A Novel Diagnostic and Prognostic Biomarker Panel for Endothelial ...

Biol Blood Marrow Transplant 22 (2016) 1573e1581

Biology of Blood and Marrow Transplantation journal homepage: www.bbmt.org

A Novel Diagnostic and Prognostic Biomarker Panel for Endothelial Cell DamageeRelated Complications in Allogeneic Transplantation Shotaro Tatekawa 1, 2, Akio Kohno 2, Kazutaka Ozeki 2, Koichi Watamoto 2, 3, Norihiro Ueda 2, 4, Yohei Yamaguchi 2, 5, Tsutomu Kobayashi 1, Isao Yokota 6, Satoshi Teramukai 6, Masafumi Taniwaki 1, Junya Kuroda 1, *, Yoshihisa Morishita 2, 7 1

Division of Hematology and Oncology, Department of Medicine, Kyoto Prefectural University of Medicine, Kyoto, Japan Division of Hematology and Oncology, JA Aichi Konan Kosei Hospital, Aichi, Japan 3 Department of Hematology, Komaki City Hospital, Aichi, Japan 4 Department of Hematology and Oncology, Nagoya University Graduate School of Medicine, Aichi, Japan 5 Department of Hematology and Oncology, Japanese Red Cross Nagoya Daini Hospital, Aichi, Japan 6 Department of Biostatistics, Graduate School of Medical Science, Kyoto Prefectural University of Medicine, Kyoto, Japan 7 Department of Internal Medicine, Seirei Hospital, Nagoya, Aichi, Japan 2

Article history: Received 8 March 2016 Accepted 20 May 2016 Key Words: Transplantation-related complications Endothelial cell damage Sinusoidal occlusive syndrome Transplantation-associated microangiopathy Biomarkers Angiopoietin 2 Allogeneic hematopoietic stem cell transplantation

a b s t r a c t Noninfectious transplantation-related complications (TRCs), such as graft-versus-host disease or TRC with endothelial cell damage (TRC-EC), remain as the major obstacle for successful allogeneic hematopoietic cell transplantation (allo-HCT). However, the diagnosis and prognosis for the emergence of these complications are difficult to define during the early post allo-HCT period. Here, we tried to generate a novel diagnostic system for TRC-EC by analyzing 188 adult patients who received allo-HCT. Our study found that the peripheral blood levels of angiopoietin 2 (ANG2), C-reactive protein (CRP), D-dimer, and thrombomodulin (TM) at the onset of TRCs were significantly associated with the development of TRC-EC. We next developed a composite biomarker panel incorporating the risk values of ANG2, CRP, D-dimer, and TM at the onset of TRCs, which classified these patients into 3 risk groups: low, intermediate, and high risk. As a result, the panel was useful not only for the diagnosis of TRC-EC with high specificity and sensitivity, but also for the prediction of the patients’ long-term outcome. The 5-year overall survival (OS) rates of patients in the low-, intermediate-, and high-risk groups since the occurrence from TRCs were 76.2%, 54.9%, and 26.9%, respectively, and the high-risk score was significantly associated with both poor OS (hazard ratio [HR], 5.60; 95% confidence interval [CI], 2.81 to 11.20; P < .01) and frequent nonrelapse mortality (HR, 19.75; 95% CI, 5.59 to 69.77; P < .01). Thus, the composite panel proposed in this study provides a powerful tool for the diagnosis of TRC-EC and for the prediction of survival for patients with TRC-EC after allo-HCT. Ó 2016 American Society for Blood and Marrow Transplantation.

INTRODUCTION Despite major progress as a curative procedure for patients with various hematological diseases, allogeneic (allo) hematopoietic cell transplantation (HCT) is still accompanied by various severe complications that are occasionally life threatening. The major complications at the early postealloHCT phase include noninfectious transplantation-related

Financial disclosure: See Acknowledgments on page 1580. * Correspondence and reprint requests: Junya Kuroda, MD, PhD, Division of Hematology and Oncology, Department of Medicine, Kyoto Prefectural University of Medicine, 465 Kajii-cho, Kamigyo-ku, Kyoto, 602-8566, Japan. E-mail address: [email protected] (J. Kuroda).

http://dx.doi.org/10.1016/j.bbmt.2016.05.018 1083-8791/Ó 2016 American Society for Blood and Marrow Transplantation.

