A search for prognostic factors in cancer of the pancreatic head: the significance of the DNA ploidy pattern

A proposal for a new and more practical grading scheme for pancreatic ductal adenocarcinoma

There is no uniformly applied grading system for pancreatic ductal adenocarcinoma (DA). The scheme advocated by the WHO is essentially that of Kloppel et al, and is based on the "highest grade" focus. Although it is precise with good prognostic value, it is unfortunately not widely applied, largely because of the lack of recognition and partly because of its complex nature (interpretation of multiple parameters). Furthermore, it is fundamentally different from the one used in Japan, which evaluates the overall pattern. To establish a more widely applicable, practical, and clinically relevant grading system, a scheme similar to Gleason's scoring system was developed and tested on 112 cases of resected pancreatic DA and was compared with the WHO system. In the grading system devised, patterns (P) of infiltration were classified as follows: P1, well-defined glands with easily discernible contours; P2, fused or poorly formed glands with ill-defined contours; P3, nonglandular patterns. A score was then obtained by the summation of the predominant and the secondary patterns. Scores < or =3 (at least some well-formed glands and no nonglandular pattern) was graded as G1, 4 as G2, and > or =5 (at least some nonglandular patterns and no well-formed glandular pattern) as G3. Seventy-three percent of the cases displayed mixed patterns, with disparate patterns (P1 with P3) in 13%, confirming the high degree of heterogeneity of DA. There was a significant correlation between grade and survival, better than the correlation between survival and either the major or minor patterns evaluated separately. The median survival for G1, G2, and G3 were 22, 14, and 8 months; 1-year survival 68%, 44%, and 33%; 2-year was 67%, 11%, and 0%; and 3-year was 23%, 4%, and 0%, respectively (P = 0.0019). In a multivariate analysis, correlating survival with grade, tumor size, and lymph node status, the grade was the strongest independent predictor of survival. Odds ratio of dying of disease were 3.56 (P < 0.0001) in G3 versus G1, 1.79 (P = 0.058) in G2 versus G1, and 1.98 (P = 0.03) in G3 versus G2. Compared with this, the same odds ratio were 1.17 (P = 0.01) in tumors >2 cm versus < or =2 cm and 1.78 (P = 0.01) in cases with positive versus negative lymph nodes. The WHO grading scheme was not found to have as good a correlation with survival in this study, with WHO grade 2 showing a better survival than 1. The reproducibility of both the proposed grading system and that of WHO were found to be moderately good (with kappa values of 0.43 and 0.44, respectively), when 32 slides of DA were graded by four independent observers. The grading scheme for pancreatobiliary adenocarcinoma proposed here is highly applicable because it is practical and readily adoptable. It reflects biologic characteristics of ductal carcinoma (prominent tubule formation and tumor heterogeneity). Most importantly, it is clinically relevant with good prognostic value. Lastly, it is also applicable for use in research, by utilizing "patterns," even in small specimens like microarrays or biopsies.