Reproducibility in retrospective grading of acute graft-versus-host disease after allogeneic marrow transplantation.

We have undertaken a formal study to evaluate the reproducibility of retrospective assessments for grading the severity of acute GVHD. Using criteria previously established by the Seattle group, three reviewers independently assigned GVHD severity grades for a set of 100 marrow transplant patients. Significant differences were found in the distribution of GVHD grades assigned by one of the reviewers as compared to the other two reviewers. In only 40% of cases did all three reviewers assign the same GVHD grade, and in only 68-71% of cases did all three reviewers assign the same grade within 0-I vs II-IV or 0-II vs III-IV categories. Despite the high rate of disagreement between any two reviewers, at least two reviewers assigned the same overall GVHD grade in 93% of cases. These results suggest that current criteria for assessing the severity of GVHD by a single reviewer are not sufficiently reliable for rigorous clinical studies. As an alternative to the original criteria, we have developed and tested simplified criteria that summarize the clinical course of GVHD as reflected by the progression of disease and the amount of immunosuppressive treatment used to control the disease. Our results suggest that the revised criteria might yield more reproducible retrospective grading than the original criteria. Although the original criteria and the revised criteria might produce different results for individual patients, the overall distributions of grades with the two systems were similar. The proposed revised criteria could be implemented without disrupting the continuity and consistency with previous grading assigned by the original criteria.