A Predictive Tool Powered by Machine Learning for Evaluating the Status of Surgical Margins After Robot-Assisted Radical Prostatectomy

A Predictive Tool Powered by Machine Learning for Evaluating the Status of Surgical Margins After Robot-Assisted Radical Prostatectomy | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article A Predictive Tool Powered by Machine Learning for Evaluating the Status of Surgical Margins After Robot-Assisted Radical Prostatectomy Gen Fan, Yang Li, Yushui Chen, Tielong Tang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8737324/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objective This research intends to create a predictive model utilizing machine learning(ML) techniques to evaluate the likelihood of positive surgical margins (PSM) in individuals receiving robot-assisted radical prostatectomy (RARP), thus aiding in clinical decision-making. Methods A retrospective study was conducted involving 301 patients who underwent RARP. The subjects were randomly assigned to two groups in a ratio of 7:3, which included 201 individuals in the training cohort and 100 in the validation cohort. A total of twenty-four clinical and oncological characteristics were gathered, initially assessed through univariate logistic regression, and later refined using feature selection facilitated by the Boruta algorithm. Utilizing the chosen features, seven distinct ML models were developed. The effectiveness of these models was comprehensively assessed through a range of metrics, including the area beneath the ROC curve and the F1 score. To conduct a detailed analysis of how features influence the optimal model, the SHAP approach was utilized for evaluating feature contributions. Results In the final analysis, there were 301 patients included, revealing a postoperative incidence of PSM at 42.0% after RARP. Through univariate logistic regression and the Boruta algorithm, five key predictive variables were recognized for the construction of the model. Of the seven ML models assessed, the ANN model demonstrated the best performance, achieving an AUC of 0.808 (95% CI: 0.702–0.899) on the validation dataset, along with superior levels of accuracy (80.11%), sensitivity (78.9%), and F1 score (77.9%). Analysis using SHAP indicated that an advanced clinical stage, increased levels of creatinine, high-risk stratification of prostate cancer, and a greater percentage of positive biopsies were strongly linked to a heightened risk of PSM. In contrast, neoadjuvant therapy demonstrated a protective influence on the occurrence of PSM. Conclusion Machine learning models demonstrate significant utility in predicting positive surgical margins after RARP. Integrating the random forest model with the SHAP interpretation framework enables precise prediction of individual PSM risk and provides intuitive insights into the impact of key features on predictive outcomes. This approach facilitates preoperative risk stratification and the development of early postoperative intervention strategies. Positive surgical margin prostate cancer robotic surgery predictive model machine learning Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Background Prostate cancer (PCa) ranks as the second most common cancer among men globally and is the fifth leading cause of cancer-related fatalities. Statistics reveal that annually, over 1.4 million new cases are identified, with around 375,000 of these resulting in mortality worldwide, highlighting a significant challenge for public health on a global scale [ 1 , 2 ]. Although it is frequently encountered, the majority of patients are identified at a localized phase. In cases of localized PCa, radical prostatectomy (RP) is still the main treatment choice, conducted in about 50% of individuals. The other patients typically undergo either radiotherapy (25%) or androgen deprivation therapy (14%) [ 3 ]. RP entails the total excision of both the prostate gland and seminal vesicles, and it is deemed the preferred standard treatment approach for localized PCa [ 4 ]. With advances in minimally invasive surgical techniques, the approach to RP has undergone significant evolution, progressing from traditional open surgery to laparoscopic surgery and subsequently to robot-assisted radical prostatectomy (RARP). RARP has progressively become the standard surgical procedure for localized pca due to its advantages in reducing intraoperative bleeding, decreasing postoperative pain, and shortening hospital stays, and is now widely performed globally [ 5 ]. Although RP is the standard treatment for localized prostate cancer, 13.2% to 46% of patients still experience positive surgical margins (PSM) postoperatively, meaning residual tumor cells are present at the margins marked with ink [ 6 – 8 ]. PSM has been recognized as a distinct risk factor for biochemical recurrence (BCR), strongly linked to a higher likelihood of disease progression and the necessity for salvage therapies, such as radiotherapy or hormone treatment. This situation not only heightens the financial strain on patients and related complications from treatment but also profoundly affects their long-term outcome [ 9 , 10 ]. Therefore, identifying predictive factors for PSM during preoperative assessment holds significant clinical importance for optimizing individualized treatment strategies and improving patient prognosis management. Recently, the use of machine learning (ML) within the field of clinical medicine has grown notably, establishing itself as a novel technological framework due to its ability to create robust risk prediction models and enhance the accuracy of predictions [ 11 ]. By employing powerful statistical methods to capture complex associations between patient characteristics and clinical outcomes, ML can objectively integrate vast amounts of electronic health record data, providing efficient support for prognosis prediction and clinical decision-making [ 12 , 13 ]. Current research data on predictors of postoperative PSM following RARP remains scarce, and systematic prognostic prediction models based on ML are lacking. Therefore, this research seeks to employ machine learning methods for developing predictive models and pinpointing crucial factors that affect postoperative PSM following RARP. We evaluated several algorithms to determine the best model and offer a dependable resource for clinical use. 2.Methods and Data 2.1 Design and Patients This retrospective analysis involved 301 individuals diagnosed with PCa who underwent RARP at the Urology Department of the Affiliated Hospital of North Sichuan Medical College, conducted between January 2022 and February 2026. Among these individuals, 127 had PSM, whereas 174 exhibited negative margins. Participants were assigned to two different groups through a computer-generated randomization process, which resulted in a distribution of 7:3 into a training cohort (n = 211) and a validation cohort (n = 90). The training cohort was employed for the development and refinement of the model, utilizing the Boruta feature selection method along with seven ML algorithms, while the validation cohort served to assess the models' predictive accuracy. Permission for this research protocol was granted by the Ethics Committee at our institution, and all processes conformed to the applicable ethical standards outlined in the Declaration of Helsinki. 2.2 Inclusion and Exclusion Criteria Exclusion criteria included: (1) significant gaps in clinical history or documentation; (2) postoperative pathology reports lacking clear documentation of surgical margin status; (3) presence of bone or visceral metastases.Inclusion criteria include: (1) Pathologically confirmed pca via transrectal or transperineal ultrasound-guided prostate biopsy; (2) Patients with primary pca undergoing RARP at this institution; (3) Complete clinical and pathological data; (4) No evidence of distant metastasis on preoperative imaging studies (including prostate MRI, whole-body bone scan, and chest CT). 2.3 Data and Variables The criteria for selecting variables included: (1) variables exhibiting a missing data rate exceeding 20%; (2) variables evaluated to have no meaningful correlation with postoperative PSM after RARP, based on existing literature and clinical judgment. The final set of variables comprised: (1) Demographic factors: age, smoking habits, history of alcohol use, body mass index (BMI), and past occurrences of hypertension and diabetes; (2) Laboratory evaluations include: total white blood cell count, counts for neutrophils and lymphocytes, platelet count, levels of serum albumin, creatinine, as well as ALT and AST; (3) Clinical characteristics associated with tumors encompass: a background of prostate surgeries, volume of the prostate, measurements of PSA, PSAD, percentage of biopsy cores that are positive, Gleason score (≥ 8), any prior neoadjuvant treatments, as well as clinical staging and risk stratification. 2.4 Statistical Analysis Continuous variables were expressed either as mean ± standard deviation or as median (interquartile range), based on the distribution of the dataset. Group comparisons were made using independent samples t-tests. Categorical variables were reported as frequency (percentage), and chi-square tests were employed for group comparisons. All statistical analyses were performed in a two-sided manner, with a significance level established at P < 0.05.. Patients were randomly assigned using R software (with a random seed set to 123), forming groups in a 7:3 proportion for the training set (n = 211) and the validation set (n = 90). Initially, potential predictors were evaluated through univariate logistic regression (P < 0.05). Subsequently, feature selection was executed by integrating the Boruta algorithm and 10-fold cross-validation, ultimately pinpointing five variables of predictive importance. Utilizing the chosen features, seven distinct machine learning models were developed: Logistic Regression (LR), Random Forest (RF), Decision Tree (DT), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Lightweight Gradient Boosting Machine (LightGBM), and Artificial Neural Network (ANN). Each model underwent five iterations of 10-fold cross-validation alongside grid search for hyperparameter tuning, aimed at enhancing their robustness and generalization capabilities. The evaluation of the models was comprehensive, utilizing various metrics such as the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), recall, and F1 score. The model with the highest performance was recognized due to its exceptional AUC and accuracy scores within the validation dataset. Calibration curves assessed how well the predicted probabilities matched the actual outcomes, whereas decision curves analyzed the clinical net benefit of the model at different risk thresholds.To enhance the model's interpretability, the SHAP approach was employed to assess the significance of features within the most effective model. This method calculates the incremental effect of each feature on prediction outcomes using Shapley values, which are derived from principles of game theory. Visualizations including feature importance plots, SHAP summary plots, and dependency plots disclose the influence and trends of significant predictive variables. Additionally, SHAP plots further depict the structural contribution of features for personalized predictions, supplying clear evidence to support clinical decision-making. 