EASIX and Mortality after Allogeneic Stem Cell Transplantation

Abstract

Allogeneic stem cell transplantation (alloSCT) is an effective immunotherapy in patients with hematological malignancies. Endothelial dysfunction was linked to major complications after alloSCT. We asked the question if the ‘Endothelial Activation and Stress Index’ (EASIX; [(creatinine × LDH) ÷ thrombocytes]) can predict mortality after alloSCT.

We performed a retrospective cohort analysis in five alloSCT centers in the USA and Germany. EASIX was assessed prior to conditioning (EASIX-pre) and correlated with mortality in 755 patients of a training cohort in multivariable models. The predictive model established in the training cohort was validated in 1267 adult allo-recipients.

Increasing EASIX-pre predicted lower overall survival (OS) after alloSCT, and successful model validation was achieved for the validation cohort. We found that EASIX-pre predicts OS irrespective of established scores. Moreover, EASIX-pre was also a significant prognostic factor for transplant-associated microangiopathy. Finally, EASIX-pre correlated with biomarkers of endothelial homeostasis such as CXCL8, interleukin-18 and insulin-like-growth-factor-1 serum levels.

This study establishes EASIX-pre based on a standard laboratory biomarker panel as a predictor of individual risk of mortality after alloSCT independently from established clinical criteria.

Introduction

Allogeneic stem cell transplantation (alloSCT) is a curative treatment option for patients suffering from leukemia and other hematological malignancies. The main clinical challenge of alloSCT is high treatment-associated mortality. For better individual assignment of alloSCT risk, considerable progress has been made by defining comorbidities, disease-specific risks and donor-related factors with indices such as the Hematopoietic Cell Transplantation-Comorbidity Index (HCT-CI),¹ the European Society for Blood and Marrow Transplantation (EBMT)-score,^{2, 3} the Dana-Farber Cancer Institute (DFCI)-score⁴ and a combination of such scores.⁵ Further improvement of pre-alloSCT risk assessment could facilitate clinical decision making.

Endothelial dysfunction plays a crucial role in the pathophysiology of major complications contributing to non-relapse mortality (NRM) of alloSCT, such as sepsis, graft-versus-host disease (GVHD), sinusoidal obstruction syndrome (SOS/VOD) and transplant-associated microangiopathy (TAM).^6–8 Recently, a positive correlation between GVHD mortality and TAM has been demonstrated.^{9, 10} Moreover, elevated serum levels of the endothelium-related protein suppressor of tumorigenicity-2 (ST2) have been associated to GVHD-related mortality and TAM after alloSCT.^10–13

Risk assessment based on quantification of endothelial dysfunction prior to alloSCT is an attractive option that could help predicting alloSCT-associated mortality. Evidence is accumulating that pre-alloSCT measurement of patient-related endothelial risk factors, such as single-nucleotide-polymorphisms (SNPs) of the thrombomodulin and the CD40 ligand genes, complement activation-related genes, and angiopoetin-2 serum levels, can be used to predict outcome after acute GVHD.^14–17 However, general clinical application of these markers for alloSCT risk assessment in the near future is hindered by a lack of standardization and cost-effectiveness.

Our current goal was to find a biomarker panel related to endothelial dysfunction for pre transplant OS prediction that consists of standardized routine laboratory parameters in order to enable broad clinical use. To this end, the ‘Endothelial Activation and Stress Index’ (EASIX; [(LDH × creatinine) / thrombocytes]), has been recently shown to predict the risk of death after acute GVHD if assessed at clinical GVHD onset.¹⁸ Here we evaluate if EASIX can predict alloSCT outcome if assessed already prior to conditioning for alloSCT (EASIX-pre).

