Prostate Cancer: A New Biopsy Risk Calculator using MRI
Study Background
In men, prostate cancer (PCa) is the most commonly diagnosed malignancy in over 50% of countries. Despite this statistic, a large number of prostate cancer cases are not clinically significant and are unlikely to lead to problems if left untreated. Differentiating high risk from low-risk prostate cancer remains difficult, leading to over diagnosis or treatment-associated morbidity. Currently, the cancer detection rate (CDR) of a prostate biopsy prompted by an elevated serum prostate-specific antigen (PSA) level or an unusual digital rectal examination (DRE) is about 40%, dropping to 25% in the setting of health screening programs.
To reduce the risk of unnecessary biopsy, a further risk assessment should be done by one of the following tools: i) a risk calculator (RC), ii) imaging, and iii) an additional PSA serum or urine test. In the past decade, several prospective studies have positioned multiparametric magnetic resonance imaging (mpMRI) as a promising exploratory imaging biomarker for prostate cancer. Some guidelines recommend that if mpMRI reveals lesions suspicious of prostate cancer, combined systemic and targeted therapies should be performed, whereas biopsy can be avoided when mpMRI is negative and clinical suspicion of prostate cancer is low.
To date, several RCs have been developed and validated; a few include the Prostate Cancer Prevention Trial (PCPT) RC (2006), The European Randomized Study of Screening for PCa (ERSPC) RCs (2010), and the Prostate Biopsy Collaborative Group (PBCG) RC (2018). Because the most recent PBCG RC is based on race, age, PSA, DRE, biopsy history, and family history of prostate cancer, it could be considered a gold standard risk prediction tool. However, mpMRI is not readily available in all PBCG RC settings, limiting its applicability.
In this paper, the authors aimed to develop and validate the PLUM RC incorporation mpMRI findings for the nominal outcomes of no cancer, grade group (GG) 1 prostate cancer (PCa), and clinically significant prostate cancer (csPCa). Furthermore, the authors compare the performance of the PLUM (Prospective Loyola University mpMRI) RC to the PBCG RC.
PATIENTS & METHODS
1010 men with a clinical suspicion of PCa without a negative biopsy and receiving mpMRI before biopsy at the Loyola University Medical Center were included in this study. Men with a prior history of PCa were excluded. Data from the University of Alabama at Birmingham (UAB) Prospective MRI-Targeted Prostate Biopsy Cohort was used as the external validation cohort. The primary outcomes were the diagnosis of no cancer, GG1 PCa, and csPCa (≥ GG2). Then, binary logistic regression was used to explore standard clinical variables and mpMRI parameters with the final PLUM RC. Receiver operating characteristics, calibration curves, and decision curve analysis (DCA) were evaluated in the training and validation cohorts. All statistical analyses were performed using STATA version 15.0.
RESULTS
A total of 1010 men were included from PLUM and grouped into training (n=674) and internal validation (n = 336) cohorts. About 322 patients (74.8%) in the training cohort were diagnosed with prostate cancer (≥ GG1) and 208 (30.9%) had clinically significant prostate cancer. Among the standard clinical variables, the univariable model AUC estimated was largest for PSA (61.1%), prior biopsy history (59%), and age (57.6%). The PLUM RC outperformed the PBCG RC in training (AUC: 85.9% vs. 66.0%, p<0.0001), internal validation (AUC: 88.2% vs. 67.8%, p<0.001), and external validation (AUC: 83.9% vs. 69.4%, p<0.001) cohorts for detection of csPCa. The PBCG RC was prone to overprediction, while PLUM RC was well calibrated for csPCa across all probabilities. In the external validation cohort, at a threshold probability of 15%, the PLUM vs. PBCG RC could avoid 13.8 vs. 2.7 biopsies per 100 patients without missing any csPCa. At a cost level of missing 7.5% of csPCa (92.5% of csPCa is detected), the PLUM RC could have avoided 41.0% (566/1381) of biopsies compared to 19.1% (264/1381) for PBCG. PLUM compared favorably to two other MRI-based RCs including the Stanford Prostate Cancer Calculator (SPCC) and the Rotterdam European Randomized Study of Screening for Prostate Cancer risk calculators (ERSPC-RC) for the outcome of csPCA. In the external validation cohort, AUC was higher for PLUM as compared to SPCC (n=356, AUC: 84.1% vs. 81.1%, p=0.002) and MRI-ERSPC (AUC: 84.5% vs. 82.6%, p=0.05).
DISCUSSION
The accurate risk stratification of patients with a clinical suspicion of prostate cancer is of paramount importance. Prostate cancer risk calculators incorporating PSA and clinical variables have been introduced to improve risk prediction and reduce over diagnosis and unnecessary biopsies. However, variation in performance among different populations demonstrated the limitations of traditional RCs. Recently, mpMRI incorporated RCs of the prostate has become widely accepted considering the superior performance compared with RCs without imaging data and mpMRI alone.
The PLUM RC discussed in this study demonstrated significant improvements in predicting PCa risk in patients receiving mpMRI over the PBCG RC in the training, internal validation, and external validation cohorts. Furthermore, the prediction of csPCa was better calibrated for PLUM RC in comparison to the PBCG RC. Most importantly, the PLUM RC performed favorably compared to SPCC and MRI-ERSPC RCs, the two popular MRI-based RCs developed recently.
PI-RADSv2.0 score improves early diagnosis and treatment of PCa. However, the risk of suspicious focal lesions in the prostate and seminal vesicles and the representation of csPCa differ based on age, ethnicity, and PSA. The authors found that the replacement of PSA with PSAD and using PI-RADS has provided significant discrimatory ability to the final model. Among the base variables evaluated, biopsy history and age were the most important ones. For patients with PI-RADS 4 lesions and PI-RADS 3 lesions , the risk of csPCa drops from 45% to 27% and from 15% 50 5%, respectively.
Several authors have reported and evaluated risk models incorporating mpMRI. However, they often did not consider patients without lesions on mpMRI, had relatively small cohorts, and never compared their models to another established prediction tool. In this study, the authors compared PLUM to two promising MRI-based RCs, SPCC described by Wang et al., and ERSPC described by Alberts et al. The results showed a relatively high AUC for PLUM as compared to SPCC and MRI-ERSPC. The coefficients for the PLUM RC should be applied in larger, biomarker cohorts to evaluate whether novel biomarkers provide additional value to PCa diagnosis over standard variables and mpMRI data.
This study has limitations. First, its retrospective design may have impacted confidence intervals due to variation between urologists and radiologists in PCa diagnosis. Second, the training and external validation data were from a single academic institution. External validation using large, international cohorts would be beneficial for further comparison of PLUM with MRI-ERSPC.
Conclusions
The present study pointed out that mpMRI-based PLUM RC could improve optimal detection of csPCa and reduce unnecessary biopsies.
Oncology Related Tools
Other
Latest Research
