Supported by a grant from the Companion Animal Health Fund at Tufts Cummings School of Veterinary Medicine. Presented in abstract form at the International Veterinary Emergency and Critical Care Society meeting, San Diego, CA, September 2004. The authors thank Dr Marjory Brooks, Associate Director–Coagulation Section, Animal Health Diagnostic Center, College of Veterinary Medicine, Cornell University for analysis of protein C and antithrombin samples.
Corresponding author: Armelle de Laforcade DVM, DACVECC, Department of Clinical Sciences, Tufts Cummings School of Veterinary Medicine, 200 Westboro Road, North Grafton, MA 01536; e-mail: firstname.lastname@example.org
Background: Protein C (PC) and antithrombin (AT) activities are decreased in humans with severe sepsis, and persistent changes are associated with decreased survival. In dogs with sepsis, PC and AT have been shown to be decreased at the time of diagnosis.
Hypothesis: PC and AT activities change significantly over time in dogs with sepsis and may be related to outcome.
Animals: Twelve dogs with naturally occurring sepsis.
Methods: Blood was collected from 12 dogs with sepsis, defined as histopathologic or microbiologic confirmation of infection and two of the following: hypo- or hyperthermia, tachycardia, tachypnea, leukopenia, leukocytosis, or >3% bands. The time of 1st sampling was considered day 1 and sampling was repeated every 24 hours for 5 days or until discharge or death. Changes over time were analyzed by ANOVA with repeated measures, and the association between PC and AT and outcome was determined by a 2-equation treatment effects model.
Results: Nine dogs and 11 dogs had decreased PC and AT activity on day 1, respectively (mean PC, 66.0 ± 25.8%; mean AT, 48.1 ± 16.5%). PC activity significantly decreased from day 1 to day 2 (P= .001), then increased over time. Changes in PC (P < .001) and AT (P < .001) over time were likely associated with outcome with nonsurvivors having lower PC and AT activities than survivors.
Conclusion: Results of this preliminary study show that PC and AT activities change significantly over time in dogs with sepsis and both are likely related to survival.
The relationship between infection and coagulation is an area of intense investigation, with studies suggesting that hypercoagulability and subsequent microvascular thrombosis contribute to multiple organ dysfunction during sepsis.1–6 Tissue factor expression and generation of proinflammatory mediators lead to the activation of coagulation. Decreased activity of anticoagulant mechanisms results from consumption and subsequent depletion of anticoagulant proteins and from cytokine-induced downregulation of anticoagulant proteins. Studies in people with severe sepsis have identified decreased activities of protein C (PC) and antithrombin (AT),7–10 and some studies suggest that persistently decreased PC activity is associated with decreased survival.9,11 At least 3 studies in animals have suggested a similar imbalance in hemostasis in dogs with naturally occurring sepsis.12–14 In 1 study of 9 dogs with parvoviral enteritis, clinical evidence of thrombosis along with increased thromboelastograph maximum amplitude and decreased AT activity suggested an imbalance in hemostasis in this patient population.12 In another study of 20 dogs with naturally occurring sepsis, both PC and AT activities were significantly decreased compared with healthy controls.13 Finally, sepsis was identified as an underlying cause of thromboembolic disease in a retrospective study of pulmonary thromboembolism in dogs.14
Several studies in people have suggested that serial measurements of PC and AT may have prognostic value.9,11 In the study by Lorente et al,11 coagulation and fibrinolysis variables were measured in 48 people with sepsis at admission and on days 1, 4, and 7 after hospital admission. At admission, PC and AT were decreased but there was no substantial difference between survivors and nonsurvivors. However, serial measurements indicated that nonsurvivors had lower activities of PC on days 4 and 7 than did survivors. PC activity was persistently decreased in nonsurvivors, whereas gradual normalization of PC activity was noted in survivors. To our knowledge, serial measurement of PC and AT in dogs with sepsis has not been reported. Therefore, the goals of this study were to measure PC and AT activities serially in dogs with naturally occurring sepsis and to determine the relationship between changes in PC and AT and short-term outcome.
Materials and Methods
All dogs admitted to the Intensive Care Unit of the Tufts Cummings School of Veterinary Medicine Foster Hospital for Small Animals between July 2002 and January 2003 that fulfilled the criteria for sepsis were considered eligible for the study. Animals were classified as having sepsis if there was histopathologic or microbiologic confirmation of infection and if two of the following criteria were met: hypo- or hyperthermia (<100°F or >103°F), tachycardia (heart rate >140 beats per minute [bpm]), tachypnea (respiratory rate >20 breaths per minute), and leukopenia (<6,000/μL), leukocytosis (>16,000/μL), or >3% bands.15 Aerobic or anaerobic bacterial cultures and sensitivities were performed on appropriate samples from all septic dogs. All dogs weighing <10 kg and those that received anticoagulant therapy or blood products before the 1st blood sampling were excluded from the study. The study protocol was approved by the Tufts University Institutional Animal Care and Use Committee and written owner consent was obtained before subject enrollment.
