Hypercoagulability in Dogs with Protein-Losing Enteropathy
This study was in part presented at the 2010 ACVIM Forum, Anaheim, CA.
Corresponding author: K. Allenspach, Department of Veterinary Clinical Sciences, Royal Veterinary College, University of London, Hawkshead Lane, North Mymms, Hatfield AL9 7TA, UK; e-mail: firstname.lastname@example.org.
Background: Dogs with protein-losing enteropathy (PLE) have previously been reported to present with thromboembolism; however, the prevalence and pathogenesis of hypercoagulability in dogs with PLE have not been investigated so far.
Hypothesis: Dogs with PLE are hypercoagulable compared with healthy control dogs.
Animals: Fifteen dogs with PLE. Thirty healthy dogs served as controls (HC).
Methods: A prospective study was performed including 15 dogs with PLE. All dogs were scored using the canine chronic enteropathy activity index (CCECAI). Thromboelastography (TEG) and other measures of coagulation were evaluated. Recalcified, unactivated TEG was performed and reaction time (R), kinetic time (K), alpha angle (α), and maximum amplitude (MA) values were recorded. Nine dogs were reassessed after initiation of immunosuppressive treatment.
Results: All dogs with PLE in the study were hypercoagulable with decreased R (PLE: median 7.8, range [2.4–11.2]; HC: 14.1 [9.1–20.3]), decreased K (PLE: 2.5 [0.8–5.2]; HC: 8.25 [4.3–13.1]), increased α (PLE: 56.7 [38.5–78.3]; HC: 25.6 [17–42.4]), and increased MA (PLE: 68.2 [54.1–76.7]; HC: 44.1, [33.5–49]) (all P < .001). Median antithrombin (AT) concentration was borderline low in PLE dogs; however, mean serum albumin concentration was severely decreased (mean 1.67 g/dL ± 5.1, reference range 2.8–3.5 g/dL). Despite a significant improvement in serum albumin and CCECAI, all 9 dogs with PLE were hypercoagulable at re-examination.
Conclusions and Clinical Importance: The hypercoagulable state in dogs with PLE cannot be solely attributed to loss of AT. Despite good clinical response to treatment, dogs remained hypercoagulable and could therefore be predisposed to thromboembolic complications.
Protein-losing enteropathy (PLE) in dogs is the result of severe small intestinal disease that allows leakage of protein into the intestinal lumen. Some authors suggest that thromboembolic disease can develop as a consequence of PLE1,2 secondary to enteric loss of antithrombin (AT), although there is little evidence to support this supposition.3,4 The prevalence of hypercoagulability in dogs with PLE is unknown.
Thromboelastography (TEG) has been used to document hypercoagulability in a number of veterinary clinical settings, including protein-losing nephropathy (PLN),a IMHA,5 parvoviral enteritis,6 neoplasia,7 and disseminated intravascular coagulation.8 TEG provides an objective evaluation of blood clot formation and dissolution, allowing hypercoagulable states to be identified and quantified. It represents an ex vivo rotatory viscoelastic test of coagulation that produces a graphical representation of the processes of clot formation and fibrinolysis in whole blood. TEG evaluates the interaction of platelets with proteins of the coagulation cascade, thereby combining evaluation of the traditional plasma components of coagulation with the cellular components. Four variables are routinely recorded from TEG traces: the reaction time (R), the clot formation time (K), the clot formation angle (α), and the maximum amplitude (MA).
The R time principally evaluates the intrinsic pathway and hence is affected by concentrations of factors VIII, IX, XI, and XII. The K time measures the speed of clot formation to a predefined clot strength and is influenced by the levels of fibrinogen, thrombin, factor VII, platelet count and function, and hematocrit (Hct). The clot formation angle evaluates the rate of clot formation and is affected by the same variables as the K time. The MA is related to final clot strength and is therefore affected by levels of fibrinogen, thrombin, factor XIII, platelet count and function, and Hct.9
It is conceivable that dogs with PLE have abnormalities of all of these 4 variables. Because of enteric protein loss, dogs with PLE might have abnormally low concentration of coagulation factors. Low-level gastrointestinal bleeding associated with PLE might lead to anemia or thrombocytosis thereby altering K time, α angle, and MA values. Dogs with PLE have an inflammatory disease that may cause an acute phase response leading to hyperfibrinogenemia which might also alter K, α, or MA values.9
We therefore hypothesized that PLE was associated with the development of a hypercoagulable state that could be identified by TEG in combination with conventional tests of coagulation and that this hypercoagulable state would resolve with a documented clinical improvement after initiation of treatment.
