Comparing the prediction of prostate biopsy outcome using the Chinese Prostate Cancer Consortium (CPCC) Risk Calculator and the Asian adapted Rotterdam European Randomized Study of Screening for Prostate Cancer (ERSPC) Risk Calculator in Chinese and European men

Development and Validation of Interpretable Machine Learning Models for Clinically Significant Prostate Cancer Diagnosis in Patients With Lesions of PI-RADS v2.1 Score ≥3

BACKGROUND

For patients with PI-RADS v2.1 ≥ 3, prostate biopsy is strongly recommended. Due to the unsatisfactory positive rate of biopsy, improvements in clinically significant prostate cancer (csPCa) risk assessments are required.

PURPOSE

To develop and validate machine learning (ML) models based on clinical and imaging parameters for csPCa detection in patients with PI-RADS v2.1 ≥ 3.

STUDY TYPE

Retrospective.

SUBJECTS

One thousand eighty-three patients with PI-RADS v2.1 ≥ 3, randomly split into training (70%, N = 759) and validation (30%, N = 324) datasets, and 147 patients enrolled prospectively for testing.

FIELD STRENGTH/SEQUENCE

3.0 T scanners/T2-weighted fast spin echo sequence and DWI with diffusion-weighted single-shot gradient echo planar imaging sequence.

ASSESSMENT

The factors evaluated for csPCa detection were age, prostate specific antigen, prostate volume, and the diameter and location of the index lesion, PI-RADSv2.1. Five ML models for csPCa detection were developed: logistic regression (LR), extreme gradient boosting, random forest (RF), decision tree, and support vector machines. The csPCa was defined as Gleason grade ≥2.

STATISTICAL TESTS

Univariable and multivariable LR analyses to identify parameters associated with csPCa. Area under the receiver operating characteristic curve (AUC), Brier score, and DeLong test were used to assess and compare the csPCa diagnostic performance with the LR model. The significance level was defined as 0.05.

RESULTS

The RF model exhibited the highest AUC (0.880-0.904) and lowest Brier score (0.125-0.133) among the ML models in the validation and testing cohorts, however, there was no difference when compared to the LR model (P = 0.453 and 0.548). The sensitivity and negative predictive values in the validation and testing cohorts were 93.8%-97.6% and 82.7%-95.1%, respectively, at a threshold of 0.450 (99% sensitivity of the RF model).

DATA CONCLUSION

The RF model might help for assessing the risk of csPCa and preventing overdiagnosis and unnecessary biopsy for men with PI-RADSv2.1 ≥ 3.

EVIDENCE LEVEL

3 TECHNICAL EFFICACY: Stage 2.

Comparison of Rotterdam and Barcelona Magnetic Resonance Imaging Risk Calculators for Predicting Clinically Significant Prostate Cancer

Background

Magnetic resonance imaging (MRI)-based risk calculators (MRI-RCs) individualise the likelihood of clinically significant prostate cancer (csPCa) and improve candidate selection for prostate biopsy beyond the Prostate Imaging Reporting and Data System (PI-RADS).

Objective

To compare the Barcelona (BCN) and Rotterdam (ROT) MRI-RCs in an entire population and according to the PI-RADS categories.

Design setting and participants

A prospective comparison of BCN- and ROT-RC in 946 men with suspected prostate cancer in whom systematic biopsy was performed, as well as target biopsies of PI-RADS ≥3 lesions.

Outcome measurements and statistical analysis

Saved biopsies and undetected csPCa (grade group ≥2) were determined.

Results and limitations

The csPCa detection was 40.8%. The median risks of csPCa from BCN- and ROT-RC were, respectively, 67.1% and 25% in men with csPCa, whereas 10.5% and 3% in those without csPCa (p < 0.001). The areas under the curve were 0.856 and 0.844, respectively (p = 0.116). BCN-RC showed a higher net benefit and clinical utility over ROT-RC. Using appropriate thresholds, respectively, 75% and 80% of biopsies were needed to identify 50% of csPCa detected in men with PI-RADS <3, whereas 35% and 21% of biopsies were saved, missing 10% of csPCa detected in men with PI-RADS 3. BCN-RC saved 15% of biopsies, missing 2% of csPCa in men with PI-RADS 4, whereas ROT-RC saved 10%, missing 6%. No RC saved biopsies without missing csPCa in men with PI-RADS 5.