After the diagnosis of sepsis, jugular venipuncture was performed and blood was collected from each dog and placed in a citrate tube (1 part 3.8% citrate: 9 parts blood). Time of 1st sampling was considered day 1. After the initial (day 1) blood sample, additional samples were collected at 24-hour intervals for 4 additional days (days 2–5) or until hospital discharge or death. Blood was centrifuged within 30 minutes of collection, separated, and citrated plasma was stored at −70 °C until analysis. PC and AT assays were performed at The Animal Health Diagnostic Center at Cornell University.
PC and AT activities were measured by functional chromogenic assays (STAchrom Protein Ca and STAchrom ATIIIb, respectively) as described previously.13 The standard curves for determination of patient and control PC and AT activities were derived from serial dilutions of a pooled, normal canine reference plasma (prepared from 15 normal healthy dogs) having an assigned value of 100% activity. The reference ranges for PC and AT activities in the dog were established at 75 to 135% of normal canine-pooled plasma.
All data were examined graphically for normalcy before analysis. Paired t-tests were used to compare PC and AT activities among days. Because PC and PT activities may affect the survival of the dogs whereas PC and AT activities are observable only when a dog is alive, the simultaneous 2-equation treatment effect model was used to analyze the changes in PC and AT activities over time in order to account for potential selection bias because of death.16,17 Similar models have been applied in biomedical research to decrease attrition bias.18 The model consists of a regression equation for PC or AT activities that relates to whether a dog survived through a given time point and a “treatment” equation for survival status that depends on time.
The general model is as follows: where Yit is the variable of interest (eg, PC or AT) for dog i at time point t, Dit is a binary variable representing survival status (1 for immediately before death and 0 for alive) for dog i at time t, δ represents the differences in outcomes between observations immediately before death and those of surviving counterparts, ɛit is the error term for the regression equation, γ measures how the probability of being the last observation immediately before death changes over time t, and υit is the error term for the treatment equation. Serial correlations within each dog were accounted for by robust sandwich estimator of variance. The model was estimated by the “treatreg” procedure in Stata SE 9.2.c
A P < .05 was considered significant. Data were analyzed by commercial statistical software.c
Twelve dogs with sepsis were enrolled in the study. Breeds included Golden Retrievers (n = 3) and one each of the following: Newfoundland, Doberman, German short hair pointer, Rottweiler, Shetland Sheepdog, Borzoi, Great Dane, and mixed breed. There were 8 male castrated dogs and 4 female dogs (3 spayed, 1 intact) and the mean age was 7.5 ± 3.0 years. Baseline data included a mean body temperature of 103.3 ± 1.6°F, heart rate of 159 ± 43 bpm, and respiratory rate of 51 ± 27/min. All 12 dogs completed the 1st 3 days of the study, 7 of the 12 dogs completed 4 days of the study before death or discharge from the hospital, and 5 of the 12 dogs completed all 5 days of the study before death or discharge from the hospital. Dogs were considered survivors if they survived to hospital discharge. Overall, 6 of the 12 dogs with sepsis died.
The most common source of sepsis was abdominal (n = 6) followed by thoracic (n=4) and urinary (n=2). Underlying causes of abdominal sepsis included foreign body with intestinal perforation (n=2), colonic ulceration secondary to neoplasia (n=2), duodenal perforation (n=1), and unknown (n=1). Respiratory causes of sepsis included pneumonia (n=3) and septic pleuritis (n=1). Urosepsis in the 2 remaining dogs was caused by purulent transmural cystitis. A positive culture result was obtained from all 12 dogs. Seven dogs had pure Gram-negative infections, 1 dog had a pure Gram-positive infection, and 4 dogs had mixed Gram-positive and Gram-negative infections. Four of the 12 septic dogs had multiple organisms isolated. The most common organism isolated was Escherichia coli (9 of 12 dogs). Eight dogs presented with sepsis, whereas 4 dogs developed sepsis during hospitalization. Of the dogs that developed sepsis, the median time period from admission to the hospital to diagnosis of sepsis was 2 days (range, 2–7 days).