complications (TRCs), such as acute graft-versus-host disease (GVHD) or TRC with endothelial cell damage (TRC-EC), including either transplantation-associated thrombotic microangiopathy (TAM) or sinusoidal obstruction syndrome (SOS) [1]. The severity of TRCs closely associates with eventual nonrelapse mortality (NRM). The diagnosis of TRCs, especially TRC-EC, has been made based on both clinical symptoms and laboratory data, including histopathology. However, the diagnosis often has to be empirical in daily practice. Indeed, although it is relatively noninvasive to obtain biopsied tissue samples from either skin or intestine, this is not always the case with TRC-EC, where the damage occurs in less-accessible organs, such as the central nervous

1574

S. Tatekawa et al. / Biol Blood Marrow Transplant 22 (2016) 1573e1581

system, liver, and kidneys, because of the patients’ poor condition. Therefore, it is necessary to identify biomarkers for the development of TRCs that are reliable, diagnostic, and predictive and that can be obtained in a noninvasive manner. The development of such a system for identification of TRCEC may provide the opportunity for prophylactic/preemptive treatment as well as provide information for an appropriate therapeutic approach. Previous studies have demonstrated that the combination of endothelial damage, impaired angiogenesis, and coagulation dysregulation play critical roles in both the development and treatment outcome of TRC-EC and GVHD [2-4]. We also previously reported that the pretransplantation plasma level of angiopoietin 2 (ANG2) is closely associated with the occurrence of TRC-EC but not with that of acute GVHD (aGVHD) [5]. In this study, our goal was to establish a biomarker panel, incorporating several markers of endothelial damage, coagulation, and inflammation, for the accurate diagnosis of TRC-EC at the beginning of the symptoms and to determine their prognostic significance for clinical outcomes. METHODS Patients One hundred eighty-eight consecutive patients who received an alloHCT between 2000 and 2013 at Konan Kosei Hospital were included in this study. The blood samples, taken between 3 days before and 3 days after the onset of symptoms of TRCs, were available from all enrolled patients. No patients had undergone any previous HCT within 6 months. Written informed consent was obtained from all patients according to the Declaration of Helsinki, and the institutional ethics committee of Konan Kosei Hospital approved this study. In terms of disease condition at allo-HCT, nonremission acute leukemia, nonremission malignant lymphoma, and myelodysplastic syndrome in refractory anemia with excess blasts II according to the World Health Organization classification were defined as the high-risk disease group [6], whereas the remaining participants were defined as the standard-risk group. The myeloablative and reduced-intensity conditioning (RIC) regimens were defined according to the criteria of the National Marrow Donor Program/Center for International Blood and Marrow Transplant Research operational definitions [7]. A mismatched graft included an HLA-allele mismatch and excluded cord blood transplantation. All cord blood transplantations were conducted with a serologically 4 to 6 HLA-A, -B, and -DR matched single cord blood unit. As the conventional prophylactic therapy for GVHD, calcineurin inhibitors (either cyclosporine A for HCT from an HLAmatched related donor or tacrolimus for other situations) and methotrexate were administered to all patients. GVHD Grading and the Definitions of Posteallo-HCT Complications TRCs occurring within 100 days were categorized into 3 groups; TRC-EC, aGVHD, and other complications (OC). TRC-EC included SOS, TAM, capillary leak syndrome (CLS), and idiopathic pulmonary syndrome (IPS), whereas OC included pre-engraftment immune reaction, engraftment syndrome, hemophagocytic syndrome, and organ dysfunction from other noninfectious causes. SOS was diagnosed according to the modified Seattle Criteria [8]; however, the classical criterion for the timing of onset was excluded. IPS, including diffuse alveolar hemorrhage, was diagnosed according to the National Heart, Lung, and Blood Institute working group criteria [9,10]. CLS was diagnosed clinically with the development of noneSOS-derived generalized fluid retention (ascites, pleural effusion, pericarditis, edema, etc.), which was refractory to diuretics and exhibited the classical criteria [11]. TAM was diagnosed according to the criteria proposed by the Blood and Marrow Transplant Clinical Trials Network [12]. Intestinal TAM (iTAM) was histopathologically diagnosed [13]. aGVHD was clinically and histologically diagnosed, and the maximum GVHD grade by day 100 was determined according to standard criteria [14,15]. We defined poor GVHD outcome as either the persistent existence of treatment-necessary active aGVHD on day 100 or death by GVHD within 100 days after allo-HCT. The definition of preengraftment immune reaction was previously described [16]. In cases where the patients experienced multiple types of TRCs, 2 TRCs were considered to be independent when the 2 events developed more than 2 weeks apart. In the cases where the TRC-ECs emerged simultaneously with aGVHD or in a sequential manner, we defined such cases as GVHD with endothelial cell damage (GVHD-EC). OC was not evaluated in cases where either TRC-EC or

