Thorgerdur Palsdottir1, Tobias Nordström2, Markus Aly3, Fredrik Jäderling4, Mark Clements1, Henrik Grönberg5, Martin Eklund6. 1. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden. 2. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden; Department of Clinical Sciences, Danderyd Hospital, Stockholm, Sweden. 3. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden; Department of Urology, Karolinska University Hospital Solna, Stockholm, Sweden. 4. Department of Molecular Medicine and Surgery, Karolinska University Hospital Solna, Stockholm, Sweden; Department of Diagnostic Radiology, Karolinska University Hospital Solna, Stockholm, Sweden. 5. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden; Department of Oncology, Capio St. Görans Sjukhus, Stockholm, Sweden. 6. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden. Electronic address: martin.eklund@ki.se.
Abstract
BACKGROUND: Risk prediction models and magnetic resonance imaging (MRI) of the prostate can reduce unnecessary biopsies and overdiagnosis of low-risk prostate cancer. However, it is unclear how these tools should be used in concert. OBJECTIVE: To develop a unified risk prediction model (S3M-MRI) that combines the Stockholm3 score (based on protein and genetic markers and clinical variables) and Prostate Imaging-Reporting and Data System v.2 scores modified for MRI without contrast (modPI-RADS). DESIGN, SETTING, AND PARTICIPANTS: We used data for 532 men from the prospective multicentre STHLM3-MRI diagnostic study to construct S3M-MRI. We compared S3M-MRI to Stockholm3 and modPI-RADS alone with respect to model discrimination, calibration, and net benefit. We also compared clinical outcomes for five diagnostic strategies according to the use of combinations of the three models. RESULTS AND LIMITATIONS: The area under the receiver operating characteristic curve (AUC) was 0.88 (95% confidence interval [CI] 0.85-0.91) for S3M-MRI, which was significantly higher (p=0.04) than for Stockholm3 (0.86, 95% CI 0.83-0.89) and modPI-RADS (0.83, 95% CI 0.79-0.87). S3M-MRI had a higher net benefit on decision curve analysis for clinically relevant probability thresholds for biopsy recommendation in comparison to Stockholm3 and modPI-RADS. However, for different diagnostic strategies, sequential use of Stockholm3 followed by MRI only for Stockholm3-positive men resulted in a similar number of unnecessary biopsies (64 vs 69) and diagnosed International Society of Urological Pathology (ISUP) grade group 1 cancers (56 vs 51) at similar sensitivity for ISUP grade group ≥2 cancers, while avoiding 38% of MRI scans. Limitations include the ethnically homogeneous study population. CONCLUSIONS: The unified S3M-MRI model was superior to the Stockholm3 model and modPI-RADS alone. However, the S3M-MRI improvement was marginal compared to sequential use of Stockholm3 followed by MRI, and resulted in 60% more MRI scans. PATIENT SUMMARY: A new risk prediction model combining clinical variables, genetic and protein biomarkers, and results from prostate magnetic resonance imaging improved the clinical outcome performance of prostate cancer diagnostics.
BACKGROUND: Risk prediction models and magnetic resonance imaging (MRI) of the prostate can reduce unnecessary biopsies and overdiagnosis of low-risk prostate cancer. However, it is unclear how these tools should be used in concert. OBJECTIVE: To develop a unified risk prediction model (S3M-MRI) that combines the Stockholm3 score (based on protein and genetic markers and clinical variables) and Prostate Imaging-Reporting and Data System v.2 scores modified for MRI without contrast (modPI-RADS). DESIGN, SETTING, AND PARTICIPANTS: We used data for 532 men from the prospective multicentre STHLM3-MRI diagnostic study to construct S3M-MRI. We compared S3M-MRI to Stockholm3 and modPI-RADS alone with respect to model discrimination, calibration, and net benefit. We also compared clinical outcomes for five diagnostic strategies according to the use of combinations of the three models. RESULTS AND LIMITATIONS: The area under the receiver operating characteristic curve (AUC) was 0.88 (95% confidence interval [CI] 0.85-0.91) for S3M-MRI, which was significantly higher (p=0.04) than for Stockholm3 (0.86, 95% CI 0.83-0.89) and modPI-RADS (0.83, 95% CI 0.79-0.87). S3M-MRI had a higher net benefit on decision curve analysis for clinically relevant probability thresholds for biopsy recommendation in comparison to Stockholm3 and modPI-RADS. However, for different diagnostic strategies, sequential use of Stockholm3 followed by MRI only for Stockholm3-positive men resulted in a similar number of unnecessary biopsies (64 vs 69) and diagnosed International Society of Urological Pathology (ISUP) grade group 1 cancers (56 vs 51) at similar sensitivity for ISUP grade group ≥2 cancers, while avoiding 38% of MRI scans. Limitations include the ethnically homogeneous study population. CONCLUSIONS: The unified S3M-MRI model was superior to the Stockholm3 model and modPI-RADS alone. However, the S3M-MRI improvement was marginal compared to sequential use of Stockholm3 followed by MRI, and resulted in 60% more MRI scans. PATIENT SUMMARY: A new risk prediction model combining clinical variables, genetic and protein biomarkers, and results from prostate magnetic resonance imaging improved the clinical outcome performance of prostate cancer diagnostics.
Authors: Katharina Beyer; Lisa Moris; Michael Lardas; Anna Haire; Francesco Barletta; Simone Scuderi; Megan Molnar; Ronald Herrera; Abdul Rauf; Riccardo Campi; Isabella Greco; Kirill Shiranov; Saeed Dabestani; Thomas van den Broeck; Sujenthiran Arun; Mauro Gacci; Giorgio Gandaglia; Muhammad Imran Omar; Steven MacLennan; Monique J Roobol; Bahman Farahmand; Eleni Vradi; Zsuzsanna Devecseri; Alex Asiimwe; Jihong Zong; Sara J Maclennan; Laurence Collette; James NDow; Alberto Briganti; Anders Bjartell; Mieke Van Hemelrijck Journal: BMJ Open Date: 2022-04-04 Impact factor: 2.692
Authors: Tae Il Noh; Chang Wan Hyun; Ha Eun Kang; Hyun Jung Jin; Jong Hyun Tae; Ji Sung Shim; Sung Gu Kang; Deuk Jae Sung; Jun Cheon; Jeong Gu Lee; Seok Ho Kang Journal: Cancer Res Treat Date: 2020-12-31 Impact factor: 4.679