Material and Methods
The study population comprised client-owned clinical patients admitted to the Queen Mother Hospital for Animals, Royal Veterinary College (RVC) between May 2008 and October 2009. The study was approved by the RVC's Ethics and Welfare Committee. Written consent was obtained from owners of all dogs included in the study. Any dogs with a diagnosis of PLE weighing >5 kg were eligible for inclusion into the study. The diagnosis of PLE was made if all of the following applied: (1) history of gastrointestinal disease (including weight loss, vomiting, diarrhea, decreased appetite); (2) panhypoproteinemia (serum albumin <2.8 g/dL and serum globulin <2.1 g/dL; reference ranges 2.8–3.9 and 2.1–4.1 g/dL, respectively); (3) histopathological confirmation of a disease process known to be associated with PLE; (4) exclusion of hepatic dysfunction; and (5) absence of proteinuria. If there were changes consistent with hepatic dysfunction on the serum biochemical profile (decreased urea, decreased glucose, elevated total bilirubin), the case could only be included on the basis of a normal bile acid stimulation test result. Proteinuria was excluded in all dogs on the basis of a negative urine dipstick or a urine protein:creatinine ratio of <0.5. In addition, all dogs had to be tapered off any steroid treatments at least 2 weeks before inclusion into the study.
Blood samples were collected by clean jugular venipuncture by veterinarians or veterinary technicians with a 21 G needle and a 5 mL syringe. All blood samples were immediately aliquoted into 1.1 or 1.3 mL nonvacuum, polypropylene tubes. CBC samples were collected in potassium EDTA. Samples for serum biochemistry and CRP were collected into serum gel separator tubes.
A total of 5.2 mL of whole blood was collected into four 1.3 mL polypropylene nonvacuum blood sample tubesb containing liquid 3.2% sodium citrate, 1 part citrate + 9 parts of blood, for TEG, prothrombin time (PT), activated partial thromboplastin time (aPTT), serum AT, fibrinogen, and D-Dimers. TEG was performed on samples that had rested at room temperature for 30 minutes by a recalcified, unactivated technique according to the manufacturer's instructions.c Values for the R, K, α, and MA were recorded from TEG tracings. PT and aPTT were performed with a point of care analyzer.d The remainder of the blood samples were centrifuged at 11,000 ×g for 5 minutes, and the plasma separated and frozen at − 80°C for batch analysis at a later time. Frozen plasma samples were sent in batches to external laboratories for analysis of fibrinogen,e AT,e and D-Dimers.f Clottable fibrinogen was measured by the Clauss method with 100 NIH/Units/mL human thrombin reagent. The fibrinogen standard curve was derived from dilutions of pooled normal canine plasma. The fibrinogen content of the canine plasma standard is determined by a quantitative, multispecies fibrinogen ELISA. AT activity was estimated by an anti-IIa chromogenic assay. A standard curve for AT was derived from dilutions of pooled normal canine plasma having an assigned value of 100% AT activity. D-Dimers were semiquantitatively measured by a commercial laboratory (Axiom) with a kit-basedg latex agglutination assay against manufacturer supplied purified high- and low-level D-Dimer controls. Where possible, a 2nd set of blood samples was collected up to 30 days after the initiation of immunosuppressive therapy. The blood samples were collected and processed as above. Other information collected included clinical signs, body condition scores (scale 1–9), CBCs, serum biochemical profiles, urinalyses, fecal analyses, diagnostic imaging, and histopathological findings. A previously published scoring system for the assessment of severity of chronic enteropathies10 was used to assign each dog a canine chronic enteropathy clinical activity index (CCECAI) score based upon clinical signs and serum albumin concentrations. Further tests were performed as indicated and included canine pancreatic lipase immunoreactivity (cPLI), trypsin-like immunoreactivity (TLI), serum folate and cobalamin concentrations, and bile acid stimulation and ACTH stimulation tests.
Recalcified, unactivated TEG tracings were produced from a control population of 30 healthy dogs (HC) that were part of the hospital blood donation program. These dogs were declared healthy on the basis of a history of preventative health care, no history of travel, a full physical examination, CBC, and serum biochemistry profile. To avoid additional venipuncture, control blood samples were collected from healthy blood donors undergoing routine annual health screening. Blood samples were collected by clean venipuncture by a single veterinary technician with a 21 G needle and a 5 mL syringe. All blood samples were immediately aliquoted into 1.3 mL nonvacuum, polypropylene tubes as used for study dogs.
