A Combination of Clinical Characteristics and Day 7 Biomarker Concentrations Predicts Graft-Versus-Host Disease Following Hematopoietic Cell Transplantation From Related Donors

We have recently described an algorithm combining biomarkers and clinical characteristics that, at the onset of acute GVHD symptoms, predicts outcomes such as treatment response and non-relapse mortality (NRM). We have now evaluated whether a combination of pre-HCT clinical characteristics and a biomarker panel measured early post-HCT can predict future GVHD.

From 2001-2011, we prospectively collected clinical data and plasma samples from 393 patients receiving related donor HCT at our institution. The cumulative incidence of grade 2-4 GVHD by day 100 was 31% (n = 121; median onset = day 40). We divided patients randomly into a training set (n=264) to develop GVHD predictive algorithms using logistic regression, and a validation set (n=129) to test the algorithms. There were no statistically significant differences in pre-HCT clinical parameters between the 2 sets. We measured 4 GVHD biomarkers (IL-2Rα, TNFR1, elafin, and REG3α) on post-HCT day 7 plasma samples by ELISA.

Grade 2-4 GVHD onset within 2 months of HCT served as the predictive endpoint. We developed 3 predictive models for GVHD in the training set: 5 validated clinical risk factors alone (patient age, myeloablative conditioning [MA]), HLA-match, GVHD prophylaxis, graft source, use of TBI); the 4 biomarker panel alone; and a combination of all 9 clinical and biomarker parameters. We compared all 3 models at a specificity of 50% for their respective sensitivities. Clinical factors alone provided 51% sensitivity, 4 biomarkers alone provided 66% sensitivity, and all 9 parameters combined provided 77% sensitivity (combined vs. clinical model, p<0.001; vs. biomarker model, p=0.07). The combined model provided similar sensitivity in the validation set (75%) which allowed us to combine the 2 sets for further analyses.

We used the algorithm to stratify patients as high (N=211) or low risk (n=182) for developing GVHD.  High risk patients were twice as likely to develop grade 2-4 GVHD (Figure 1A; 38% vs. 20%, p<0.001) which developed almost 1 month earlier than in the low risk group (median day = 39 vs. 65). The greater incidence of GVHD in high risk patients resulted in significantly greater NRM by day 180 post-HCT (Figure 1B; 12% vs. 3%; p=0.001). The relapse rate was identical in both groups (24%), thus overall survival (OS) was significantly better in the low risk group (84% vs. 73%, p = 0.004). The differences between high and low risk patients remained significant at 1-yr post-HCT for NRM (17% vs. 6%, p<0.001) and OS (61% vs. 72%, p=0.01).

In conclusion, combining a panel of 4 biomarkers at day 7 post-HCT and 5 pre-HCT clinical parameters produced the best algorithm to predict GVHD following related donor HCT. The algorithm successfully stratified patients into high and low risk groups for GVHD, NRM and OS. We hypothesize the use of such an algorithm may permit preemptive therapy for patients who are at greatest risk in the early transplant course.