Conclusions

ROT-RC provided a lower and narrower range of csPCa probabilities than BCN-RC. BCN-RC showed a net benefit over ROT-RC in the entire population. However, BCN-RC was useful in men with PI-RADS 3 and 4, whereas ROT-RC was useful only in those with PI-RADS 3. No RC seemed to be helpful in men with negative MRI and PI-RADS 5.

Patient summary

Barcelona risk calculator was more helpful than Rotterdam risk calculator to select candidates for prostate biopsy.

Magnetic Resonance Imaging-Based Predictive Models for Clinically Significant Prostate Cancer: A Systematic Review

Simple Summary

Magnetic resonance imaging (MRI) has allowed the early detection of PCa to evolve towards clinically significant PCa (csPCa), decreasing unnecessary prostate biopsies and overdetection of insignificant tumours. MRI identifies suspicious lesions of csPCa, predicting the semi-quantitative risk through the prostate imaging report and data system (PI-RADS), and enables guided biopsies, increasing the sensitivity of csPCa. Predictive models that individualise the risk of csPCa have also evolved adding PI-RADS score (MRI-PMs), improving the selection of candidates for prostate biopsy beyond the PI-RADS category. During the last five years, many MRI-PMs have been developed. Our objective is to analyse the current developed MRI-PMs and define their clinical usefulness through a systematic review. We have found high heterogeneity between MRI technique, PI-RADS versions, biopsy schemes and approaches, and csPCa definitions. MRI-PMs outperform the selection of candidates for prostate biopsy beyond MRI alone and PMs based on clinical predictors. However, few developed MRI-PMs are externally validated or have available risk calculators (RCs), which constitute the appropriate requirements used in routine clinical practice.

Abstract

MRI can identify suspicious lesions, providing the semi-quantitative risk of csPCa through the Prostate Imaging-Report and Data System (PI-RADS). Predictive models of clinical variables that individualise the risk of csPCa have been developed by adding PI-RADS score (MRI-PMs). Our objective is to analyse the current developed MRI-PMs and define their clinical usefulness. A systematic review was performed after a literature search performed by two independent investigators in PubMed, Cochrane, and Web of Science databases, with the Medical Subjects Headings (MESH): predictive model, nomogram, risk model, magnetic resonance imaging, PI-RADS, prostate cancer, and prostate biopsy. This review was made following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) criteria and studied eligibility based on the Participants, Intervention, Comparator, and Outcomes (PICO) strategy. Among 723 initial identified registers, 18 studies were finally selected. Warp analysis of selected studies was performed with the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool. Clinical predictors in addition to the PI-RADS score in developed MRI-PMs were age, PCa family history, digital rectal examination, biopsy status (initial vs. repeat), ethnicity, serum PSA, prostate volume measured by MRI, or calculated PSA density. All MRI-PMs improved the prediction of csPCa made by clinical predictors or imaging alone and achieved most areas under the curve between 0.78 and 0.92. Among 18 developed MRI-PMs, 7 had any external validation, and two RCs were available. The updated PI-RADS version 2 was exclusively used in 11 MRI-PMs. The performance of MRI-PMs according to PI-RADS was only analysed in a single study. We conclude that MRI-PMs improve the selection of candidates for prostate biopsy beyond the PI-RADS category. However, few developed MRI-PMs meet the appropriate requirements in routine clinical practice.

External validation of two mpMRI-risk calculators predicting risk of prostate cancer before biopsy

Purpose

Risk calculators (RC) aim to improve prebiopsy risk stratification. Their latest versions now include multiparametric magnetic resonance imaging (mpMRI) findings. For their implementation into clinical practice, critical external validations are needed.

Methods

We retrospectively analyzed the patient data of 554 men who underwent ultrasound-guided targeted and systematic prostate biopsies at 2 centers. We validated the mpMRI-RCs of Radtke et al. (RC-R) and Alberts et al. (RC-A), previously shown to predict prostate cancer (PCa) and clinically significant PCa (csPCa). We assessed these RCs’ prediction accuracy by analyzing the receiver-operating characteristics (ROC) curve and evaluated their clinical utility using Decision Curve Analysis (DCA), including Net-Benefit and Net-Reduction curves.

Results

We found that the Area Under the ROC Curve (AUC) for predicting PCa was 0.681 [confidence interval (CI) 95% 0.635–0.727] for RC-A. The AUCs for predicting csPCa were 0.635 (CI 95% 0.583–0.686) for RC-A and 0.676 (CI 95% 0.627–0.725) for RC-R. For example, at a risk threshold of 12%, RC-A needs to assess 334 and RC-R 500 patients to detect one additional true positive PCa or csPCa patient, respectively. At the same risk threshold of 12%, RC-A only needs to assess 6 and RC-R 16 patients to detect one additional true negative PCa or csPCa patient.

