MacLeod, Jana B. A. MD, MSc; Lynn, Mauricio MD; McKenney, Mark G. MD; Cohn, Stephen M. MD; Murtha, Mary RN

Mortality from trauma remains a major public health issue. Early death has been shown to be associated with hemorrhage in a large percentage of patients.¹ Acquired coagulopathy is known to occur in patients with multiple injuries. Kapsch et al., Attar et al., and McNamara et al. have all shown in small cohorts of patients that both civilian and combat trauma is associated with coagulation and fibrinolytic derangements.^2‐4 Ferrara et al. in 1990 and Garrison et al. in 1996 both showed prothrombin time to be independently associated with mortality in bleeding trauma patients who were massively transfused.^5,6 Therefore, current explanations for coagulopathy in the patient with multiple injuries are described as secondary to an initial or developing insult or intervention. These primary causes of coagulopathic derangements include closed head injury, massive blood transfusion, and significant fluid resuscitation.

The actual relationship of prothrombin time (PT) and partial thromboplastin time (PTT) to mortality in the general trauma population is unknown. Our objective was to measure the predictive value of the initial coagulation profile, as measured by PT, PTT, and platelet count, on trauma‐related mortality.

PATIENTS AND METHODS

The trauma registry of prospectively collected data from a Level I trauma center was reviewed. All patients admitted to the University of Miami/Jackson Memorial Hospital, Ryder Trauma Center, Miami, Florida, from January 1, 1995, to December 31, 2000, were included in the cohort. The initial values for hemoglobin, platelet count, base deficit, PT, and PTT were retrieved. The registry information allowed calculation of the time interval from injury to arrival at the trauma bay and the interval from arrival to laboratory investigation. The disposition of the patient at the time of discharge from Ryder Trauma Center was classified for this study as either death or survival.

A multiple logistic regression model was performed to ascertain the association between the initial coagulation parameters and all‐cause hospital mortality. The following confounders were retrieved from the database and entered into the regression model: age, sex, Injury Severity Score (ISS), initial scene and resuscitation bay systolic blood pressure, presence of acidosis, time intervals, and traumatic brain injury (TBI). TBI was defined by the following variables: Glasgow Coma Scale (GCS) score, positive computed tomographic scan of the head, and a positive history for loss of consciousness.

The quantitative laboratory variables were categorized into abnormal versus normal with an abnormal PT value above 14.0 seconds and an abnormal PTT value above 34 seconds. Laboratory calibration was taken into consideration and all values above these limits that were coded by the laboratory as abnormal were categorized as abnormal for the analysis. All laboratory evaluations were performed in the same central laboratory.

Statistical Analysis

A Kaplan‐Meier survival analysis for time to death stratified for PT and PTT results was generated. Using a Wilcoxon log‐rank sum test, the difference between the two survival curves for patients with a normal versus abnormal PT and PTT was tested for statistical significance. Significance was defined as a value of p < 0.05. All statistical analysis was performed using Stata Version 6.0.

RESULTS

There were 20,103 patients entered in the trauma registry from January 1, 1995, to December 31, 2000. Of these patients, 14,397 had complete disposition data. Of these patients with disposition data, 10,790 had a PT value, 10,453 had a PTT value, and 11,290 had a platelet count value (Table 1). For the final multiple logistic regression analysis, a sample size of 7,638 patients had complete data for all variables included in the model. In Table 1, the frequency distribution of each variable with its corresponding sample size is listed.

Table 1

The cohort was 76.2% male, with a mean age of 38 years (range, 1‐108 years). The median ISS was 9 (range, 1‐75). The all‐cause hospital mortality of the cohort was 8.9% (1,276 of 14,397).

The median time to laboratory investigation was 31 minutes, with 75% of the patients having their laboratory investigations obtained by 54 minutes. The median length of stay was 2.8 days for the entire cohort, with a range from less than 1 day to 747 days. Of those who died, the mean length of stay was 1.1 days (range, 60 minutes‐626 days) as opposed to survivors whose median length of stay was 5.6 days (range, 12 hours‐747 days). Of the 1,276 deaths, 25% died within 3 hours, 50% by 26 hours, and 75% by 8.5 days. Only 7.4% (94 of 1,276) of the patients died after 30 days of hospitalization.

