Validation of PI-RADS Version 2 in Transition Zone Lesions for the Detection of Prostate Cancer

Abstract

Purpose To determine the association between Prostate Imaging Reporting and Data System (PI-RADS) version 2 scores and prostate cancer (PCa) in a cohort of patients undergoing biopsy of transition zone (TZ) lesions. Materials and Methods A total of 634 TZ lesions in 457 patients were identified from a prospectively maintained database of consecutive patients undergoing prostate magnetic resonance imaging. Prostate lesions were retrospectively categorized with the PI-RADS version 2 system by two readers in consensus who were blinded to histopathologic findings. The proportion of cancer detection for all PCa and for clinically important PCa (Gleason score ≥3+4) for each PI-RADS version 2 category was determined. The performance of PI-RADS version 2 in cancer detection was evaluated. Results For PI-RADS category 2 lesions, the overall proportion of cancers was 4% (one of 25), without any clinically important cancer. For PI-RADS category 3, 4, and 5 lesions, the overall proportion of cancers was 22.2% (78 of 352), 39.1% (43 of 110), and 87.8% (129 of 147), respectively, and the proportion of clinically important cancers was 11.1% (39 of 352), 29.1% (32 of 110), and 77.6% (114 of 147), respectively. Higher PI-RADS version 2 scores were associated with increasing likelihood of the presence of clinically important PCa (P < .001). Differences were found in the percentage of cancers in the PI-RADS category between PI-RADS 3 and those upgraded to PI-RADS 4 based on diffusion-weighted imaging for clinically important cancers (proportion for clinically important cancers for PI-RADS 3 and PI-RADS 3+1 were 11.1% [39 of 352] and 30.8% [28 of 91], respectively; P < .001). Conclusion Higher PI-RADS version 2 scores are associated with a higher proportion of clinically important cancers in the TZ. PI-RADS category 2 lesions rarely yield PCa, and their presence does not justify targeted biopsy.

Transition zone prostate cancer: Logistic regression and machine‐learning models of quantitative ADC, shape and texture features are highly accurate for diagnosis

“…The second subgroup analysis evaluated TZ PCa to only stromal BPH nodules for lesions that measured <15 mm in size (eg, PI‐RADS v2 assessment category 4 lesions), which have a much lower yield for clinically significant cancer at targeted biopsy compared with lesions measuring >15 mm in size . There was no difference comparing size of <15 mm TZ tumors and stromal BPH nodules (13 ± 2 mm vs. 11 ± 2 mm, P = 0.78).…”

“…We have noted that specific features of TZ PCa described in PI‐RADS v2 including lenticular shape, a noncircumscribed (blurred, indistinct, irregular, or spiculated) border and the absence of a complete hypointense rim are limited when assessing smaller lesions in the TZ, where these features may be absent or difficult to apply subjectively. This was confirmed by Thai et al, who showed a substantial disparity in the detection of significant cancer when comparing lesions measuring over or under 15 mm in size . Furthermore, the described imaging findings of TZ cancer on T 2 W and DWI in PI‐RADS v2 are subjective and may not be entirely reproducible …”

“…Studies have consistently shown substantially lower clinically significant cancer detection rates for diagnosis of TZ tumors compared with peripheral zone (PZ) tumors stratified PI‐RADS v2 assessment category at targeted biopsy . This was recently highlighted in the study by Thai et al, which reported a low proportion of clinically significant cancer diagnosis for PI‐RADS v2 assessment category 4 TZ lesions (clinically significant cancer likely to be present) of 30% compared with a moderate‐to‐high proportion of significant tumors detected for PI‐RADS v2 assessment category 5 TZ lesions (clinically significant cancer highly likely to be present) of 78% . Limitations in the diagnosis of TZ PCa using subjective assessment of mp‐MRI with PI‐RADS v2 relate to overlapping features between cancers and stromal benign prostatic hyperplasia (BPH) nodules, which may both show homogeneous low T 2 W signal intensity and restricted diffusion …”

“…[9][10][11] This was recently highlighted in the study by Thai et al, which reported a low proportion of clinically significant cancer diagnosis for PI-RADS v2 assessment category 4 TZ lesions (clinically significant cancer likely to be present) of 30% compared with a moderate-to-high proportion of significant tumors detected for PI-RADS v2 assessment category 5 TZ lesions (clinically significant cancer highly likely to be present) of 78%. 12 Limitations in the diagnosis of TZ PCa using subjective assessment of mp-MRI with PI-RADS v2 relate to overlapping features between cancers and stromal benign prostatic hyperplasia (BPH) nodules, which may both show homogeneous low T 2 W signal intensity and restricted diffusion. 9 We have noted that specific features of TZ PCa described in PI-RADS v2 including lenticular shape, a noncircumscribed (blurred, indistinct, irregular, or spiculated) border and the absence of a complete hypointense rim 13,14 are limited when assessing smaller lesions in the TZ, where these features may be absent or difficult to apply subjectively.…”

“…This was confirmed by Thai et al, who showed a substantial disparity in the detection of significant cancer when comparing lesions measuring over or under 15 mm in size. 12 Furthermore, the described imaging findings of TZ cancer on T 2 W and DWI in PI-RADS v2 are subjective and may not be entirely reproducible. 15 The use of quantitative analysis of mp-MRI may be of particular value for the diagnosis of TZ tumors.…”