Acute graft-versus-host disease (aGVHD) affects at least 40–60 percent of allogeneic hematopoietic stem cell transplant (HSCT) patients after conditioning with myeloablative preparative regimens and is the major cause of early morbidity and nonrelapse mortality (NRM) in this population. One barrier to improving strategies for aGVHD prophylaxis and therapy is the lack of a clinical grading system that reliably predicts outcomes attributable to aGVHD across all severity grades and that adjusts the mortality risk based on changes in aGVHD activity and treatment over time. To provide a more precise and dynamic prognostic tool, Leisenring and colleagues developed an aGVHD activity index (aGVHDAI) to predict the risk of day-200 NRM. Using training and validation clinical data sets, the aGVHDAI proved to be accurate over a full range of threshold values when calculated either as an average aGVHDAI score across time or as a current aGVHDAI score at specific time points during the first 100 days after transplant. The data sets were retrospectively obtained from patients ≤ 55 years old with chronic myeloid leukemia who underwent allogeneic HSCT from unrelated donors after myeloablative conditioning with total body irradiation and cyclophosphamide. For each 10-day period after transplant, aGVHD activity was quantitatively staged according to the extent of skin rash, serum bilirubin elevation, and diarrhea volume. In addition, daily caloric intake, immunosuppressive drug usage, fever, and performance status were scored. Logistic regression models were fitted to these clinical parameters from a training data set of 193 patients and coefficients for each factor were used to assign a weight that optimized the ability of the aGVHDAI score to predict NRM by day 200. The six most significant factors for both the average aGVHDAI score and the real-time, current aGVHDAI score included two different ranges of serum hyperbilirubinemia, oral caloric intake < 40 percent of requirements with poorly controlled anorexia, nausea, and vomiting, any prednisone dose or secondary immunosuppressive therapy, and two different levels of reduced performance status. Notably, patient age, HLA disparity, severity of skin GVHD, and diarrhea volume did not improve the models. The final scaled weight factors and the aGVHDAI were then validated with an independent data set of 193 patients. When using the validation cohort data set to predict day-200 NRM, the average aGVHDAI score significantly outperformed the prognostic accuracy of both the conventional Glucksberg scale1,2 and IBMTR Severity Index3 . The validation data set was also used to generate three discrete contour lines with 95 percent confidence intervals that reflected the probability of day-200 NRM as a function of the current aGVHDAI score calculated at days 0-19, 20-39, or 40-100 after transplant.
In Brief
The aGVHDAI offers a number of improvements over the conventional aGVHD clinical grading systems. Most exciting is the ability to extrapolate a real-time mortality risk estimate based on the current aGVHDAI score. This prognostic information could greatly facilitate clinical decision making, especially in regard to tailoring treatment choices, and considering experimental interventions. While not directly applicable for day-to-day management, the average aGVHDAI score may also prove to be a more useful research tool than the conventional systems in assessing aGVHD severity, treatment responses, and outcomes. The aGVHDAI must still be validated in prospective studies and at other institutions. It must also be determined whether the scaled weight factors defined for the aGVHDAI are applicable to other diseases, age groups, donor sources, and conditioning regimens. Because it has been 30 years since the introduction of the original aGVHD grading system and almost 10 years since the last modifications, the aGVHDAI is a welcome and innovative advancement.
References
Competing Interests
Dr. Linenberger indicated no relevant conflicts of interest.