In patients with a normal PT, 489 of 7,796 died, as compared with 579 of 2,994 with an abnormal PT (6.3% vs. 19.3%; Χ² = 414.1, p < 0.001). The univariate analysis generated a crude odds ratio of 3.6 for death with an abnormal PT (95% confidence interval [CI], 3.15‐4.08; p < 0.001) and 7.81 for death with an abnormal PTT (95% CI, 6.65‐9.17; p < 0.001) (Table 2). The PT and PTT remained independent predictors of mortality in a multiple logistic regression model with an adjusted odds ratio of 1.35 (95% CI, 1.11‐1.68; p < 0.001) and 4.26 for PTT (95% CI, 3.23‐5.63; p < 0.001), respectively. The platelet count predicted death only in the univariate analysis (Table 2). In Table 3, the odds ratio for the other independent predictors included in the model are listed: age, ISS, base deficit, scene and trauma‐bay systolic blood pressure, hemoglobin, GCS score, and head computed tomographic scan result. Platelet count, gender of the patient, loss of consciousness, time from injury to arrival at the trauma bay, and time from arrival at trauma bay to laboratory investigation did not improve the model for prediction of mortality and were therefore not included in the final statistical model. Statistical analysis repeated with different cutoffs for abnormal for PT and PTT did not change the independent prediction of these variables for death in the regression model.

Table 2

Table 3

Kaplan‐Meier curves using data from length of hospital stay was performed and generated survival estimates for patients with and without abnormal PT and PTT (Figs. 1 and 2). The curves show that most deaths occurred early in the hospital stay. The median survival was lowest, 0.5 days, for patients with an abnormal PTT and 1.1 days for patients who had an abnormal PT. The overall median survival for the cohort in the final analysis (n = 7,638) was 6.9 days. The log‐rank test for equality of survivor functions showed a statistically significant difference between the curves for both PT (Χ² = 334.23, p < 0.001) and PTT (Χ² = 761.01, p < 0.001). The greatest difference in survival between the two curves for each laboratory value is seen in the initial postinjury period. Following the curves over time shows that the probability of survival, for patients with or without an abnormal coagulation profile, parallels after the initial time period.

Fig. 1

Fig. 2

DISCUSSION

In a cohort of all trauma patients arriving at a Level I trauma center, the initial prothrombin time and partial thromboplastin time independently predicted all‐cause mortality. Other than core body temperature, which was not available in this trauma registry data, other known prognostic indicators were controlled for and independently associated with mortality. An elevated initial prothrombin time is associated with a crude increase in mortality of 260%, and after controlling for the other prognostic indicators, an increased initial prothrombin time was still associated with a 35% increase in mortality when elevated above 14.0 seconds. The partial thromboplastin time was an even stronger predictor of mortality, with patients who had an abnormal PTT having a 326% increase in all‐cause mortality when controlling for other prognostic indicators. Neither the time factor of a prolonged prehospital period nor protracted time to laboratory investigation was shown to account for the correlation between death and an elevated PT and PTT.

There are certain limitations to this study. The retrospective design of the study limits the ability to infer a causeand‐effect relationship between death and coagulopathy. Mortality in this database was defined as all‐cause, and specific pathophysiologically plausible causes of death related to a coagulation defect are unknown. However, if an elevated PT or PTT were only related to hemorrhage‐associated death, then using all‐cause mortality would only weaken the detection of an association. Our model's ability to detect an independent link suggests that an elevated PT and PTT are either more general prognostic indicators for death or their association with hemorrhage‐related death is even stronger than we have detected here. Any additional information about the patient's medication usage, such as warfarin, or other medical problems such as liver failure that may account for a predisposition to coagulopathy is unavailable in this cohort. In the trauma registry database, there is no information on preexisting medical conditions or medications such as Coumadin, which could have accounted for an elevated PT in some patients. However, in light of the size of the cohort and the mean age, it is unlikely that this relationship could be explained by Coumadin use or preexisting medical conditions alone.