4. Results 4.1 Basic Characteristics The research involved 301 participants, consisting of 127 individuals in the PSM cohort and 174 participants in the negative margin cohort. Table 1 displays the baseline characteristics of each group. Compared with the negative margin cohort, patients in the PSM cohort exhibited a higher proportion of Gleason scores ≥ 8 (63.78% vs 44.25%), high-risk pca stratification (88.98% vs 70.11%), and a higher proportion of positive needle cores. Conversely, the proportion receiving preoperative neoadjuvant therapy was significantly lower (6.30% vs 20.69%, P = 0.001). Regarding laboratory tests, the PSM group exhibited significantly higher PSA, PSAD, and serum creatinine levels compared to the negative margin cohort ( P < 0.05). Table 1 Baseline characteristics of patients with or without positive resection margins. Variables Overall No-PSM PSM p-value Hypertension (%) 107 (35.55) 65 (37.36) 42 (33.07) 0.519 Diabetes (%) 41 (13.62) 26 (14.94) 15 (11.81) 0.54 Smoking (%) 172 (57.14) 103 (59.20) 69 (54.33) 0.469 Drinking (%) 236 (78.41) 138 (79.31) 98 (77.17) 0.76 Prior TURP history (%) 24 ( 7.97) 17 ( 9.77) 7 ( 5.51) 0.258 Clinical T stage (%) < 0.001 T2 187 (62.13) 139 (79.89) 48 (37.80) T3a 60 (19.93) 26 (14.94) 34 (26.77) T3b 51 (16.94) 9 ( 5.17) 42 (33.07) T4 3 ( 1.00) 0 ( 0.00) 3 ( 2.36) Gleason score ≥ 8 (%) 158 (52.49) 77 (44.25) 81 (63.78) 0.001 High risk prostate cancer (%) 235 (78.07) 122 (70.11) 113 (88.98) < 0.001 Neoadjuvant therapy (%) 44 (14.62) 36 (20.69) 8 ( 6.30) 0.001 Percentage of positive biopsy cores (%) < 0.001 1 48 (15.95) 39 (22.41) 9 ( 7.09) 2 40 (13.29) 30 (17.24) 10 ( 7.87) 3 37 (12.29) 28 (16.09) 9 ( 7.09) 4 55 (18.27) 27 (15.52) 28 (22.05) 5 23 ( 7.64) 12 ( 6.90) 11 ( 8.66) 6 18 ( 5.98) 6 ( 3.45) 12 ( 9.45) 7 5 ( 1.66) 3 ( 1.72) 2 ( 1.57) 8 13 ( 4.32) 4 ( 2.30) 9 ( 7.09) 9 20 ( 6.64) 8 ( 4.60) 12 ( 9.45) 10 23 ( 7.64) 8 ( 4.60) 15 (11.81) Age (years) 70.00 [67.00, 74.00] 70.00 [67.00, 74.75] 70.00 [67.00, 74.00] 0.652 BMI (kg/m²) 24.09 [22.31, 26.06] 24.12 [22.31, 25.95] 23.94 [22.26, 26.08] 0.972 Prostate volume (cm³) 34.88 [24.20, 47.36] 32.62 [23.21, 48.68] 35.28 [24.90, 46.16] 0.385 PSA (ng/mL) 18.00 [11.20, 33.70] 15.65 [10.12, 28.25] 21.00 [14.00, 48.70] < 0.001 PSAD (ng/mL/cm³) 0.59 [0.33, 1.10] 0.50 [0.28, 0.99] 0.68 [0.42, 1.37] 0.004 Leukocyte count (×10⁹/L) 5.46 [4.53, 6.54] 5.68 [4.65, 6.66] 5.17 [4.40, 6.44] 0.098 Neutrophil count (×10⁹/L) 3.33 [2.55, 4.00] 3.45 [2.75, 4.01] 3.19 [2.43, 3.97] 0.119 Lymphocyte count (×10⁹/L) 1.44 [1.20, 2.00] 1.48 [1.19, 1.93] 1.42 [1.22, 2.04] 0.878 Neutrophil count (×10⁹/L) 3.33 [2.55, 4.00] 3.45 [2.75, 4.01] 3.19 [2.43, 3.97] 0.119 Lymphocyte count (×10⁹/L) 1.44 [1.20, 2.00] 1.48 [1.19, 1.93] 1.42 [1.22, 2.04] 0.878 Platelet count (×10⁹/L) 169.00 [135.00, 198.00] 165.50 [134.00, 196.00] 173.00 [143.00, 199.50] 0.269 Neutrophil percentage (%) 60.50 [55.10, 66.70] 60.40 [54.80, 66.60] 61.40 [55.55, 66.70] 0.993 Albumin (g/L) 40.70 [38.70, 44.00] 41.10 [38.70, 44.50] 40.20 [38.70, 43.10] 0.179 Creatinine (µmol/L) 74.60 [67.30, 84.60] 72.90 [65.93, 83.35] 77.40 [69.50, 85.65] 0.015 ALT (U/L) 20.00 [16.00, 28.00] 20.00 [17.00, 28.00] 20.00 [16.00, 28.00] 0.766 AST (U/L) 24.00 [21.00, 28.00] 24.00 [20.00, 28.00] 25.00 [21.00, 28.00] 0.613 PSM: Positive Surgical Margin; No-PSM: Negative Surgical Margin; TURP: Transurethral Resection of the Prostate; PSA: Prostate-Specific Antigen; PSAD: PSA Density; BMI: Body Mass Index; ALT: Alanine Aminotransferase; AST: Aspartate Aminotransferase. 4.2 Model Development To determine the key predictors of PSM after RARP, we utilized the Boruta algorithm for feature selection (see Fig. 1 ). From the identified significant predictors—clinical stage, high-risk prostate cancer, neoadjuvant therapy, the proportion of positive cores, and serum creatinine level—we developed seven machine learning models: LR, RF, DT, SVM, XGBoost, ANN, and LightGBM. The optimization of hyperparameters for each model was conducted through five rounds of 10-fold cross-validation combined with grid search, aimed at enhancing generalization and reducing overfitting. 4.3 Model Evaluation Table 2 and Fig. 2 illustrate the comparative performance of seven different machine learning models. Within the validation dataset, the ANN model exhibited exceptional outcomes, attaining an AUC of 0.808, an accuracy of 0.811, a sensitivity of 0.789, and a specificity of 0.827 (see Table 2 and Fig. 2 A), significantly surpassing the results of the other models. The calibration curves demonstrated a significant relationship between the anticipated probabilities generated by the ANN model and the real incidence rates (refer to Fig. 2 B). The DCA showed that the predictive model based on ANN provided the highest net benefit for anticipating positive margins, surpassing the effectiveness of other models (Fig. 2 C). Following a thorough evaluation of its discriminative capacity, calibration, and practical application, the ANN model emerged as the top-performing model and was employed in the ensuing SHAP interpretability analysis. Table 2 The prediction performance of each model. Model AUC Accuracy Precision Sensitivity Specificity F1 Score Kappa Youden's J PPV NPV Logistic 0.791 0.688 0.879 0.756 0.7 0.737 0.769 0.718 0.503 0.506 Decision Tree 0.684 0.572 0.789 0.622 0.542 0.684 0.577 0.605 0.252 0.261 Random Forest 0.808 0.702 0.899 0.811 0.769 0.789 0.827 0.779 0.614 0.616 XGBoost 0.763 0.656 0.855 0.722 0.71 0.579 0.527 0.638 0.416 0.406 LightGBM 0.727 0.617 0.825 0.689 0.639 0.605 0.75 0.622 0.358 0.355 SVM 0.781 0.674 0.873 0.7 0.641 0.658 0.731 0.649 0.387 0.389 ANN 0.773 0.673 0.863 0.733 0.706 0.632 0.808 0.667 0.446 0.439 4.4 Model Interpretation SHAP was used to evaluate how each predictor affects the outcomes, applying a game-theoretic framework to determine the significance of each feature. SHAP significance analysis using the ANN model visualized feature importance rankings, as shown in Fig. 3 A. Our analysis identified five predictors associated with positive surgical margins: clinical stage, high-risk prostate cancer, neoadjuvant therapy, percentage of positive biopsies, and serum creatinine level. The ranking is further enhanced by the SHAP summary plot (Fig. 3 B), which visually demonstrates the impact of each feature on the model's output. Features with positive Shapley values signify a heightened risk of achieving positive margins, whereas features exhibiting negative values imply a diminished risk. As shown in Fig. 4 , interpretative analysis based on SHAP dependency plots indicates that elevated clinical T stage, high-risk prostate cancer, increased biopsy positive percentage, and elevated creatinine levels are associated with increased margin positivity risk, exhibiting linear or near-linear relationships. Conversely, neoadjuvant therapy demonstrates a negative influence in the model, suggesting it may be associated with reduced margin positivity risk. 5. Discussion This retrospective study employed ML methods to create a clinical prediction model for the probability of PSM following RARP. Out of the seven algorithms assessed, the ANN exhibited the best predictive accuracy. Analysis using SHAP interpretability indicated that factors such as clinical T stage, serum creatinine levels, high-risk PCa, and a greater percentage of positive biopsy cores were key independent predictor for PSM, whereas neoadjuvant therapy appeared to have a protective influence. This predictive tool assists clinicians in preoperatively identifying high-risk patients, providing evidence-based support for developing individualized treatment strategies and thereby optimizing clinical decision-making processes. Earlier research has established that pathological staging serves as a crucial predictor of PSM. A comprehensive meta-analysis revealed that the incidence of positive margins for pT2, pT3, and pT4 pca were 9%, 37%, and 50%, respectively, indicating a rising trend [16]. Since pathological staging requires postoperative confirmation, this study employed preoperatively accessible clinical staging as a surrogate indicator. Multiple studies support the association between clinical staging and positive margins: Ramos et al. reported significantly higher positive margin rates for T2b tumors compared to T1c (29% vs. 20%, p < 0.05), with T1c-staged patients exhibiting lower risks of biochemical recurrence [17]; Ficarra et al. confirmed in 322 RARP cases that clinical staging is an independent predictor of positive margins (HR = 2.217, p < 0.05) [18]. This may stem from advanced tumors exhibiting extensive local infiltration with indistinct boundaries from critical structures like neurovascular bundles. Intraoperative efforts to preserve function or avoid complications often compromise adequate resection margins, thereby elevating the risk of positive margins. Elevated preoperative serum creatinine is associated with an increased risk of PSM following RARP. The mechanisms may include: on one hand, elevated creatinine indicates renal insufficiency and poorer systemic condition, prompting surgeons to adopt conservative resection strategies to reduce surgical risks, thereby limiting the attainment of safe margins; on the other hand, elevated creatinine may be secondary to obstructive nephropathy caused by advanced tumors (such as ureteral compression or bladder outlet obstruction) [19, 20]. In such cases, elevated creatinine serves as an indirect marker of locally advanced or high-burden tumors, where extensive tumor invasion itself directly causes positive margins. Tollefson et al.'s study of 10,099 patients further confirmed that eGFR is an independently predictive factor for overall mortality in patients with PCA [21], underscoring the prognostic value of renal function indicators. The positive core rate is defined