Proteomic Pattern-Based Risk Stratification of Outcome Shows Significant Higher Accuracy Compared to HCT-CI in Patients with AML and MDS,

Allogeneic stem cell transplantation (SCT) is an established treatment for many severe disorders of hematopoiesis. Although SCT has considerable curative potential, its application is limited by transplant-related complications such as infections and graft-versus host disease (GvHD) which could lead to high mortality rates especially in older or less fit patients. Therefore, a careful pre-SCT assessment of risk and benefit is mandatory and different scores have recently emerged as helpful tools. We have previously applied proteomics to identify a specific urinary polypeptide patterns (PP) predictive for developing acute GvHD (aGvHD) (Weissinger EM et al, Blood 2007;109:5511–5519). The aim of this study was to investigate whether the PPs can predict overall outcome after allo-SCT and to compare these findings to those of the hematopoietic cell transplantation comorbidity index (HCT-CI) (Sorror M et al, Blood 2005;106:2912–2919).

In this retrospective analysis from Hannover Medical School, the datasets from all patients (pts) with acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS), who were allo-transplanted from a fully matched donor (matched related/unrelated donor (MRD/MUD)) between 2003–2008 and for whom relevant PP data were available, were included. Pts with a pt-donor HLA-mismatch constellation were excluded from this study. PP data from urine samples which were prospectively collected by day ≥ +7 after allo-SCT were correlated with overall survival (OS), aGvHD, non-relapse mortality (NRM), relapse rate and mortality (RM), and compared to the predictive value of the HCT-CI.

PP data were available from 111 pts (97 pts with AML, 14 with MDS; median age 52y; median EBMT score 4; 59 male/52 female; 69 MUD/42 MRD). They were grouped in high (PP-HRG), low (PP-LRG) or intermediate risk groups (PP-IRG). Forty-three pts (39%) belonged to the PP-LRG for aGvHD compared to 47 pts (42%) who were classified PP-HRG. Patient characteristics of PP-LRG and PP-HRG were similar in terms of age, sex and EBMT score (median 4 in both groups).

OS compared favorably for the PP-LRG with an OS of 72% vs. 49% for the PP-HRG (p=0.03), also if only reduced intensity conditioning (RIC) was considered (73% vs 42%; p=0.01), respectively. There was a trend for higher incidence of NRM in the PP-HRG compared to PP-LRG (30% vs 14%, p=0.07) for the whole cohort, and a significant higher NRM rate, if only RIC was evaluated (35% vs 11%, p=0.01).

However, if risk stratification was based on the HCT-CI, there was no significant difference between high risk (S-HRG) and low risk group (S-LRG) in terms of OS and NRM regardless of intensity of conditioning (OS for whole cohort: 57% vs 45%, p=0.4; OS for RIC: 56% vs 36%, p=0.2; NRM for whole cohort: 20% vs 23%, p=0.8; NRM for RIC: 18% vs 29%, p=0.4).

Concerning the PP-IRG, there was a difference in OS between PP-IRG and PP-LRG (38% vs 73%, p=0.02). However, there was no significant difference in OS of the PP-IRG compared to the other PP-based risk groups nor between the HCT-CI based risk groups. Further, NRM did not show a significant difference neither for PP-based nor HCT-CI-based intermediate risk group compared to the other risk groups.

Thirty vs 15 pts developed aGvHD in PP-HRG and PP-LRG (64% vs 35%, p<0.01) compared to 48% vs 64% (p=0.2) for S-HRG and S-LRG of the whole cohort, respectively. Incidence of aGvHD differed also significantly in the RIC cohort for PP-HRG and PP-LRG (65% vs 32%, p=0.01), but not for HCT-CI-based risk groups (47% vs 64%, p=0.1). Relapse rates and RM were not significantly different between high and low risk groups, neither for PP-based nor HCT-CI based (whole cohort and RIC subgroup), respectively.

Risk stratification according to GvHD-match based PP, which has previously been shown to predict aGvHD, now also allows the identification of patient groups with significantly different OS and NRM. In comparison to the HCT-CI, PP-based prediction shows significantly higher accuracy in this rather homogeneous cohort of patients. Since proteomics is a new method which has been available only at a few centers, further multicenter analyses are essential to determinate the value of PP-based prediction of complications and outcome in SCT.

Metzger:Mosaiques Diagnostics GmbH: Employment.