Mynderse, Lance

Objective To define the natural history, patterns of recurrence and treatment modalities for local prostate cancer (PCa) recurrence following radical prostatectomy (RP) and radiation therapy (RT), and to investigate factors that could predict metastasis‐free survival (MFS) in this unique patient population. Methods We queried a prospectively maintained PCa registry to identify men developing in‐field recurrence (IFR) following RP and RT from 2008 to 2021 at a single institution. IFR was defined as biopsy‐proven recurrent PCa or the presence of persistent positron emission tomography‐avid lesions in the prior radiation field without evidence of metastasis. Cox regression was conducted to determine predictors of MFS. Kaplan–Meier methods were used to calculate MFS, cancer‐specific survival (CSS) and overall survival (OS) for patients in three primary therapy categories: cryoablation, androgen deprivation therapy (ADT) alone, and surveillance. Results Of 4575 patients from our registry, 108 (2.3%) with IFR were identified. The median (interquartile range [IQR]) time to IFR from salvage treatment was 78 (50–126) months. A total of 29 patients (26%) were managed with cryoablation, 23 (21%) with ADT, and 28 (25%) with surveillance. The median (IQR) follow‐up was 76 (48–100) months. There were no statistically significant differences in MFS ( P = 0.67) or OS ( P = 0.07) among the three primary treatment cohorts. Patients treated with ADT or cryoablation had longer CSS compared to patients managed with surveillance ( P = 0.047). Conclusions We found that IFR may present years after completion of primary treatment for PCa. While curative management strategies may be attempted, local and distant metastatic recurrence is common and often requires systemic therapy.

PURPOSE Literature on clinical note mining has highlighted the superiority of machine learning (ML) over hand-crafted rules. Nevertheless, most studies assume the availability of large training sets, which is rarely the case. For this reason, in the clinical setting, rules are still common. We suggest 2 methods to leverage the knowledge encoded in pre-existing rules to inform ML decisions and obtain high performance, even with scarce annotations. METHODS We collected 501 prostate pathology reports from 6 American hospitals. Reports were split into 2,711 core segments, annotated with 20 attributes describing the histology, grade, extension, and location of tumors. The data set was split by institutions to generate a cross-institutional evaluation setting. We assessed 4 systems, namely a rule-based approach, an ML model, and 2 hybrid systems integrating the previous methods: a Rule as Feature model and a Classifier Confidence model. Several ML algorithms were tested, including logistic regression (LR), support vector machine (SVM), and eXtreme gradient boosting (XGB). RESULTS When training on data from a single institution, LR lags behind the rules by 3.5% (F1 score: 92.2% v 95.7%). Hybrid models, instead, obtain competitive results, with Classifier Confidence outperforming the rules by +0.5% (96.2%). When a larger amount of data from multiple institutions is used, LR improves by +1.5% over the rules (97.2%), whereas hybrid systems obtain +2.2% for Rule as Feature (97.7%) and +2.6% for Classifier Confidence (98.3%). Replacing LR with SVM or XGB yielded similar performance gains. CONCLUSION We developed methods to use pre-existing handcrafted rules to inform ML algorithms. These hybrid systems obtain better performance than either rules or ML models alone, even when training data are limited.