Accurate prediction of the onset and severity of GvHD and treatment-related mortality (TRM) can lead to improved outcomes through optimized donor selection and GvHD prophylaxis. Current assessments are typically made after bone marrow transplantation (BMT), and often require inherently unreliable qualitative measures. The persistence of host antigen presenting cells (APC), particularly dendritic cells (DC) has been implicated in the onset of GvHD, and a recent clinical trial sought to confirm previous reports that extracorporeal photopheresis (ECP) can prevent GvHD after allogeneic BMT. We examined blood from a subset of 32 ECP-treated patients to identify peripheral blood cell surface markers that might predict GvHD and TRM. Samples were drawn immediately prior to ECP (at baseline) and immediately after ECP, but prior to myeloablative conditioning, then assayed by flow cytometry. Cytometry data were grouped and modeled to assess their predictive accuracy alone or in combination with clinical laboratory values. Logistic regression showed that markers for specific DC subsets present prior to myeloablation were the best predictors of outcomes. The likelihood of grade II-IV GvHD (aGvHD) increases when baseline lin− HLA-DR+ CD11c+ myeloid cells make up a smaller proportion of circulating mononuclear cells with a predictive accuracy, reflected in the area under the receiver-operator curve (ROC) of 0.83. The best predictive model for TRM was a lower absolute abundance of circulating lin− HLA-DR+ CD123+ plasmacytoid DC at baseline (ROC=0.86). No additional predictive power arose with respect to aGvHD or TRM after including laboratory values. However, models that combined certain clinical laboratory results and demographic factors also predicted these clinical outcomes, and offer a possible alternative to complex cytometric assays. The best such model included baseline measurements of BUN/Creatinine ratio, serum albumin, and the match/relatedness of the graft donor and recipient (ROC=0.82, n=59) for TRM, while baseline neutrophil counts were most predictive of aGvHD (ROC=0.69, n=60). In summary, we have identified biomarkers including host DC subsets, readily measured clinical tests, and demographic factors that, at least in patients receiving ECP, are present before conditioning and can predict outcomes. Although this study does not address the impact of ECP on antigen presenting cells directly, the DC profiles described here highlight the potential role of ECP in modulating DC, and thereby decreasing the risks associated with BMT. These results encourage larger studies with non-ECP control populations to directly address this issue.
Disclosures: ECP for Graft v Host Disease.; Dr. Abhyankar: Speaker.; Therakos does not have a speakers bureau.