Peritransplant Serum Albumin Decline Predicts Subsequent Severe Acute Graft-versus-Host Disease after Mucotoxic Myeloablative Conditioning

Conditioning-related gut toxicity can result in a protein-losing enteropathy manifesting as a decline in serum albumin in the peritransplant period. Inspired by the pathogenesis of acute graft-versus-host disease (aGVHD), we hypothesized that the magnitude of decline in serum albumin from the day of conditioning initiation until its nadir in the first 2 weeks after hematopoietic cell transplantation HCT (DeltaAlb) predicts the risk for subsequent severe aGVHD. We reviewed the medical records of all 88 patients with acute myeloid leukemia or myelodysplastic syndrome who underwent highly mucotoxic myeloablative (busulfan/cyclophosphamide or cyclophosphamide/total body irradiation) allogeneic HCT from a matched related donor (MRD) or matched unrelated donor (MUD) at our institution between January 1, 2012 and January 1, 2015. Severe aGVHD was associated with MUD (47% versus 14% with MRD; P = .001) and DeltaAlb, which was significantly greater among patients who developed versus did not develop severe aGVHD (1.2 ± .5 versus .8 ± .4 g/dL, respectively; P < .001). In multivariate analysis DeltaAlb remained a significant predictor of severe aGVHD (odds ratio, 5.68; 95% CI, 1.65 to 19.64; P = .006; area under the ROC curve, .74; 95% CI, .63 to .86; P < .001). The best cutoff for DeltaAlb to predict severe aGVHD was .9, with a sensitivity, specificity, and overall classification accuracy of 77%, 66%, and 69%, respectively. The model was validated using the bootstrap technique, with no significant change in its performance. These results were not generalizable to a cohort of 30 patients who received less mucotoxic myeloablative or reduced-intensity conditioning. In conclusion, with mucotoxic myeloablative HCT, each .1-g/dL increase in DeltaAlb was associated with an approximately 23% increase in the odds of developing severe aGVHD. As an early biomarker of gut damage, DeltaAlb can be incorporated in composite risk models for aGVHD prediction, with hopes for ultimately allowing for individualized GVHD prophylaxis and potential intervention according to the predicted risk.