Real-world evidence on the diagnostic performance of PI-RADSv2.1 using both mpMRI and bpMRI is insufficient because prospective design studies are scarce, and the results of conventional US-guided cognitive biopsies are uncertain. In the present prospective study, we confirmed that PI-RADS v2.1 in mpMRI and bpMRI could stratify prostate lesions according to their risk of csPCa, based on specimens obtained by MRI/US fusion-guided biopsy in a Japanese cohort. Although the number of lesions evaluated using bpMRI was small, the csPCa detection rate of bpMRI was comparable to that of mpMRI using PI-RADSv2.1 categories ≥ 3 or ≥ 4. Our data help solidify the evidence for the feasibility and utility of PI-RADSv2.1 using mpMRI for PCa-suspected patients and bpMRI for patients who cannot be administered a Gd-based contrast agent in clinical practice.

As observed in two studies using prospective cohorts [6, 7], the current study also showed that a higher PI-RADSv2.1 category increased the detectability of csPCa in mpMRI. The detection rate of PI-RADSv2.1 category 4 of mpMRI (60%, 43/72) and bpMRI (61%, 19/31) was relatively higher than that in previous studies (37–44%). The higher detection rate of PI-RADSv2.1 category 4 in the presented data might result from the difference in the study cohort or less years of experience of the evaluator [15].

Understanding the differences between prospective and retrospective studies on csPCa detection rates in each category is necessary for credible use of the PI-RADS system in clinical practice. Interestingly, the range of lesion-based csPCa detection rates for PI-RADSv2.1 category 5 of mpMRI in current and previous prospective studies was 65–80% [6, 7], somewhat lower than the 89% reported in a meta-analysis mainly composed of retrospective studies [5]. This indicates that the csPCa detection rate in retrospective studies might overestimate category 5 owing to problems with the relative inaccuracy of the evaluation method in the retrospective design. In retrospective studies, where analyzed groups were derived from patients who have undergone biopsies or surgeries to match pathological outcomes with lesions identified on MRI, there was a possibility that non-cancerous lesions, which might have been rated as PI-RADSv2.1 category 5 in a prospective setting, were excluded from the analysis due to escaping biopsy. As a result, the detection rate of PI-RADSv2.1 category 5 lesions in a retrospective design may be higher than that observed in a prospective design. It could be considered that more experienced evaluators might be less likely to erroneously judge non-cancerous lesions as PI-RADSv2.1 category 5 in a prospective setting. Further evidence from prospective studies reviewed by radiologists with different experiences is needed to establish the csPCa risk rate of PI-RADSv2.1 in category 5.

Although the total number of lesions evaluated by bpMRI was small, a significant trend was observed where higher PI-RADSv2.1 categories in bpMRI increased the detectability of csPCa. This result aligns closely with a previous study that demonstrated higher csPCa detection rates in higher PI-RADSv2.0 categories using bpMRI [16]. The only exception was the reversed trend in detection rates observed in PI-RADSv2.1 categories 2 (33%, 1/3) and 3 (25%, 3/12) in bpMRI, which could be attributed to the small sample size in category 2. In addition, although concerns regarding being underpowered for the detection of difference still remain, no significant differences in detection rate between mpMRI and bpMRI in lesions with PI-RADS v2.1 positive (score ≥ 3 or ≥ 4) and negative (score ≤ 2 or ≤ 3) were found. If the detection rate remains consistent, bpMRI can be considered a more efficient test than mpMRI because of its shorter testing time and lower performing cost [17]. Additionally, even if the rate of side effects to Gd-based contrast agents is low, patients would not be exposed to unnecessary risks when prostate MRI can omit DCE imaging. However, it should be noted that even with our positive bpMRI results, adopting the bpMRI approach was not perfectly acceptable. DCE imaging may prevent radiologists with less experience from missing findings, such as lesions located at the apex of the prostate and TZ lesions that can be first noticed by early enhancement. Furthermore, DCE imaging contributes to locoregional staging decisions related to treatment strategies [18]. However, in actual clinical practice, it is also true that urologists order to perform bpMRI for a variety of patient-specific reasons, even if mpMRI has advantages. In our cohort, the types of MRI were determined by the attending urologists. Patients with bpMRI have significantly lower renal function than those with mpMRI, and the patients with a history of asthma and allergy to Gd-based contrast agents underwent bpMRI. The current results showed no problems with the detection rate of csPCa using bpMRI in specific patients; however, it was premature to apply bpMRI in all patients. Two ongoing prospective multicenter clinical trials have evaluated whether bpMRI is non-inferior to mpMRI in the detection of csPCa [19, 20]. These clinical trials may provide evidence supporting the omission of Gd-based contrast agents in prostate MRI.