Table 1 Patient Characteristics Characteristics

No. of patients Age at transplantation 50 yr > 50 yr Sex Female Male Disease Myeloid malignancy Others Disease risk Standard High HCT-CI 3 >3 Graft type Matched (included CB) Mismatched Conditioning intensity MAC RIC Conditioning regimen 1 BU 8 mg/kg > 8 mg/kg MEL 100 mg/m2 > 100 mg/m2 TBI 6 Gy >6 Gy Conditioning regimen 2 RIC FLU-CY-TBI FLU-MEL-TBI Other MAC CY-TBI BU-CY CA-CY-TBI MEL-TBI FLU-MEL140 mg/m2 FLU-MEL-TBI FLU-BU Other

Total

TRCs Cohort (n ¼ 149) TRCs without EC

TRC-EC

P Value

188

92

57

112 (60) 76 (40)

57 (62) 35 (38)

34 (60) 23 (40)

61 (32) 127 (68)

32 (35) 60 (65)

19 (33) 38 (67)

139 (74) 49 (26)

69 (75) 23 (25)

40 (70) 17 (30)

.86

116 (62) 72 (38)

59 (64) 33 (36)

29 (51) 28 (49)

.13

156 (83) 32 (17)

76 (83) 16 (17)

45 (79) 12 (21)

.67

151 (80) 37 (20)

76 (83) 16 (17)

39 (68) 18 (32)

.07

161 (86) 27 (14)

82 (89) 10 (11)

46 (81) 11 (19)

.23

152 (81) 36 (19)

76 (83) 16 (17)

51 (89) 6 (11)

.34

127 (68) 61 (32)

67 (73) 25 (27)

28 (49) 29 (51)

50 versus 50 years), sex, disease type (myeloid disease versus other), disease risk (high versus standard), hematopoietic cell transplantation-specific comorbidity index score (>3 versus 0 to 3), conditioning regimen (myeloablative versus RIC and the use of busulfan, melphalan, or total-body irradiation), and graft type (mismatched donor versus matched donor). Results are expressed as hazard ratios (HRs) with 95% confidence intervals (CIs). All tests were 2 sided and P < .05 was considered to be statistically significant. All statistical analyses were performed with EZR ver.1.21 (Saitama Medical Center, Jichi Medical University, Saitama, Japan) [17], which is a graphical user interface for R 3.0.2 (The R Foundation for Statistical Computing, Vienna, Austria).

RESULTS Patient Characteristics The patient characteristics are summarized in Table 1. Briefly, the median age was 46 years (range, 16 to 68 years) and the ratio of female to male was approximately 1:2. The median follow-up duration for survivors was 55 months (range, 4 to 176 months). The risks of primary diseases at the time of the HCT were standard in 116 and high in 72 patients. Three cases with T cell depletion by pretransplantation antithymocyte globulin treatment were included in the present study (1 acute myelogenous leukemia and 2 severe aplastic anemia). Twenty-seven of the 188 patients received a RIC regimen. The conditioning regimen included busulfan (>8 mg/kg) for 36 patients, melphalan (>100 mg/m2) for 61 patients, and total-body irradiation (>6 Gy) for 92 patients. Incidence of TRCs Overall, the cumulative incidence of TRCs by day 100 was 77.6%, including 30.2% of TRC-EC (57 patients), 31.4% of aGVHD, and 16.0% of OC. TRC-EC, aGVHD, and OC occurred at a median of 16.5 days (range, 3 to 53 days), 23 days (range, 10 to 50 days), and 11 days (range, 3 to 33 days) after HCT,

patients without TRC-EC

patients with TRC-EC 50

0 1

40

30

20

10

0

10

20

30

40

50

Low

2 3

Int

4 5 6

High

7 Total score

Figure 3. The frequency distribution of patients according to the total score based on the composite panel. The number of patients with TRC-EC (n ¼ 57) and without TRC-EC (n ¼ 131) are shown in the bars based on the total score of the composite panel.