as the proportion of needles with cancerous tissue in the total number of needles during preoperative prostate biopsy. This indicator directly reflects the spatial distribution and invasive burden of the tumor within the prostate. Consequently, a greater rate of positive cores is linked to a heightened risk of positive surgical margins after surgery. The study carried out by Tuliao and colleagues further confirmed that the number of positive cores found in preoperative biopsies serves as a separate predictor of PSM, which is consistent with the findings of this research [22]. Neoadjuvant therapy plays a crucial role in minimizing the likelihood of PSM following RARP. Research conducted by Hu and colleagues, which included 48 patients classified as having intermediate to high-risk PCA, revealed that those in the neoadjuvant hormone therapy (NHT) cohort, who underwent treatment for durations between 2 to 12 months, showed significantly reduced rates of PSM and instances of BCR when contrasted with the control cohort that did not receive any intervention [23]. Furthermore, a prospective analysis involving 69 patients who had RP after 3 months of androgen deprivation therapy (ADT) against 72 patients who had immediate surgery revealed a markedly higher percentage of negative margins in the neoadjuvant cohort (87% versus 64%, p < 0.01) along with a greater proportion of tumors classified as organ-confined (74% vs. 49%, p < 0.01) [24], aligning with the results of this investigation. This is because neoadjuvant therapy induces tumor cell apoptosis, leading to tumor volume reduction, pathological downstaging, and decreased extracapsular invasion. Simultaneously, it improves surgical field exposure and anatomical visualization, facilitating the achievement of adequate safe margins during surgery and thereby reducing the risk of PSM. Individuals classified as having high-risk PCA are characterized by a PSA ≥ 20 ng/mL, a Gleason score ≥ 8, or a clinical stage of cT2c or more advanced. Research by D'Amico and associates demonstrated that the risk of mortality linked to PCA in this population is 14.2 times greater compared to those considered low-risk[25]. High-risk stratification inherently incorporates multiple aggressive features that increase surgical difficulty: higher clinical stage often indicates extracapsular invasion or seminal vesicle involvement, elevated Gleason score reflects poor tumor differentiation and high aggressiveness, while increased PSA levels signify greater tumor burden. Collectively, these factors substantially elevate the risk of PSM in high-risk patients. This study employed a single-center retrospective design, with cases performed by different surgeons whose varying experience and technical skills may have influenced positive margin rates. Previous research indicates that surgeons' learning curves, annual case volumes, and expertise levels may correlate with the risk of PSM [26]. However, the model did not account for these factors, potentially introducing bias. Furthermore, the model lacks external validation, leaving its generalizability unconfirmed. Future research should involve multicenter prospective studies that take into account surgical factors, including the experience of the surgeon, while also ensuring external validation across various healthcare institutions to improve the model's clinical relevance and applicability.. 6. CONCLUSION The ML based predictive model developed in this study effectively assesses the risk of PSM following RARP. The ANN model showed exceptional predictive capabilities, helping in the preoperative recognition of patients at high risk and offering guidance for personalized clinical decision-making. Declarations Ethical Approval and Informed Consent: The study protocol was approved by the Ethics Committee of the Affiliated Hospital of North Sichuan Medical College (Approval No.: 2022-NSMCAH-073). As this study involved retrospective analysis, all data were anonymized and did not contain any personally identifiable patient information. Therefore, written informed consent was waived. All research procedures adhered to the ethical principles outlined in the Declaration of Helsinki. Funding: The authors declare that they received no funding, grants, or other support during the preparation of this manuscript. Author Contribution All authors contributed to the study conception and design. Writing - original draft preparation: Gen Fan; Writing - review and editing: Gen Fan, Yushui Chen; Conceptualization: Gen Fan; Methodology: Gen Fan, Yang Li; Formal analysis and investigation: Gen Fan, Yushui Chen, Yang Li; Funding acquisition: Tielong Tang; Resources: Yushui Chen, Yang Li; Supervision: Tielong Tang, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Acknowledgments: Not applicable. Publication Permission: Not applicable. Data Availability The study protocol was approved by the Ethics Committee of the Affiliated Hospital of North Sichuan Medical College (Approval No.: 2022-NSMCAH-073). As this study involved retrospective analysis, all data were anonymized and did not contain any personally identifiable patient information. Therefore, written informed consent was waived. All research procedures adhered to the ethical principles outlined in the Declaration of Helsinki. References Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A (2018) Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. Cancer J Clin 68:394–424 Sung H, Ferlay J, Siegel RL et al (2021) Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. Cancer J Clin 71:209–249 Cooperberg MR, Broering JM, Carroll PR (2010) Time trends and local variation in primary treatment of localized prostate cancer. J Clin oncology: official J Am Soc Clin Oncol 28:1117–1123 Martini A, Falagario UG, Villers A et al (2020) Contemporary Techniques of Prostate Dissection for Robot-assisted Prostatectomy. Eur Urol 78:583–591 Hamdy FC, Donovan JL, Lane JA et al (2016) 10-Year Outcomes after Monitoring, Surgery, or Radiotherapy for Localized Prostate Cancer. N Engl J Med 375:1415–1424 Patel VR, Coelho RF, Rocco B et al (2011) Positive surgical margins after robotic assisted radical prostatectomy: a multi-institutional study. J Urol 186:511–516 Carbonell E, Matheu R, Muní M et al (2022) The Effect of Adverse Surgical Margins on the Risk of Biochemical Recurrence after Robotic-Assisted Radical Prostatectomy. Biomedicines. ;10 Su SH, Chang YH, Huang LK et al (2022) Clinical predictors for biochemical failure in patients with positive surgical margin after robotic-assisted radical prostatectomy. Tumori 108:270–277 Wu S, Lin SX, Wirth GJ et al (2019) Impact of Multifocality and Multilocation of Positive Surgical Margin After Radical Prostatectomy on Predicting Oncological Outcome. Clin Genitourin Cancer 17:e44–e52 Lu J, Wirth GJ, Wu S et al (2012) A close surgical margin after radical prostatectomy is an independent predictor of recurrence. J Urol 188:91–97 Haug CJ, Drazen JM (2023) Artificial Intelligence and Machine Learning in Clinical Medicine, 2023. N Engl J Med 388:1201–1208 Deo RC (2020) Machine Learning in Medicine: Will This Time Be Different? Circulation 142:1521–1523 Liang H, Tsui BY, Ni H et al (2019) Evaluation and accurate diagnoses of pediatric diseases using artificial intelligence. Nat Med 25:433–438 Guan C, Ma F, Chang S, Zhang J (2023) Interpretable machine learning models for predicting venous thromboembolism in the intensive care unit: an analysis based on data from 207 centers. 27:406 Critical care (London, England) Song Y, Zhang D, Wang Q et al (2024) Prediction models for postoperative delirium in elderly patients with machine-learning algorithms and SHapley Additive exPlanations. Translational psychiatry 14:57 Novara G, Ficarra V, Mocellin S et al (2012) Systematic review and meta-analysis of studies reporting oncologic outcome after robot-assisted radical prostatectomy. Eur Urol 62:382–404 Ramos CG, Carvalhal GF, Smith DS, Mager DE, Catalona WJ (1999) Clinical and pathological characteristics, and recurrence rates of stage T1c versus T2a or T2b prostate cancer. J Urol 161:1525–1529 Ficarra V, Novara G, Secco S et al (2009) Predictors of positive surgical margins after laparoscopic robot assisted radical prostatectomy. J Urol 182:2682–2688 Lienert A, Ing A, Mark S (2009) Prognostic factors in malignant ureteric obstruction. BJU Int 104:938–941 AlOmeir OK (2023) Risk of Prostate Cancer in Chronic Kidney Disease Patient: A Meta-Analysis using Observational Studies. J Pharm bioallied Sci 15:21–28 Tollefson MK, Boorjian SA, Gettman MT, Rangel LJ, Bergstralh EJ, Karnes RJ (2013) Preoperative estimated glomerular filtration rate predicts overall mortality in patients undergoing radical prostatectomy. Urol Oncol 31:1483–1488 Tuliao PH, Koo KC, Komninos C et al (2015) Number of positive preoperative biopsy cores is a predictor of positive surgical margins (PSM) in small prostates after robot-assisted radical prostatectomy (RARP). BJU Int 116:897–904 Hu JC, Hung SC, Ou YC (2017) Assessments of Neoadjuvant Hormone Therapy Followed by Robotic-Assisted Radical Prostatectomy for Intermediate- and High-Risk Prostate Cancer. Anticancer Res 37:3143–3150 Cookson MS, Sogani PC, Russo P et al (1997) Pathological staging and biochemical recurrence after neoadjuvant androgen deprivation therapy in combination with radical prostatectomy in clinically localized prostate cancer: results of a phase II study. Br J Urol 79:432–438 D'Amico AV, Whittington R, Malkowicz SB et al (1998) Biochemical outcome after radical prostatectomy, external beam radiation therapy, or interstitial radiation therapy for clinically localized prostate cancer. JAMA 280:969–974 Lee RS, Ma R, Pham S et al (2022) Machine Learning to Delineate Surgeon and Clinical Factors That Anticipate Positive Surgical Margins After Robot-Assisted Radical Prostatectomy. J Endourol 36:1192–1198 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8737324","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":586035019,"identity":"15c16610-3cc3-4e5f-b551-e4cbb241631a","order_by":0,"name":"Gen Fan","email":"","orcid":"","institution":"North Sichuan Medical College/Affiliated Hospital of North Sichuan Medical College","correspondingAuthor":false,"prefix":"","firstName":"Gen","middleName":"","lastName":"Fan","suffix":""},{"id":586035020,"identity":"cf689553-02c1-47e1-a538-065dc41c8e75","order_by":1,"name":"Yang Li","email":"","orcid":"","institution":"North Sichuan Medical College/Affiliated Hospital of North Sichuan Medical College","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Li","suffix":""},{"id":586035021,"identity":"6089e15d-61ed-406c-a6e6-bd092d224469","order_by":2,"name":"Yushui Chen","email":"","orcid":"","institution":"North Sichuan Medical College/Affiliated Hospital of North Sichuan Medical College","correspondingAuthor":false,"prefix":"","firstName":"Yushui","middleName":"","lastName":"Chen","suffix":""},{"id":586035022,"identity":"e02acf7b-fe8f-48f5-b794-2b5bb50d3848","order_by":3,"name":"Tielong Tang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAwklEQVRIie3RsQrCMBCA4YSALtGuKS4+wkmhU97EJaVQJ/cOBVMK6eiqz+ELVA6cAr5CJl0Lrg7W0UEaN4d88/3D3RESBH9JEdKDYBFj6HwTWh9KOY3bSQHeScNtEcGVL4VXELWbu5sZXCTICZBKrkcTYV2mY4NJirPOkUux1WMJCHXWK4N5inMFVKNPktU6M7g7NRyEZ5JT3dmCAfNNhL3Roy4lEzgcWfnsMlzMPZ7vV+4RXV/J8eST+m08CIIg+OYF8iRAPDs85h8AAAAASUVORK5CYII=","orcid":"","institution":"North Sichuan Medical College/Affiliated Hospital of North Sichuan Medical College","correspondingAuthor":true,"prefix":"","firstName":"Tielong","middleName":"","lastName":"Tang","suffix":""}],"badges":[],"createdAt":"2026-01-30 05:23:22","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8737324/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8737324/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":102389312,"identity":"541d147e-cfae-452f-b39e-82e9b5f2d490","added_by":"auto","created_at":"2026-02-11 08:27:55","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":17161,"visible":true,"origin":"","legend":"\u003cp\u003eBourta Screening.