Counselling of patients with AML about allogeneic hematopoietic stem cell transplantation (alloHSCT) is still an ambitious task in the light of the potential curative perspective after alloHSCT, poor outcomes after non-transplant approaches but a high risk of transplant-associated complications and still a significant risk of relapse even after HSCT. Several scores have been developed to predict outcome after HSCT, such as the HCT-CI and the Pre-transplant Assessment of Mortality (PAM) score. The PAM score has been revised recently, thereby acknowledging the shift to more frequently used reduced intensity conditioning. This score utilizes information on pts.age, donor type, disease risk, theserostatus for the CMV of pts.and donor, and the forcedexspiratory volume in the 1 second (FEV1). The aim of this study was to analyze the predictive power of the PAM score in an independent, large cohort of AML pts.who receivedalloHSCT within the last 12 years.
Patients and Methods
We selected all adult AMLpts.whoreceived the firstalloHSCTat the University Hospital of Dresden, a tertiary care hospital with a large transplant program, from January, 1, 2003 to July, 1, 2015.Pts.withhaplo-identical donors or after cord-blood transplantation were excluded. All patients gave their informed consent on analysing data. The PAM score was calculated as published (Au et al., BBMT 2015) and stratified into 4 groups: scores <17,scores17 to <24, scores 24 through 30, and scores >30.
Overall survival (OS), event-free survival (EFS), cumulative incidence of relapse (CIR) and non-relapse-mortality (NRM) after alloHSCTwereanalyzed according to several factors known to impact outcome using the log-rank test for univariate comparison. Age, AML type (de novo vs. sAMLvs. t-MN), sex match (female donor/male recipient vs. all other), CMV match (negative/negative vs. all other), donor type (sibling vs. matched unrelated vs. mismatched unrelated), ELN risk classification, type of conditioning (RIC vs. MAC), disease stage (CR1 vs. primary induction failure vs. >= first relapse) and the PAM score as a continuous variable were selected a priori for multivariate Cox regression analyses.
Overall, 544 pts.metthe inclusion criteria and were analyzed,the median age was 57 years (range, 18 to 76). Two-hundred-three pts.(37%) were treated with standard myeloablative conditioning (MAC) regimens while the remaining pts. received reduced intensity conditioning (RIC). Donors were siblings in 120 (22%), matched unrelated in 295 (54%) and mismatched unrelated donors in 129 (24%) pts. With a median follow up of 47 months (range, 1 to 161), the estimated OS for the whole cohort at five years was 43%, with a CIR of 30% and a NRM of 31% up to that time-point. The probability for OS at five years for pts.in PAM score group 0, 1, 2, and 3 was 65%, 50%, 33%, 22%, respectively (log-rank test, p= <.001). Both the CIR and NRM increased with increasing PAM scores (gray-tests, p= .005 and p= <.001, respectively). Notably, the PAM score contributed significantly to the prediction of OS even when added to a multivariate regression model which contained the single components of the score. In the final multivariate model, age (HR 1.02 per year, p= .004), disease stage (primary induction failure versus CR1, HR 1.5, p= .03), and the PAM score (HR 1.04, p= .03) had a significant impact on OS.
We validated the revised PAM for the prediction of OS after HLA-compatiblealloHSCTin a large, well characterised cohort of AMLpts.treatedat a large German transplantcenter. To the best of our knowledge, this is the first external validation of the revised PAM score. OS prediction based on this tool will be useful for counselling of futurepts.withAML.
Middeke:Sanofi: Honoraria. Thiede:AgenDix: Employment, Other: Ownership. Schetelig:Sanofi: Honoraria.
Asterisk with author names denotes non-ASH members.