Grading of invasive cribriform carcinoma on prostate needle biopsy: an interobserver study among experts in genitourinary pathology

The distinction between cribriform Gleason pattern 3 and 4 prostate cancer is controversial. Out of 3590 prostate cancers sent to one of the authors over 7 months, 30 needle biopsy cases were selected that possibly represented cribriform Gleason pattern 3 cancer. Thirty-six digital images were taken and sent to 10 experts in prostate pathology. Consensus was defined when at least 7/10 experts agreed on the grade. Sixty-seven percent (n = 24) of images reached consensus (23 pattern 4; 1 pattern 3). Of the 12 nonconsensus images, 7 were favor pattern 4 (6/10 experts agreed), 1 was favor pattern 3 (6/10 experts agreed), and 4 were equivocal ( < 6 experts agreed). The most common criteria used to call pattern 4 in the 23 consensus pattern 4 images were in frequency: irregular contour, irregular distribution of lumens, slit-like lumens, large glands, number of glands, and small lumens. In the only consensus pattern 3 image, criteria used were regular contour, small glands, regular distribution of lumens, and uniform round lumens. Discrepancy between experts was qualified as primarily objective (different criteria present) in 38%, subjective (different interpretation of the same criteria) in 12%, and mixed (both objective and subjective) in 50%. The most frequent situation with different interpretations of the same criteria were regular versus irregular contour and small versus large glands, with the former more common. Even in this highly selected set of images thought to be the best candidates for cribriform pattern 3 from a busy consult service, most experts interpreted the cribriform patterns as pattern 4. Moreover, most of the cribriform foci investigated (73%) were associated with more definitive pattern 4 elsewhere on the needle biopsy specimen. In conclusion, most of the small cribriform cancer foci seen on needle biopsy should be interpreted as Gleason pattern 4 and not pattern 3.