Likert vs PI‐RADS v2: a comparison of two radiological scoring systems for detection of clinically significant prostate cancer

Abstract

Objective To compare the clinical validity and utility of Likert assessment and the Prostate Imaging Reporting and Data System (PI‐RADS) v2 in the detection of clinically significant and insignificant prostate cancer. Patients and Methods A total of 489 pre‐biopsy multiparametric magnetic resonance imaging (mpMRI) scans in consecutive patients were subject to prospective paired reporting using both Likert and PI‐RADS v2 by expert uro‐radiologists. Patients were offered biopsy for any Likert or PI‐RADS score ≥4 or a score of 3 with PSA density ≥0.12 ng/mL/mL. Utility was evaluated in terms of proportion biopsied, and proportion of clinically significant and insignificant cancer detected (both overall and on a ‘per score’ basis). In those patients biopsied, the overall accuracy of each system was assessed by calculating total and partial area under the receiver‐operating characteristic (ROC) curves. The primary threshold of significance was Gleason ≥3 + 4. Secondary thresholds of Gleason ≥4 + 3, Ahmed/UCL1 (Gleason ≥4 + 3 or maximum cancer core length [CCL] ≥6 or total CCL≥6) and Ahmed/UCL2 (Gleason ≥3 + 4 or maximum CCL ≥4 or total CCL ≥6) were also used. Results The median (interquartile range [IQR]) age was 66 (60–72) years and the median (IQR) prostate‐specific antigen level was 7 (5–10) ng/mL. A similar proportion of men met the biopsy threshold and underwent biopsy in both groups (83.8% [Likert] vs 84.8% [PI‐RADS v2]; P = 0.704). The Likert system predicted more clinically significant cancers than PI‐RADS across all disease thresholds. Rates of insignificant cancers were comparable in each group. ROC analysis of biopsied patients showed that, although both scoring systems performed well as predictors of significant cancer, Likert scoring was superior to PI‐RADS v2, exhibiting higher total and partial areas under the ROC curve. Conclusions Both scoring systems demonstrated good diagnostic performance, with similar rates of decision to biopsy. Overall, Likert was superior by all definitions of clinically significant prostate cancer. It has the advantages of being flexible, intuitive and allowing inclusion of clinical data. However, its use should only be considered once radiologists have developed sufficient experience in reporting prostate mpMRI.

Does prostate magnetic resonance imaging (MRI) reporting system affect performance of MRI in men with a clinical suspicion of prostate cancer?

“…Overall, this retrospective analysis by Khoo et al , comparing Likert scores reported using clinical variables vs PIRADSv2.0, provides further evidence that good quality prostate MRI can be used as a risk‐stratification and biopsy targeting tool in men with a clinical suspicion of prostate cancer. Each centre needs to develop its own quality control process and continually review its own performance measures of prostate MRI and MRI‐targeted biopsy.…”

“…In this issue of the BJUI , Khoo et al retrospectively analysed reports from a multicentre prostate cancer pathway registry, Rapid Assessment and Prostate Imaging for Diagnosis (RAPID). Men with a clinical suspicion of prostate cancer were enrolled based on various clinical criteria such as: age, performance status, and PSA level.…”

Clinico-pathological Outcomes of Men with Initial Likert 2 Multiparametric Magnetic Resonance Imaging of the Prostate: Findings from a Case Series in a Non-teaching Hospital in UK

“…All MP-MRI imaging in this study was based on the LIKERT scoring system. There have been some recent studies comparing the LIKERT system with the PI-RADS system, and perhaps this is an area we could look to develop this current work [12,13].…”

Contrast Medium or No Contrast Medium for Prostate Cancer Diagnosis. That Is the Question

“…In order to address these multiple sources of heterogeneity, the Cochrane Handbook for Systematic Reviews of Interventions suggests that linked receiver operating characteristic (ROC) comparisons be undertaken to get a sense of the change in accuracy within studies where paired comparisons are performed, in order to note the consistency of observations 37 . The Niu et al 18 meta‐analysis compared head‐to‐head studies using linked ROC analyses for the detection of all prostate cancers in 11 of 33 studies with paired data. They showed that mpMRI had a modestly higher pooled sensitivity (85; 95% CI, 78–93) than bpMRI (80%; 95% CI, 71–90) ( P = 0.01) due to trends observed in 7 of the 11 studies included.…”

“…For example, when mpPI‐RADS category 2 lesions are upgraded into Likert category 4 on the basis of focal enhancement without corresponding focal DWI abnormalities 17 . These detection and categorization uses of DCE‐MRI can lead to seemingly different results between PI‐RADS and Likert systems of mpMRI data, 17,18 and are the major reasons for discussions among supporters and opponents of DCE‐MRI usage.…”