DISCUSSION
In this study, the correlation of prostate gland diseases with patient age, gland volume, serum PSA value and, PSA density, the effectiveness of the MpMRI and PI-RADSv2.1 scoring system in detecting prostate gland diseases, PI-RADSv2.1 and ISUP score correlation, and interobserver agreement were assessed. In our study, serum PSA did not show a significant difference for malignant or benign disease. PSAd was observed significantly high in the malignant group. Joshua Stephen Jueve et al. reported a sensitivity of 90-95% for PSAd and considering the 0.15 ng/ml2 threshold value, they supported that the high NPV may prevent unnecessary biopsy in patients with proportional PSA increase compared to prostate volume11. There was a negative correlation found between prostate volume and malignancy diagnosis. This result is similar to the results of studies by Shadi Al-Khalil et al.12 and Tang et al.13. The causes for the increase of prostate volume may be interpreted as due to benign causes like hyperplasia and prostatitis. The study of Haas et al. presented that patients with prostate cancer were of advanced ages14. Droz et al. showed high mean age in the cancer group15. In our study, the mean age in the cancer group was consistent with the literature and was higher compared to benign diseases of the prostate gland; however, the difference was not statistically significant(p:0.053). PI-RADSv2 is a scoring system widely used for the detection of PCa and its reliability has been demonstrated by numerous studies. When we examine these studies in the literature, the cut-off value for detection of clinically significant prostate cancer on MpMRI of PI-RADS 3 or 4 had ranges of 85.7-94.5% for sensitivity, 23-71% for specificity, and 34-97% and 50-92% for PPV and NPV, respectively16–21. Venderink et al.22 determined the clinically significant PCa rates (GS≥3+4) for PI-RADS-3, 4, and 5 lesions were 17%, 34%, and 67%, respectively. A study by Mathur et al. found the detection rates for clinically significant PCa were 6.1%, 33.3%, and 64.4% for PI-RADS 3, 4, and 5 respectively, and increased in proportion to the score23. A study assessing 737 lesions with MpMRI-targeted TRUS-bx found the PCa rates for PI-RADS 1, 2, 3, 4 and 5 lesions were 0%, 10%, 12%, 22% and 72% respectively24. In our study, 22.2%, 56%, and 94.45% cancers were detected in PI-RADS 3, 4, and 5 groups respectively. None of the malignant lesions in the PI-RADS3 group had ISUP>1 pathology results. As in all PI-RADS versions, disease management after scoring is not specified for patients in v2.1. In PI-RADS v2.1, ”Category 3 lesions are of intermediate status with an equivocal risk of presenting clinically significant prostate cancer (csPCa)” risk is stated and there are limited studies in the literature regarding the selection of follow-up-biopsy and it has not been clarified yet9,25. Therefore, all PI-RADS-3 lesions were biopsied with the clinician’s preference.
There was a positive correlation found between the PI-RADSv2.1 score with the ISUP score(Table-6) (p<0.001). A study by Walker et al.26 found a positive correlation between PI-RADSv2.1 categories and ISUP groups with a correlation value of 0.5 and with the increase of the PI-RADS category, clinically significant cancer rates were shown to increase. Additionally, as shown in the study by Walker et al., also in our study, in PZ when lesions with DWI score 3 upgraded to PI-RADS4 group with DCE positivity and PI-RADS4 lesions with DWI score 4 are compared, the PI-RADS4 lesions with DWI score 4 were observed to have higher ISUP grades26. These results clearly show that as the PI-RADSv2.1 score increases, the clinically-significant cancer detection rate increases, and can be interpreted as the tumors have more aggressive histopathology. However, in our study, the histopathological evaluation of lesions observed 25.47% had ISUP>1, while 40.56% of lesions had PI-RADS scores of 4 or 5. It is clear that PI-RADSv2.1 also needs improvements and more objective recommendations, and further research may contribute to achieving this aim. In our study, when cut-off values for PZ and whole gland are accepted as PI-RADS≥3, the NPV for malignancy was 100.00%. For cut-off value PI-RADS≥4 lesions, this value was 76.47% for PZ, 83.33% for TZ, and 80.65% for the whole gland, compatible with the literature27–29. The high NPV is very important in terms of excluding cancer for patients without performing bx. The sensitivity, specificity, PPV, and NPV analysis in terms of PI-RADSv2.1 sequences and zones are summarized inTable-4 andTable-5 . However, no study in the literature separately evaluated the sequences in PI-RADSv2.1. When we compare with meta-analyses performed for PI-RADSv2, the sensitivity, specificity, PPV, and NPV values for the sequences are compatible with the previous studies30. A study comparing PI-RADSv2 and 2.1 calculated the diagnostic sensitivity, specificity, PPV, and NPV for PI-RADSv2.1 according to PI-RADS≥3 are taken as positive for the detection of GS≥7 tumors according to zones as 94.3%, 24.2%, 46.1%, and 86.1% for PZ and 93.8%, 42.1%, 45% and 93% for TZ, respectively31. In our study, taking the PI-RADS score cut-off value as ≥3 positive for PZ, the sensitivity for PCa was 100%, specificity was 11.11%, PPV was 46.67% and NPV was 100.00%, similar to levels to the literature for PZ was observed. Although the PI-RADSv2 system is well standardized and expanded for MpMRI use, studies have reported interobserver agreement is highly variable from low to high32–34. A study with 3 observers by Popita et al.35 found the interobserver agreement kappa coefficient(κ) was 0.643, 0.664, and 0.568. A study which two radiologists examined 170 patients, determined the interobserver agreement for PI-RADS≥3 was substantial (all zones κ=0.63 PZ κ=0.62, TZ κ=0.53) and for PI-RADS≥4 was almost perfect (all zones κ=0.91, PZ κ=0.91, TZ κ=0.87)36. Smith et al.37found the interobserver agreement was fair with κ=0.24. The compatibility degree between experienced observers was higher for the whole gland and PZ lesions compared to the observers with moderate levels of experience. When the sequence-specific interobserver agreement is assessed, values were κ=0.24, 0.24, and 0.23 for T2W, DWI, and DCE respectively. Between two radiologists with different levels of experience, we observed moderate compatibility for the use of PI-RADSv2.1 without the cut-off value(all zones κ=0.562) and the cut-off value of PI-RADS≥4(all zones κ=0.77). Our data show that the use of the new update to PI-RADSv2.1 increases interobserver agreement with more specific definitions. Increasing observers’ experience and future PI-RADS updates will increase the agreement power between inexperienced observers or observers with similar experience.