Clinical Utility of Negative Multiparametric Magnetic Resonance Imaging in the Diagnosis of Prostate Cancer and Clinically Significant Prostate Cancer

Background Multiparametric magnetic resonance imaging (MRI) is increasingly used to diagnose prostate cancer (PCa). It is not yet established whether all men with negative MRI (Prostate Imaging-Reporting and Data System version 2 score <3) should undergo prostate biopsy or not. Objective To develop and validate a prediction model that uses clinical parameters to reduce unnecessary prostate biopsies by predicting PCa and clinically significant PCa (csPCa) for men with negative MRI findings who are at risk of harboring PCa. Design, setting, and participants This was a retrospective analysis of 200 men with negative MRI at risk of PCa who underwent prostate biopsy (2014–2020) with prostate-specific antigen (PSA) >4 ng/ml, 4Kscore of >7%, PSA density ≥0.15 ng/ml/cm3, and/or suspicious digital rectal examination. The validation cohort included 182 men from another centre (University of Miami) with negative MRI who underwent systematic prostate biopsy with the same criteria. Outcome measurements and statistical analysis csPCa was defined as Gleason grade group ≥2 on biopsy. Multivariable logistic regression analysis was performed using coefficients of logit function for predicting PCa and csPCa. Nomogram validation was performed by calculating the area under receiver operating characteristic curves (AUC) and comparing nomogram-predicted probabilities with actual rates of PCa and csPCa. Results and limitations Of 200 men in the development cohort, 18% showed PCa and 8% showed csPCa on biopsy. Of 182 men in the validation cohort, 21% showed PCa and 6% showed csPCa on biopsy. PSA density, 4Kscore, and family history of PCa were significant predictors for PCa and csPCa. The AUC was 0.80 and 0.87 for prediction of PCa and csPCa, respectively. There was agreement between predicted and actual rates of PCa in the validation cohort. Using the prediction model at threshold of 40, 47% of benign biopsies and 15% of indolent PCa cases diagnosed could be avoided, while missing 10% of csPCa cases. The small sample size and number of events are limitations of the study. Conclusions Our prediction model can reduce the number of prostate biopsies among men with negative MRI without compromising the detection of csPCa. Patient summary We developed a tool for selection of men with negative MRI (magnetic resonance imaging) findings for prostate cancer who should undergo prostate biopsy. This risk prediction tool safely reduces the number of men who need to undergo the procedure.


Introduction
Multiparametric magnetic resonance imaging (mpMRI) has emerged as a promising tool for guiding prostate biopsy decision-making. The introduction of mpMRI-targeted prostate biopsy has increased the detection of clinically significant disease and reduced the number of unnecessary biopsies and the detection of clinically indolent cancers [1][2][3][4]. The most recent European Association of Urology guidelines [5] recommend against biopsy for men with an abnormal prostate-specific antigen (PSA) and negative mpMRI findings, provided the suspected risk of aggressive cancer is low and the patient has discussed the pros and cons of forgoing biopsy with a doctor. The American Urological Association protocol for prostate MRI [6] raises concern regarding the risk of missing clinically significant prostate cancer (csPCa) on negative MRI examinations. Whether men with negative MRI findings can safely avoid unnecessary prostate biopsies remains unclear. Here we describe the development and validation of a novel risk prediction tool for PCa and csPCa among men with negative MRI. This tool will help to identify men who may safely avoid biopsy, reducing the burden of unnecessary biopsies and overtreatment.

2.
Patients and methods Patient summary: We developed a tool for selection of men with negative MRI (magnetic resonance imaging) findings for prostate cancer who should undergo prostate biopsy. This risk prediction tool safely reduces the number of men who need to undergo the procedure. calculated a matrix of correlation coefficients between the 4Kscore and these predictors. We also conducted variance inflation factor analysis (the inflation in the variance for the parameter estimates due to collinearities among predictors) to evaluate the potential presence of substantial multicollinearity between these predictors in our model.

Regression analysis
In univariate analysis, family history of PCa, PSA density, and the 4Kscore emerged as significant predictors of PCa. In the multivariable analysis, family history of PCa, PSA density,     Tables 1 and 2).

3.2.
Nomogram to estimate the risk of PCa and csPCa Figure 1A shows the nomogram built for prediction of PCa in the development cohort. Family history of PCa, PSA density, and the 4Kscore were significant contributors to the total score determining the probability of PCa in the nomogram. Figure 1B shows the nomogram built for prediction of csPCa in the development cohort. Family history of PCa, PSA density, and the 4Kscore contributed significantly to the total score.

