Abstract
Abstract 1980
Chronic GVHD is one of the most serious consequences of allo-SCT associated with high morbidity and mortality. Therefore, identification of potential predictors of cGVHD is crucial. At present, there is no validated prognostic blood test for cGVHD. The aim of this study was to expand the search for cGVHD biomarkers, to validate candidate proteins using high-throughput assays in a series of 152 consecutive patients treated in a single center, and to determine a composite prognostic score for prediction of extensive cGVHD.
Patients included in the analysis were treated between 2005 and 2008 and had a median follow-up of 2.3 y. Per study definition, patients survived at least 4 months after allo-SCT. Median age at time of allo-SCT was 49 y (range, 17–70) and 55% were males. 87 (57%) recipients were treated for myeloid malignancies. Seventy patients (53%) received allo-SCT from an HLA-matched related donor; while 60 patients (40%) received allo-SCT from an HLA-matched related donor and 22 (14%) received an HLA-mismatched unrelated graft. Prior to allo-SCT, 48 patients (32%) underwent a myeloablative conditioning (MAC) regimen, while 104 cases (68%) received a reduced-intensity conditioning (RIC). G-CSF-mobilized peripheral blood stem cells (PBSCs) were used in 108 cases (72%), bone marrow in 28 cases (18%) and unrelated cord blood cells in 18 cases (12%). In this series, 70 patients were diagnosed with cGVHD (23 with a localized/limited form and 47 with an extensive form) after a median of 7.14 months after allo-SCT. At 2 years, the cumulative incidence of extensive cGVHD was 40% and overall survival of the whole cohort was 69% (95%CI, 62–77 %). For the purpose of this study, serum samples were collected around day 100 (range, 83–119) after allo-SCT. Forty-one cytokines were studied using Luminex xMAP technology (Millipore, Billerica, USA).
In multivariate analysis, 2 clinical factors were associated with extensive cGVHD occurrence: history of prior acute GVHD (P=0.0019, HR=2.6, CI95%, 1.429–4.839) and the use of PBSCs as graft source (P=0.0067, HR=3.1, CI95%, 1.365–6.896). Independently from these clinical factors, 10 cytokines were found to be significantly correlated with the incidence of extensive cGVHD. High levels of IP10 (CXCL10) (p=0.0010), IL15 (p=0.0280), IL10 (p=0.0112), IL2RA (p=0.0261) and MIP-1beta (CCL4) (p=0.0108) were associated with higher incidence of extensive cGVHD, while high levels of Fractalkine (CX3CL1) (p=0.0081), MDC (CCL22) (p=0.0005), RANTES (CCL5) (p=0,0083), TARC (CCL17) (p=0.0041), IL12p40 (p=0.0155) were associated with a lower risk of developing extensive cGVHD.
Based on these findings, we then sought to establish a practical prognostic score capable of predicting extensive cGVHD. This score was calculated using the traditional multivariate Cox model combined with the statistical approach called “time-dependent receiver-operator characteristic (ROC) curves”, making it possible to assess the predictive capacity marker that is being evaluated, as described bu Heagerty PJ et al., (biometrics 2000). Here, we have focused on the 2 significant clinical variables (prior history of acute GVHD and PBSCs as graft source) and the most significant cytokines found in the multivariate analysis in order to obtain a useful tool in the daily medical practice. Based on 0,632 bootstrap resampling method for repeated cross-validation, the area under the time-dependent ROC curve was 0.80 (95%CI, 0.72–0.87) indicating that such composite score is a powerful predictor of the risk of extensive cGVHD at 2 years.
In summary, results from this study allowed to build a new noninvasive score to accurately predict the risk of extensive cGVHD occurrence after allo-SCT. Such score could be used as a decision tool in the clinical management of allo-SCT. We are currently undertaking an additional validation of this score on another independent cohort.
No relevant conflicts of interest to declare.
Author notes
Asterisk with author names denotes non-ASH members.
