Plasma Proteomic Profiles May Predict Early Acute Graft-vs.-Host Disease (aGVHD) Following Reduced Intensity Conditioning (RIC) Allogeneic HLA-Identical Sibling Transplantation (allo-SCT).

GVHD remains a matter of concern after RIC allo-SCT. Our knowledge of aGVHD risk factors is still primarily based on historical analyses performed in the myeloablative allo-SCT setting. Thus, modern approaches to predict the occurrence/severity of aGVHD are needed. This study aimed to identify a plasma protein signature correlating with occurrence of early aGVHD. We performed Surface-Enhanced Laser Desorption/Ionization-time (SELDI) of flight mass spectrometry profiling of plasma from 88 patients who received a RIC allo-SCT from HLA-identical siblings. Patients characteristics were: median age was 51 (range, 18–70) y. 41 patients (47%) had a myeloid malignancy, whereas 30 patients (34%) had a lymphoid malignancy. The remaining 17 patients (19%) were treated for metastatic non-hematological malignancies. The majority of patients (n=70, 80%) had an advanced disease with high risk clinical features precluding the use of standard myeloablative allo-SCT. All patients (100%) received G-CSF-mobilized PBSCs. The RIC regimen consisted of fludarabine, busulfan and ATG in 53 patients (60%) and low dose irradiation in 35 patients (40%). With a median follow-up of 400 (range, 127–829) d, 20 patients (23%) experienced early (prior to day 35 after allo-SCT) grade 2–4 acute GVHD (12 grade 2 and 8 grade 3–4). Denatured plasma samples (collected at a median of 28 days after allo-SCT) were incubated with H50 and CM10 ProteinChip arrays and subjected to SELDI analysis. Patient population was divided into a training (n=59) and a validation set (n=29). In the training set, 36 protein peaks were differentially expressed according to early aGVHD occurrence. By combining partial least squares and logistic regression methods, we built a multiprotein model that correctly predicted outcome in 96% of patients (14/14 patients with early aGVHD; specificity, 96%). The observed correct prediction rate in the validation set was 69% with a sensitivity of 67%, and a specificity of 70%. While negative predictive value of the model was only 36%, predictive positive value was estimated to 89% in the validation set. The performances of the model remained very similar after iterative (500 times) random resampling (correct prediction rate: 74%, median sensitivity: 48%, median specificity: validation set: 83%). Univariate and multivariate analyses of known risk factors (demographic features, diagnoses and transplant procedures) for occurrence of an early grade 2–4 acute GVHD did not show any statistically significant difference between the group of 20 patients who had early grade 2–4 aGVHD as compared to the remaining patients, and suggested that the multiprotein index is likely to be the only independent prognostic parameter. Major components of this multiprotein index are currently being characterized and will be presented. Obviously, larger prospective studies are still needed, but our results already suggest that proteomic analysis of plasma will prove increasingly important in the early and clinical diagnosis of aGVHD allowing improving patient outcome.