1577

respectively (Figure 1A). More precisely, 57 TRC-ECs included 17 SOS, 10 CLS, 25 TAM/iTAM, and 5 IPS. The median onset days of SOS, CLS, TAM/iTAM, and IPS were 13 (range, 3 to 17), 11 (range, 5 to 38), 19 (range, 6 to 53), and 13 (range, 7 to 18), respectively. GVHD-EC developed in 14 patients, and 13 of these patients were diagnosed with aGVHD with TAM/iTAM. The cumulative incidences and the numbers of individual events at any time points are also shown in Figure 1B-E. Identification of Diagnostic Biomarkers for TRC-EC First, we evaluated the diagnostic values of candidate biomarkers for TRC-EC by comparing 57 patients with TRCEC and 131 patients without TRC-EC. As the result, the blood levels of ANG2, CRP, D-dimer, sIL2R, and TM at the onset of TRCs were significantly higher in patients who developed TRC-EC, whereas those of PIC or VEGF were not higher, based on a univariate analyses of logistic regression (Figure 2A). The cut-off values of 5 diagnostic biomarkers (ANG2, CRP, D-dimer, sIL2R, and TM) were determined by the ROC curves (Figure 2B). Furthermore, a combination of the values for ANG2, D-dimer, CRP, and TM levels, excluding the sIL2R levels, was found to be the best model for diagnosing the occurrence of TRC-EC using a multivariate analysis of logistic regression. The odds ratios of ANG2 (>4.3 ng/mL), Ddimer (>3.7 mg/mL), CRP (>4.3 mg/dL), and TM (>3.0 ng/mL) for TRC-EC were 18.40, 9.86, 4.14 and 3.12, respectively (Supplemental Table S1). Development of a Composite Panel for Diagnosing TRC-EC We next tried to develop a composite panel by incorporating risk values of the levels of ANG2, D-dimer, CRP, and TM. According to the odds ratio of each biomarker, we assigned a score of 3 to high ANG2 (>4.3 ng/mL), a score of 2 to high Ddimer (>3.7 mg/mL), and a score of 1 to high CRP (>4.3 mg/dL) and to high TM (>3.0 ng/mL). We divided the patients into 3 groups according to their total score based on the composite panel: low- (L) score group with a total score 0 or 1, intermediate- (Int) score group with a total score 2 to 4, and high- (H) score group with total score 5 to 7. The H-score group demonstrated high specificity (.99; 95% CI, to 95 to 1.00) and a high positive predictive value (.96; 95% CI, .85 to 1.00) for the diagnosis of TRC-EC (Table 2). Conversely, the L-score group reliably excluded the emergence of TRC-EC with high sensitivity (.97; 95% CI, .88 to 1.00) and with high negative predictive value (.98; 95% CI, .93 to 1.00). Indeed, 43 of the 45 patients in the H-score group developed TRC-EC, whereas only 2 of the 94 patients in L-score group developed TRC-EC (Figure 3). Role of the Composite Panel in the Diagnosis of aGVHD When we evaluated the role of the composite panel for the diagnosis of aGVHD, the sensitivity was .71 in patients with a total score 2 (Int-score plus H-score groups), and the specificity was .78 in patients with total score of 5 (H-score group) for the composite panel to diagnose grade III/IV GVHD in 188 patients. These findings suggest that the panel did not appear to be diagnostically relevant for aGVHD (Supplemental Table S2). However, in 75 patients who developed aGVHD as the first episode of TRCs, the composite panel successfully distinguished GVHD-EC from aGVHD without endothelial cell damageerelated events (Table 3). Indeed, the high sensitivity in patients with total score 2 (Int-score plus H-score groups) and the high specificity in patients with total score 5 (H-score group) for the complication of endothelial cell damage-related events were observed both in the entire GVHD cohort and in 21 patients

1578


Table 3 Composite Panel for the Diagnosis of GVHD-EC Grades III and IV aGVHD (n ¼ 21)

Subject

GVHD Cohort (n ¼ 75) Total Score 2

Total Score 5

Total Score 2

Total Score 5

Sensitivity Specificity PPV NPV DA

.86 .71 .40 .96 .73

.57 1.00 1.00 .91 .92

.92 .56 .73 .83 .76

.67 1.00 1.00 .69 .81

(.57-.98) (.57-.82) (.23-.59) (.85-1.00) (.62-.83)

(.29-.82) (.91-1.000) (.52-1.000) (.82-.97) (.83-.97)

(.62-1.00) (.21-.86) (.45-.92) (.36-1.00) (.53-.92)

(.35-.90) (.56-1.00) (.52-1.00) (.39-.91) (.58-.95)

The sensitivity and specificity of the composite panel for the diagnosis of GVHD-EC are displayed in the whole GVHD cohort (n ¼ 75) and in the grades III and IV GVHD patients (n ¼ 21).