Data for both PC and AT were normally distributed for all time points of the study. All dogs lived for the 1st 3 days of hospitalization. Therefore, changes in PC and AT activities over this period of time were assessed. PC and AT activities for all dogs were 66.0 ± 25.8% and 48.1 ± 16.5%, respectively, on day 1; 48.5 ± 28.4% and 38.0 ± 16.2% on day 2; and 56.3 ± 32.4% and 46.9 ± 17.5% on day 3. PC and AT activities were decreased on day 1 in 9/12 and in 11/12 dogs, respectively (Figs 1, 2). Average PC activity decreased significantly from days 1 to 2 (P=.001), and then increased over time (Fig 1). PC and AT activities were significantly associated with outcome, with nonsurvivors having lower PC (P<.001) and AT (P<.001) activities than survivors (Figs 1, 2).
This preliminary study documented a substantial association between changes in PC and AT activity and survival. Studies of sepsis in people have found a significant correlation between PC activity and outcome. In the study by Yan et al,8 a trend toward an increase in 30-day mortality was observed with lower baseline PC activity, with significance achieved using measurements taken 44 hours after admission. Other studies failed to detect a correlation between mortality and baseline PC activity, but found subsequent changes in PC to be significantly related to outcome.11 In the study of septic dogs by deLaforcade et al,13 PC activity at admission was not significantly associated with outcome. However, the results of the present study are similar to findings in humans as changes in PC activity were significantly associated with outcome.
The results of the present study also confirm previous findings of reduced AT and PC activities in dogs with naturally occurring sepsis and support the predominance of a hypercoagulable state in this disease process. To date, few prospective studies investigating changes in coagulation associated with naturally occurring sepsis in dogs have been performed. However sepsis has been identified as an underlying cause of thrombosis in several retrospective studies of thrombotic disorders.14,19,20 Fewer studies have been performed in cats, but sepsis was identified as an underlying disease in a retrospective study of pulmonary thromboembolism in cats.21 Despite the recurrent association of sepsis and thrombosis, anticoagulation of dogs with sepsis currently is not routinely employed. Reasons for not using anticoagulation may include lack of an overt association between death from sepsis and thromboembolic disease, and fear of a bleeding diathesis from disseminated intravascular coagulation. In 1 study investigating heparin administration in rabbits with sepsis-induced multiple organ failure and disseminated intravascular coagulation, heparin administration was associated with prevention of thrombocytopenia, leukopenia, and increases in plasma bilirubin and creatinine concentration.22 A randomized study of unfractionated heparin administration in humans with sepsis currently is under way.23
Finally, the present study also identified a significant decrease in PC activity from days 1 to 2, followed by a gradual increase in PC activity over time. In a study of humans with sepsis, serial PC measurements were significantly different between survivors and nonsurvivors. Survivors exhibited a progressive normalization in PC activity, whereas the PC activity of nonsurvivors remained persistently decreased.10 Fourrier et al9 also documented persistent decreases in PC activity in nonsurviving patients with sepsis and septic shock. In a study of disseminated intravascular coagulation, a decrease in PC activity was documented in the 1st 24–48 hours followed by a gradual increase in survivors.24
The implications of decreased PC and AT activity may extend beyond their impact on coagulation, because both PC and AT are also known to possess anti-inflammatory properties in addition to their roles as endogenous anticoagulants.25–32 The administration of recombinant human activated PC has been shown to decrease mortality in human patients with severe sepsis,33 and the recombinant form of this protein (rhAPC) is approved for use in human patients with severe sepsis. Whether it is the anticoagulant nature of PC or its ability to modulate inflammation that is responsible for the impact of PC administration on survival is still a subject of debate. To date, few clinically relevant markers of sepsis have been identified in dogs. Although routine administration of PC is unlikely in dogs because of a lack of a recombinant canine form of the protein, the high antigenicity of the recombinant human activated PC in dogs, as well as exorbitant cost, PC activity may remain a useful marker of disease severity in dogs with naturally occurring sepsis. Larger studies are required to further investigate the value of these proteins as prognostic indicators in sepsis in dogs.
There are a number of limitations to the current study, and the results should be viewed as preliminary. One potential limitation is the statistical model used. This model was selected over conventional regression models to avoid the risk of neglecting potential bias because of informative dropouts associated with death. Other limitations are the potential for bias in case selection with the definition of sepsis and the fact that the study included only 12 dogs. Despite the limitation of small sample size, the relationship between PC and AT activities and survival seems to mirror that observed in humans with sepsis. A study of larger size is needed, however, to confirm these relationships.
Data from the current study suggest that PC and AT activities change substantially over the course of hospitalization and that both are related to survival. Additional studies of these anticoagulant proteins as possible clinically relevant markers of sepsis are warranted.
aSTAchrom Protein C American Bioproducts, Parsippany, NJ