Our Kaplan‐Meier survival curves suggest that most deaths resulting from an elevated PT or PTT occur early in the hospital stay, with the probability of survival paralleling between the two groups as time goes on. This is consistent with a hypothesis that coagulopathy as defined by an elevated PT or PTT plays a major role in early trauma‐related deaths, attributed mainly to hemorrhage as is seen in previous epidemiologic series.¹

Numerous authors have documented in cohorts of head injury patients a high prevalence of coagulation abnormalities.^7‐11 Selladurai et al. showed that coagulation defects such as the level of the partial thromboplastin time and fibrin degradation products correlated with poor outcome in patients with an acute closed head injury and varied with the patient's GCS score.¹² May et al., in their retrospective review, found that patients with a GCS score of 7 or greater had no PT/PTT abnormalities at admission.¹³ However, 81% were coagulopathic if the GCS score was less than or equal to 6 and 100% were coagulopathic if the GCS score was 3 or 4, with a trend toward a higher mortality in the coagulopathic patients. In our cohort, we were able to show that, even when controlling for the GCS score, PT and PTT were highly associated with mortality. Head injury has been shown to affect the coagulation system by means of release of tissue thromboplastin and other mechanisms.^4,7,10,11,14 In children with TBI and coagulopathy, Chiarettii et al. also showed a poorer outcome that was independent of the GCS score.¹⁵ Our data are consistent with this finding by Chiarettii and colleagues. We have shown in this cohort that, even when controlling for head injury, the coagulation profile of injured patients is still an independent predictor of mortality. This suggests that coagulopathy after injury may not be solely attributable to head injury, as is often done. An elevated PT or PTT in this cohort impacts death separately from its elevation with head injury alone.

Massive transfusion is known to be related to significant coagulopathy in the population of trauma patients.¹⁶ However, in our cohort, the PT and PTT were measured at the initial laboratory investigation, which was drawn within 31 minutes for 50% and by 54 minutes for 75% of the patients. Therefore, even though we were unable to control for the amount of replacement these patients received because of its retrospective nature, the short time from injury to arrival and to laboratory investigation makes massive replacement unlikely. Other authors such as Harvey et al., who reviewed a heterogeneous group of bleeding patients who were massively transfused (>10 U/24 h), found no correlation with the number of units transfused and the degree of coagulopathy that developed.¹⁷ Their conclusion hypothesized that duration of tissue hypoperfusion was likely more important than simple hemodilution of coagulation factors. In another study of patients without TBI receiving massive transfusions, Cosgriff et al. found that the development of coagulopathy was independently linked with acidosis, hypothermia (<34°C), ISS > 25, and systolic blood pressure < 70 mm Hg.¹⁸ Other than the patient's body temperature, which was not available for analysis, our analysis controlled for these factors and coagulopathy, in our patient cohort, still increased the risk of death. This may reflect a primary process rather than a process secondary to the presence of acidosis, hypotension, or high ISS as is the current thinking.

Other studies lend support to the hypothesis that coagulopathy in the injured patient may not always arise from fluid replacement. In 1985, Ordog et al. showed that, in a cohort of 180 trauma patients who died, 97% had evidence of a coagulation defect before either fluid or blood replacement treatment.¹⁹ In a recent report from Israel, soldiers who had sustained a combat injury showed no correlation between the PT or PTT measured on hospital arrival and volume of prehospital fluid treatment.²⁰ Because our coagulation profile is measured early in the postinjury period, an abnormal finding at this time may reflect a primary influence on an increased risk of trauma‐related death rather than being secondary to commonly held mechanisms such as massive fluid resuscitation.

Is there a cellular mechanism occurring early in these trauma patients that activates the intrinsic/extrinsic coagulation pathway before the conventionally‐thought‐of causes of blood transfusion, fluid replacement, progressive acidosis, and ongoing hypothermia? Or is this early prolongation of the PT and PTT simply a marker of death as opposed to a possible origin of the cause? As we continue to look for correctable abnormalities and their treatment, we need to investigate prospectively the possible independent role of coagulation defects in mortality after multiple trauma. It will also allow us to better define the role of new treatments, such as recombinant factor VIIa, antithrombin III, and aprotinin, presently being investigated for their usage in bleeding caused by coagulopathy.^21‐25

We conclude that coagulopathy occurs early in the postinjury period, with 28% of patients having an abnormal PT and 8% of patients having an abnormal PTT shortly after arrival at the trauma bay. Furthermore, a trauma patient with an abnormal initial prothrombin time has a 35% increased risk of mortality and a patient with an abnormal initial partial thromboplastin time has a 326% increased risk of mortality over patients whose initial PT and PTT are normal, and it is independent of the other known risk factors the patient may exhibit.

REFERENCES