Background

Prostate cancer, the most frequently diagnosed cancer in males, contributes to approximately 20% of cancer-related deaths. Among men, prostate cancer ranks as the second leading cause of death [1]. There is a lack of consensus regarding the definition of clinically significant prostate cancer [2]. However, it is commonly defined either histopathologically based on the criteria proposed by Wolters et al. [3] or radiologically [4].

The Prostate Imaging Reporting and Data System (PIRADS) was introduced by the European Society of Urogenital Radiology in 2012 as a means to provide guidance and to standardize the acquisition, interpretation, and reporting of prostate magnetic resonance imaging (MRI) [5]. In March 2019, the guidelines were revised in PIRADS version 2.1 (PIRADSv2.1), incorporating changes to enhance interobserver agreement and simplify interpretation while maintaining the fundamental acquisition and scoring guidelines [6].

Multiparametric MRI is crucial in clinical practice and is valuable for detecting prostate cancer [7, 8]. It offers high sensitivity for identifying clinically significant tumors and has been employed to guide prostate biopsies [9]. By pinpointing the location of suspicious lesions, multiparametric MRI enables real-time co-registration with transrectal ultrasound (TRUS) images, facilitating targeted biopsy using MRI-TRUS fusion guidance. This technique allows for selecting lesions more likely to contain clinically significant cancers while reducing the detection of insignificant tumors [10].

Several guidelines highlight the significance of implementing quality assurance programs for MRI of the prostate, including comparing MRI findings with pathology results [7]. This study aimed to determine the prevalence of prostate cancer at a tertiary center and establish a correlation between the findings obtained from 3T multiparametric MRI of the prostate using PIRADSv2.1 and the subsequent results of the prostate biopsy, which served as the reference standard. Furthermore, sensitivity, specificity, positive predictive value, negative predictive value, and the receiver operating characteristic (ROC) curve were calculated for 3T multiparametric MRI of the prostate using PIRADSv2.1.

Methods

Study design

A retrospective study was conducted at King Faisal Specialist Hospital and Research Center (KFSHRC) in Jeddah, Saudi Arabia, a renowned tertiary and referral center for prostate cancer. The target population was identified through an electronic search of the prostate biopsy database, and data were collected from various sources, including the hospital database and the radiology system.

We included patients who underwent a prostate biopsy between June 2019 and June 2021 and had a prostate MRI conducted at the same hospital within a 6-month window before or after the biopsy. Patients with a history of recurrent prostate cancer and those with an MRI study performed on a 1.5T machine were excluded from the study (Table 1).

Table 1 PIRADSv2.1 criteria
Full size table
Fig. 1
figure 1

These images represent PIRADS 3. T2 shows a left central zone heterogeneous lesion with an obscured margin (a). Diffusion-weighted images score less than or equal to 4 (b and c)

Full size image
Fig. 2
figure 2

These images represent PIRADS 4, as evidenced by 1.2 cm left peripheral zone lesion on T2 (a) with marked restricted diffusion (b)

Full size image
Fig. 3
figure 3

These images represent PIRADS 5, as evidenced by marked restricted diffusion (a and b). Diffusion-weighted images show marked hyperintensity (a), while the Apparent Diffusion Coefficient (ADC) map displays marked hypointensity (b). T2 reveals a 3.6 cm mass involving the left peripheral zone with intermediate signal intensity (c)

Full size image

MRI interpretation

The interpretation of all prostate MRI studies was performed by abdominal imaging radiologists. The classification of prostate nodules and image interpretation was conducted using the PIRADSv2.1 criteria [11].

Prostate biopsy

The biopsy samples were examined by the pathologist at the same hospital using the Gleason classification system [12]. Histopathology results were categorized as positive if the Gleason classification system yielded a score of 7 or higher and negative if it was 6 or lower. The MRI findings were then compared and correlated with the pathology outcomes obtained from each site (Table 2).

Table 2 Gleason classification system
Full size table

Correlation between prostate MRI and biopsy

Patient information such as age, prostate-specific antigen (PSA) level, prostate biopsy date, and the reason for the biopsy were collected. Furthermore, specific details from the prostate biopsy reports were gathered, including biopsy location, biopsy method, the number of cores obtained from each site, and the operator of the biopsy. The MRI lesion’s zonal location was classified as peripheral zone, central gland, or a combination of both. Information such as the MRI procedure date, PIRADS scores with corresponding nodule sites, and the MRI reader was collected from the record of the MRI reports.

Ethical consideration

This study was conducted in accordance with the Helsinki Declaration, and approval to conduct the study was obtained from the Research Committee at KFSHRC—Jeddah. Informed consent was waived as the study involved a retrospective chart review. To ensure the confidentiality of the subjects, access to the completed forms and the database was limited to the research team. The data were securely stored in an office and were not shared with any unauthorized individuals.