Patients and Methods

Patients

Patients from five independent cohorts were included in this retrospective study. The training cohort (cohort I) contained 755 adult patients who had undergone alloSCT at the University of Heidelberg between 09/2001 and 06/2014. Statin based endothelial protection (SEP) with pravastatin 20 mg/d and ursodeoxycholic acid (UDA) 3×250 mg/d was introduced in 01/2010 in the training cohort only. The validation cohort comprised 1267 adult patients who were allografted in 3 independent centers (pooled cohorts II-IV): Cohort II was transplanted at the Charité, Campus Benjamin Franklin, Berlin between 08/1995 and 12/2011. Cohort III consisted of adult patients who had undergone alloSCT at the Seattle Fred Hutchinson Cancer Research Centre between 01/2010 and 12/2013. Cohort IV consisted of adult patients transplanted between 01/2009 and 12/2013 at the University Hospital Essen. In addition, 262 pediatric and young adult patients who had undergone alloSCT at the Cincinnati Children’s Hospital Medical Centre between 01/2010 and 09/2014 were analyzed (cohort V). Written informed consent according to the declaration of Helsinki was obtained in all eligible patients and the study was approved by the responsible Institutional Review Boards.

Diagnosis of transplant-associated thrombotic microangiopathy (TAM)

The diagnosis of TAM was assigned based on laboratory values by performing chart review (cohorts I and V). TAM was diagnosed on the basis of BMT/CTN Toxicity Committee Consensus Definition for TAM⁶ if all of the following parameters were present: a 50% increase in serum lactate dehydrogenase (LDH) levels (or a pre-existing LDH above 400 U/L), a 50% drop in platelet counts (or a pre-existing platelet count below 50/nl), an otherwise unexplained 50% rise in creatinine or neurological symptoms along with at least 2 schistocytes/HPF, and a negative Coombs test.

Determination of serum levels of endothelium-related factors

A detailed description of the methodology is given in the Supplementary online material.

Statistical analysis

EASIX was calculated by the formula::LDH (U/l) x Creatinine (mg/dL) / Thrombocytes (nL). For the primary statistical analysis, we used the log2 transformed index, log2(EASIX) = log2(LDH) + log2(Creatinine) - log2(Thrombocytes). For univariable and multivariable analysis of overall survival (OS) the proportional hazards model of Cox was used. Competing risks of TAM, acute GVHD, time to relapse (TTR), and non-relapse mortality (NRM) were analyzed by univariable and multivariable proportional cause specific hazards regression. Hazard ratios (HR) for Cox models and cause-specific hazard ratios (CSHR) for the competing risks models were computed to describe the prognostic effect of EASIX. For the evaluation of prediction accuracy of prognostic models, prediction error estimates were calculated for all event times using a time-dependent adaption of the Brier score and the concordance index.¹⁹ The univariable Heidelberg-model was assessed on the validation set directly, whereas for the models “No EASIX (multi)” and “EASIX trained (multi)” .632 (concordance index) respectively .632+ (Brier score) bootstrap estimates were used. Detailed statistical methods are given in the online Supplementary.

Data sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to anonymized data may be granted following review.

Results

Patient characteristics

Patient characteristics by cohort are given in Table 1. For the adult cohorts, these were largely similar except for the Essen cohort which consisted exclusively of patients with acute myeloid leukemia and myelodysplastic syndromes. Patient characteristics in the pediatric cohort were profoundly different with a median age of seven years and more than 70% non-malignant diseases as underlying indication for alloSCT.

Table 1:

Patient characteristics by cohort.

