Head-to-head Comparison of Conventional, and Image- and Biomarker-based Prostate Cancer Risk Calculators

• Published in: European urology focus 2021-05, Vol.7 (3), p.546-553
• Main Authors: Mortezavi, Ashkan, Palsdottir, Thorgerdur, Eklund, Martin, Chellappa, Venkatesh, Murugan, Sarath Kumar, Saba, Karim, Ankerst, Donna P., Haug, Erik S., Nordström, Tobias
• Format: Article
• Language: English
• Subjects: Biomarker; Biomarkers; Humans; Magnetic resonance imaging; Male; Prospective Studies; Prostate cancer; Prostate-Specific Antigen; Prostatic Neoplasms - diagnostic imaging; Prostatic Neoplasms - epidemiology; Risk Assessment - methods; Risk prediction model
• ISSN: 2405-4569; 2405-4569
• DOI: 10.1016/j.euf.2020.05.002

A new generation of risk calculators (RCs) for prostate cancer (PCa) incorporating magnetic resonance imaging (MRI) data have been introduced. However, these have not been validated externally, and their clinical benefit compared with alternative approaches remains unclear. To assess previously published PCa RCs incorporating MRI data, and compare their performance with traditional RCs (European Randomized Study of Screening for Prostate Cancer [ERSPC] 3/4 and Prostate Biopsy Collaborative Group [PBCG]) and the blood-based Stockholm3 test. RCs were tested in a prospective multicenter cohort including 532 men aged 45–74 yr participating in the Stockholm3-MRI study between 2016 and 2017. The probabilities of detection of clinically significant PCa (csPCa) defined as Gleason score ≥3 + 4 were calculated for each patient. For each RC and the Stockholm3 test, discrimination was assessed by area under the curve (AUC), calibration by numerical and graphical summaries, and clinical usefulness by decision curve analysis (DCA). The discriminative ability of MRI RCs 1–4 for the detection of csPCa was superior (AUC 0.81–0.87) to the traditional RCs (AUC 0.76–0.80). The observed prevalence of csPCa in the cohort was 37%, but calibration-in-the-large predictions varied from 14% to 63% across models. DCA identified only one model including MRI data as clinically useful at a threshold probability of 10%. The Stockholm3 test achieved equivalent performance for discrimination (AUC 0.86) and DCA, but was underpredicting the actual risk. Although MRI RCs discriminated csPCa better than traditional RCs, their predicted probabilities were variable in accuracy, and DCA identified only one model as clinically useful. Novel risk calculators (RCs) incorporating imaging improved the ability to discriminate clinically significant prostate cancer compared with traditional tools. However, all but one predicted divergent compared with actual risks, suggesting that regional modifications be implemented before usage. The Stockholm3 test achieved performance comparable with the best MRI RC without utilization of imaging. Predicted risks by novel risk calculators incorporating multiparametric magnetic resonance imaging data have a broad range, and only one model by van Leeuwen et al (van Leeuwen PJ, Hayen A, Thompson JE, et al. A multiparametric magnetic resonance imaging-based risk model to determine the risk of significant prostate cancer prior to biopsy. BJU Int 2017;120:774–81) showed c