Old family photos going back two or more generations are often treasured mementos that lend depth and clarity to explorations into one’s ancestry. Inevitably, there will be some where the focus is not crisp, or there’s been water damage that obscures the image, raising the question, “Who is this person?” A magnifying glass may reveal some details, but it’s not unusual to consult with other family members to get their opinion. A sister says, “Oh, that’s daddy’s great uncle Samuel,” while a cousin opines, “I’m pretty sure it’s great uncle Samuel’s business partner when they opened a diner.” There’s agreement on the era, but divergence of opinion means a confident identity can’t be achieved.
Numerous clinical studies involving multiparametric MRI of the prostate are designed to evaluate what is called interobserver agreement. This means that ideally, two or more radiologists (observers) who interpret the same prostate scan would arrive at the same conclusion. Often this is not the case. One would expect that the more specialized and experienced the observer, the more authoritative the merits of his/her opinion—and generally, this is the case. However, what happens if two experts have different opinions?
Historically, lack of interobserver agreement was a bigger challenge nearly a decade ago, when multiparametric MRI (mpMRI) of the prostate was still proving itself to be a game changer in detecting and diagnosing prostate cancer (PCa). Radiologists a reference, or lexicon, that would provide consistent guidelines and a common language for reading prostate mpMRI. A standard taken from breast imaging was already in existence. It’s called the Breast Imaging-Reporting and Data System (BI-RADS), first created in 1993 but refined and improved with later versions in 1995, 1998 and 2003. It served as a model for today’s PI-RADS.
PI-RADS was published in 2013 after nearly 6 years of committee work to design it and gain consensus. Like BI-RADS, it also has later versions that significantly improved interobserver agreement, and the latest is PI-RADS v2.1 (see this post for more information).
New study tests accuracy of PI-RADS v2.1
While interobserver agreement is of key importance, it’s relatively meaningless if the imaging itself is inaccurate. The latest version 2.1 improved technical specifications for 3 parameters, clarified measurement of gland volume, revised the mapping of prostate zones (peripheral, transition and central) and other revisions. Therefore, Bhayana et al. (2020)[i] carried out a study with the goal of comparing PI-RADS v2.1 with the earlier PI-RADS v2 in terms of interobserver agreement in the peripheral zone (PZ) and transition zone (TZ), and to test their diagnostic accuracy in both zones.
For the evaluation of interobserver agreement, they had 6 radiologists with different levels of experience. Each of them had a separate session using v2 and v2.1 to assess 40 patient MRIs with PCa in the PZ, and 40 with PCa in the TZ. For accuracy, they compared the radiologists’ evaluations of 50 MRIs for which there were also pathology records (tissue samples).
After all analyses were complete, the study team concluded that PI-RADS v2.1 improved interobserver agreement in the PZ but not the TZ (note about 80% of PCa tumors occur in the PZ). As for more accurate diagnosis, v2.1 increased accuracy more for experienced readers, and only in the PZ.
On balance, the authors found that v2.1 offered similar to or slightly better than v2 performance, with the most striking improvement occurring in the PZ accuracy—and only for more experienced readers.
To return to the analogy of old family pictures, it’s a little like saying that the best people to identify murky photos of ancestors would be those family members who had the most familiarity with or memories of them, likely to be the older living family members. We would expect that their ability to name the subjects in the photos would be better than young siblings or cousins who never knew the people in the pictures. The takeaway is, anyone interested in figuring out visual images of the family tree should turn to the most experienced family members.
In a similar way, a patient with suspected PCa who undergoes an mpMRI of the prostate should turn to an experienced reader, even for a second opinion of his image CD. For best results, the radiologist reading the image should be familiar with PI-RADS v2.1 as well as being an “old hand” at interpreting prostate mpMRI.
NOTE: This content is solely for purposes of information and does not substitute for diagnostic or medical advice. Talk to your doctor if you are experiencing pelvic pain, or have any other health concerns or questions of a personal medical nature.
[i] Bhayana R, O’Shea A, Anderson MA, Bradley WR et al. PI-RADS versions 2 and 2.1: Interobserver Agreement and Diagnostic Performance in Peripheral and Transition Zone Lesions Among Six Radiologists. AJR Am J Roentgenol. 2020 Sep 9.