	Cohort I Heidelberg (training) n= 755	Cohort II Berlin n=386	Cohort III Seattle n=446	Cohort IV Essen n=435	Cohort V (mainly paediatric) Cincinnati n=262
Year of alloSCT	2001–2014	1992–2011	2010–2013	2009–2013	2010–2014
Age (years) at alloSCT (median, range)	53 (17–75)	51 (18–77)	52 (18–78)	57 (17–73)	7 (0–33)
Recipient sex	(%)	(%)	(%)	(%)	(%)
female	295 (39)	155 (40)	176 (39)	216 (50)	101 (39)
male	460 (61)	231 (60)	270 (61)	219 (50)	161 (61)
Donor sex
female	255 (34)	152 (39)	206 (46)	148 (34)	n.a.
male	500 (66)	234 (61)	240 (54)	287 (66)	n.a.
Donor
MRD	238 (32)	118 (31)	147 (33)	81 (19)	69 (27)
MUD	346 (46)	243 (63)	249 (56)	236 (54)	133 (51)
MMUD	152 (20)	25 (6)	50 (11)	103 (24)	56 (22)
Haplo/MMRD	19 (3)		0	15 (3)	2 (1)
n.a.					2
Disease
AML, MDS	309 (41)	219 (57)	297 (66)	435 (100)	44 (17)
ALL	63 (8)	45 (12)	75 (16)	0	24 (9)
Lymphoma, CLL	219 (29)	60 (15)	53 (12)	0	3 (1)
MPN	27 (4)	28 (7)	21 (5)	0	2 (1)
MM, Amyloidosis	128 (17)	24 (6)	0	0	0
AA, PNH, BM failure	8 (1)	10 (3)	0	0	66 (25)
Immune deficiency	0	0	0	0	110 (42)
Genetic	0	0	0	0	13 (5)
n.a.	1	0	0	0
Disease stage before alloSCT (EBMT Risk Score criteria)2
0	234 (32)	158 (42)	292 (65)	197 (45)
1	196 (27)	141 (38)	70 (16)	97 (22)	n.a.
2	300 (41)	77 (20)	84 (18)	140 (32)
n.a.	25	10		1
Stem cell source
Peripheral blood	706 (94)	377 (98)	388 (87)	395 (91)	44 (17)
Bone marrow	49 (6)	9 (2)	58 (13)	40 (9)	218 (83)
Conditioning
RIC	602 (80)	242 (63)	243 (54)	384 (89)	120 (46)
MAC, Aplasia conditioning	150 (20)	144 (37)	203 (46)	49 (11)	142 (54)
n a	3			2

Abbreviations:

AA: aplastic anaemia, ALL: acute lymphoblastic leukaemia, AML: acute myelogenous leukaemia, Aplasia conditioning: sequential intermediate intensity chemotherapy followed by RIC, CLL: chronic lymphocytic leukaemia, MAC: myeloablative conditioning, MDS: myelodysplastic syndrome, MM: multiple myeloma, MMUD: mismatched unrelated donor, MPS: myeloproliferative syndrome, MUD: matched unrelated donor, NA: not available; NRM: non-relapse mortality, PD: progressive disease, PNH: paroxysmal nocturnal haemoglobinuria, MRD: matched related donor, RIC: reduced intensity conditioning, SCT: stem cell transplantation, MMRD: mismatched related donor

EASIX-pre predicts risk of TAM after alloSCT

Data on TAM incidence were available exclusively in cohorts I and V. In the Heidelberg cohort, EASIX-pre was significantly associated with increased hazards of TAM in univariable and multivariable analysis (CSHR 1.24, 95% CI 1.03–1.50, for a 2-fold change, p=0.03; CSHR 1.21, 95% CI 1.00–1.47, for a 2-fold change, p=0.05; Table 2, descriptively depicted in Suppl. Figure 1). In multivariable Cox regression, statin based endothelial protection (SEP) and male donor sex significantly associated with lower risk of TAM. In the pediatric cohort, EASIX-pre was a significant predictor of risk of TAM in univariable analysis (CSHR 1.12, 95% CI 1.02–1.23, for a 2-fold change, p=0.02) but not in multivariable analysis (CSHR 1.09, 95% CI 0.97–1.21, for a 2-fold change, p=0.14, Suppl. Table 1). As a probable explanation, we found a strong correlation of EASIX and age in children (Suppl. Figure 2A) but not in adults (Suppl. Figure 2B).

Table 2:

Prognostic factors for TAM (Heidelberg cohort I, n=755, events=41)

A) univariable
	TAM
	CSHR	95% CI	p
EASIX pre conditioning (2-fold change)	1.24	1.03–1.50	0.03
B) multivariable
	TAM
	CSHR	95% CI	p
EASIX pre-conditioning (2-fold change)	1.21	1.00–1.47	0.05
SEP (yes vs. no)	0.37	0.18–0.75	0.006
Age at alloSCT (per year)	1.02	0.99–1.06	0.14
disease score (2 vs. other)	1.80	0.94–3.44	0.08
Donor (MRD vs other)	0.96	0.49–1.89	0.91
Donor sex (male vs. female)	0.52	0.28–0.99	0.05
Recipient sex (male vs. female)	1.40	0.71–2.77	0.33
Conditioning (RIC vs. MAC)	0.70	0.33–1.51	0.36
Stem cell source (BM vs. PBSC)	0.75	0.18–3.13	0.69