\u003c/p\u003e","description":"","filename":"OnlineFigure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8737324/v1/79e41d9acc617d135864498b.png"},{"id":102389153,"identity":"23d2b6be-9a48-4d28-a8a8-ff6d8557f16c","added_by":"auto","created_at":"2026-02-11 08:27:34","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":5713,"visible":true,"origin":"","legend":"\u003cp\u003eComprehensive Evaluation of Machine Learning Models. (A) ROC Curve and AUC Value of the Validation Set. (B) Calibration Curve of the Validation Set. (D) Decision Curve Analysis of the Validation Set.\u003c/p\u003e","description":"","filename":"placeholderimage.png","url":"https://assets-eu.researchsquare.com/files/rs-8737324/v1/f0c8a3b5e396f8a480ead94a.png"},{"id":102389126,"identity":"48cf1d43-5656-4242-b0a9-000ec2599a1f","added_by":"auto","created_at":"2026-02-11 08:27:26","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":5713,"visible":true,"origin":"","legend":"\u003cp\u003eFeature Importance Analysis of ANN Model Based on SHAP Method. (A) Feature Importance Ranking Based on Mean SHAP Significance Analysis. (B) SHAP Summary Plot for ANN Model.\u003c/p\u003e","description":"","filename":"placeholderimage.png","url":"https://assets-eu.researchsquare.com/files/rs-8737324/v1/4e8395ed8e97608fcc08bb55.png"},{"id":102389152,"identity":"ddf1e320-461c-4e79-805b-1dca37d7b0a2","added_by":"auto","created_at":"2026-02-11 08:27:32","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":52296,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP Dependency Plot of Features in the ANN Model.\u003c/p\u003e","description":"","filename":"OnlineFigure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8737324/v1/8a74828f629a36980e9adcfe.png"},{"id":102640934,"identity":"fc2a41e8-abb2-42f5-8f2c-2196c4687d6f","added_by":"auto","created_at":"2026-02-14 01:54:20","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":791341,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8737324/v1/3ab8fc57-05f9-4217-88de-587342205e5e.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Predictive Tool Powered by Machine Learning for Evaluating the Status of Surgical Margins After Robot-Assisted Radical Prostatectomy","fulltext":[{"header":"1. Background","content":"\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eProstate cancer (PCa) ranks as the second most common cancer among men globally and is the fifth leading cause of cancer-related fatalities. Statistics reveal that annually, over 1.4\u0026nbsp;million new cases are identified, with around 375,000 of these resulting in mortality worldwide, highlighting a significant challenge for public health on a global scale [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Although it is frequently encountered, the majority of patients are identified at a localized phase. In cases of localized PCa, radical prostatectomy (RP) is still the main treatment choice, conducted in about 50% of individuals. The other patients typically undergo either radiotherapy (25%) or androgen deprivation therapy (14%) [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. RP entails the total excision of both the prostate gland and seminal vesicles, and it is deemed the preferred standard treatment approach for localized PCa [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. With advances in minimally invasive surgical techniques, the approach to RP has undergone significant evolution, progressing from traditional open surgery to laparoscopic surgery and subsequently to robot-assisted radical prostatectomy (RARP). RARP has progressively become the standard surgical procedure for localized pca due to its advantages in reducing intraoperative bleeding, decreasing postoperative pain, and shortening hospital stays, and is now widely performed globally [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e].\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eAlthough RP is the standard treatment for localized prostate cancer, 13.2% to 46% of patients still experience positive surgical margins (PSM) postoperatively, meaning residual tumor cells are present at the margins marked with ink [\u003cspan additionalcitationids=\"CR7\" citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. PSM has been recognized as a distinct risk factor for biochemical recurrence (BCR), strongly linked to a higher likelihood of disease progression and the necessity for salvage therapies, such as radiotherapy or hormone treatment. This situation not only heightens the financial strain on patients and related complications from treatment but also profoundly affects their long-term outcome [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Therefore, identifying predictive factors for PSM during preoperative assessment holds significant clinical importance for optimizing individualized treatment strategies and improving patient prognosis management.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eRecently, the use of machine learning (ML) within the field of clinical medicine has grown notably, establishing itself as a novel technological framework due to its ability to create robust risk prediction models and enhance the accuracy of predictions [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. By employing powerful statistical methods to capture complex associations between patient characteristics and clinical outcomes, ML can objectively integrate vast amounts of electronic health record data, providing efficient support for prognosis prediction and clinical decision-making [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Current research data on predictors of postoperative PSM following RARP remains scarce, and systematic prognostic prediction models based on ML are lacking. Therefore, this research seeks to employ machine learning methods for developing predictive models and pinpointing crucial factors that affect postoperative PSM following RARP. We evaluated several algorithms to determine the best model and offer a dependable resource for clinical use.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e"},{"header":"2.Methods and Data","content":"\u003cdiv id=\"Sec3\"\u003e\n \u003ch2\u003e2.1 Design and Patients\u003c/h2\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\"\u003e\n \u003cp\u003eThis retrospective analysis involved 301 individuals diagnosed with PCa who underwent RARP at the Urology Department of the Affiliated Hospital of North Sichuan Medical College, conducted between January 2022 and February 2026. Among these individuals, 127 had PSM, whereas 174 exhibited negative margins. Participants were assigned to two different groups through a computer-generated randomization process, which resulted in a distribution of 7:3 into a training cohort (n\u0026thinsp;=\u0026thinsp;211) and a validation cohort (n\u0026thinsp;=\u0026thinsp;90). The training cohort was employed for the development and refinement of the model, utilizing the Boruta feature selection method along with seven ML algorithms, while the validation cohort served to assess the models\u0026apos; predictive accuracy. Permission for this research protocol was granted by the Ethics Committee at our institution, and all processes conformed to the applicable ethical standards outlined in the Declaration of Helsinki.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\"\u003e\n \u003ch2\u003e2.2 Inclusion and Exclusion Criteria\u003c/h2\u003e\n \u003cp\u003eExclusion criteria included: (1) significant gaps in clinical history or documentation; (2) postoperative pathology reports lacking clear documentation of surgical margin status; (3) presence of bone or visceral metastases.Inclusion criteria include: (1) Pathologically confirmed pca via transrectal or transperineal ultrasound-guided prostate biopsy; (2) Patients with primary pca undergoing RARP at this institution; (3) Complete clinical and pathological data; (4) No evidence of distant metastasis on preoperative imaging studies (including prostate MRI, whole-body bone scan, and chest CT).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\"\u003e\n \u003ch2\u003e2.3 Data and Variables\u003c/h2\u003e\n \u003cp\u003eThe criteria for selecting variables included: (1) variables exhibiting a missing data rate exceeding 20%; (2) variables evaluated to have no meaningful correlation with postoperative PSM after RARP, based on existing literature and clinical judgment. The final set of variables comprised: (1) Demographic factors: age, smoking habits, history of alcohol use, body mass index (BMI), and past occurrences of hypertension and diabetes; (2) Laboratory evaluations include: total white blood cell count, counts for neutrophils and lymphocytes, platelet count, levels of serum albumin, creatinine, as well as ALT and AST; (3) Clinical characteristics associated with tumors encompass: a background of prostate surgeries, volume of the prostate, measurements of PSA, PSAD, percentage of biopsy cores that are positive, Gleason score (\u0026ge;\u0026thinsp;8), any prior neoadjuvant treatments, as well as clinical staging and risk stratification.