National Institutes of Health chronic graft-versus-host disease staging in severely affected patients: organ and global scoring correlate with established indicators of disease severity and prognosis

Abstract

Between 2004 and 2010, 189 adult patients were enrolled on the National Cancer Institute's cross-sectional chronic graft-versus-host disease (cGVHD) natural history study. Patients were evaluated by multiple disease scales and outcome measures, including the 2005 National Institutes of Health (NIH) Consensus Project cGVHD severity scores. The purpose of this study was to assess the validity of the NIH scoring variables as determinants of disease severity in severely affected patients in efforts to standardize clinician evaluation and staging of cGVHD. Out of 189 patients enrolled, 125 met the criteria for severe cGVHD on the NIH global score, 62 of whom had moderate disease, with a median of 4 (range, 1-8) involved organs. Clinician-assigned average NIH organ score and the corresponding organ scores assigned by subspecialists were highly correlated (r = 0.64). NIH global severity scores showed significant associations with nearly all functional and quality of life outcome measures, including the Lee Symptom Scale, Short Form-36 Physical Component Scale, 2-minute walk, grip strength, range of motion, and Human Activity Profile. Joint/fascia, skin, and lung involvement affected function and quality of life most significantly and showed the greatest correlation with outcome measures. The final Cox model with factors jointly predictive for survival included the time from cGVHD diagnosis (>49 versus ≤49 months, hazard ratio [HR] = 0.23; P = .0011), absolute eosinophil count at the time of NIH evaluation (0-0.5 versus >0.5 cells/μL, HR = 3.95; P = .0006), and NIH lung score (3 versus 0-2, HR = 11.02; P < .0001). These results demonstrate that NIH organs and global severity scores are reliable measures of cGVHD disease burden. The strong association with subspecialist evaluation suggests that NIH organ and global severity scores are appropriate for clinical and research assessments, and may serve as a surrogate for more complex subspecialist examinations. In this population of severely affected patients, NIH lung score is the strongest predictor of poor overall survival, both alone and after adjustment for other important factors.

Figure 1

Patients with cGVHD involvement per organ. A score of ≥1 on the NIH scale is reflected by the blue bar, and a score of 3 (severe) is represented by the red bar. Although eye, oral and skin are the most commonly involved organs, severity in the overall cohort was driven by severe skin disease and a high percentage of lung involvement.

Figure 2

Comparison of average NIH Organ score to the expert subspecialist evaluator (SSE) average score, shows a moderately strong correlation Spearman r = 0.64 (strong: r>0.70). The scores show impressive correlation considering the SSE score is a strictly objective assessment vs. the NIH score, which integrates patient reported symptoms and/or need for treatment.

Figure 3

Comparison of the NIH Organ Score to 10 outcome variables that are related to disease severity in patients with chronic disease. Scores with highly significant association with the outcome measure are indicated in the solid blocks (p<0.001) and those with a trend toward significance are represented by striped blocks (p<0.05).

Figure 4

Kaplan-Meier survival curves for the entire cohort (A) from the time enrolled on study. Karnofsky performance score (KPS) was associated with survival (B) in this cohort as was eosinophil count (C), but the best predictor of outcome was the NIH Lung Score (D), in which a severe score (=3) was the most likely indicator of poor survival p<0.0001.

Figure 4

Kaplan-Meier survival curves for the entire cohort (A) from the time enrolled on study. Karnofsky performance score (KPS) was associated with survival (B) in this cohort as was eosinophil count (C), but the best predictor of outcome was the NIH Lung Score (D), in which a severe score (=3) was the most likely indicator of poor survival p<0.0001.

Figure 4

Kaplan-Meier survival curves for the entire cohort (A) from the time enrolled on study. Karnofsky performance score (KPS) was associated with survival (B) in this cohort as was eosinophil count (C), but the best predictor of outcome was the NIH Lung Score (D), in which a severe score (=3) was the most likely indicator of poor survival p<0.0001.

Figure 4

Kaplan-Meier survival curves for the entire cohort (A) from the time enrolled on study. Karnofsky performance score (KPS) was associated with survival (B) in this cohort as was eosinophil count (C), but the best predictor of outcome was the NIH Lung Score (D), in which a severe score (=3) was the most likely indicator of poor survival p<0.0001.