In this study, the likelihood of malignancy increased with increasing PI-RADSv2.1 category on pre-biopsy MRI. PI-RADSv2.1 categories and ISUP grade groups showed a correlation with Kendall tau-b rank statistics (0.42 for both), which was comparable between mpMRI and bpMRI. Walker et al. also reported a Kendall tau-b rank correlation of 0.50 between PI-RADSv2.1 categories and ISUP grade groups in their prospective cohort [6]. The increase in csPCa detection rate and malignancy with increasing categories, both in mpMRI and bpMRI, showed that the PI-RADS achieved its objective, which is the development of assessment categories that summarize levels of suspicion or risk and can be used to select patients for biopsies and beyond management (e.g., immediate interventions or observation strategy).

Our study has several limitations. First, this was a single-center study with a relatively small cohort, and all mpMRI and bpMRI scans were interpreted by a single urogenital radiologist owing to its prospective design. Although the inter-reader reliability of mpMRI and bpMRI using PI-RADS v2.1 was reported to be comparable in a retrospective study [21], further prospective evidence is needed to confirm its reproducibility. Second, mpMRI or bpMRI assignment was left to the attending physician depending on the individual patient’s situation. Except for the factors that prevented the administration of Gd-based contrast agents, there was no significant association between the representative information for suspected PCa, such as PSA, and the availability of performing DCE imaging. However, the patient’s other factors that cannot be accurately analyzed (e.g. subjectivity of the attending physician in the digital rectal exam, fluctuations in PSA levels) may have influenced how the prostate MRI was planned. As a result, there might be an unrecognizable selection bias, leading to the preferential assignment of patients whose PCa lesions were more easily depicted on MRI for bpMRI. This could result in an overestimation of the detection rate of csPCa in bpMRI. Third, because 3D volume data of the prostate were essential for conducting the MRI/US fusion-guided biopsies, T2WI was obtained using a 3D fast spin-echo sequence in the current study, not the 2D fast spin-echo sequence that is recommended for PI-RADSv2.1 assessment. Several published studies have reported that the PI-RADS category using 3D T2WI showed a diagnostic performance comparable to that of the usual PI-RADS category using 2D T2WI [22, 23]. In practice, though 3D T2WI can depict high-resolution imaging, its signal-to-noise ratio and contrast-to-noise ratio tend to be lower than those of 2D T2WI. This difference might have made it somewhat more difficult to detect csPCa lesions, particularly in the TZ, in our study. Finally, only the results of MRI/US fusion-guided biopsies were considered, and systematic biopsies in some participants may have demonstrated the diagnosis of csPCa with negative MRI/US fusion-guided biopsies. However, the interpretation of the discrepancy in results between target and systematic biopsies was complicated, and we decided to employ only the results of target biopsies to clarify the correlation of csPCa detection rates across different PI-RADSv2.1 categories of both mpMRI and bpMRI.

In conclusion, our prospective studies demonstrated that PI-RADSv2.1 using mpMRI and bpMRI could stratify the risk of csPCa, and csPCa detection rate of bpMRI in specific patients was comparable to that of mpMRI using cut-offs of PI-RADSv2.1 categories ≥ 3 or ≥ 4. PI-RADSv2.1 category 5 had lower csPCa detection rates than previous prospective studies or systemic reviews. Our findings may help PI-RADS users to properly understand the advantages and limitations of PI-RADS’s diagnostic capabilities and enable the accurate use of PI-RADSv2.1 in bpMRI with specific patients, in addition to mpMRI in clinical practice.