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\"\u003e\n \u003ch2\u003e2.4 Statistical Analysis\u003c/h2\u003e\n \u003cp\u003eContinuous variables were expressed either as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation or as median (interquartile range), based on the distribution of the dataset. Group comparisons were made using independent samples t-tests. Categorical variables were reported as frequency (percentage), and chi-square tests were employed for group comparisons. All statistical analyses were performed in a two-sided manner, with a significance level established at P\u0026thinsp;\u0026lt;\u0026thinsp;0.05..\u003c/p\u003e\n \u003cp\u003ePatients were randomly assigned using R software (with a random seed set to 123), forming groups in a 7:3 proportion for the training set (n\u0026thinsp;=\u0026thinsp;211) and the validation set (n\u0026thinsp;=\u0026thinsp;90). Initially, potential predictors were evaluated through univariate logistic regression (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Subsequently, feature selection was executed by integrating the Boruta algorithm and 10-fold cross-validation, ultimately pinpointing five variables of predictive importance. Utilizing the chosen features, seven distinct machine learning models were developed: Logistic Regression (LR), Random Forest (RF), Decision Tree (DT), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Lightweight Gradient Boosting Machine (LightGBM), and Artificial Neural Network (ANN). Each model underwent five iterations of 10-fold cross-validation alongside grid search for hyperparameter tuning, aimed at enhancing their robustness and generalization capabilities. The evaluation of the models was comprehensive, utilizing various metrics such as the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), recall, and F1 score. The model with the highest performance was recognized due to its exceptional AUC and accuracy scores within the validation dataset. Calibration curves assessed how well the predicted probabilities matched the actual outcomes, whereas decision curves analyzed the clinical net benefit of the model at different risk thresholds.To enhance the model\u0026apos;s interpretability, the SHAP approach was employed to assess the significance of features within the most effective model. This method calculates the incremental effect of each feature on prediction outcomes using Shapley values, which are derived from principles of game theory. Visualizations including feature importance plots, SHAP summary plots, and dependency plots disclose the influence and trends of significant predictive variables. Additionally, SHAP plots further depict the structural contribution of features for personalized predictions, supplying clear evidence to support clinical decision-making.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"4. Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003e4.1 Basic Characteristics\u003c/h2\u003e\n \u003cp\u003eThe research involved 301 participants, consisting of 127 individuals in the PSM cohort and 174 participants in the negative margin cohort. Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e displays the baseline characteristics of each group. Compared with the negative margin cohort, patients in the PSM cohort exhibited a higher proportion of Gleason scores\u0026thinsp;\u0026ge;\u0026thinsp;8 (63.78% vs 44.25%), high-risk pca stratification (88.98% vs 70.11%), and a higher proportion of positive needle cores. Conversely, the proportion receiving preoperative neoadjuvant therapy was significantly lower (6.30% vs 20.69%, P\u0026thinsp;=\u0026thinsp;0.001). Regarding laboratory tests, the PSM group exhibited significantly higher PSA, PSAD, and serum creatinine levels compared to the negative margin cohort ( P\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eBaseline characteristics of patients with or without positive resection margins.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"5\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eVariables\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eOverall\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNo-PSM\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePSM\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ep-value\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHypertension (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e107 (35.55)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e65 (37.36)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e42 (33.07)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.519\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDiabetes (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e41 (13.62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e26 (14.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e15 (11.81)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.54\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSmoking (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e172 (57.14)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e103 (59.20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e69 (54.33)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.469\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDrinking (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e236 (78.41)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e138 (79.31)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e98 (77.17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePrior TURP history (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e24 ( 7.97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e17 ( 9.77)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7 ( 5.51)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.258\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eClinical T stage (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e187 (62.13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e139 (79.89)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e48 (37.80)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT3a\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e60 (19.93)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e26 (14.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e34 (26.77)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT3b\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e51 (16.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e9 ( 5.17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e42 (33.07)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eT4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3 ( 1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0 ( 0.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3 ( 2.36)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGleason score\u0026thinsp;\u0026ge;\u0026thinsp;8 (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e158 (52.49)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e77 (44.25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e81 (63.78)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHigh risk prostate cancer (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e235 (78.07)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e122 (70.11)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e113 (88.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNeoadjuvant therapy (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e44 (14.62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e36 (20.69)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e8 ( 6.30)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePercentage of positive biopsy cores (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e48 (15.95)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e39 (22.41)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e9 ( 7.09)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e40 (13.29)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e30 (17.24)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e10 ( 7.87)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e37 (12.29)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e28 (16.09)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e9 ( 7.09)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e55 (18.27)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e27 (15.52)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e28 (22.05)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e23 ( 7.64)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e12 ( 6.90)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e11 ( 8.66)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e18 ( 5.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6 ( 3.45)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e12 ( 9.45)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5 ( 1.66)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3 ( 1.72)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2 ( 1.57)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e13 ( 4.32)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4 ( 2.30)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e9 ( 7.09)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e20 ( 6.64)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e8 ( 4.60)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e12 ( 9.45)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e23 ( 7.64)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e8 ( 4.60)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e15 (11.81)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e70.00 [67.00, 74.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e70.00 [67.00, 74.75]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e70.00 [67.00, 74.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.652\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBMI (kg/m\u0026sup2;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e24.09 [22.31, 26.06]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e24.12 [22.31, 25.95]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e23.94 [22.26, 26.08]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.972\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eProstate volume (cm\u0026sup3;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e34.88 [24.20, 47.36]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e32.62 [23.21, 48.68]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e35.28 [24.90, 46.16]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.385\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePSA (ng/mL)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e18.00 [11.20, 33.70]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e15.65 [10.12, 28.25]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e21.00 [14.00, 48.70]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePSAD (ng/mL/cm\u0026sup3;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.59 [0.33, 1.10]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.50 [0.28, 0.99]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.68 [0.42, 1.37]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLeukocyte count (\u0026times;10⁹/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.46 [4.53, 6.54]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.68 [4.65, 6.66]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.17 [4.40, 6.44]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.098\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNeutrophil count (\u0026times;10⁹/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.33 [2.55, 4.