Dear Editor,

Endothelial complications are important causes of mortality after allogeneic stem cell transplantation (alloSCT) and can be predicted by endothelial activation and stress index (EASIX) [1,2,3]. EASIX integrates basic laboratory parameters (LDH, creatinine, and platelets) characterizing transplant-associated thrombotic microangiopathy and is prognostic for outcome in a variety of clinical settings, including coronavirus disease 2019 (COVID-19) and chimeric antigen receptor T (CAR-T) cell therapy [4, 5].

Sepsis is a dysfunctional endothelial response to harmful microorganisms. We hypothesized that EASIX may predict hazard of sepsis.

In this retrospective evaluation, 1290 adult patients allografted at our institution between 2004 and 2018 were assessed for presence of sepsis, neutropenic fever, and infectious pathogens within 50 days after transplantation. Neutropenic fever was graded according to the 2010 Infectious Diseases Society of America (IDSA) guidelines, while sepsis and septic shock were defined according to the modified Sepsis-3 guidelines by the Intensive Care Working Party (iCHOP) of the German Society of Hematology and Medical Oncology (DGHO) for neutropenic cancer patients (see supplements). Established pre-transplant scores and additional serum markers [Ferritin, interferon-gamma (IFNγ), CXCL9, interleukin (IL)-18, suppressor of tumorigenicity 2 (ST-2)], and soluble thrombomodulin (sCD141) were assessed longitudinally before and after transplantation (supplements) and correlated with outcome. EASIX was calculated as LDH(U/L)*creatinine(mg/dl)/platelets (per nl). For risk factor evaluation, patients were randomly allocated 1:1 to a training and a validation cohort, respectively.

Neutropenic fever occurred in 77%, while infectious pathogens could be identified in 31% of patients. Sepsis was diagnosed in 7.2% of patients (Suppl. Tables 1, 2). Patients who developed sepsis had higher median EASIX values before conditioning (EASIX-pre) and at any later time point until day+28 irrespective of pathogen detection (Fig. 1A).

Fig. 1
figure 1

A Box plot comparison of EASIX before and after alloSCT in patients who did or did not develop sepsis within 50 days after transplantation (full cohort). Sepsis patients were subgrouped by pathogen detection. B Prediction error curves for multivariate cause-specific Cox regression in the validation cohort (n = 613, complete case analysis). All multivariate models were established in the training cohort (n = 609) and included EASIX contrasted with multivariate models without EASIX. C Exploratory, time-dependent, competing risk ROC analysis with sepsis as endpoint (full dataset included). D Univariate cause-specific Cox regression with endpoint sepsis in the validation cohort (n = 525, complete case analysis), models established in the training cohort (n = 539). E Comparison of univariate time-dependent area under the curve (AUC) curves of different risk scores for end point NRM within the first 6 months after alloSCT in the full dataset. C–E EBMT, European Society for Blood and Marrow Transplantation, HCT-CI, Hematopoietic Cell Transplant-Co-morbidity Index, CIBMTR-VOD, Center for International Blood & Marrow Transplant Research—Veno-occlusive disease score)

Full size image

In the training cohort, EASIX-pre was the only marker significantly associated with hazard of sepsis in multivariable cause-specific Cox regression analyses (HR 2.3 for a two-fold change, p < 0.001, Suppl. Table 3). When EASIX-pre was replaced by its three single parameters in a separate multivariable analysis, LDH and platelets but not creatinine were significantly associated with sepsis risk (suppl. Table 4).

Based on Gray’s maximally selected rank statistics performed in the training cohort, EASIX was dichotomized into a "low-risk" group (EASIX < 2.32) and a "high-risk" group (EASIX > 2.32). Using this cut-off for multivariable Cox regression analysis in the validation cohort, we observed a cause-specific HR of 16.1 (7.0–36.9) for EASIX > 2.32 with respect to time to sepsis, corresponding to 7/462 sepsis events (1.5%) in the low-risk group vs 40/172 (23%) in the high-risk group (Suppl. Table 5). Prediction errors measured via time-dependent Brier score were smaller for the model including EASIX (Fig. 1B).

Sepsis risk was also significantly associated with post-, but not with pre-transplant levels of the endothelial vulnerability markers ST-2 and IL-18, but not with IFNγ, CXCL9, and sCD141 (Suppl. Figure 2). EASIX was superior to other established transplantation risk scores for predicting sepsis and 6-month non-relapse mortality (Fig. 1C–E).

Limitations of the study are its retrospective design and the single-center analysis.

In conclusion, EASIX-pre is a powerful prognostic marker of sepsis after alloSCT and might be used for guiding risk-adapted prevention strategies.