Prediction of Non-Relapse Mortality in Recipients of Reduced Intensity Conditioning Allo-HSCT with Acute Myeloid Leukemia in First Complete Remission: Integrating the Seattle Comorbidity Index (HCT-CI) and EBMT Scoring Systems

Abstract 734

Acute myeloid leukemia (AML) in first complete remission (CR1) has become the dominant indication for alloHSCT and an increasing number of patients currently receive reduced- intensity conditioning (RIC), because of age or pre-existing comorbidities. While relapse after alloHSCT is mainly determined by the underlying cytogenetic and molecular AML profile, non-relapse mortality (NRM) is mainly determined by pre-existing comorbidities in addition to patients' characteristics and transplant-related complications. In order to assess the risk of NRM prior to transplantation, predictive scores, including the Seattle hematopoietic cell transplantation comorbidity index (HCT-CI) and the European Group for Blood and Marrow Transplantation (EBMT)-score, were developed. In the present study, the EBMT acute leukemia working party set out to evaluate the predictive value of both scores and, next, to integrate the predominant parameters of each score into a novel score.

Five hundred and sixty recipients of RIC-alloHSCT between 2000 and 2010 with AML in CR1, for whom both scores were assessed prior to transplantation, were included. The median follow-up was 35 (range 2–124) months. In this cohort, leukemia free survival (LFS) at 2 years was 55±2%, and the cumulative incidences (CI) of relapse, and NRM were 29%±2% and 16±2%, respectively. Of note, NRM continued to increase beyond 2 years and reached 21±2% at 5 years. The 2-year NRM split by HCT-CI score estimated 13±3%, 16±2%, and 17±3%, for patients with scores 0, 1–2, and >3, respectively, with relatively poor predictive value. The 2-year NRM split by EBMT-score estimated 13±2%, 17±2%, and 27±8% for patients with scores 2, 3, and >4, respectively, also bearing weak predictive value.

Next, we assessed the relative value of each of the contributing individual parameters by multivariable analysis (according to the Fine & Gray method), which resulted in the identification of 13 parameters with a Hazard Ratio (HR) >1. These parameters included 9 specific comorbidities from the Seattle HCT-CI, but also age, donor-type, interval from diagnosis to transplant, and CMV-serology as delineated in the EBMT scoring system. The integration of these parameters into a novel score was evaluated first in a development cohort of 373 patients and subsequently validated in the remaining 187 patients. Patient's characteristics in the development and validation cohorts were not statistically different.

Overall, the median age was 58 (range 23–76) years, and median interval from diagnosis to transplant was 164 days. 229 patients received grafts from HLA-identical sibling donors and 144 patients from unrelated donors. 268 patients had a positive CMV-serology. Comorbidities with HR >1 included infection history (n=142), peptic ulcer (n=15), cerebrovascular disease (n=17), arrhythmia (n=25), severe liver function abnormalities (n=15), obesity (n=25), rheumatological disease (n=15), heart valve disease (n=29), and renal disease (n=23). Of note, co-occurrence of comorbidities was observed frequently in patients with a history of infection, pulmonary disease and cardiac disease, occurring in 161, 100, and 82 patients respectively, in addition to one or more comorbidities.

Per study definition, each significant parameter was attributed 1, 2 or 3 points, as determined by its quantified HR. Combining all significant parameters into an integrated score, NRM estimated 9±2% at 2 years for low-risk patients with scores 0–2, 15±4% for intermediate-risk patients with score 3, and 28±4% for high-risk patients with a score of 4 points or more. NRM increased to 11±2%, 24±4% and 36±4% in the low-, intermediate- and high-risk cohorts, respectively, at 5 years from transplant. A further refined statistical analysis indicated relatively strong predictive values both in the sample and validation cohort. The integrated score did not predict for relapse, but proved to be significantly associated with LFS.

Collectively, this analysis demonstrates that both HCT-CI and EBMT risk scores were relatively weak predictors of NRM in a large cohort of 560 AML-CR1 patients receiving RIC-alloHSCT. However, integration of the most dominant parameters from each score into a new, integrated score yielded a stronger risk categorization that significantly predicted for NRM at 2 years post-transplantation, and also at 3 and 5 years post-transplantation with increasing NRM at those time points.