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.45 [2.75, 4.01]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.19 [2.43, 3.97]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.119\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLymphocyte count (\u0026times;10⁹/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.44 [1.20, 2.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.48 [1.19, 1.93]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.42 [1.22, 2.04]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.878\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNeutrophil count (\u0026times;10⁹/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.33 [2.55, 4.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.45 [2.75, 4.01]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.19 [2.43, 3.97]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.119\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLymphocyte count (\u0026times;10⁹/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.44 [1.20, 2.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.48 [1.19, 1.93]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.42 [1.22, 2.04]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.878\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePlatelet count (\u0026times;10⁹/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e169.00 [135.00, 198.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e165.50 [134.00, 196.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e173.00 [143.00, 199.50]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.269\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNeutrophil percentage (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e60.50 [55.10, 66.70]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e60.40 [54.80, 66.60]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e61.40 [55.55, 66.70]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.993\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAlbumin (g/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e40.70 [38.70, 44.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e41.10 [38.70, 44.50]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e40.20 [38.70, 43.10]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.179\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCreatinine (\u0026micro;mol/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e74.60 [67.30, 84.60]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e72.90 [65.93, 83.35]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e77.40 [69.50, 85.65]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.015\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eALT (U/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e20.00 [16.00, 28.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e20.00 [17.00, 28.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e20.00 [16.00, 28.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.766\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAST (U/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e24.00 [21.00, 28.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e24.00 [20.00, 28.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e25.00 [21.00, 28.00]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.613\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\"\u003ePSM: Positive Surgical Margin; No-PSM: Negative Surgical Margin; TURP: Transurethral Resection of the Prostate; PSA: Prostate-Specific Antigen; PSAD: PSA Density; BMI: Body Mass Index; ALT: Alanine Aminotransferase; AST: Aspartate Aminotransferase.\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n \u003ch2\u003e4.2 Model Development\u003c/h2\u003e\n \u003cp\u003eTo determine the key predictors of PSM after RARP, we utilized the Boruta algorithm for feature selection (see Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). From the identified significant predictors\u0026mdash;clinical stage, high-risk prostate cancer, neoadjuvant therapy, the proportion of positive cores, and serum creatinine level\u0026mdash;we developed seven machine learning models: LR, RF, DT, SVM, XGBoost, ANN, and LightGBM. The optimization of hyperparameters for each model was conducted through five rounds of 10-fold cross-validation combined with grid search, aimed at enhancing generalization and reducing overfitting.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003e\u003cstrong\u003e4.3 Model Evaluation\u003c/strong\u003e\u003c/h2\u003e\n \u003cp\u003eTable \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e and Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e illustrate the comparative performance of seven different machine learning models. Within the validation dataset, the ANN model exhibited exceptional outcomes, attaining an AUC of 0.808, an accuracy of 0.811, a sensitivity of 0.789, and a specificity of 0.827 (see Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e and Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eA), significantly surpassing the results of the other models. The calibration curves demonstrated a significant relationship between the anticipated probabilities generated by the ANN model and the real incidence rates (refer to Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eB). The DCA showed that the predictive model based on ANN provided the highest net benefit for anticipating positive margins, surpassing the effectiveness of other models (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eC). Following a thorough evaluation of its discriminative capacity, calibration, and practical application, the ANN model emerged as the top-performing model and was employed in the ensuing SHAP interpretability analysis.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eThe prediction performance of each model.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"11\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eModel\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAUC\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAccuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePrecision\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eF1 Score\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eKappa\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eYouden\u0026apos;s J\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePPV\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNPV\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLogistic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.791\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.688\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.879\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.756\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.737\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.769\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.718\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.503\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.506\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDecision Tree\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.684\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.572\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.789\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.622\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.542\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.684\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.577\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.605\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.252\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.261\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRandom Forest\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.808\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.702\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.899\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.811\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.769\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.789\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.827\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.779\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.614\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.616\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eXGBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.763\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.656\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.855\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.722\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.579\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.527\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.638\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.416\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.406\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLightGBM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.727\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.617\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.825\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.689\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.639\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.605\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.622\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.358\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.355\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSVM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.781\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.674\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.873\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.641\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.658\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.731\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.649\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.387\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.389\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eANN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.773\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.673\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.863\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.733\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.706\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.632\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.808\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.667\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.446\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.439\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003e\u003cstrong\u003e4.4 Model Interpretation\u003c/strong\u003e\u003c/h2\u003e\n \u003cp\u003eSHAP was used to evaluate how each predictor affects the outcomes, applying a game-theoretic framework to determine the significance of each feature. SHAP significance analysis using the ANN model visualized feature importance rankings, as shown in Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eA. Our analysis identified five predictors associated with positive surgical margins: clinical stage, high-risk prostate cancer, neoadjuvant therapy, percentage of positive biopsies, and serum creatinine level. The ranking is further enhanced by the SHAP summary plot (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eB), which visually demonstrates the impact of each feature on the model\u0026apos;s output. Features with positive Shapley values signify a heightened risk of achieving positive margins, whereas features exhibiting negative values imply a diminished risk. As shown in Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e, interpretative analysis based on SHAP dependency plots indicates that elevated clinical T stage, high-risk prostate cancer, increased biopsy positive percentage, and elevated creatinine levels are associated with increased margin positivity risk, exhibiting linear or near-linear relationships. Conversely, neoadjuvant therapy demonstrates a negative influence in the model, suggesting it may be associated with reduced margin positivity risk.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"5. Discussion","content":"\u003cp\u003eThis retrospective study employed ML methods to create a clinical prediction model for the probability of PSM following RARP. Out of the seven algorithms assessed, the ANN exhibited the best predictive accuracy. Analysis using SHAP interpretability indicated that factors such as clinical T stage, serum creatinine levels, high-risk PCa, and a greater percentage of positive biopsy cores were key independent predictor for PSM, whereas neoadjuvant therapy appeared to have a protective influence. This predictive tool assists clinicians in preoperatively identifying high-risk patients, providing evidence-based support for developing individualized treatment strategies and thereby optimizing clinical decision-making processes.\u003c/p\u003e\n\u003cp\u003eEarlier research has established that pathological staging serves as a crucial predictor of PSM. A comprehensive meta-analysis revealed that the incidence of positive margins for pT2, pT3, and pT4 pca were 9%, 37%, and 50%, respectively, indicating a rising trend [16]. Since pathological staging requires postoperative confirmation, this study employed preoperatively accessible clinical staging as a surrogate indicator. Multiple studies support the association between clinical staging and positive margins: Ramos et al. reported significantly higher positive margin rates for T2b tumors compared to T1c (29% vs. 20%, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), with T1c-staged patients exhibiting lower risks of biochemical recurrence [17]; Ficarra et al. confirmed in 322 RARP cases that clinical staging is an independent predictor of positive margins (HR\u0026thinsp;=\u0026thinsp;2.217, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) [18]. This may stem from advanced tumors exhibiting extensive local infiltration with indistinct boundaries from critical structures like neurovascular bundles. Intraoperative efforts to preserve function or avoid complications often compromise adequate resection margins, thereby elevating the risk of positive margins.\u003c/p\u003e\n\u003cp\u003eElevated preoperative serum creatinine is associated with an increased risk of PSM following RARP. The mechanisms may include: on one hand, elevated creatinine indicates renal insufficiency and poorer systemic condition, prompting surgeons to adopt conservative resection strategies to reduce surgical risks, thereby limiting the attainment of safe margins; on the other hand, elevated creatinine may be secondary to obstructive nephropathy caused by advanced tumors (such as ureteral compression or bladder outlet obstruction) [19, 20]. In such cases, elevated creatinine serves as an indirect marker of locally advanced or high-burden tumors, where extensive tumor invasion itself directly causes positive margins. Tollefson et al.\u0026apos;s study of 10,099 patients further confirmed that eGFR is an independently predictive factor for overall mortality in patients with PCA [21], underscoring the prognostic value of renal function indicators.\u003c/p\u003e\n\u003cp\u003eThe positive core rate is defined as the proportion of needles with cancerous tissue in the total number of needles during preoperative prostate biopsy. This indicator directly reflects the spatial distribution and invasive burden of the tumor within the prostate. Consequently, a greater rate of positive cores is linked to a heightened risk of positive surgical margins after surgery. The study carried out by Tuliao and colleagues further confirmed that the number of positive cores found in preoperative biopsies serves as a separate predictor of PSM, which is consistent with the findings of this research [22].\u003c/p\u003e\n\u003cp\u003eNeoadjuvant therapy plays a crucial role in minimizing the likelihood of PSM following RARP. Research conducted by Hu and colleagues, which included 48 patients classified as having intermediate to high-risk PCA, revealed that those in the neoadjuvant hormone therapy (NHT) cohort, who underwent treatment for durations between 2 to 12 months, showed significantly reduced rates of PSM and instances of BCR when contrasted with the control cohort that did not receive any intervention [23]. Furthermore, a prospective analysis involving 69 patients who had RP after 3 months of androgen deprivation therapy (ADT) against 72 patients who had immediate surgery revealed a markedly higher percentage of negative margins in the neoadjuvant cohort (87% versus 64%, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) along with a greater proportion of tumors classified as organ-confined (74% vs. 49%, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) [24], aligning with the results of this investigation. This is because neoadjuvant therapy induces tumor cell apoptosis, leading to tumor volume reduction, pathological downstaging, and decreased extracapsular invasion. Simultaneously, it improves surgical field exposure and anatomical visualization, facilitating the achievement of adequate safe margins during surgery and thereby reducing the risk of PSM.\u003c/p\u003e\n\u003cp\u003eIndividuals classified as having high-risk PCA are characterized by a PSA\u0026thinsp;\u0026ge;\u0026thinsp;20 ng/mL, a Gleason score\u0026thinsp;\u0026ge;\u0026thinsp;8, or a clinical stage of cT2c or more advanced. Research by D\u0026apos;Amico and associates demonstrated that the risk of mortality linked to PCA in this population is 14.2 times greater compared to those considered low-risk[25]. High-risk stratification inherently incorporates multiple aggressive features that increase surgical difficulty: higher clinical stage often indicates extracapsular invasion or seminal vesicle involvement, elevated Gleason score reflects poor tumor differentiation and high aggressiveness, while increased PSA levels signify greater tumor burden. Collectively, these factors substantially elevate the risk of PSM in high-risk patients.\u003c/p\u003e\n\u003cp\u003eThis study employed a single-center retrospective design, with cases performed by different surgeons whose varying experience and technical skills may have influenced positive margin rates. Previous research indicates that surgeons\u0026apos; learning curves, annual case volumes, and expertise levels may correlate with the risk of PSM [26]. However, the model did not account for these factors, potentially introducing bias. Furthermore, the model lacks external validation, leaving its generalizability unconfirmed. Future research should involve multicenter prospective studies that take into account surgical factors, including the experience of the surgeon, while also ensuring external validation across various healthcare institutions to improve the model\u0026apos;s clinical relevance and applicability..\u003c/p\u003e"},{"header":"6. CONCLUSION","content":"\u003cp\u003eThe ML based predictive model developed in this study effectively assesses the risk of PSM following RARP. The ANN model showed exceptional predictive capabilities, helping in the preoperative recognition of patients at high risk and offering guidance for personalized clinical decision-making.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eEthical Approval\u003c/h2\u003e \u003cp\u003e and Informed Consent: The study protocol was approved by the Ethics Committee of the Affiliated Hospital of North Sichuan Medical College (Approval No.: 2022-NSMCAH-073). As this study involved retrospective analysis, all data were anonymized and did not contain any personally identifiable patient information. Therefore, written informed consent was waived. All research procedures adhered to the ethical principles outlined in the Declaration of Helsinki.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding:\u003c/h2\u003e \u003cp\u003eThe authors declare that they received no funding, grants, or other support during the preparation of this manuscript.