Abbreviations:

MRD, matched related donor, RIC, reduced intensity conditioning, MAC: myeloablative conditioning, BM: bone marrow, PBSC: peripheral blood stem cells, SEP: statin-based endothelial protection

EASIX-pre predicts overall mortality and associates with NRM in adult alloSCT recipients

In the training cohort, EASIX-pre was significantly correlated with overall survival (OS), NRM, but not TTR on uni- and multivariable Cox regression analyses (univariable per log2 increase: OS HR 1.14 95% CI 1.06–1.23, p<0.001; NRM HR 1.23 95% CI 1.10–1.38, p<0.001; TTR HR 0.98, 95% CI 0.93–1.11, p=0.744; visualized by quartiles in Figure 1A and Suppl. Figure 3, multivariable: Table 3).

Figure 1. Visualization of the univariable outcome analysis according to EASIX-pre quartiles A) training cohort, n=755, B) validation cohort (adult cohorts II-IV, n=1267).

Overall survival (OS), non-relapse mortality (NRM) and time to relapse (TTR) are shown according to the EASIX-pre quartiles raised in the respective cohorts. The high quartiles 4 (blue) and 3 (green) associate with lower OS and higher NRM in the training cohort and with lower OS, higher NRM and higher TTR in the validation cohort.

Table 3:

EASIX before conditioning therapy is a prognostic factor for OS and NRM after alloSCT the training cohort. 313 out of 755 patients died within the study. 743 patients were evaluable for the competing risk analysis of NRM and TTR, where 241 patients relapsed and 131 died in remission.

A. Univariable analysis	OS			NRM			TTR
	HR	95% CI	p	CSHR	95% CI	p	CSHR	95% CI	p
EASIX pre conditioning (2-fold change)	1.14	1.06–1.23	<0.001	1.23	1.10–1.38	<0.001	1.02	0.93–1.11	0.744
B. Multivariable analysis	OS			NRM			TTR
	HR	95% CI	p	CSHR	95% CI	p	CSHR	95% CI	p
EASIX pre-conditioning (2-fold increase)	1,09	1,01–1,18	0,033	1,21	1,07–1,36	0,002	0,98	0,90–1,08	0,738
Age at alloSCT	1,01	1,00–1,02	0,080	1,02	1,01–1,04	0,004	1,00	0,99–1,01	0,838
Disease stage before alloSCT (EBMT Risk Score criteria)2 (high risk 2 vs. lower risk 0 or 1)	1,78	1,41–2,25	<0,001	1,18	0,82–1,69	0,363	2,06	1,58–2,69	<0,001
Donor (RD vs other)	1,21	0,94–1,55	0,147	1,14	0,77–1,69	0,500	1,14	0,86–1,52	0,356
Donor sex (male vs. female)	1,03	0,81–1,31	0,822	0,82	0,57–1,18	0,279	1,13	0,85–1,50	0,388
Recipient sex (male vs. female)	1,02	0,81–1,30	0,840	1,09	0,75–1,58	0,636	0,95	0,72–1,24	0,688
Conditioning (RIC vs. MAC)	0,73	0,54–0,99	0,041	0,63	0,40–1,00	0,048	0,91	0,64–1,30	0,616
Diagnosis (lymphoid vs myeloid)	1,02	0,79–1,32	0,857	1,33	0,90–1,96	0,157	1,09	0,82–1,46	0,556

Abbreviations: OS: overall survival, NRM: non-relapse mortality, TTR: time to relapse, RD, related donor, RIC, reduced intensity conditioning, MAC: myeloablative conditioning, BM: bone marrow, PBSC: peripheral blood stem cells

In the pooled validation cohort, similar uni- and multivariable effects on OS and NRM were observed (univariable per log2 increase: OS HR 1.25 95% CI 1.20–1.31, p<0.001; NRM HR 1.34 95% CI 1.24–1.42, p<0.001, visualized by quartiles in Figure 1B, multivariable: Table 4). In contrast to the training cohort, TTR was also associated with EASIX-pre in the validation cohort: (univariable per log2 increase HR 1.13, 95% CI 1.06–1.20, p<0.001, visualized by quartiles in Figure 1B; multivariable: Table 4).