Nomogram validation
The area under the receiver operating characteristic curve (AUC) for predicting PCa and csPCa was 0.80 and 0.87, respectively ( Fig. 2A,B). Here again, the 4Kscore, PSA density, and a family history of PCa were significant contributors to the AUC. We evaluated the nomogram calibration by comparing predicted and actual probabilities of PCa in the validation cohort. There was agreement between the predicted and actual rate of probabilities for PCa, as seen by points on the diagonal line in Figure 3.

Discussion
We developed a multivariable risk prediction tool for patients with negative MRI comprising age, family history of PCa, history of prior negative biopsy, 4Kscore, DRE finding, and PSA density to predict PCa and csPCa. Our model confers three key benefits.
(1) The model can be used to avoid a significant number of biopsies among men with negative MRI without compromising detection of csPCa.
(2) The model shows the efficacy of the 4Kscore, PSA density, and family history of PCa for predicting PCa and csPCa for men with negative MRI. (3) MRI misses 8% of csPCa cases. As diagnosis of csPCa is critical, the model addresses this gap. Our model results support the avoidance of a substantial number of biopsies without significantly missing csPCa among men with negative MRI. With the number of prostate biopsy procedures increasing every year, the complications associated with biopsy have attracted greater attention. Common nonfatal complications after biopsy include pain, bleeding, and voiding dysfunction. Postbiopsy fever and infection are less common, but can be potentially fatal complications [8] as there is increasing prevalence of biopsy-related antibiotic-resistant bacterial infections [9]. In addition, standard systematic prostate biopsy is associated with greater detection of indolent or clinically insignificant PCa [10]. Use of our model can help clinicians to reduce unnecessary biopsies among men with negative MRI.
A number of prediction calculators for diagnosing csPCa have been developed. Most of these tools include mpMRI as a variable, yet it remains unclear whether patients with negative MRI should undergo biopsy. Our findings show that men with elevated PSA density, elevated 4Kscore, and a family history of PCa should undergo prostate biopsy regardless of negative MRI findings. Elevated PSA density (!0.15 ng/ml/cm 3 ) in men with negative MRI predicts csPCa [11]. Our previously published study [12] demonstrated the significance of an elevated 4Kscore and PSA density in predicting csPCa in patients with negative MRI. Similar to our study, Buisset et al [13] showed the significance of family history of PCa and PSA density for predicting csPCa in men with negative MRI.
In the literature, the number of csPCa cases missed on negative MRI ranges from 4% to 18% [11,14]. In our development cohort, 8% of csPCa cases would have been missed without a biopsy based on negative MRI findings, which is within the current expected range for missed csPCa. The long-term prognosis for PCa with deferred treatment is well predicted by the Gleason grade. Egevad et al [15] studied 305 men diagnosed with PCa for whom there was no curative treatment. They found that mean disease-specific survival was 5-10 yr for patients with csPCa (Gleason score !7), and 16-20 yr for men with clinically indolent PCa. Moreover, the cribriform pattern on histology seen in csPCa is a strong predictor of distant metastases and disease-specific death [16], with median time to diseasespecific death of 120 mo. Clearly, we should make every attempt to diagnose csPCa.
We recognize that our study has some limitations. First, the small sample size and the number of events in the development (16 csPCa cases) and validation (11 csPCa cases) cohorts is a limitation of the study and may affect its applicability. Second, this is a retrospective, single-center study with a single biopsy expert and our outcomes may not be reproducible. In addition, the small number of events and the risk of overfitting for the proposed model with six variables may affect its generalizability and applicability. The p values were not corrected for multiple-hypothesis testing. However, our prediction tool was validated in an entirely different cohort to show the robustness of the risk estimation. Finally, we have not revised mpMRI in men with csPCa.

Conclusions
We developed an easily accessible tool to help clinicians in biopsy decision-making and in counseling patients at risk of PCa with negative MRI. Use of this novel prediction model can significantly reduce the number of biopsies without markedly missing csPCa in men with a negative MRI examination. Our results show the significance of the 4Kscore, PSA density, and family history of PCa for predicting PCa and csPCa in men with negative MRI findings. The prediction model we have developed could provide valuable information for physicians and patients in assessing an individual's risk for csPCa with negative MRI.
Author contributions: Vinayak G. Wagaskar had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Critical revision of the manuscript for important intellectual content: Wagaskar, Wiklund.
Administrative, technical, or material support: None.
Financial disclosures: Vinayak G. Wagaskar certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.