The effect of using a morphological image scale on inter-observer variation in Gleason grading of adenocarcinoma of prostate

Introduction: The Gleason grade (GG) and grade group (Gg) are the most important prognostic indicators in histopathology report of a prostate core biopsy for adenocarcinoma. These indicators influence in determining the treatment options and prognosis. However, due to the presence of complex architectural patterns, inter observer variability (IOV) in assessing GG and Gg of prostate is unavoidable. A morphological image scale (MIS) of varied architectural patterns of GG was used in the current study to aid in reporting. The aim of the study was to assess its influence on IOV. 
Methods: The study included 101 Haematoxylin and Eosin (H&E) stained histopathological slides of transrectal ultrasound scan guided prostate core biopsies of adenocarcinoma, reported over a period of two years in a Teaching Hospital in Sri Lanka. Two general histopathologists assessed and scored the slides, according to modified Gleason grading system (MGGs), independently without the use of MIS in the first round. Reassessment and scoring were done in the second round after a period of 3 months, using the MIS. MIS was prepared using the microscopic images published in a reference article.
The inter-rater reliability was assessed using the Kappa statistics (K) to determine consistency between two investigators for Gleason sum score (GSS) and Gg separately, for first and second rounds, to assess the effect of the MIS on IOV.
Results: Exact percentage agreement on GSS achieved in the first and second rounds were 48.5% (K=0.264, p< 0.001) and 55.4% (K=0.368, p<0.001) respectively. Exact percentage agreement on Gg achieved in first and second rounds were 45.5% (K=0.273, p<0.001) and 47.5% (K=0.324, p< 0.001) respectively. 
The inter-rater reliability for GSS (55.4% vs. 48.5%, p=0.322) and Gg (47.5% vs. 45.5%, p=0.779) were found to be higher with the assistance of MIS. However, the improvement observed after introducing MIS was not statistically significant for both GSS and Gg.
Conclusions: There was a non-significant improvement in percentage inter-observer agreement with MIS assistance. Replication of the study in a larger sample is suggested to further evaluate the usefulness of MIS.