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAll authors contributed to the study conception and design. Writing - original draft preparation: Gen Fan; Writing - review and editing: Gen Fan, Yushui Chen; Conceptualization: Gen Fan; Methodology: Gen Fan, Yang Li; Formal analysis and investigation: Gen Fan, Yushui Chen, Yang Li; Funding acquisition: Tielong Tang; Resources: Yushui Chen, Yang Li; Supervision: Tielong Tang, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgments:\u003c/h2\u003e \u003cp\u003eNot applicable.\u003c/p\u003e \u003cp\u003ePublication Permission: Not applicable.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe study protocol was approved by the Ethics Committee of the Affiliated Hospital of North Sichuan Medical College (Approval No.: 2022-NSMCAH-073). As this study involved retrospective analysis, all data were anonymized and did not contain any personally identifiable patient information. Therefore, written informed consent was waived. All research procedures adhered to the ethical principles outlined in the Declaration of Helsinki.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eBray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A (2018) Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. Cancer J Clin 68:394\u0026ndash;424\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSung H, Ferlay J, Siegel RL et al (2021) Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. Cancer J Clin 71:209\u0026ndash;249\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCooperberg MR, Broering JM, Carroll PR (2010) Time trends and local variation in primary treatment of localized prostate cancer. J Clin oncology: official J Am Soc Clin Oncol 28:1117\u0026ndash;1123\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMartini A, Falagario UG, Villers A et al (2020) Contemporary Techniques of Prostate Dissection for Robot-assisted Prostatectomy. Eur Urol 78:583\u0026ndash;591\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHamdy FC, Donovan JL, Lane JA et al (2016) 10-Year Outcomes after Monitoring, Surgery, or Radiotherapy for Localized Prostate Cancer. N Engl J Med 375:1415\u0026ndash;1424\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePatel VR, Coelho RF, Rocco B et al (2011) Positive surgical margins after robotic assisted radical prostatectomy: a multi-institutional study. J Urol 186:511\u0026ndash;516\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarbonell E, Matheu R, Mun\u0026iacute; M et al (2022) The Effect of Adverse Surgical Margins on the Risk of Biochemical Recurrence after Robotic-Assisted Radical Prostatectomy. Biomedicines. ;10\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSu SH, Chang YH, Huang LK et al (2022) Clinical predictors for biochemical failure in patients with positive surgical margin after robotic-assisted radical prostatectomy. Tumori 108:270\u0026ndash;277\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu S, Lin SX, Wirth GJ et al (2019) Impact of Multifocality and Multilocation of Positive Surgical Margin After Radical Prostatectomy on Predicting Oncological Outcome. Clin Genitourin Cancer 17:e44\u0026ndash;e52\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLu J, Wirth GJ, Wu S et al (2012) A close surgical margin after radical prostatectomy is an independent predictor of recurrence. J Urol 188:91\u0026ndash;97\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaug CJ, Drazen JM (2023) Artificial Intelligence and Machine Learning in Clinical Medicine, 2023. N Engl J Med 388:1201\u0026ndash;1208\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDeo RC (2020) Machine Learning in Medicine: Will This Time Be Different? Circulation 142:1521\u0026ndash;1523\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiang H, Tsui BY, Ni H et al (2019) Evaluation and accurate diagnoses of pediatric diseases using artificial intelligence. Nat Med 25:433\u0026ndash;438\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuan C, Ma F, Chang S, Zhang J (2023) Interpretable machine learning models for predicting venous thromboembolism in the intensive care unit: an analysis based on data from 207 centers. 27:406 Critical care (London, England)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSong Y, Zhang D, Wang Q et al (2024) Prediction models for postoperative delirium in elderly patients with machine-learning algorithms and SHapley Additive exPlanations. Translational psychiatry 14:57\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNovara G, Ficarra V, Mocellin S et al (2012) Systematic review and meta-analysis of studies reporting oncologic outcome after robot-assisted radical prostatectomy. Eur Urol 62:382\u0026ndash;404\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRamos CG, Carvalhal GF, Smith DS, Mager DE, Catalona WJ (1999) Clinical and pathological characteristics, and recurrence rates of stage T1c versus T2a or T2b prostate cancer. J Urol 161:1525\u0026ndash;1529\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFicarra V, Novara G, Secco S et al (2009) Predictors of positive surgical margins after laparoscopic robot assisted radical prostatectomy. J Urol 182:2682\u0026ndash;2688\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLienert A, Ing A, Mark S (2009) Prognostic factors in malignant ureteric obstruction. BJU Int 104:938\u0026ndash;941\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlOmeir OK (2023) Risk of Prostate Cancer in Chronic Kidney Disease Patient: A Meta-Analysis using Observational Studies. J Pharm bioallied Sci 15:21\u0026ndash;28\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTollefson MK, Boorjian SA, Gettman MT, Rangel LJ, Bergstralh EJ, Karnes RJ (2013) Preoperative estimated glomerular filtration rate predicts overall mortality in patients undergoing radical prostatectomy. Urol Oncol 31:1483\u0026ndash;1488\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTuliao PH, Koo KC, Komninos C et al (2015) Number of positive preoperative biopsy cores is a predictor of positive surgical margins (PSM) in small prostates after robot-assisted radical prostatectomy (RARP). BJU Int 116:897\u0026ndash;904\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu JC, Hung SC, Ou YC (2017) Assessments of Neoadjuvant Hormone Therapy Followed by Robotic-Assisted Radical Prostatectomy for Intermediate- and High-Risk Prostate Cancer. Anticancer Res 37:3143\u0026ndash;3150\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCookson MS, Sogani PC, Russo P et al (1997) Pathological staging and biochemical recurrence after neoadjuvant androgen deprivation therapy in combination with radical prostatectomy in clinically localized prostate cancer: results of a phase II study. Br J Urol 79:432\u0026ndash;438\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eD'Amico AV, Whittington R, Malkowicz SB et al (1998) Biochemical outcome after radical prostatectomy, external beam radiation therapy, or interstitial radiation therapy for clinically localized prostate cancer. JAMA 280:969\u0026ndash;974\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee RS, Ma R, Pham S et al (2022) Machine Learning to Delineate Surgeon and Clinical Factors That Anticipate Positive Surgical Margins After Robot-Assisted Radical Prostatectomy. J Endourol 36:1192\u0026ndash;1198\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Positive surgical margin, prostate cancer, robotic surgery, predictive model, machine learning","lastPublishedDoi":"10.21203/rs.3.rs-8737324/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8737324/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjective\u003c/h2\u003e \u003cp\u003eThis research intends to create a predictive model utilizing machine learning(ML) techniques to evaluate the likelihood of positive surgical margins (PSM) in individuals receiving robot-assisted radical prostatectomy (RARP), thus aiding in clinical decision-making.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eA retrospective study was conducted involving 301 patients who underwent RARP. The subjects were randomly assigned to two groups in a ratio of 7:3, which included 201 individuals in the training cohort and 100 in the validation cohort. A total of twenty-four clinical and oncological characteristics were gathered, initially assessed through univariate logistic regression, and later refined using feature selection facilitated by the Boruta algorithm. Utilizing the chosen features, seven distinct ML models were developed. The effectiveness of these models was comprehensively assessed through a range of metrics, including the area beneath the ROC curve and the F1 score. To conduct a detailed analysis of how features influence the optimal model, the SHAP approach was utilized for evaluating feature contributions.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eIn the final analysis, there were 301 patients included, revealing a postoperative incidence of PSM at 42.0% after RARP. Through univariate logistic regression and the Boruta algorithm, five key predictive variables were recognized for the construction of the model. Of the seven ML models assessed, the ANN model demonstrated the best performance, achieving an AUC of 0.808 (95% CI: 0.702\u0026ndash;0.899) on the validation dataset, along with superior levels of accuracy (80.11%), sensitivity (78.9%), and F1 score (77.9%). Analysis using SHAP indicated that an advanced clinical stage, increased levels of creatinine, high-risk stratification of prostate cancer, and a greater percentage of positive biopsies were strongly linked to a heightened risk of PSM. In contrast, neoadjuvant therapy demonstrated a protective influence on the occurrence of PSM.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eMachine learning models demonstrate significant utility in predicting positive surgical margins after RARP. Integrating the random forest model with the SHAP interpretation framework enables precise prediction of individual PSM risk and provides intuitive insights into the impact of key features on predictive outcomes. This approach facilitates preoperative risk stratification and the development of early postoperative intervention strategies.\u003c/p\u003e","manuscriptTitle":"A Predictive Tool Powered by Machine Learning for Evaluating the Status of Surgical Margins After Robot-Assisted Radical Prostatectomy","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-11 08:26:55","doi":"10.21203/rs.3.rs-8737324/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e941aacc-54ea-4c83-a338-2e66971e96f8","owner":[],"postedDate":"February 11th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-02-14T01:53:58+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-11 08:26:55","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8737324","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8737324","identity":"rs-8737324","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}