Table 4:

EASIX before conditioning therapy is a prognostic factor for OS and NRM after alloSCT the validation cohort. 597 out of 1267 patients died within the study. 1265 patients were evaluable for the competing risk analysis of NRM and TTR, where 366 patients relapsed and 310 died in remission.

A. Univariable analysis	OS			NRM			TTR
	HR	95% CI	p	CSHR	95% CI	p	CSHR	95% CI	p
EASIX pre conditioning (2-fold change)	1.25	1.20–1.31	<0.001	1.34	1.26–1.42	<0.001	1.13	1.06–1.20	<0.001
B. Multivariable analysis	OS			NRM			TTR
	HR	95% CI	p	CSHR	95% CI	p	CSHR	95% CI	p
EASIX pre-conditioning (2-fold increase)	1,23	1,17–1,28	<0,001	1,31	1,23–1,40	<0,001	1,11	1,04–1,18	0,003
Age at alloSCT	1,01	1,00–1,02	0,003	1,02	1,01–1,01	<0,001	1,00	0,99–1,01	0,494
Disease stage before alloSCT (EBMT Risk Score criteria)2 (high risk 2 vs. lower risk 0 or 1)	1,27	1,06–1,53	0,011	1,21	0,94–1,56	0,147	1,13	0,89–1,44	0,325
Donor (RD vs other)	1,00	0,84–1,20	0,997	1,12	0,87–1,46	0,380	0,76	0,61–0,95	0,014
Donor sex (male vs. female)	1,25	1,01–1,55	0,037	1,18	0,87–1,58	0,285	1,41	1,08–1,84	0,011
Recipient sex (male vs. female)	0,97	0,82–1,15	0,747	1,12	0,88–1,41	0,354	0,90	0,73–1,11	0,317
Conditioning (RIC vs. MAC)	0,83	0,71–0,98	0,030	0,72	0,58–0,91	0,006	0,99	0,80–1,22	0,897
Diagnosis (lymphoid vs myeloid)	1,05	0,85–1,30	0,622	1,22	0,92–1,62	0,171	0,92	0,70–1,20	0,528

Abbreviations: OS: overall survival, NRM: non-relapse mortality, TTR: time to relapse, RD, related donor, RIC, reduced intensity conditioning, MAC: myeloablative conditioning, BM: bone marrow, PBSC: peripheral blood stem cells

Although Figures 1A and 1B suggest the existence of a clinically relevant cut-off (4^th quartile), using maximally selected log rank statistics we weren’t able to identify a cut-off consistent across all adult cohorts. Therefore, the data analysis was pursued with the continuous EASIX-pre score.

The prediction errors with the offset of the univariable training model with endpoint OS showed lower values with EASIX-pre as compared to the models without EASIX-pre (Figure 2A), and a concordance index of 0.629 at 12 months (Figure 2C). The univariable Heidelberg model was thus validated and forms the basis of the EASIX-pre calculator (http://biostatistics.dkfz.de/EASIX/). For multivariable analysis comparing training and validation cohort, the confounders were adjusted to the validation cohort and only the effect of EASIX was assessed. In this setting, prediction errors and concordance index showed evidence for model validation of the EASIX-pre effect (endpoint OS, Figure 2 B, C).

Figure 2. Model validation of EASIX as predictor of overall mortality.

(A) Univariable model: The prediction error curve for EASIX-pre (red curve) developed in the Heidelberg training cohort of adult alloSCT recipients was computed for 36 months after alloSCT and compared with the reference (marginal Kaplan-Meier estimate, black curve). (B) Multivariable model: Confounders were adjusted to the validation cohort. The prediction error of EASIX-pre (green curve) developed in the multivariable Heidelberg training cohort was computed for 36 months and compared to the prediction error of the multivariable model without EASIX-pre (blue curve).

A lower prediction error curve in the model including EASIX (red curve below black curve / green curve below blue curve) supports the usefulness of EASIX for predicting prognosis.

C) Time-dependent concordance indices for overall survival. Color coding as in A and B: Highest concordance indices are found in the models including EASIX-pre (red: univariable, green: multivariable model with confounders adjusted to validation cohort and EASIX-pre, blue: multivariable model with confounders adjusted to validation cohort and without EASIX-pre). A concordance index of 0·5 (dotted line) implies random concordance. A concordance index above 0·6 is regarded as acceptable.

In the pediatric cohort, increasing EASIX-pre was significantly associated with low OS and high NRM but not TTR only in univariable analysis (OS: HR 1.17, 95%CI 1.05–1.31, p=0.004; NRM: CSHR 1.19, 95%CI 1.06–1.34, p=0.004; TTR: CSHR 1.16, 95%CI 0.97–1.39, p=0.10) (Suppl. Figure 3D) but not on multivariable analyses (Suppl. Table 2). Again, the probable explanation is the strong correlation of EASIX and age in children (Suppl. Figure 2A) but not in adults (Suppl. Figure 2B).

In order to assess the prognostic value of the single EASIX components we tested LDH, creatinine and thrombocyte counts in a multivariate Cox regression model with endpoint NRM for the combined cohort of all adult patients (n=2022). All three components were significantly associated with the hazard of NRM (suppl. Table 3).

EASIX-pre and acute GVHD risk

We investigated if EASIX-pre associates with acute GVHD grade in the combined adult training and validation cohorts (n=2026). Although not significant in uni- and multivariable analyses, higher EASIX-pre tended to be associated with higher risk of grade 3–4 acute GVHD (p=0.08 univariable, p=0.10 multivariable) (Suppl. Table 4, Suppl. Figure 4).

EASIX-pre predicts mortality independent of established scores (HCT-CI and EBMT)

The HCT-CI score was exclusively available for patients from cohort III (Seattle) and was categorized in 4 groups: 1=score 0; 2= score 1+2, 3= score 3–4, 4=score ≥5. Suppl. Figure 5 demonstrates that EASIX predicted OS after alloSCT independently from the HCT-CI score. The EBMT-score was available for adult patients of the training cohort and was categorized in three groups: 1= score 1+2, 2= score 3, 3= score >3.² We found that EASIX predicted OS after alloSCT independently from the EBMT score (Suppl. Figure 6).

Correlation of EASIX-pre with endothelium-related serum factors

In a subset of patients from the Heidelberg and Essen cohorts, stored serum samples were available for measurement of serum concentrations of endothelium-related serum factors before alloSCT. A positive correlation of serum levels of CXCL8/IL8 and free IL-18, but an inverse correlation of IGF-1 serum levels and the IGF1/IGFBP4 ratio with EASIX-pre was found in both cohorts (Suppl. Figures 7 and 8). In contrast, angiopoietin-2, serum nitrates and ST2 did not correlate with EASIX-pre (Suppl. Figure 9). As these markers have been shown to predict outcome exclusively in patients developing acute GVHD,^{13, 15, 20} we assessed if the predictive value of EASIX-pre also extended to those patients who never developed acute GVHD. Suppl. Figure 10 shows that EASIX was also associated with NRM risk in this cohort of patients without acute GVHD.

Discussion

In this study we establish EASIX-pre, calculated prior to alloSCT by the simple formula ‘(LDH[U/L] × creatinine[mg/dL] ) / thrombocytes/nl]’, for prediction of mortality after alloSCT based on clinical data of 5 independent patient cohorts. EASIX is a simple score based on three laboratory parameters used worldwide. All three parameters are independently associated with the risk of NRM (suppl. Table 3). The present large international data set provides the opportunity to generate a generally accepted prognostic score that focuses on endothelial complications in patients with alloSCT.

