BACKGROUND: Prostate‐specific antigen (PSA) screening and concomitant treatment can be implemented in several ways. The authors investigated how the net benefit of PSA screening varies between common practice versus “good practice.”METHODS: Microsimulation screening analysis (MISCAN) was used to evaluate the effect on quality‐adjusted life‐years (QALYs) if 4 recommendations were followed: limited screening in older men, selective biopsy in men with elevated PSA, active surveillance for low‐risk tumors, and treatment preferentially delivered at high‐volume centers. Outcomes were compared with a base model in which annual screening started at ages 55 to 69 years and were simulated using data from the European Randomized Study of Screening for Prostate Cancer.RESULTS: In terms of QALYs gained compared with no screening, for 1000 screened men who were followed over their lifetime, recommended good practice led to 73 life‐years (LYs) and 74 QALYs gained compared with 73 LYs and 56 QALYs for the base model. In contrast, common practice led to 78 LYs gained but only 19 QALYs gained, for a greater than 75% relative reduction in QALYs gained from unadjusted LYs gained. The poor outcomes for common practice were influenced predominantly by the use of aggressive treatment for men with low‐risk disease, and PSA testing in older men also strongly reduced potential QALY gains.CONCLUSIONS: Commonly used PSA screening and treatment practices are associated with little net benefit. Following a few straightforward clinical recommendations, particularly greater use of active surveillance for low‐risk disease and reducing screening in older men, would lead to an almost 4‐fold increase in the net benefit of prostate cancer screening. Cancer 2016;122:3386–3393. © 2016 American Cancer Society.

OBJECTIVE: To report the results from Switzerland's participation in the ERSPC from 1998; importantly, epidemiological data showed that Switzerland has one of the highest rates of morbidity and mortality from prostate cancer in the world. The local study protocol was accepted by the ethical committee and after the successful pilot study phase, the centre joined the ERSPC. SUBJECTS AND METHODS: From September 1998 to June 2003 10 300 men accepted an invitation for the study and were then randomized 1 : 1 into an active screening arm (assessed by testing prostate‐specific antigen, PSA) or a control group (no intervention). The re‐screening interval is 4 years and is ongoing (beginning in September 2002). The study protocol includes offering a prostate biopsy when the total PSA is > 3.0 ng/mL (the main study protocol in agreement with ERSPC requirements) or when the total PSA is 1–3 ng/mL and the free‐to‐total PSA ratio < 20% (side study protocol). RESULTS: During the first 3 years of screening 3562 men aged 55–70 years were screened; 395 (11.1%) of all participants had a total PSA of > 3 ng/mL and 251 (7.4%) were eligible for the side‐study. In all, 599 (17.2%) of 3562 accepted a prostate biopsy (93% of 646); 120 cases of prostate cancer were detected (3.4% detection rate). The incidence was 2.5% in main study group (positive predictive value, PPV, 24%) and 0.9% in side study group (PPV 13.6%). In radical prostatectomy specimens the cancers were mostly ‘significant’ (92% in main study group and 87% in side‐study group). CONCLUSIONS: A randomized screening study for prostate cancer is feasible in Switzerland. A longer follow‐up is needed to address within the ERSPC the primary hypothesis (that there will be a reduction in mortality in the active screening arm) and to determine the level of over/under‐diagnosis and over/under‐treatment in the active screening and control arms, respectively.

In this work, several machine learning (ML) algorithms, both classical ML and modern deep learning, were investigated for their ability to improve the performance of a pipeline for the segmentation and classification of prostate lesions using MRI data. The algorithms were used to perform a binary classification of benign and malignant tissue visible in MRI sequences. The model choices include support vector machines (SVMs), random decision forests (RDFs), and multi-layer perceptrons (MLPs), along with radiomic features that are reduced by applying PCA or mRMR feature selection. Modern CNN-based architectures, such as ConvNeXt, ConvNet, and ResNet, were also evaluated in various setups, including transfer learning. To optimize the performance, different approaches were compared and applied to whole images, as well as gland, peripheral zone (PZ), and lesion segmentations. The contribution of this study is an investigation of several ML approaches regarding their performance in prostate cancer (PCa) diagnosis algorithms. This work delivers insights into the applicability of different approaches for this context based on an exhaustive examination. The outcome is a recommendation or preference for which machine learning model or family of models is best suited to optimize an existing pipeline when the model is applied as an upstream filter.