Grading Acute GVHD: Getting It Right Every Time!

Introduction: The acute GVHD grading for the Center for International Blood and Marrow Transplant Research (CIBMTR) reporting is based on criteria published by Przepiorka et al (Bone Marrow Transplant,1995;15:825). While the CIBMTR's FormsNet3^SM application allows electronic data submission, data professionals still need to manually assign organ stages and overall grade, and track data over time.  We report the design and validation of an electronic grader tool to automate acute GVHD severity scoring.

Methods: Tables from the Przepiorka et al and the Rowlings et al (Br J Haematol 1997; 97: 855) papers were used to build logic for the "Consensus" and the IBMTR grading schemata respectively. This free tool, aGVHD Grader (agvhd.com), was developed using HTML, CSS and JavaScript. We also reviewed the acute GVHD data on the NMDP and CIBMTR forms for 30 patients with severe acute GVHD and tested data entry in a prototype database version of the aGVHD Grader.

Results: The 30 randomly selected allo-BMT patients were transplanted between 1999 and 2014 with median age of 11±6.1 y and average follow-up of 2.3±3 y. Maximal acute GVHD overall grade on submitted forms was recorded as II in 2 (6.6%), III in 13 (43.3%) and IV in 15 (50%) cases. On this retrospective review, the maximal grade was concordant in 22 cases but was revised from IV to III in eight cases; these eight cases were scored based on institutional practice of assigning stage 4 lower GI as overall grade IV. Additional cause for diarrhea and elevated bilirubin was considered in 22 (73%) cases and downstaging done in 18 of them. CIBMTR forms do not record downstaging and we were unable to find consistent documentation of downstaging process in the scorer's internal data records. Histologic evidence of GVHD in stomach or duodenum was found in 19 out of 21 cases with upper GI endoscopy; 11 of these were scored as upper GI involved and 10 as not involved, clinically.

We entered clinical data and bilirubin values into the aGVHD Grader which automatically assigned organ stage, downstaged for additional cause, and derived maximal Consensus overall grade and IBMTR severity score accurately in 100% cases when compared with manual scores using the published criteria. We also generated a sample one-screen view of daily acute GVHD assessments using the prototype database version of the Grader (Figure 1).

Conclusions: Acute GVHD is a clinical syndrome with considerable within-grade heterogeneity with different patterns of skin, liver or gut involvement. Data professionals commonly face situations when there is ambiguity in scoring and in judging GVHD severity. The software program reported here automates the organ staging and grading of acute GVHD and enables tracking change over time. This may help with consistency in data capture and error reduction and serve as a useful record for outcomes reporting and clinical research.

Figure 1.