An acute graft-versus-host disease activity index to predict survival after hematopoietic cell transplantation with myeloablative conditioning regimens

Abstract

Algorithms for grading acute graft-versus-host disease (GVHD) are inaccurate in assessing mortality risk. We developed a method to predict mortality by using data from 386 patients with acute GVHD. From the onset of GVHD to day 100, GVHD manifestations were scored for the skin, liver, and upper and lower gastrointestinal tract, and data were recorded for immunosuppressive treatment, performance, and fever. Logistic regression models predicting nonrelapse mortality (NRM) at day 200 were developed with data from 193 randomly selected patients and then validated in the remaining 193 patients. Clinical parameters were grouped to optimize predictive accuracy measured as the area under a receiver-operator characteristic (ROC) curve. The optimal model included the total serum bilirubin concentration, oral intake, need for treatment with prednisone, and performance score. When the overall burden of GVHD was measured by using average Acute GVHD Activity Index (aGVHDAI) scores for each patient in training and validation data sets, areas under ROC curves were 0.87 and 0.85, respectively. Contour lines were generated to reflect the predicted NRM at day 200 as a function of current aGVHDAI scores. These results demonstrate that clinical manifestations of GVHD severity can be used to accurately predict the risk of NRM in real time.

Figure 1.

ROC curves illustrating the accuracy of the aGVHDAI across the full range of threshold values. All graphics used the common weighting system shown in Table 2. Panels A and C illustrate ROC curves when average aGVHDAI scores for each patient were used to predict day-200 NRM in the training data set and validation data set, respectively. Panels B and D illustrate ROC curves when aGVHDAI scores for each time interval were used to predict day-200 NRM in the training data set and validation data set, respectively.

Figure 2.

Distribution of values for average aGVHDAI, according to the modified GVHD grading system in 386 patients. The boxes display values for the average aGVHDAI from the 25th to the 75th percentile; the bars within the boxes display the median value; and the vertical bars display the upper and lower adjacent values.23 Data points outside this range are plotted as individual circles.

Figure 3.

ROC curves depicting the accuracy of 3 GVHD scoring systems for predicting day-200 NRM in 193 patients from the validation data set. The average aGVHDAI line (small ○) is identical to that in Figure 1C. Data points for grading according to Przepiorka et al4 and the International Bone Marrow Transplant Registry5,10 are shown in panels A and B, respectively. The numbers adjacent to each ROC curve are the areas under the respective curves.

Figure 4.

Predicted NRM by day 200 as a function of current aGVHDAI scores at different points in time after transplantation. The top panel shows this relationship for days 0 to 19, the middle panel for days 20 to 39, and the bottom panel for days 40 to 100. Within each panel, the solid line represents the predicted NRM by day 200 across a range of values for current aGVHDAI scores, and the shaded areas represent 95% confidence intervals.