Distribution of data was assessed graphically, and normally distributed data are presented as mean (standard deviation) and nonnormally distributed data are presented as median (range). Nonparametric univariate statistics were used to compare TEG results of PLE dogs with those of healthy controls (unpaired Wilcoxon's signed-rank test) and to compare pre- and posttreatment TEG, albumin, globulin, AT, fibrinogen, D-Dimers, and CCECAI scores (paired Wilcoxon's signed-rank test). Pearson's correlation coefficient was used to examine linear relationships between any of the TEG parameters (R, K, α angle, MA) with white blood cell (WBC) numbers, Hct, platelet numbers, fibrinogen, AT, albumin, globulin, and CCECAI. In addition, multivariate logistic regression analysis for factors predicting a decrease in R and K or an increase in α angle and MA was performed, including either the factors WBC, Hct, platelets, fibrinigen, AT, albumin, globulin, and CCECAI in the model or only including WBC and Hct in the model. Statistical significance was set at P < .05. All statistical analyses were conducted by commercially available software.h
Fifteen dogs of various breeds were enrolled in the patient group, including 2 mixed breed dogs and 1 each of Airedale Terrier, Beagle, Boxer, Cavalier King Charles Spaniel, Giant Schnauzer, Greyhound, Lurcher, Rottweiler, Scottish Terrier, Staffordshire Bull Terrier, Tibetan Terrier, Toy Poodle, and Yorkshire Terrier. There were 8 male dogs in the study population, of which 6 were neutered, and 7 females, of which 5 were spayed. The mean age of dogs in the study population was 8.5 (2.4) years. Mean body weight was 16.6 (9.5) kg with a median body condition score of 3/9 (range 1–6).
The 30 healthy control dogs consisted of 12 breeds including Golden Retriever (n = 5), Labrador Retriever (n = 5), Newfoundland (n = 3), Italian Spinone (n = 2), Rhodesian Ridgeback (n = 2), and 1 each of Bloodhound, Boxer, German Shorthaired Pointer, German Shepherd Dog, Great Dane, Staffordshire Bull Terrier, and Schipperke. There were also 6 crossbred dogs. There were 11 castrated males, 6 entire males, 8 entire females, and 5 spayed females. The age at the time of sampling was 4.3 years (2.0).
For all dogs, clinical signs had been present for between 10 days and 3 years before presentation. Only 3 dogs had been treated with steroids before referral, and these dogs were tapered off their steroid treatments at least 2 weeks before inclusion into the study.
Diarrhea and weight loss were the most common clinical signs reported, with 14 of 15 (93%) dogs being affected by both. Lethargy (n = 12; 80%), vomiting (n = 10; 67%), and inappetence (n = 9; 60%) were also commonly reported. Other clinical signs and physical examination findings included hemorrhagic diarrhea (n = 2), ascites (n = 2), subcutaneous edema (n = 2), pleural effusion (n = 1), neurological signs (seizure episode followed by circling, n = 1), intermittent lameness because of femoral arterial thrombus (n = 1), polydipsia/polyuria (n = 1), and prostatomegaly (n = 1).
Two of the dogs in this study presented with signs compatible with possible or confirmed thromboembolism. One of the dogs developed severe dyspnea and died 145 days after diagnosis, which was interpreted as a possible pulmonary thromboembolism. Another dog presented with signs compatible with a femoral arterial thrombus (absent femoral pulse, hyperesthesia of the foot, and intermittent lameness), which was confirmed ultrasonographically.
The median serum albumin concentration was 1.56 g/dL (0.98–2.56 g/dL, reference range 2.8–3.9 g/dL), and the median serum globulin concentration was 1.51 g/dL (1.01–2.01 g/dL, reference range 2.1–4.1 g/dL). The median CCECAI score was 12 (range 6–17), consistent with very severe disease.10
Mean platelet numbers before treatment were 408.12 × 109/μL (SD 280.4; range 33–856, reference range 150–700 × 109/μL). One dog had platelet numbers below the reference range (33 × 109/μL); however, numbers seemed adequate on smear evaluation by a clinical pathologist. One had platelet numbers above the reference range (856 × 109/μL). Platelet numbers after treatment were 531 × 109/μL (SD 133, range 357–717) with 1/9 dog having a platelet count of 717 × 109/μL. There was no statistically significant difference in platelet numbers before and after treatment.
