Simultaneous Sessions IV: Graft-versus-Host Disease
Predicting the occurrence of chronic graft-versus-host disease (cGVHD) would be of significant clinical interest. We have previously reported a scoring system based on laboratory values on day 100 (EHA Stockholm 2013)
We have now refined this analysis on a larger group of patients, and are presenting a final model, capable of identifying patients at risk of moderate/severe cGvHD, and of predicting transplant related mortality (TRM) and survival (OS).
We enrolled 1292 patients allografted between 1990 and 2016. Inclusion criteria were: patients alive on day +100 from transplant; aGvHD max grade I on day 100; cGvHD max grade “minimal” on at day 100. End points were: development of cGvHD grade “moderate/severe” (here referred to as extensive), TRM and survival. Lab tests included full blood counts and full chemistry. Patients were divided into a training (n = 646) and a validation cohort (n = 646) with comparable characteristics, transplant conditions, stem cell source, risk of cGvHD, TRM and survival. Univariate and multivariate Cox models were used to identify predictors of cGvHD, TRM, OS. Gray's test was used to compare cumulative incidence (CI) curves for cGvHD and TRM, whereas the log rank test was used to compare different survival curves. The cut off for lab values was determined using ROC curve.
TRAINING COHORT. Multivariate analysis for cGvHD identified platelets (PLTs) (HR 1.40, p = 0.01), cholinesterase (CHE) (HR 1.39, p = 0.03) and lactic dehydrogenase (LDH) (HR 1.66, p = 0.02) as independent variables. Using ROC curve a cut-off was defined as follows: < 139*10^9/L for PLTs, <3389 IU/L for CHE and ≥550 IU/L for LDH. A score of 1 was assigned to each variable, producing a low risk group (score 0-1, n = 396) and a high risk group (score 2-3, n = 250). The CI of extensive cGvHD at 32 years, was 42% vs 70% for low vs high risk patients (p < 0.0001, Fig 1A). In multivariate analysis the lab score predicted OS (HR 1.45, p = 0.005) and TRM (HR 2.57, p < 0.0001). The 2 year OS was 79% vs 60% (p < 0.0001, Fig. 1B), and the CI of TRM of 6% vs 24% (p < 0.0001) respectively for low vs high risk patients (Fig. 1C).
VALIDATION COHORT. Prognostic Lab-Score was then tested in the validation cohort. Multivariate analysis confirmed Lab-Score to be an independent predictor of extensive cGvHD (HR 1.79, p < 0.0001), OS (HR 1.43, p = 0.008) and TRM (HR 1.62, p = 0.004). The 2 year CI of extensive cGvHD was 43% vs 65% in the low and high risk patients (p < 0.0001). The 2 year OS was 77% vs 59% for low vs high risk patient (p < 0.0001). Finally, the 2 year CI of TRM was of 7% vs 20% respectively (p = 0.0001).
Figure 1: Data from the training cohort of patients A) Cumulative incidence of extensive cGvHD in low vs high risk patients, as identified with the Lab-Score, B) 2 year OS; C) 2 year cumulative incidence of TRM.
we confirm that a simple Lab-Score on day +100, predicts extensive cGvHD, together with TRM and survival.