the impact of various other prognostic factors involved in the allo-HCT outcome in our cohort. Among the various variables listed in Table 4, univariate analysis revealed that the high disease risk, HLA-mismatched donor, and H-score and Intscore groups defined by the composite panel were significantly associated with shorter OS (Table 4). The adjustment of potential variables by multivariate analysis revealed that high disease risk (HR, 4.02; 95% CI, 2.27 to 7.14; P < .01) and the H-score group defined by the composite panel (HR, 5.60; 95% CI, 2.81 to 11.20; P < .01) were significantly associated with shorter OS. In addition, the high disease risk and the composite paneledefined H-score group were significantly associated with higher NRM by both univariate and multivariate analyses (high disease risk; HR, 3.14; 95% CI, 1.49 to 6.64; P < .01; and H-score group; HR, 19.75; 95% CI, 5.59 to 69.77; P < .01) (Table 4). In contrast, the 5-year relapse rates of H-score, Int-score, and L-score groups were not significantly different at 12.7%, 28.7%, and 26.9%, respectively (Supplemental Figure 1), and this finding was obviously opposite to the results evaluated by the disease risk (high [33.7%] versus standard [16.1%], P < .01).

with grades III or IV GVHD. Furthermore, the total score of the composite panel at the onset of aGVHD was significantly higher in patients with a poorer GVHD outcome than in those with a better GVHD outcome based on a logistic regression model (median score, 5 versus 1; P < .01). Thus, the composite panel developed here was also useful for diagnosing the co-occurrence of both aGVHD and TRC-EC as well as for the prediction of poor outcome in patients with aGVHD. The Relevance of the Composite Panel for Survival We next evaluated the value of the risk categorization along with the composite panel for predicting survival period after the onset of TRCs in 149 patients with any type of TRC. Overall, 5-year OS rate was 54.9% and the 5-year NRM rate was 27.1%. The 5-year OS rate was significantly lower in the H-score group compared with those of the Int-score and L-score groups, and 5year OS rate of Int-score group was significantly lower than that of the L-score group (Figure 4A). In addition, the 5-year NRM rate in the H-score group was significantly higher compared with those of the Int-score and L-score groups, whereas that of the Int-score group tended to be higher than that of the L-score group, although this was not statistically significant (Figure 4B).

DISCUSSION The combined dysregulation of molecules for pleiotropic biologic processes, such as inflammation, tissue damage and repair, T cell alloreactivity, endothelial cell damage, and coagulation, underlies the complex pathophysiology of TRCs.

The Prognostic Impact of the Composite Panel in Association with Other Factors Involved in Allo-HCT Finally, we determined whether the composite panel maintained its prognostic power even when accounting for

OS

1.0

Probability

* † †

0.4

High Int Low

1.0

0.8 0.6

NRM

B

High Int Low

Cumulative incidence

A

0.8 0.6

†

0.4 0.2

0.2

†

P = .08

0.0

0.0 0

10

20 30 40 50 Months after any TRCs

Number at risk High 44 13 10 30 24 Int 44 Low 61 41 34

8 21 31

7 20 28

7 17 25

60 5 12 24

0

10

20 30 40 50 Months after any TRCs

Number at risk High 44 13 10 44 27 20 Int Low 61 40 31

8 18 28

7 18 25

7 16 23

60

5 12 22

* P < .05, † P < .01 Figure 4. OS and NRM based on the composite panel. The probability of overall survival (A) and cumulative incidence of nonrelapse mortality (B) are plotted according to the 3 groups; low (n ¼ 61), intermediate (n ¼ 44), and high (n ¼ 44). The 5-year OS was 76.2%, 54.9%, and 26.9%, and the 5-year NRM was 7.1%, 16.3% and 61.3% for TRC patients in the low-, int-, and high-score groups, respectively. * ¼ P < .05 and y ¼ P < .01.


1579

Table 4 Univariate and Multivariate Analyses of Risk Factors for Both OS and NRM Variables

OS

NRM

Univariate

Age: > 50 yr versus 50 yr Sex: male versus female Disease: myeloid versus Other Disease risk: high versus standard HCT-CI: > 3 versus 3 Conditioning intensity: MAC versus RIC Conditioning regimen BU: > 8 mg/kg versus 8 mg/kg MEL: >100 mg/m2 versus 100 mg/m2 TBI: >6 Gy versus 6 Gy Graft type: mismatch versus match Composite panel High versus low Int versus low

Multivariate

HR

95% CI

P Value HR

1.59 1.15 1.13 3.65 1.23 .98

.97-2.60 .06 .66-1.89 .69 .66-1.96 .65 2.18-6.11