Skip to main content
. 2020 Jul 23;6(9):1–9. doi: 10.1001/jamaoncol.2020.2485

Table 3. Mean Absolute Difference in Gleason Pattern Quantitation Relative to Subspecialistsa.

Gleason pattern No. Subspecialist discordance, % (95% CI)
Deep learning system Pathologist
3 (Tumor-containing specimens) 498 9.2 (8.0-10.5)b 14.0 (12.4-15.6)
4 (Tumor-containing specimens) 498 10.0 (8.6-11.2)b 16.3 (14.6-18.1)
5 (Tumor-containing specimens) 498 1.5 (0.9-2.1)b 3.2 (2.2-4.3)
4 (Grade group 2 specimens only) 122 12.0 (10.4-13.6)b 22.0 (19.6-24.6)
a

Gleason pattern quantitation reflects the proportion of tumor in each biopsy specimen that is characterized as each Gleason pattern. The mean absolute differences in Gleason pattern quantitation are measured against the mean of subspecialist quantitation results for all tumor-containing biopsy specimens (rows 1-3) or grade group 2 biopsy specimens only (row 4).

b

Lower absolute differences (higher agreement rate in Gleason pattern quantitation).