Thousands of men take a PSA test to investigate whether they have prostate cancer every year. For the vast majority, the test is normal and further investigations are not necessary. In others, the test is sufficiently elevated that men are referred for further investigations. Most men with an elevated PSA are offered an MRI examination of the prostate gland, and for some, a tissue sample of the prostate is also recommended if the suspicion of cancer is high enough. Although this comprehensive investigation reveals most clinically significant , or dangerous cases of cancer, many indolent, or "harmless" cancer cases are also detected, which would not have caused the man any harm during his lifetime (approximately 20% of all prostate cancer diagnoses diagnosed in current clinical practive). Cancer treatment is not recommended for such cases, but for many men, the diagnosis and subsequent follow-up can cause him and his family anxity and concern. In addition, prostate biopsies are unpleasant for the patient and the investigation process is resource-intensive for both the man and the health care service.

Risk stratification uses machine learning methods to better identify the men who require further investigations with MRI and tissue samples. In this project, the investigators investigate whether the best risk stratification tools are non-inferior in detecting clinically significant prostate cancer compared to current practice, and whether they lead to fewer tissue samples, MRI scans, less health anxiety, and better cost-effectiveness.

Background Prostate cancer (PCa) is the most common non-skin cancer among men. Most PCa cases are identified after an elevated prostate-specific antigen (PSA) level has been identified on blood testing. In most men this leads to a hospital referral, magnetic resonance imaging (MRI) of the prostate, and biopsy (tissue sampling) of the prostate in men where MRI is suggestive of a prostate cancer.

Although the PSA-test is the main method for identifying men who have prostate cancer, the test has low specificity, meaning that many men with an elevated PSA-level do not harbour PCa that will cause the man harm within his lifetime. In current clinical practice, such "false" PSA elevations can be difficult to distinguish from the "true" elevations found in men with clinically significant PCa (csPCa, Gleason >3+3) and therefore still lead to an MRI, and in many cases even a biopsy procedure is necessary to exclude csPCa.

Multivariable risk stratification uses the statistical methods of machine learning to quantify a man's probability of having prostate cancer based on other information that is available about the patient, which is known to affect the likelihood of him having prostate cancer. The aim of such stratification is to improve the identification of men who are likely to have prostate cancer, beyond the capacity of the unspecific PSA test, with its many false positives. This in order to improve identification of men who actually need further diagnostic testing, and to better distinguish these men from those men where the probability of significant disease is so low that further investigations can safely be omitted, without missing cases of significant disease.

Although such risk stratification tools have previously been developed, are readily available, and in retrospective studies have been demonstrated to reduce unnecessary MRIs and biopsies considerably, clinical uptake continues to be low due to the lack of prospective trials demonstrating efficacy and safety. As a result, at present, almost all men referred with a suspicion of prostate cancer undergo MRI and subsequent clinical assessment for biopsy. In health care settings where PCa-screening is implemented, as is a stated goal in the EU Council cancer screening guidelines, the number of men requiring diagnostic work-up will likely increase dramatically in ensuing years.

Knowledge gaps The risk calculators (RCs) stemming from the European Randomised Study of Screening for Prostate Cancer (ERSPC) are the currently best known and most utilised multivariable risk stratification tools. Their use has been retrospectively validated in multiple studies. All presently available development, and almost all validating studies are, however, hampered by their retrospective design and varying size which presently hinder widespread clinical implementation of this promising new method. In the only prospective study known to us, 47% of MRI investigations could be avoided compared with today's practice of performing an MRI in all men referred with an elevated PSA (>3ng/ml). However, this study was not designed to answer the primary research question of the present proposal, namely whether this RC approach gives csPCa detection that is non-inferior to performing an MRI in all referred men. As a result, although multivariable risk stratification of men with suspected PCa holds promise, whether the method gives non-inferior cancer detection to current practice represents a key knowledge gap which presently limits its implementation.

Study aims The primary objective of the current proposal is to perform the first prospective, adequately powered investigation of whether risk stratification using the risk stratification calculators from the European Randomised Study of Screening for Prostate Cancer study gives non-inferior detection of clinically significant prostate cancer compared with performing an MRI in all men referred with a suspicion of prostate cancer.

How clinical decision-making in men suspected of harbouring PCa affects men's health-related quality of life also has not been thoroughly investigated. And similarly, whether multivariable risk stratification is in fact cost-efficient compared with an MRI-all approach has not been conclusively determined in a prospective, clinical trial. The secondary aims of the study are to examine how the decision of whether to perform biopsy affects the health-related quality of life of men who undergo such consideration, and to prospectively determine the cost-efficiency of performing multivariable risk stratification, compared with performing an MRI in all men referred with a suspicion of prostate cancer.

Inclusion Criteria:

• more than 10 years remaining life expectancy
• Suspected localized prostate cancer
• Suspicious DRE (cT2)
• PSA 3-20 ng/ml

Exclusion Criteria:

• cT3 and/or cT4 (on DRE)
• PSA >20 ng/ml
• Prior diagnosis of prostate cancer
• Contraindications to MRI or to prostate biopsy
• Medications known to affect serum PSA levels

Sharing of anonmous data for scientific purposeswill be considered upon reasonable request

The care provider is blinded to information from the MRI when performing a DRE and calculating risk by the ERSPC 3/4 risk calculators.