WBC before treatment were 18.48 × 109/μL (SD 9.39; range 4.1–40.1, reference range 6–17.1 × 109/μL). Seven out of 15 dogs had WBC above the reference range (range 21.5–40.1 × 109/μL), and 1 dog had a low WBC (4.1 × 109/μL). WBC after treatment were 20.7 × 109/μL (SD 6.4, range 10.4–25.8) with 6/9 dogs having WBC above the reference range (range 19.8–27.8 × 109/μL). There was no statistically significant difference in WBC before and after treatment.
Hct before treatment was 45% (SD 9; range 37–52%, reference range 37–55%). Two out of 15 dogs had a high Hct at 60 and 55%, respectively. Hct after treatment was 47% (SD 12; range 30–62%). Two of 9 dogs had a low Hct at 30 and 31%, respectively, and 1 dog had a high Hct at 61%. There was no statistically significant difference between mean Hct before and after treatment.
Histopathology of endoscopic biopsies confirmed a diagnosis of severe IBD in 14/15 dogs and intestinal lymphoma in 1/15 dogs. Of the dogs with IBD, 2/14 had concurrent lymphangiectasia (Yorkshire Terrier and Tibetan Terrier) and 1/14 had concurrent ulcerative colitis (Greyhound).
All 15 dogs had moderate hypocholesterolemia (median 2.5 mmol/L, range 1.3–4.5 mmol/L; reference range 3.3–8.9 mmol/L). Seven of 15 dogs had cPLI assayed, with 3 having values greater than the reference range (241, 330, and 782 μg/L; reference range <200 μg/L). Six dogs had serum TLI concentration measured, with 1 value being outside the reference range (>50 μg/L; reference range 6–35 μg/L). Thirteen dogs had serum folate and cobalamin concentrations assessed. Four of 13 dogs had decreased and 1/13 had increased serum folate concentrations; and 8/13 dogs had decreased serum cobalamin concentrations. Bile acid stimulation tests were performed in 3 dogs and were within reference range in all 3. Six dogs had ACTH stimulation tests performed, 4 of which were normal and 2 of which were consistent with a stress response. Fecal analysis was negative for endoparasites, Giardia, and Cryptosporidia in all cases, and fecal culture yielded growth of Campylobacter spp. in 2/9 dogs.
When compared with HC, all dogs in the study were significantly hypercoagulable with decreased R (PLE: median 7.8 [2.4–11.2]; HC: 14.1 [9.1–20.3]), decreased K (PLE: 2.5 [0.8–5.2]; HC: 8.25 [4.3–13.1]), increased α (PLE: 56.7 [38.5–78.3]; HC: 25.6 [17–42.4]), and increased MA (PLE: 68.2 [54.1–76.7]; HC: 44.1 [33.5–49]) (all P < .001).
The AT concentration in the 14 dogs with available results was borderline low (65% [46–121]%, reference range 65–145%) with 9 dogs having AT ≤ 65% (range 46–65%). Fibrinogen concentration was moderately increased (587 mg/dL [404–1359 mg/dL], reference range 147–479 mg/dL). Four of 12 dogs had D-Dimer concentration in the equivocal range (250–500 ng/mL), with the remaining 8 dogs having values within the reference range (<250 ng/mL). Results of PT and aPTT were available for 10/15 and 9/15 dogs, respectively. All were normal apart from 1 mild aPTT prolongation (124.5% of the upper limit of the reference range).
Linear correlation of the TEG parameters R, K, α, and MA with WBC, Hct, platelet numbers, fibrinogen, AT, albumin, globulin, and CCECAI was performed by Pearson's correlation. A moderate linear correlation of WBC before treatment with MA before treatment (Pearson's r2= 0.68, P= .0045) was detected. Similarly, a moderate linear correlation of Hct pretreatment values with K, α, and MA before treatment was present (Hct and K: Pearson's r2= 0.35, P= .01; Hct with α: r2= 0.43, P= .008; Hct with MA: r2= 0.43, P= .01). When WBC, Hct, platelet numbers, fibrinogen, AT, albumin, globulin, and CCECAI were included in the multivariate model, there were no statistically significant associations detected. Similarly, if a model including only WBC and Hct was used, no statistically significant association was identified between potential risk factors and TEG parameters.
