Increased levels of plasma thrombomodulin are associated with vascular and all-cause mortality in patients on long-term anticoagulant treatment. J.-H. Jansson* ...
European Heart Journal (1996) 17, 1503-1505
Increased levels of plasma thrombomodulin are associated with vascular and all-cause mortality in patients on long-term anticoagulant treatment J.-H. Jansson*, K. Boman*, M. Brannstrom* and T. K. Nilssonf *Department of Medicine, Skelleftea1 Hospital and ^Department of Clinical Chemistry, Umed University Hospital, Urned, Sweden
lar mortality (P=0-042) was significantly and independently associated with increased levels of thrombomodulin.
Introduction
that increased levels of tPA, PAI-1, and vWF may be indicators of endothelial dysfunction, perhaps due to the atherosclerotic process. It has been suggested that thrombomodulin, an endothelial membrane protein with antithrombotic action mediated via the thrombin-thrombomodulin complex, is a marker of endothelial cell dysfunction or lesions, for instance in disseminated intravascular coagulation'81. In general, the highest increases in thrombomodulin levels as compared to controls have been seen in patients with various neoplastic diseases (lymphoma, leukaemia, cancer) and liver diseases (hepatitis, cholestasis) except for chronic hepatitis; patients with venous or arterial thrombotic disease had less pronounced increases in thrombomodulin levels'81. Weak positive correlations between thrombomodulin and tPA or vWF mass concentrations have been reported'81, as well as between thrombomodulin and PAI-1 and the prostacyclin Pgl2[9]. However, in a population-based sample, subjects with electrocardiographic (ECG) signs of ischaemic heart disease did not have increased thrombomodulin levels compared to subjects lacking such signs'101. It has also been found that pre-menopausal females have slightly lower levels than males, and that this difference disappears at higher ages. In contrast, thrombomodulin seems not to correlate with anthropometric or lipid metabolic variables'101.
Conclusions The level of plasma thrombomodulin can predict all-cause and vascular mortality in patients on Methods and Results The thrombomodulin level was long-term anticoagulant treatment. 60-9 ± 29-8 ug . 1 - ' for all 45 patients who died and (Eur Heart J 1996; 17: 1503-1505) 60-5 ± 30-5 ug . 1"~' for the 38 vascular deaths, compared to 52-3 ± 20-7 ug . 1 ~ ' for the 164 survivors. We found that, Key words: Anticoagulants, vascular mortality, risk in Cox regression analyses, all-cause (/>=0025) and vascu- factors, thrombomodulin.
Anticoagulant treatment with coumarin derivates and warfarin is widely used in the prevention of thromboembolic events on both the venous and the arterial sides of the vascular system. Nevertheless, vascular mortality and serious complications are common in these patients. Vascular mortality is, in general, the consequence of intravascular thrombus formation. It is therefore reasonable to assume that measurements of components of the haemostatic system could give an indication of the risk of vascular mortality. We and others have previously suggested that high plasma mass concentrations of tissue plasminogen activator (tPA), plasminogen activator inhibitor type 1 (PAI-1), and von Willebrand factor (vWF) are associated with an increased risk of myocardial reinfarctions or death'1"61. We have shown earlier that in the population studied there was a significant association between all-cause mortality and increased levels of vWF, tPA, and between vascular mortality and tPA, PAI-1 and vWF171. Since these substances are all synthesized and secreted from the vascular endothelial cells it is thought Revision submitted 6 February 1996, and accepted 12 February 19%. Correspondence: Dr J.-H. Jansson, Department of Medicine, Skelleftea Hospital, S-93I86 Skelleftea Sweden. 0195-668X/96/101503+03 $18.00/0
In the present study we evaluated thrombomodulin mass concentration as a predictor of mortality 1996 The European Society of Cardiology
Downloaded from eurheartj.oxfordjournals.org by guest on July 15, 2011
Background The aim of the study was to test if plasma levels of thrombomodulin could predict mortality in 209 patients on long-term anticoagulant treatment followed up for 3-8 years.