Conclusion

The mpMRI-RCs, RC-R and RC-A, are robust and valuable tools for patient counseling. Although they do not improve PCa and csPCa detection rates by a clinically meaningful margin, they aid in avoiding unnecessary prostate biopsies. Their implementation could reduce overdiagnosis and reduce PCa screening morbidity.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00345-022-04119-8.

Comparative Analysis of PSA Density and an MRI-Based Predictive Model to Improve the Selection of Candidates for Prostate Biopsy

This study is a head-to-head comparison between mPSAD and MRI-PMbdex. The MRI-PMbdex was created from 2432 men with suspected PCa; this cohort comprised the development and external validation cohorts of the Barcelona MRI predictive model. Pre-biopsy 3-Tesla multiparametric MRI (mpMRI) and 2 to 4-core transrectal ultrasound (TRUS)-guided biopsies for suspicious lesions and/or 12-core TRUS systematic biopsies were scheduled. Clinically significant PCa (csPCa), defined as Gleason-based Grade Group 2 or higher, was detected in 934 men (38.4%). The area under the curve was 0.893 (95% confidence interval [CI]: 0.880−0.906) for MRI-PMbdex and 0.764 (95% CI: 0.774−0.783) for mPSAD, with p < 0.001. MRI-PMbdex showed net benefit over biopsy in all men when the probability of csPCa was greater than 2%, while mPSAD did the same when the probability of csPCa was greater than 18%. Thresholds of 13.5% for MRI-PMbdex and 0.628 ng/mL2 for mPSAD had 95% sensitivity for csPCa and presented 51.1% specificity for MRI-PMbdex and 19.6% specificity for mPSAD, with p < 0.001. MRI-PMbdex exhibited net benefit over mPSAD in men with prostate imaging report and data system (PI-RADS) <4, while neither exhibited any benefit in men with PI-RADS 5. Hence, we can conclude that MRI-PMbdex is more accurate than mPSAD for the proper selection of candidates for prostate biopsy among men with suspected PCa, with the exception of men with a PI-RAD S 5 score, for whom neither tool exhibited clinical guidance to determine the need for biopsy.

Prostate Cancer Risk Calculators for Healthy Populations: Systematic Review

Background

Screening for prostate cancer has long been a debated, complex topic. The use of risk calculators for prostate cancer is recommended for determining patients’ individual risk of cancer and the subsequent need for a prostate biopsy. These tools could lead to better discrimination of patients in need of invasive diagnostic procedures and optimized allocation of health care resources

Objective

The goal of the research was to systematically review available literature on the performance of current prostate cancer risk calculators in healthy populations by comparing the relative impact of individual items on different cohorts and on the models’ overall performance.

Methods

We performed a systematic review of available prostate cancer risk calculators targeted at healthy populations. We included studies published from January 2000 to March 2021 in English, Spanish, French, Portuguese, or German. Two reviewers independently decided for or against inclusion based on abstracts. A third reviewer intervened in case of disagreements. From the selected titles, we extracted information regarding the purpose of the manuscript, analyzed calculators, population for which it was calibrated, included risk factors, and the model’s overall accuracy.

Results

We included a total of 18 calculators from 53 different manuscripts. The most commonly analyzed ones were the Prostate Cancer Prevention Trial (PCPT) and European Randomized Study on Prostate Cancer (ERSPC) risk calculators developed from North American and European cohorts, respectively. Both calculators provided high diagnostic ability of aggressive prostate cancer (AUC as high as 0.798 for PCPT and 0.91 for ERSPC). We found 9 calculators developed from scratch for specific populations that reached a diagnostic ability as high as 0.938. The most commonly included risk factors in the calculators were age, prostate specific antigen levels, and digital rectal examination findings. Additional calculators included race and detailed personal and family history.

Conclusions

Both the PCPR and ERSPC risk calculators have been successfully adapted for cohorts other than the ones they were originally created for with no loss of diagnostic ability. Furthermore, designing calculators from scratch considering each population’s sociocultural differences has resulted in risk tools that can be well adapted to be valid in more patients. The best risk calculator for prostate cancer will be that which has been calibrated for its intended population and can be easily reproduced and implemented.

Trial Registration

PROSPERO CRD42021242110; https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=242110