Interaction of Age and Comorbidities and Their Impacts on Hematopoietic Cell Transplantation (HCT) Outcomes

Abstract 665

Historically, age has been the main patient (pt)-specific decision-making factor for allogeneic HCT. The HCT comorbidity index (CI) was developed to capture pretransplant comorbidities. The index predicts non-relapse mortality (NRM) and has revolutionized outcome analysis for allogeneic HCT. Whether calendar age adds additional level of information to the HCT-CI in outcome prediction is unknown. Here, we investigated 1) how well the HCT-CI predicts outcomes across different age groups and 2) whether age could be incorporated into the HCT-CI. Data from 3033 consecutive pts treated with allogeneic HCT between January 2000 and December 2006 from HLA-matched related or unrelated donors at five collaborating institutions were used for this study. All data were collected by a single investigator, who was blinded from the final outcomes of pts, to ensure consistent comorbidity coding. Median age was 45 (range 0.1–74.5) years.

Overall, there was a weak correlation between increasing age and increasing HCT-CI scores (r=0.26). Pts were randomly divided into training (n=1853) and validation (n=1180) sets. In the training set, the HCT-CI predicted increased cumulative incidence rates of NRM and worsening of overall survival (OS) rates consistently in the 5 separate age groups (Table 1). Pulmonary function tests were not performed in 51% of pts <20 years of age, which might have affected the assignment of HCT-CI scores. Scores of 0, 1–2, and ≥3 were assigned to 28%, 32%, and 40%, respectively, of pts ≥20 years of age compared to 55%, 27%, and 18% of pts <20 years of age. Nevertheless, HCT-CI scores predicted OS of 73%, 61%, and 41% (p<0.0001), respectively, among pts <20 years of age.

A proportional hazards model was used to estimate the hazard ratios (HRs) for NRM and OS associated with different age intervals and other covariates, including the HCT-CI scores (Table 2). In this model, tests of homogeneity of HRs associated with HCT-CI scores of 1–2 and ≥3 across age groups were not rejected for either NRM (p=0.66 and p=0.86, respectively) or OS (p=0.76 and p=0.24, respectively). Increasing HCT-CI scores were associated with the highest HRs for NRM compared to other covariates. Pts in age groups 40–50, 50–60, and >60 years had HRs for NRM ranging between 1.48–1.84 compared to pts <20 years of age. Accordingly, age >40 years was assigned a score of 1 to be added to the HCT-CI scores. In the validation set, although we continued to observe increases in HRs for NRM with increasing age, only minor improvement in c-statistics for NRM (0.66 versus 0.68) was detected when age was added to the HCT-CI.

These results indicate that the HCT-CI is valid for outcome prediction across all age groups and that age per se has a relatively minor impact on HCT outcome prediction in models that account for comorbidities. Age >40 years had an impact equivalent to a single comorbidity with a weight of 1, and therefore should be assigned a score of 1 when using the HCT-CI/Age composite index.