The HCT-Specific Comorbidity Index Fails to Predict Survival After Allogeneic HCT In High Risk AML Patients - A Single Center Experience

Assessment of the individual risk of mortality after allogeneic hematopoietic cell transplantation (HCT) is essential for decision-making and patient counseling. For patients with AML several risk factors have been identified, including patient age, comorbidity, unfavorable cytogenetics, prior history of myelodysplasia and the remission status at HCT. The hematopoietic cell transplantation – specific comorbidity index (HCT-CI) has been designed and validated to predict mortality by concurrent disease and organ dysfunction prior to HCT at the Fred Hutchinson Cancer Research Center. Local data-dependence may affect the transportability of a prognostic system. Cross center validation of prognostic systems is therefore essential. In order to check the performance of the HCT-CI for our center, we applied the HCT-CI in a well defined cohort of AML patients. Our unit represents a large German transplant center characterised by referral of a large proportion of elderly patients with advanced myeloid leukemias.

We performed a retrospective cohort analysis in patients with AML according to the WHO definition who were transplanted between January 2000 and December 2008. Second allogeneic HCTs and haploidentical HCTs were excluded. Baseline variables were extracted from the local database by a trained data manager. Supplementary data were collected by comprehensive review of medical records by the investigator. For each patient we assessed the HCT-CI prior to transplant. Overall survival (OS) was estimated using the Kaplan-Meier method. Non-relapse mortality (NRM) was calculated with cumulative incidence statistics. The effect of the HCT-CI on the OS and NRM was analyzed in a multivariate Cox regression model, considering age, Karnofsky score, type of AML, cytogenetic risk (according to the ELN classification), treatment and remission status, sex match, CMV match, donor type, and conditioning intensity.

340 patients were eligible for the analysis. The median age of our patients was 53 years (range, 11 to 76 years). Eighty-six patients (26%) were in CR-1 of standard risk AML and 45 patients (14%) were in CR-1 of high risk AML while the remainder had more advanced disease. The median HCT-CI was 4 (range, 0 to 10). Donors were HLA-identical siblings in 116 patients (34%), matched unrelated donors in 130 patients (38%) and partially matched unrelated donors in 94 patients (28%). Overall survival at 2 years was 41±12%, 49±6% and 49±3% in the low, intermediate and high-risk HCT-CI groups (p=0.7), respectively. The corresponding NRM at 2 years was 33±23%, 26±10% and 23±5% (p=0.6). In multivariate analysis the HCT-CI failed to predict OS and NRM. In multivariate regression modeling neither its inclusion as a categorical variable (low, intermediate or high risk) nor as a numerical variable demonstrated adequate discrimination.

Comparing our population with the original validation cohort of the index hepatic impairment (51% vs. 20%, p=0.06), infection (47% vs. 4%, p=<0.001) and heart valve disease (44% vs. 2%, p=<0.001) were significantly more frequent in our population. These differences led to a statistically highly different distribution among the three HCT-CI risk categories of our patients (5% low risk, 21% intermediate risk, and 74% high risk patients) compared to the Seattle cohort of AML patients (51% low risk, 29% intermediate risk, and 21% high risk patients) (chi square test, p<0.0001). Referring to heart valve disease, we strictly applied the definitions from the original publication and considered even trace or mild regurgitation as heart valve disease. The fact that the resulting frequency of valvular dysfunction in our cohort is within the expected range for elderly patients (∼35% according to the Framingham Heart Study) leads to the suggestion that different working definitions had been applied. However, even after correction of this potential misclassification the modified HCT-CI did not discriminate OS and NRM.

We found no predictive value of the HCT-CI for OS and NRM in our patients. Our results suggest that different working definitions for the assignment of comorbidities had been applied. Refined definitions could therefore help to improve the performance of the index. Until its validity has been shown across different patient populations, its use as decision making tool for transplantation eligibility needs to be reconsidered.

Platzbecker:Celgene: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding.