Prediction of Veno-Occlusive Disease Using Biomarkers of Endothelial Injury

Article Outline

• Abstract
• Introduction
• Methods
  • Statistical Analysis
• Results
  • Comparison of Biomarkers
  • Analysis of Receiver Operator Characteristic Curves
• Discussion
• Acknowledgments
• Authorship Statement
• References
• Copyright

Predicting the development of veno-occlusive disease (VOD) of the liver remains challenging. We hypothesized that biomarkers of endothelial injury in myeloablative allogeneic transplantation recipients could predict VOD occurrence. We evaluated 4 biomarkers—von Willebrand Factor (vWF), thrombomodulin, E-selectin, and soluble intercellular adhesion molecule-1 (sICAM-1)—weekly in the peritransplantation period in an attempt to predict VOD. In the patients who received sirolimus, vWF, thrombomodulin, and sICAM-1 levels were significantly elevated in patients with VOD compared with those without VOD on day −1 (P ≤ .035), day +7 (P ≤ .0001), and day +14 (P ≤ .004). E-selectin was predictive on day +7 (P = .007). Levels of vWF ≥1400 IU/mL and thrombomodulin ≥100 ng/mL on day +7 were both 100% sensitive and 100% specific in predicting VOD. These biomarkers were informative when adjusted for other risk factors for VOD in regression analysis. Among patients not receiving sirolimus, biomarkers of endothelial injury were not informative. We conclude that vWF, thrombomodulin, and sICAM-1 elevations before and early after transplantation may be useful in predicting VOD in patients receiving sirolimus.

Key Words: Veno-occlusive disease, Endothelium, Sirolimus

Back to Article Outline

Introduction 

Veno-occlusive disease (VOD, also referred to as sinusoidal obstruction syndrome) of the liver occurs in 5%-15% of patients after myeloablative allogeneic hematopoietic stem cell transplantation (HSCT). VOD is thought to result from conditioning-related injury to hepatic sinusoidal endothelium and hepatocytes, compounded by cytokine-mediated effects related to allogenicity [1]. Although clinical risk factors for VOD are well established, precise prediction of VOD in individuals remains elusive.

In previous work, we demonstrated an increased frequency of VOD after sirolimus-based graft-versus-host disease (GVHD) prophylaxis (relative risk [RR], 1.55, P = .33 without concomitant methotrexate [MTX]; RR, 2.35, P = .005 with concomitant MTX) [2]. Sirolimus may act as an endothelial toxin or may prevent endothelial repair after conditioning-related or mechanical injury. It is commonly used to coat endovascular stents to prevent restenosis [3] and has been associated with another endothelial injury syndrome, thrombotic microangiopathy, after transplantation [4].

We hypothesized that the occurrence of VOD can be predicted by measuring biomarkers of endothelial injury, particularly in patients receiving sirolimus therapy.

Back to Article Outline

Methods 

We performed a retrospective analysis of biomarkers of endothelial injury using banked plasma and serum samples collected weekly in the peritransplantation period, with clinical VOD as the outcome of interest. We selected 4 biomarkers—von Willebrand factor (vWF), thrombomodulin, soluble intracellular adhesion molecule-1 (sICAM-1), and E-selectin—based on their association with VOD, known endothelial expression pattern, and ability to be measured in stored plasma or serum. The biomarkers were measured using commercially available ELISA kits (vWF: American Diagnostica, Greenwich, CT; thrombomodulin: Diagnostica Stago, Parsippany, NJ; sICAM-1 and E-selectin: R&D Systems, Minneapolis, MN). vWF and thrombomodulin were assayed in plasma, and sICAM-1 and E-selectin were assayed in serum.

All patients in the study group underwent myeloablative HSCT using cyclophosphamide and total body irradiation, as described previously [5]. In brief, cyclophosphamide (1800 mg/m² on days −5 and −4) was administered, followed by total body irradiation at 14.0 Gy, delivered in 7 fractions at a dose rate of 10 cGy/min. Lead blocks were used to compensate for lung absorption. Tacrolimus was administered at 0.02 mg/kg/day i.v. by continuous infusion beginning on day −3, with a target serum concentration of 5-10 ng/mL. Sirolimus was administered as a 12-mg oral loading dose on day -3, followed by a 4-mg/day single dose, with a target serum concentration of 3-12 ng/mL as assessed by high-performance liquid chromatography. Recipients of matched related and matched unrelated grafts were included.

Patients were selected to represent 2 GVHD prophylaxis groups: sirolimus/tacrolimus (SIR⁺) and tacrolimus/MTX (SIR^-) with or without VOD (VOD⁺/VOD^-). A sufficient number of patients were randomly selected from our database to ensure the availability of least 10 samples for assay at each of 3 time points (days -1, +7, and +14); however, not all patients in groups other than the SIR⁺VOD⁺ reference group had serum and plasma measurements at each time point. Assays were performed before the clinical development of VOD. VOD was diagnosed based on the Baltimore criteria [6], with diagnosis confirmed by Doppler ultrasonography and/or liver biopsy with wedged hepatic venous pressure gradient measurement.

Statistical Analysis 

All assays were performed in duplicate, and the results presented here are the mean of 2 assays. The 2-sided exact Wilcoxon rank-sum test was used for comparison of continuous variables, and the 2-sided Fisher exact test was used for comparison of categorical variables. All biomarkers were first evaluated at each time point. To establish a cutoff value for predictive biomarkers, analysis of the receiver operator characteristic (ROC) curve was performed at each time point. To assess whether the cutoff value determined in the ROC analysis predicted the occurrence of VOD in the presence of other prognostic factors, exact logistic regression analysis was performed at each time point, adjusting for age, recipient–donor sex mismatch, and donor type. In addition, to test for a group difference (ie, VOD⁺ vs VOD^-) over time, a mixed model for repeated measures was explored for each biomarker using PROC MIXED in SAS version 9.2 (SAS Institute, Cary, NC). The level of each biomarker was log-transformed for modeling. All tests were 2-sided. Testing for multiple biomarkers was not adjusted for in the level of significance.

Back to Article Outline

Results 

Table 1 summarizes characteristics of the study patients. Significant intragroup differences in baseline characteristics can be seen, with SIR⁺ patients engrafting earlier than SIR^- patients (14 days vs 16 days; P <.01) and SIR⁺VOD⁺ patients being more likely to receive an unrelated donor graft and to experience delayed platelet recovery. Only 2 patients (1 SIR⁺VOD⁺ and 1 SIR⁺VOD^-) were exposed to gentuzumab ozogomycin before HSCT. A total of 61 patients were needed for the analysis to ensure the availability of 10 samples at each of the 3 analysis time points; however, in the SIR^-VOD⁺ group, only 9 patients with banked samples were ultimately identified. In the SIR⁺VOD⁺ group, all 10 patients had samples at all time points. The median time to development of VOD was 17 days (range, 11-28 days) in the SIR⁺ group and 21 days (range, 10-40 days) in the SIR^- group (P = .35).

Table 1. Clinical Characteristics of Patients Analyzed

	SIR+VOD+	SIR+VOD-	P	SIR−VOD+	SIR-VOD-	P
Sample size	12	26		9	14
Age, years, median (range)	45 (19-59)	42.5 (29-56)	.85	34 (19-51)	48.5 (31-58)	.03
Males sex, n (%)	9 (75)	8 (31)	.02	33	79	.006
Matched related donor, n (%)	2 (17)	22 (85)	<.001	33	64	<.001
Previous transplantation, n (%)	1 (8)	0	.32	0	0	1
Disease, n (%)			NS			NS
Acute myelogenous leukemia	4 (33)	9 (35)		5 (56)	6 (43)
Acute lymphoblastic leukemia	3 (25)	2 (8)		3 (33)	5 (36)
Chronic lymphocytic leukemia/small lymphocytic leukemia/prolymphocytic leukemia	0	0		0	1 (7)
Chronic myelogenous leukemia	0	8 (31)		0	1 (7)
Myelodysplastic syndrome	2 (17)	2 (8)		0	1 (7)
Multiple myeloma	0	1 (4)		0	0
Myeloproliferative disease	1 (8)	1 (4)		0	0
Non-Hodgkin lymphoma	2 (17)	3 (12)		1 (11)	0
Time from diagnosis to transplant, months, median (range)	5.7 (3.2-26.2)	4.9 (2.1-31.6)	.94	7.3 (2.0-34.3)	4.5 (2.7-17.1)	.25
Time to neutrophil recovery, days, median (range)	14 (13-20)	13 (10-21)	.13	17.5 (13-27)	16 (13-20)	.44
Time to platelet recovery, days, median (range)	29 (14-138)	15 (8-39)	.007	20 (15-102)	18 (14-26)	.33
Grade II-IV acute GVHD, n (%)	3 (27)	8 (32)	1	5 (63)	6 (43)	.66

NS indicates not significant.; SIR+, sirolimus/tacrolimus; SIR−, tacrolimus/methotrexate; VOD+, with veno-occlusive disease; VOD−, without veno-occlusive disease; GVHD, graft-versus-host disease

Comparison of Biomarkers 

Significant differences in biomarker levels were detected between SIR⁺ patients with VOD and those without VOD. vWF, thrombomodulin, and sICAM-1 levels were significantly elevated in VOD⁺ patients compared with VOD^- patients on days -1, +7, and +14 (Table 2 and Figure 1). E-selectin level was significantly elevated only on day +7 (P = .007). A repeated-measures analysis performed for each predictive biomarker found that each was significantly associated with the occurrence of VOD when measured over time (vWF, P = .003 mU/mL; thrombomodulin, P = .002 ng/mL; sICAM1, P = .004 ng/mL).

Table 2. Summary of Biomarkers in Patients Receiving Sirolimus Therapy

SIR⁺ indicates sirolimus/tacrolimus; SIR−, tacrolimus/methotrexate; VOD+, with veno-occlusive disease; VOD−, without veno-occlusive disease; sICAM1, soluble intercellular adhesion molecule-1; VWF, von Willebrand factor.

• 
  • View Large Image
  • Download to PowerPoint

• Figure 1 

  Biomarker levels stratified by sirolimus exposure and VOD outcome. Line plots of biomarker levels and time from transplantation, stratified into 4 clinical groups. Confidence intervals represent interquartile ranges.

In contrast, biomarker levels did not differ significantly between the SIR^-VOD⁺ and SIR^-VOD^- groups at any of the time points tested (Figure 1), except for thrombomodulin level on day +7 (median, 46 vs 16; P = .0003) and day +14 (median, 43 vs 21.5; P = .02). This difference was not seen on the repeated-measures analysis, at least in partly because of the small sample size. Even though the significant differences in thrombomodulin level were seen between the SIR^-VOD⁺ and SIR^-VOD^- groups, the thrombomodulin level in the SIR^-VOD⁺ group was much lower than that in the SIR⁺VOD⁺ group (median, 227.5; P <.001).

Among patients without VOD, no differences in biomarker levels were detected between SIR⁺ and SIR^- patients. This indicates that in the absence of VOD, biomarkers of endothelial injury are not elevated even when sirolimus is administered.

Analysis of Receiver Operator Characteristic Curves 

A receiver operating characteristic (ROC) analysis of SIR⁺ patients found that a vWF level ≥1200 mU/mL was 100% sensitive and 90% specific for the development of VOD when measured on day -1 (area under the curve [AUC] = 0.99). At day +7, a vWF level ≥1400 mU/mL was 100% sensitive and specific for the development of VOD (AUC = 1). A thrombomodulin level ≥80 ng/mL on day -1 was 70% sensitive and 90% specific (AUC = 0.78), whereas a level ≥100 ng/mL by day +7 was 100% sensitive and 100% specific for the development of VOD (AUC = 1). Because each individual biomarker was very predictive, the simultaneous use of these 2 biomarkers did not add to the predictive capability of these assays compared with vWF alone (AUC = 0.99 at day −1 and 1 at day +7). sICAM1 also was useful for the prediction of VOD; a level ≥500 ng/mL on day -1, was 80% sensitive and 100% specific for VOD (AUC = 0.88), whereas on day +7, a level ≥400 ng/ml was 100% sensitive and 90% specific for VOD (AUC = 0.99). The simultaneous use of these 2 biomarkers did not add to the predictive capability of these assays.

In the SIR^- patients, a thrombomodulin level ≥35 ng/mL on day +7 was 100% sensitive and 100% specific for VOD (AUC = 1), whereas a vWF level ≥1200 ng/mL on day +7 was 100% sensitive and 70% specific for VOD (AUC = 0.72). But, because of the sample size constraints and the lack of reproducibility in repeated-measures analyses, these results must be interpreted with caution.

To investigate whether the proposed cutoff values are independently predictive of VOD in the presence of other risk factors, we performed multivariate exact logistic regression analysis, adjusting for such prognostic factors as age, recipient–donor sex mismatch, and donor type. Because of sample size constraints, we contructed 2 models, one with donor type and biomarker and the other with age, recipient–donor sex mismatch, and biomarker; both models used VOD as the outcome of interest. In the first model, the odds of VOD occurrence at day +7 in patients with a vWF level ≥1400 mU/mL or a thrombomodulin level ≥100 ng/mL was 5.65 (P = .004). For sICAM1, the odds of VOD occurrence in patients with a day +7 sICAM1 level ≥400 ng/mL was 4.45 (P = .036). Results of the regression analysis are presented in Table 3.

Table 3. Summary of Logistic Regression Models for the Occurrence of VOD

Back to Article Outline

Discussion 

Numerous previous attempts have been made to identify predictive markers for VOD. In the present study, we have identified markers of endothelial injury that can predict the occurrence of VOD in patients treated with sirolimus. Most previous studies focused on parameters of hemostasis and coagulation, because one of the prominent clinical features of VOD is microthrombus formation in the hepatic sinusoid. Several studies have demonstrated changes in hemostatic parameters at the time of VOD diagnosis, with elevated plasminogen activator inhibitor-1 (PAI-1) level the most common marker at the time of (but not before) the diagnosis of VOD 7, 8. PAI-1 has been shown to have both diagnostic and prognostic value in VOD [9]. We were unable to measure PAI-1 because we had no platelet-poor plasma cryopreserved for assay. Gerecitano et al. [10] suggested that homocysteine and prothrombin fragments 1 and 2 could predict VOD occurrence with modest sensitivity, and Scrobohaci et al. [11] found an association between low baseline levels of factor VII and protein C with increased risk for VOD. Other prospective studies found a decreased protein C level before the onset of VOD 12, 13, 14; however, in many cases, the protein C level was diminished even before conditioning began, suggesting that a low protein C level is a risk factor for VOD. Similarly, impaired vWF cleaving protease activity assessed before conditioning also may be a risk factor for VOD [15].

Several groups have attempted to use endothelial markers to predict VOD. Most were not able to demonstrate the utility of the same markers that we identified (vWF 11, 16, thrombomodulin [17], and E-selectin and sICAM-1 [18]) in predicting VOD. Of note, none of these studies included patients who had received sirolimus therapy.

The primary mechanism of injury in VOD is thought to be conditioning-related injury to the hepatic sinusoidal endothelium. The mechanism of injury in sirolimus-exposed patients may be more complicated, however. Conditioning-related injury may be potentiated by sirolimus (or a combination of sirolimus and tacrolimus), because sirolimus may accelerate the senescence of hepatic endothelial cells after conditioning-related injury [19]. The healing process after conditioning-related injury may be altered as well. During recovery after VOD, excess vascular endothelial growth factor may be released by the hepatic endothelium to promote healing and endothelial recovery [20]. Sirolimus may reduce vascular endothelial growth factor levels, hindering repair in this scenario [21]. Finally, the possibility exists that the pathogenesis of VOD is multifactorial, and that only a subset of patients sustain endothelial injury as the primary mechanism, with sirolimus accentuating this injury. This may explain why previous attempts to identify endothelial markers of injury have not been more fruitful.

In summary, we have demonstrate that VOD can be predicted as early as the time of stem cell infusion using markers of endothelial injury in patients receiving sirolimus. Our findings suggest that endothelial injury predates clinical VOD by at least several days to as much as weeks. The high sensitivity and specificity of these assays make them useful for real-time clinical testing and early clinical intervention. Using these markers, specific strategies for preemptive therapy (with agents that target endothelial injury 22, 23) in these patients should be explored, which in turn could improve outcomes in this disease. We plan on validating this analysis with a prospective evaluation of these biomarkers in the future.

Back to Article Outline

Acknowledgments 

Financial disclosure: This work was supported by the Ted and Eileen Pasquarello Research Fund and National Institutes of Health Grants AI29530, CA142106, and HL070149. Corey Cutler is supported by the Stem Cell Cyclists of the Pan-Mass Challenge.

Back to Article Outline

Authorship Statement 

Corey Cutler designed the study, cared for research subjects, collected data, analyzed data, and wrote and reviewed the manuscript. Haesook T. Kim analyzed data and wrote and reviewed the manuscript. Shaké Ayanian performed laboratory analysis and reviewed the manuscript. Gary Bradwin performed laboratory analysis and reviewed the manuscript. Carolyn Revta collected data and reviewed the manuscript. Julie Aldridge analyzed data and reviewed the manuscript. Vincent Ho cared for research subjects, collected data, and reviewed the manuscript. Edwin Alyea cared for research subjects and reviewed the manuscript. John Koreth cared for research subjects and reviewed the manuscript. Philippe Armand cared for research subjects and reviewed the manuscript. Robert Soiffer cared for research subjects and reviewed the manuscript. Jerome Ritz performed laboratory analysis and reviewed the manuscript. Paul G. Richardson designed the study, cared for research subjects, collected data, and reviewed the manuscript. Joseph H. Antin designed the study, cared for research subjects, and reviewed the manuscript.

Back to Article Outline

References 

1. Ho VT, Revta C, Richardson PG. Hepatic veno-occlusive disease after hematopoietic stem cell transplantation: update on defibrotide and other current investigational therapies. Bone Marrow Transplant. 2008;41:229–237
   • View In Article
   • CrossRef
2. Cutler C, Stevenson K, Kim HT, et al. Sirolimus is associated with veno-occlusive disease of the liver after myeloablative allogeneic stem cell transplantation. Blood. 2008;112:4425–4431
   • View In Article
   • CrossRef
3. Ardissino D, Cavallini C, Bramucci E, et al. Sirolimus-eluting vs uncoated stents for prevention of restenosis in small coronary arteries: a randomized trial. JAMA. 2004;292:2727–2734
   • View In Article
   • CrossRef
4. Cutler C, Henry NL, Magee C, et al. Sirolimus and thrombotic microangiopathy after allogeneic hematopoietic stem cell transplantation. Biol Blood Marrow Transplant. 2005;11:551–557
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (125 KB)
   • CrossRef
5. Cutler C, Li S, Ho VT, et al. Extended follow-up of methotrexate-free immunosuppression using sirolimus and tacrolimus in related and unrelated donor peripheral blood stem cell transplantation. Blood. 2007;109:3108–3114
   • View In Article
   • MEDLINE
6. Jones RJ, Lee KS, Beschorner WE, et al. Veno-occlusive disease of the liver following bone marrow transplantation. Transplantation. 1987;44:778–783
   • View In Article
   • MEDLINE
   • CrossRef
7. Park YD, Yasui M, Yoshimoto T, et al. Changes in hemostatic parameters in hepatic veno-occlusive disease following bone marrow transplantation. Bone Marrow Transplant. 1997;19:915–920
   • View In Article
   • MEDLINE
8. Salat C, Holler E, Kolb HJ, et al. Plasminogen activator inhibitor-1 confirms the diagnosis of hepatic veno-occlusive disease in patients with hyperbilirubinemia after bone marrow transplantation. Blood. 1997;89:2184–2188
   • View In Article
   • MEDLINE
9. Kaleelrahman M, Eaton JD, Leeming D, et al. Role of plasminogen activator inhibitor-1 (PAI-1) levels in the diagnosis of BMT-associated hepatic veno-occlusive disease and monitoring of subsequent therapy with defibrotide (DF). Hematology. 2003;8:91–95
   • View In Article
   • MEDLINE
   • CrossRef
10. Gerecitano J, Mathias C, Mick R, et al. Homocysteine and prothrombin fragment 1+2 levels in patients with veno-occlusive disease after stem cell transplantation. J Hematother Stem Cell Res. 2003;12:215–223
    • View In Article
    • MEDLINE
11. Scrobohaci ML, Drouet L, Monem-Mansi A, et al. Liver veno-occlusive disease after bone marrow transplantation changes in coagulation parameters and endothelial markers. Thromb Res. 1991;63:509–519
    • View In Article
    • Abstract
    • Full-Text PDF (785 KB)
    • CrossRef
12. Lee JH, Lee KH, Kim S, et al. Relevance of proteins C and S, antithrombin III, von Willebrand factor, and factor VIII for the development of hepatic veno-occlusive disease in patients undergoing allogeneic bone marrow transplantation: a prospective study. Bone Marrow Transplant. 1998;22:883–888
    • View In Article
    • MEDLINE
13. Faioni EM, Krachmalnicoff A, Bearman SI, et al. Naturally occurring anticoagulants and bone marrow transplantation: plasma protein C predicts the development of veno-occlusive disease of the liver. Blood. 1993;81:3458–3462
    • View In Article
    • MEDLINE
14. Tanikawa S, Mori S, Ohhashi K, et al. Predictive markers for hepatic veno-occlusive disease after hematopoietic stem cell transplantation in adults: a prospective single center study. Bone Marrow Transplant. 2000;26:881–886
    • View In Article
    • MEDLINE
    • CrossRef
15. Park YD, Yoshioka A, Kawa K, et al. Impaired activity of plasma von Willebrand factor–cleaving protease may predict the occurrence of hepatic veno-occlusive disease after stem cell transplantation. Bone Marrow Transplant. 2002;29:789–794
    • View In Article
    • MEDLINE
    • CrossRef
16. Collins PW, Gutteridge CN, O'Driscoll A, et al. von Willebrand factor as a marker of endothelial cell activation following BMT. Bone Marrow Transplant. 1992;10:499–506
    • View In Article
    • MEDLINE
17. Catani L, Gugliotta L, Vianelli N, et al. Endothelium and bone marrow transplantation. Bone Marrow Transplant. 1996;17:277–280
    • View In Article
    • MEDLINE
18. Sudhoff T, Heins M, Sohngen D, et al. Plasma levels of D-dimer and circulating endothelial adhesion molecules in veno-occlusive disease of the liver following allogeneic bone marrow transplantation. Eur J Haematol. 1998;60:106–111
    • View In Article
    • MEDLINE
    • CrossRef
19. Imanishi T, Kobayashi K, Kuki S, et al. Sirolimus accelerates senescence of endothelial progenitor cells through telomerase inactivation. Atherosclerosis. 2006;189:288–296
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (647 KB)
    • CrossRef
20. Iguchi A, Kobayashi R, Yoshida M, et al. Vascular endothelial growth factor (VEGF) is one of the cytokines causative and predictive of hepatic veno-occlusive disease (VOD) in stem cell transplantation. Bone Marrow Transplant. 2001;27:1173–1180
    • View In Article
    • MEDLINE
    • CrossRef
21. Sartelet H, Toupance O, Lorenzato M, et al. Sirolimus-induced thrombotic microangiopathy is associated with decreased expression of vascular endothelial growth factor in kidneys. Am J Transplant. 2005;5:2441–2447
    • View In Article
    • MEDLINE
    • CrossRef
22. Richardson PG, Murakami C, Jin Z, et al. Multi-institutional use of defibrotide in 88 patients after stem cell transplantation with severe veno-occlusive disease and multisystem organ failure: response without significant toxicity in a high-risk population and factors predictive of outcome. Blood. 2002;100:4337–4343
    • View In Article
    • MEDLINE
    • CrossRef
23. Benimetskaya L, Wu S, Voskresenskiy AM, et al. Angiogenesis alteration by defibrotide: implications for its mechanism of action in severe hepatic veno-occlusive disease. Blood. 2008;112:4343–4352
    • View In Article
    • CrossRef