging in the Methods section. The revised sentence reads: “Ultrasound sonofeedback was used during all sessions to provide real-time visualization of pelvic floor muscle contractions, enabling patients to learn correct activation and relaxation and to coordinate this activity with breathing.” (See Methods, Rehabilitation protocol) Comment 5: “Were surface electrodes placed on the rectus abdominis and gluteal muscles to check for the absence of accessory muscle contraction? How many channels did the EMG device have, and what was the device model name?” Response: These details have now been added. Surface electrodes were placed on the rectus abdominis and right gluteal muscle to prevent accessory muscle activation. We also specified the EMG system used. The text now reads: “Surface electrodes were placed over the rectus abdominis and right gluteal muscle to monitor and prevent the activation of accessory muscles during pelvic floor contractions. The electromyography system used in this study was a Noraxon Ultium® EMG device equipped with 16 channels, of which four were activated for the purposes of this protocol.” (See Methods, Rehabilitation protocol) Comment 6: “In the exercises section, how were patients able to maintain a 30-second contraction during the first treatment session?” Response: We appreciate this question. Patients attempted a 30-second sustained contraction according to the standardized Glazer protocol implemented via Noraxon software. Not all were able to sustain the contraction for the full duration, as this task was used for functional assessment rather than as a therapeutic objective. The sentence has been added: “While not all patients were able to maintain a full 30-second contraction during the initial session, this step served as a standardized functional assessment rather than a clinical outcome measure in this study.” (See Methods, Detailed first session procedure) Comment 7: “Were the exercises in the second and third phases similar to those in the first phase? How many times per week were the treatment sessions conducted? Please provide a more complete description of the treatment protocol, including how many sessions were in phases two and three and their weekly frequency.” Response: Thank you for pointing this out. The text has been expanded to describe the exercise protocol in phases II and III. The revised sentence reads: “In phases II and III of the rehabilitation program, patients attended five supervised physiotherapy sessions per week. In cases requiring adjunct neuromuscular electrostimulation, a total of 10 sessions were conducted over two weeks, also at a frequency of five sessions per week. The exercise protocol in these phases was similar to that used in phase I and included pelvic floor muscle activation and relaxation in lying, sitting, and standing positions, synchronizing contractions with breathing, and incorporating functional movement patterns.” (See Methods, Rehabilitation protocol) Tables Comment 8: “Some abbreviations in Table 1 are not defined (e.g., PSA). Please verify that all abbreviations are defined in all tables.” Response: Thank you for your careful review. We have updated the Abbreviations section under Table 1 to include the definition of PSA as “prostate-specific antigen.” (See Table 1, Abbreviations) Comment 9: “The full name of the ‘ISUP’ is also not provided in the text. Please check and define all abbreviations upon first use.” Response: This has been corrected. The full name “International Society of Urological Pathology (ISUP)” is now provided upon first use in the text and included in the Abbreviations list under Table 1. (See Methods and Table 1, Abbreviations) We hope that these revisions adequately address all concerns. We are grateful for your valuable feedback, which has helped strengthen the clarity and rigor of our manuscript. Sincerely, Prof. Paweł Macek, PhD, MSc Corresponding Author On behalf of all co-authors Collegium Medicum, Jan Kochanowski University of Kielce, Poland Scientific Research, Epidemiology and R&D Centre, Holycross Cancer Centre, Kielce, Poland pawel.macek@onkol.kielce.pl on behalf of all authors Submitted filename: Response_to_Reviewer_2_Round_3_Pone-D-25-33254.docx 30 Nov 2025 Phenotypic Stratification Predicts the Pace, but Not the Outcome, of Continence Recovery after Radical Prostatectomy PONE-D-25-33254R2 Dear Dr. Paweł Macek We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements. Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication. An invoice will be generated when your article is formally accepted. Please note, if your institution has a publishing partnership with PLOS and your article meets the relevant criteria, all or part of your publication costs will be covered. Please make sure your user information is up-to-date by logging into Editorial Manager at Editorial Manager® and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact billing support . If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org. Kind regards, Stanisław Jacek Wroński, M.D., Ph.D, FEBU Academic Editor PLOS ONE Dear Authors, Considering that it took a long time to find a reviewer for this article I suggest that the article ultimately meets the conditions for publication in PLOS ONE. Personally, I believe that the authors' approach to the issue of urinary incontinence after radical prostatectomy fully deserves attention and presentation in the journal. With compliments Stanisław Wroński PLOS ONE academic editor Reviewers' comments: Reviewer's Responses to Questions Comments to the Author Reviewer #2: All comments have been addressed ********** 2. Is the manuscript technically sound, and do the data support the conclusions??> Reviewer #2: Yes ********** 3. Has the statistical analysis been performed appropriately and rigorously? -->?> Reviewer #2: I Don't Know ********** 4. Have the authors made all data underlying the findings in their manuscript fully available??> The PLOS Data policy Reviewer #2: Yes ********** 5. Is the manuscript presented in an intelligible fashion and written in standard English??> Reviewer #2: Yes ********** Reviewer #2: The authors have responded to the review and made corrections to the text. The manuscript appears to be suitable for acceptance. ********** what does this mean? ). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy Reviewer #2: No ********** PONE-D-25-33254R2 PLOS One Dear Dr. Macek, I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS One. Congratulations! Your manuscript is now being handed over to our production team. At this stage, our production department will prepare your paper for publication. This includes ensuring the following: * All references, tables, and figures are properly cited * All relevant supporting information is included in the manuscript submission, * There are no issues that prevent the paper from being properly typeset You will receive further instructions from the production team, including instructions on how to review your proof when it is ready. Please keep in mind that we are working through a large volume of accepted articles, so please give us a few days to review your paper and let you know the next and final steps. Lastly, if your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org. You will receive an invoice from PLOS for your publication fee after your manuscript has reached the completed accept phase. If you receive an email requesting payment before acceptance or for any other service, this may be a phishing scheme. Learn how to identify phishing emails and protect your accounts at https://explore.plos.org/phishing. If we can help with anything else, please email us at customercare@plos.org. Thank you for submitting your work to PLOS ONE and supporting open access. Kind regards, PLOS ONE Editorial Office Staff on behalf of Dr. Stanisław Jacek Wroński Academic Editor PLOS One