Statistical analysis

The data were coded and entered into SPSS version 26 for analysis. Descriptive statistics were used to summarize the quantitative variables, presenting means and standard deviations, while qualitative variables were described using frequencies and percentages. Sensitivity, specificity, positive predictive value, and negative predictive value were calculated. Cutoffs for the scoring systems were determined using the Youden index. The significance level was set at a p value of less than 0.05, with a 95% confidence interval (CI).

Results

We identified a total of 152 patients, out of which 63 had PIRADSv2.1 category 3 and above. However, 13 patients were excluded from the analysis as histopathology was not performed for them. As a result, only 50 patients met the criteria for inclusion in the final analysis. The prevalence of prostate cancer among the patients who underwent biopsy was determined to be 69% (95% CI 54–81%). The mean age (standard deviation [SD]) of the participants was 66.48 (7.5), and the mean PSA level (SD) was 44.5 (88.4). Further details on other variables can be found in Table 3.

Table 3 Characteristic features of participants
Full size table

Diagnostic accuracy of 3T multiparametric MRI

ROC curves were generated to evaluate the diagnostic performance of 3T multiparametric MRI of the prostate using PIRADSv2.1 for prostate cancer. The area under the ROC curves was calculated to be 0.81 (95% CI 0.68–0.95, p < 0.001, Fig. 4). The optimal cutoff value for diagnosis, determined by the Youden index, was found to be 5.

Fig. 4
figure 4

Receiver operating characteristic curves for 3T multiparametric prostate MRI using PIRADSv2.1 to diagnose prostate cancer

Full size image

The sensitivity, specificity, positive predictive value, and negative predictive value of 3T multiparametric prostate MRI using PIRADSv2.1 were determined to be 74.0%, 87.0%, 92.9%, and 59.1%, respectively.

Discussion

Currently, the standard diagnostic approach for prostate cancer involves TRUS-guided systematic prostate biopsy [13]. While this method is generally considered safe, it is an invasive procedure with certain risks. These risks include rectal bleeding, vasovagal symptoms, genitourinary system infection, hematospermia, fever, dysuria, and macroscopic hematuria [13].

On the other hand, multiparametric prostate MRI offers a noninvasive approach to detect prostate cancer. In 2012, the European Society of Urogenital Radiology developed the first standardized guidelines for prostate MRI, aiming to enhance the accuracy of radiologic interpretations [4]. However, subsequent evaluations revealed practical limitations in this system, leading to the release of PIRADSv2.0 in 2015. Further refinements and assessments resulted in the latest version, PIRADSv2.1, which was introduced in 2019 [9].

This study evaluated the diagnostic accuracy of 3T multiparametric MRI using PIRADSv2.1 for detecting prostate cancer, with prostate biopsy as a reference. Our findings indicated that at a cutoff point of a PIRADS score of 5, the overall accuracy of 3T multiparametric prostate MRI in diagnosing prostate cancer was 0.81. The corresponding sensitivity, specificity, positive predictive value, and negative predictive value were 74.0%, 87.0%, 92.9%, and 59.1%, respectively.

A previous validation study of PIRADSv2 reported an overall accuracy of 82.2%, sensitivity of 90.0%, specificity of 80.1%, positive predictive value of 83.3%, and negative predictive value of 81.8%. The study noted that the highest accuracy rate was observed for a PIRADS score of 5. [14]

In another study, the sensitivity, specificity, positive predictive value, and negative predictive value of PIRADSv2 were reported as 89.0%, 76.5%, 89.7%, 31.7%, and 98.4%, respectively [7]. When comparing our study to that previous one, we observed a similar overall accuracy, specificity, and positive predictive value. However, our study demonstrated lower sensitivity and negative predictive value.

This research has several limitations. First, it relied solely on the interpretation of a single radiologist, which may limit the generalizability of the findings. However, this approach does reflect routine practice. Second, setting a PIRADS score of 5 as the cutoff value may result in a higher incidence of clinically significant prostate cancer being missed. Third, the study focused on assessing specific zones of the gland rather than the entire prostate. Additionally, using biopsy specimens as the reference standard instead of radical prostatectomy specimens is another limitation. However, utilizing whole-mount prostate specimens for validation would have introduced a selection bias toward patients at higher risk, as only those who underwent prostatectomy would have been included. Lastly, because this study was retrospective, it cannot be ensured that the radiologists were blinded to the biopsy results.

Conclusions

PIRADSv2.1 exhibited favorable performance in diagnosing prostate cancer when compared to the gold standard of biopsy results. However, to confirm these findings, conducting larger-scale studies with a prospective design is advisable.