A subset of 9 dogs had repeat blood samples performed 4–24 days after initiation of treatment. Despite a statistically significant improvement in median albumin concentration (pretreatment 1.46 g/dL, posttreatment 2.22 g/dL; P= .013) and median CCECAI score (pretreatment 12, posttreatment 3; P= .003), there was no significant difference between pre- and posttreatment TEG variables or AT, fibrinogen, or D-Dimer concentrations.
Ten dogs died or were euthanized between 6 and 148 days from the time of presentation. Nine dogs were euthanized because of deterioration of the primary disease process (including the dog with lymphoma) and 1 dog died shortly after developing acute onset dyspnea.
These data suggest that dogs with PLE are hypercoagulable when compared with healthy dogs. All 4 of the routinely recorded TEG tracing variables from affected dogs were significantly different from those of healthy controls. This implies that the prevalence of hypercoagulabilty in dogs with PLE is high, which could have implications for treatment decisions.
Four principal TEG assays have been reported in dogs to date, namely native, recalcified unactivated, TF-activated, and kaolin activated. In the native assay whole blood is sampled and placed immediately into the TEG assay cup. The higher interindividual variation in native blood samples means that use of citrated whole blood is preferred for TEG in dogs.11 The recalcified unactivated assay used in our study of PLE dogs involves recalcification of citrated blood before analysis and no activator was used to accelerate the assay. This assay carries an intrinsically higher preanalytical variation.
In the TF-activated TEG assay, a low concentration of human recombinant tissue factor combined with phospholipidsi is used to initiate coagulation ex vivo. Tissue factor is the principal initiator of the hemostatic process in vivo via the TF-VII (extrinsic) pathway. As such, some consider the TF assay to be most applicable to the in vivo situation, although there is some debate regarding the optimal TF concentration for this assay.
In the kaolin-activated TEG assay, the contact (intrinsic) pathway is used to initiate the assay.11 Once adsorbed to negatively charged agents such as glass, kaolin, celite, or ellagic acid Hageman factor (factor XII) is activated, initiating the intrinsic pathway by activating F XI, a process accelerated by HMWK and plasma prekallikrein.11
The optimal TEG assay for clinical use remains unclear. Although all of the TEG assays produce an integrative analysis of ex vivo coagulation, the various assays evaluate different aspects of the coagulation system and do not produce directly comparable results. The activated assays produce results more rapidly and more reproducibly, that is with lower coefficients of variation. If kaolin is capable of inducing strong activation of clotting in native whole-blood samples, mild hypercoagulability caused by the incomplete inhibition of thrombin activation in citrated specimens would not have been detected.
As mentioned above, unactivated assays such as the one used in our study are more vulnerable to preanalytical variation.12,13 We have made an attempt to reduce the variability in these factors to a minimum by keeping venipuncture procedure and collection into citrated tubes as constant as possible. In addition, TEG was run within 30 minute after collection.
The dogs with PLE in our study had borderline low AT concentrations, which could in part explain the hypercoagulability identified with TEG. AT acts with its endogenous cofactors, heparins, heparan, and dermatan sulfates, to bind to and inactivate thrombin (FIIa) and factors Xa, VIIa, IXa, XIa, and XIIa. A deficiency of AT would potentially increase the rate and the quantity of activated clotting factors, most importantly thrombin. AT deficiency might therefore affect all 4 TEG values.9 Although hereditary AT deficiency is well recognized in people, whole blood TEG in patients with AT deficiency has not been reported. A recent abstract reporting TEG in dogs with PLN identified alterations in MA values but unfortunately did not report AT activities.a In this abstract, a kaolin-activated assay was used. We assume this was after recalcification although this is not stated in the abstract itself. Measurement of Thrombin-Antithrombin complex (TAT) would possibly be able to clarify the relationship of AT and TEG parameters in the dogs of our study, but unfortunately was not available.
In PLN, AT deficiency is thought to be the major contributory factor to the development of a hypercoagulable state. In the previously reported pilot study,a dogs with PLN had mildly prolonged R times, normal K times, and increased α and MA values. In contrast, the dogs with PLE reported here had decreased R and K times as well as increased α and MA values. This suggests that the scope of hypercoagulability identified in our dogs with PLE was greater than that reported for dogs with PLN. Conclusions drawn from comparing these 2 studies must be tempered as it is possible that the observed differences could stem from methodological differences in the assay rather than the disease processes themselves.