1504 J.-H. Jansson el al.
in a prospective study of patients undergoing oral anticoagulant treatment. 5.28
Material and methods
Each patient was monitored by prothrombin time assay at least every 8th week. The recommended therapeutic range of prothrombin was 10-25%, corresponding to an international normalized ratio, INR, of 20-4-0. Prothrombin time was measured in a P&P assay using Nycotest (Nycomed, Oslo). Blood samples from the included subjects were obtained for preparation of citrated plasma, and assayed for thrombomodulin with a commercially available ELISA method, which employs two monoclonal antibodies reacting with different epitopes on the molecule. The reagents were purchased from Diagnostica Stago, France"11. In March 1992 any deaths that had occurred among the patients studied were noted from the patients' records. Death certificates were obtained and vascular death was defined according to ICD-9 (390-459). No patient was lost to follow-up. Statistical analysis was performed using the SAS computer statistical package. To illustrate the relationship between a possible prognostic factor and the incidence of mortality, baseline variables were divided into quartiles, and the incidence of vascular events in each quartile (Q1-Q4) was calculated per 1000 patient months. However, these quartiles were not used to test relationships: for this, a Cox regression analysis was performed. A two-tailed test was performed and a /"-value of 0-05 or less was regarded as statistically significant. Relative risk (RR) was calculated as incidence Q4/incidence Ql. Expert statistical advice was given by Kjell Pennert, Gothenburg. Eur Heart J, Vol. 17, October 1996
§
3'
Ql-39
Q2 40-50
Q3 51-61
Q4 62-
Ql-39
Q2 40-50
Q3 51-61
Q4 62-
Figure 1 Bar graph showing incidence per 1000 patientmonths of all-cause mortality (upper panel) and vascular mortality (lower panel) in relation to quartiles Q, through Q 4 of thrombomodulin level (ug . 1 ~ ' ) .
Results In the course of 3-8-years follow-up, there were 45 deaths of which 38 were from vascular causes. These comprised 17 from acute myocardial infarction, three from coronary heart disease, two from sudden death, eight from congestive heart failure, four from ischaemic stroke, two from subdural haemorrhage, one from subarachnoidal haemorrhage, and one from aortic rupture. The thrombomodulin level was 60-9 ± 29-8 ug . 1" ' for all 45 patients who died and 60-5 ± 30-5 ug . 1 ~ ' for the 38 vascular deaths, as compared to 52-3 ± 20-7 ug . 1 ~' for the 164 survivors. We found that all-cause mortality (f=0-025, regression coefficient=0-012) and vascular mortality (/>=0-042, regression coefficient=0011) was significantly and positively associated with thrombomodulin in univariate analyses. The relative risk (Q4/Q1) for thrombomodulin levels for all-cause mortality was 1 -62 and for vascular mortality 1-56 (Fig. 1). In multivariate Cox regression analysis (including: age, sex, smoking habits, body mass index, diabetes mellitus, hypertension, and thrombomodulin), thrombomodulin (/)=OOO36) and smoking (/>=0029) were independently and significantly associated with all-cause mortality and thrombomodulin (P=0-0028) and age (P=0-0335) with vascular mortality.
Downloaded from eurheartj.oxfordjournals.org by guest on July 15, 2011
Recruitment for this study took place in March 1987, when 212 of 258 patients registered at the out-patient clinic of the Department of Medicine, Umea University Hospital, attended the clinic for their regular prothrombin sampling. At the time of sampling, the patients had been treated with anticoagulants from 1 to 134 months. In three subjects, blood samples were not sufficient for analyses of thrombomodulin; the remaining 209 subjects formed the basis of the present study. The main indications for anticoagulant treatment in this unselected cohort of anticoagulant-treated patients were; (i) valvular heart disease (45%), which comprises both operated and non-operated patients (ii) arterial thromboembolism (33%), which includes cases with ischaemic stroke, transient ischaemic attack and peripheral arterial thromboembolism (iii) atrial fibrillation (29%), (iv) deep vein thrombosis or pulmonary embolism (22%). In 59 (28%) subjects there was more than one indication: each is recorded here as a separate indication. There are thus more indications than patients. Two hundred patients were treated with warfarin and 12 with dicumarol.
Thrombomodulin predicts vascular mortality
Discussion
therapy has the same predictive value, and most importantly, whether the association between thrombomodulin levels and mortality can be confirmed, are essential future issues to be addressed. The present study shows for the first time that the level of plasma thrombomodulin can predict allcause and vascular mortality in patients on long-term anticoagulant treatment. These findings introduce a new haemostatic predictor to a patient group with a high risk of atherothrombotic complications. This work was supported by grants from the Swedish Medical Research Council (No. 08267), The National Association for Heart and Lung Patients, Anny and Ragnar Wikstens foundation, and the Regional Councils of Northern Sweden.