EASIX-pre has to be put in perspective with already well-established alternatives to forecast alloSCT mortality. Currently, three clinical scores to predict alloSCT-associated mortality are in clinical use. The HCT-CI focuses exclusively on patient-related factors and includes the history or presence of different pathologic conditions.¹ In contrast, the EBMT score consists of patient and donor data including histocompatibility, stage of disease, age and sex of donor and recipient, and time from diagnosis to transplantation.^{2, 3} Both scores are helpful in clinical practice and a combination of both may even increase accuracy.⁵ Finally, the DFCI score focusses on disease and disease status to predict mortality.⁴ Comparing EASIX with the HCT-CI and EBMT scores, respectively, we observed an independent prognostic value of EASIX. Interestingly, there was a tendency of improved prediction when these scores were applied in combination (Suppl. Figures 5 and 6). However, further analyses are needed for precisely defining the synergies and overlaps between the scores.

Endothelial dysfunction is a common pathomechanism of several severe infectious and non-infectious alloSCT-related complications. There are no standard laboratory parameters available that are specific for endothelial dysfunction. Nevertheless, we chose creatinine, LDH and thrombocyte counts for EASIX because they were already established for TAM diagnosis, making them the most obvious routine laboratory parameters that are connected to endothelial pathology.^{6, 8} High creatinine levels link endothelial pathology to renal function, as endothelial pathomechanisms play a role in conditions such as acute renal failure, glomerulonephritis, diabetic nephropathy and transplant glomerulonephropathy.²¹ LDH is involved because endothelial activation leads to release of LDH from endothelial cells as well as to a higher turnover of circulating cells, such as thrombocytes and leukocytes, resulting in elevated LDH serum levels.²² Low platelet counts are a result of endothelial damage and complement activation in many diseases, such as thrombotic microangiopathies,²³ chronic GVHD,²⁴ adult respiratory distress syndrome (ARDS)²⁵ and other critically ill conditions. It is undisputed that a multitude of explanations exists for isolated alterations of each creatinine, LDH, and thrombocyte counts, including side effects of chemotherapy, infections, supportive medication, transfusion, fluid intake, spleen size or other manifestations of the underlying disease. This caveat also applies for TAM, illuminating the challenge of diagnosing the disease and the need for additional diagnostic parameters such as schistocytes or low haptoglobin.⁶ However, EASIX is a prognostic rather than a diagnostic tool designed to be applicable with minimal costs in all transplantation centers.

Moreover, this study suggests that EASIX-pre might be a prognostic factor for TAM, a condition characterized by extensive endothelial dysfunction and high mortality.^{12, 26} Interestingly, statin based endothelial protection remained a significant factor associating with TAM in the training cohort also in the context of EASIX-pre supporting our previous observation.¹⁰ Furthermore, EASIX-pre correlates with serum markers of endothelial dysfunction, such as CXCL8 (IL-8), free IL18, and (inversely) IGF-1. IL-8/CXCL8 is relevant for attraction of leukocytes to the endothelium as well as for transmigration.^{27, 28} IL-18 has been demonstrated to inhibit endothelial function and has been connected to endothelial pathology in vascular diseases.^29–32 In contrast, IGF-1 promotes physiologic function of endothelial cells including migration, tube formation and production of nitric oxide.³³ Indeed, low pre-alloSCT serum levels of IGF-1 have been found to be associated with an increased risk of SOS/VOD.³⁴ Thus, the link between EASIX-pre and various markers of endothelial dysfunction as well as TAM risk suggests that EASIX-pre is at least partly driven by endothelial activation or dysfunction.

A strength of our study is the simplicity of the approach as well as the validation of findings in large cohorts of alloSCT recipients in Europe and the USA. A limitation is that the applicability of EASIX-pre could not be proven for pediatric transplants. One possible explanation might be that 72% of pediatric patients suffered from non-malignant diseases. Age-specific normal values for LDH and creatinine as reflected by the strong correlation between of age and EASIX-pre in children may be one explanation for this. Larger cohorts of children of different age subgroups are required to understand the impact of EASIX in pediatric patients. Therefore, the current version of the EASIX online calculator can exclusively be used to predict mortality in adult alloSCT recipients. It must also be acknowledged that small differences in LDH cut-off levels exist in different laboratories.

EASIX-pre is a predictor of mortality based on readily available standard routine lab parameters. In clinical routine, pre-transplant EASIX determination supports the assessment of the individual transplant risk. To design clinical trials on prophylaxis or early treatment of endothelial complications after alloSCT (e.g. by statins, defibrotide, or others), EASIX can be used to define a high risk cohort.