The link between inflammation and coagulation is well established14–18 and it is possible that the inflammatory process itself, in combination with loss of AT, promotes hypercoagulability in dogs with PLE. A mild linear correlation of WBC numbers with MA and Hct with K, α, and MA was found, but was not detectable anymore after inclusion of all parameters or WBC and Hct only, in a model of multivariate analysis. It is possible that these factors could influence TEG values; however, our small study numbers make meaningful assessment of multivariate analysis including 8 different factors difficult as too many confounding factors may exist. Our dogs did have moderately increased fibrinogen concentrations, which is compatible with the inflammatory nature of PLE. It is possible that despite an apparent clinical improvement, ongoing subclinical intestinal inflammation could contribute to a persistent hypercoagulable state. Previous longitudinal studies in dogs with inflammatory bowel disease have shown persistent inflammation of the intestines, despite clinical improvement after treatment.19 In people, Crohn's disease and ulcerative colitis, the 2 major clinical forms of IBD, are associated with a hypercoagulable state18 and systemic thromboembolism is reported as an important cause of morbidity and mortality in this patient population.14–17 Alternative markers of hypercoagulability such as TAT or thrombin generation curves were not assessed in our study but might have been of value in corroborating the TEG findings. Assessment of C-reactive protein or other acute phase proteins may have been useful in better characterizing the inflammatory response in these dogs.
As has been reported for dogs with PLN, the degree of hypercoagulability did not appear to be associated with the severity of hypoalbuminemia in our PLE dogs. The improvements seen in serum albumin concentration and CCECAI scores after treatment were not associated with an improvement in AT, fibrinogen concentration or in the degree of hypercoagulability as measured by TEG—an unexpected finding. It is possible that this is because of the fact that all dogs investigated in this study were treated with immunosuppressive doses of prednisolone, which has been associated with development of a hypercoagulable state.20–22 It is possible that hypercoagulability would not have persisted with the use of other treatments such as ciclosporin.
The prevalence of clinically evident thromboembolic disease was low in our study: clot formation was only confirmed in 1 case and suspected in another. It is possible that minor thromboembolic events occurred unnoticed; however, it is reasonable to conclude that demonstration of vitro hypercoagulability does not inevitably lead to a thromboembolic event in cases with PLE.
D-Dimer concentrations are highest 2 hours after embolism and remained elevated for 24 hours.23 None of the dogs in this study had significantly elevated D-Dimers with only 4/12 dogs having results within the equivocal range. Normal D-Dimer results could have been true negative or false-negative results in these dogs. False negatives could result from an intrinsic lack of assay sensitivity to canine D-Dimer. True negative results indicate that fibrinolysis of cross-linked fibrin was not occurring when the sample was taken. However, we must be cautious in concluding that these patients were not at increased risk of thrombosis simply because there was limited evidence of fibrinolysis occurring at the time of sampling. Indeed, at the time of presentation, 1 dog with an arterial thrombus had D-Dimers within the reference range, potentially because of the chronicity of the disease process.
The small sample size in this study precludes meaningful interpretation of possible breed associations.
The number of dogs in this study who had repeated blood sampling performed was small, which would have limited our ability to identify the effects of treatment on coagulation variables. Additionally, because this was an observational study, treatment was prescribed according to attending clinician preference, preventing standardization, and reducing our ability to generalize to a larger population. Future prospective studies in which therapy and follow-up are standardized might allow us to better identify the effects of therapy on coagulation variables in dogs with PLE and are warranted based on our results.
a Hilling KM, Labato MA, de Laforcade AM, Shaw S. Documentation of hypercoagulability in protein-losing nephropathy via thromboelastography in 10 dogs. J Vet Intern Med 2009;23:690 (abstract)
b Pediatric Tube, International Scientific Supplies, Bradford, UK
c TEG 5000 Thrombelastograph Hemostasis Analyzer, Haemoscope Corporation, Niles, IL
d CoagDx, IDEXX Laboratories, Southwater, Horsham, UK
e Coagulation Section-Animal Health Diagnostic Center, Cornell University College of Veterinary Medicine, Ithaca, NY
f Helena Biosciences, Gateshead, Tyne and Wear, UK
g IDEXX Laboratories
h SPSS 16 for Windows, SPSS Inc, Chicago, IL
i Innovin, Dade Behring, Deerfield, IL
This study was supported by a grant from the Waltham Foundation.