References [1] Hamsten A, DeFaire U, Walldius G et al. Plasminogen activator inhibitor in plasma: risk factor for recurrent myocardial infarction. Lancet 1987; ii: 3-9. [2] Meade TW, Ruddock V, Stirling Y, Chakrabarti R, Miller GJ. Fibrinolytic activity, clotting factors, and long-term incidence of ischaemic heart disease in the Northwick Park Heart study. Lancet 1993; 342: 1076-9. [3] Cortellaro M, Boschetti C, Cofrancesco E et al. The PLAT study: Hemostatic function in relation to atherothrombotic ischemic events in vascular disease patients. Principal results. Arterioscler Thromb 1992; 12: 1063-70. [4] Ridker PM, Vaughan DE, Stampfer MJ, Manson JE, Hennekens CH. Endogenous tissue-type plasminogen activator and risk of myocardial infarction. Lancet 1993, 341: 1165-8. [5] Thompson SG, Kienast MAG, Pyke SDM, Haverkate F, van de Loo CW. Hemostatic factors and the risk of myocardial infarction or sudden death m patients with angina pectoris. N Engl J Med 1995; 332: 635-41. [6] Jansson JH, Nilsson TIC, Johnson O. von Willebrand factor in plasma: a novel risk factor for recurrent myocardial infarction and death. Br Heart J 1991; 66: 351-5. [7] Brannstrom M, Jansson J-H, Boman K, Nilsson TK.. Haemostatic factors can predict mortality in patients on long-term anticoagulant treatment. Thromb Haem 1995; 74: 612-15. [8] Takahashi H, Ito S, Hanano M et al. Circulating thrombomodulin as a novel endothelial cell marker: Comparison of its behaviour with von Willebrand factor and tissue-type plasminogen activator. Am J Hematol 1992; 41: 32-9. [9] Seigneur M, Dufourcq P, Conri C et al. Levels of plasma thrombomodulin are increased in atheromatous arterial disease. Thromb Res 1993; 71: 423-43. [10] Nilsson TK, Hellsten G, Amiral J. Plasma thrombomodulin concentrations in relation to cardiovascular risk factors in a population sample. Blood Coag Fibrin 1993; 4: 455-8. [11] Arniral J, Adam M, Mimilia F, Larrivaz I, Chambrette B, Boffa MC. A new assay for soluble forms of thrombomodulin in plasma. Thromb Haemostas 1991; 947: 65. [12] Takahashi H, Hanano M, Wada K et al. Circulating thrombomodulin in thrombotic thrombocytopenic purpura. Am J Hematol 1991; 38: 174-7. [13] Tanaka A, Ishii H, Hiraishi S et al. Increased thrombomodulin values in plasma of diabetic men with microangiopathy. Clin Chem 1991; 37: 269-72. [14] Hamsten A. HemostaU'c function and coronary artery disease. N Engl J Med 1995; 332: 677-8.
Eur Heart J, Vol. 17, October 1996
Downloaded from eurheartj.oxfordjournals.org by guest on July 15, 2011
A high thrombomodulin mass concentration has previously been found in some patients with thrombotic thrombocytopenic purpura'12', diabetic microangiopathy1'31 and atheromatous arterial disease1'1, suggesting a relationship with endothelial damage. The plasma thrombomodulin level depends on the biosynthesis rate, the integrity of the endothelium, and the clearance rate of the molecule. Data from the literature indicate that plasma levels of thrombomodulin are increased in a number of pathological states when the endothelium is likely to have been damaged, but also in kidney and liver disease where one can suspect altered clearance rates. Little is known whether increased levels of thrombomodulin are related to the extent of the endothelial lesion and if it is independent of other markers of endothelial cell activation and/or lesion. In some case-control studies, patients with varying degrees of atherosclerotic disease have had higher thrombomodulin levels than control groups'91, but in a crosssectional population-based study, thrombomodulin did not differ between subjects with and without ECG signs of ischaemic heart disease1'01. It is not inconceivable that the extent of the disease process, the presence or absence of concomitant disease, and the extent of ongoing acute-phase reaction might confound this type of group comparison, thus accounting for the discrepant results in these studies. We studied total and cardiovascular mortality in different groups of patients on anticoagulant therapy. Although Seigneur et a/.'91 found no correlation between the thrombomodulin level and the different groups of atherosclerotic disease, we suggest that anticoagulanttreated subjects with signs of endothelial damage, reflected in a high thrombomodulin, are particularly prone to die. This hypothesis on the relationship between the extent of endothelial damage and the thrombomodulin level should be addressed in future studies. Recently, other endothelial-derived haemostatic factors such as tPA and its inhibitor (PAI-1) and von Willebrand factor have been shown to predict cardiovascular complications in patients with established atherosclerotic disease and in apparently healthy males1'"61, suggesting that 'basic research on the role of haemostasis in atherosclerosis and its thrombotic complications should be given high priority, because the modulation of haemostatic function is likely to become an important approach to prevention''141 So far, in many patients it is difficult to predict the risk benefit ratio of anticoagulant treatment. The number of patients included in this study does not allow subgroup analyses of treatment indication, sex, age, concomitant diseases or drug treatment. Whether sampling before initiating anticoagulant or antiplatelet
1505
