Development, multi-institutional external validation, and algorithmic audit of an artificial intelligence-based Side-specific Extra-Prostatic Extension Risk Assessment tool (SEPERA) for patients undergoing radical prostatectomy: a retrospective cohort study.

BACKGROUND: Accurate prediction of side-specific extraprostatic extension (ssEPE) is essential for performing nerve-sparing surgery to mitigate treatment-related side-effects such as impotence and incontinence in patients with localised prostate cancer. Artificial intelligence (AI) might provide robust and personalised ssEPE predictions to better inform nerve-sparing strategy during radical prostatectomy. We aimed to develop, externally validate, and perform an algorithmic audit of an AI-based Side-specific Extra-Prostatic Extension Risk Assessment tool (SEPERA). METHODS: Each prostatic lobe was treated as an individual case such that each patient contributed two cases to the overall cohort. SEPERA was trained on 1022 cases from a community hospital network (Trillium Health Partners; Mississauga, ON, Canada) between 2010 and 2020. Subsequently, SEPERA was externally validated on 3914 cases across three academic centres: Princess Margaret Cancer Centre (Toronto, ON, Canada) from 2008 to 2020; L'Institut Mutualiste Montsouris (Paris, France) from 2010 to 2020; and Jules Bordet Institute (Brussels, Belgium) from 2015 to 2020. Model performance was characterised by area under the receiver operating characteristic curve (AUROC), area under the precision recall curve (AUPRC), calibration, and net benefit. SEPERA was compared against contemporary nomograms (ie, Sayyid nomogram, Soeterik nomogram [non-MRI and MRI]), as well as a separate logistic regression model using the same variables included in SEPERA. An algorithmic audit was performed to assess model bias and identify common patient characteristics among predictive errors. FINDINGS: Overall, 2468 patients comprising 4936 cases (ie, prostatic lobes) were included in this study. SEPERA was well calibrated and had the best performance across all validation cohorts (pooled AUROC of 0·77 [95% CI 0·75-0·78] and pooled AUPRC of 0·61 [0·58-0·63]). In patients with pathological ssEPE despite benign ipsilateral biopsies, SEPERA correctly predicted ssEPE in 72 (68%) of 106 cases compared with the other models (47 [44%] in the logistic regression model, none in the Sayyid model, 13 [12%] in the Soeterik non-MRI model, and five [5%] in the Soeterik MRI model). SEPERA had higher net benefit than the other models to predict ssEPE, enabling more patients to safely undergo nerve-sparing. In the algorithmic audit, no evidence of model bias was observed, with no significant difference in AUROC when stratified by race, biopsy year, age, biopsy type (systematic only vs systematic and MRI-targeted biopsy), biopsy location (academic vs community), and D'Amico risk group. According to the audit, the most common errors were false positives, particularly for older patients with high-risk disease. No aggressive tumours (ie, grade >2 or high-risk disease) were found among false negatives. INTERPRETATION: We demonstrated the accuracy, safety, and generalisability of using SEPERA to personalise nerve-sparing approaches during radical prostatectomy. FUNDING: None.

The Cancer of the Bladder Risk Assessment (COBRA) score accurately predicts cancer-specific survival after radical cystectomy: external validation and lymphovascular invasion assessment value to improve its performance.

The Cancer of the Bladder Risk Assessment (COBRA) score is a predictive tool for estimating Cancer Specific Survival (CSS) after Radical Cystectomy (RC) for urothelial carcinoma. COBRA score variables are: age at RC, Tumor stage and Lymph Node Density (LND). We sought to externally validate the COBRA score and to improve its performance in estimating CSS adding Lymphovascular Invasion (LVI) as a further variable (Modified COBRA score). Clinicopathological and survival data from 789 patients who underwent RC and Pelvic Lymph Node Dissection (PLND) between January 2007 and December 2020 in two European referral centers (Paris, France and Badalona, Spain) were analyzed. COBRA score was applied to our cohort and CSS Kaplan-Meier curves were performed. Univariable and Multivariable analysis was performed in order to identify risk factors for Cancer Specific Mortality (CSM) and a score was assigned for any statistically significant risk factor; afterward, c-index calculation was performed and CCS curves have been plotted for the model after having integrated LVI variable to the COBRA score. Finally, we compared both COBRA score and Modified COBRA score models with the established AJCC model. A total of 789 patients underwent RC during the observation period. Complete data were available for 731 patients with a median follow-up of 32 months (8-47). CSM was 27.6% (no. 218 patients) at follow-up. When COBRA score was applied to our cohort, c-index was 0.76. Regression COX analysis has shown HR 0.36, CI 95% (0.16-0.83), P = .016 for patients with COBRA score 1; HR 0, CI 95% (0-1.77), P =.94 for score 2; HR 0.51, CI 95% (0.39 -0.67), P =.001 for score 3; HR 1.67, CI 95% (1.23-2.27), P =.001 for score 4; HR 2.45, CI 95% (1.51-3.99), P =.001 for score 5; HR 2.01, CI 95% (1.42-2.85), P =.001 for score 6 and HR 0.66, CI 95% (0.09-4.73), P =.682 for score 7. When the LVI variable was added to the CSS predictive model the discriminatory power increased to a c-index of 0.78. COBRA score adequately identifies those patients with a higher risk of CSM, with a c-index of 0.76. Moreover, LVI variable further improves its predictive accuracy from c-index of 0.76 to c-index of 0.78. LVI variable could be integrated in the COBRA score to optimizing prognosis stratification for patients who undergo RC.