It is estimated that more than 1.5 million people in the United States sustain traumatic head injury each year [¹]. Radiologic evidence of intracranial hemorrhage at the time of presentation is present in up to 45% of cases and is associated with a markedly poorer prognosis [²,³]. Traumatic intracranial hemorrhage encompasses cerebral contusion, subdural hematoma, subarachnoid hemorrhage, epidural hematoma and intracerebral hemorrhage. These are characterized by a relatively high risk of bleeding progression, especially within the first 24 hours [²,^4-6].

Traumatic intracranial hemorrhage is also associated with a high risk of thromboembolic complications [⁷]. This risk is related to the immobility of head-injured patients arising from the underlying neurologic lesion itself, concomitant injuries following trauma, or the use of sedatives and neuromuscular blocking agents. The reported incidence of deep vein thrombosis (DVT) ranges between 18 and 58% in the absence of anticoagulant prophylaxis [^8-11]. DVT is associated with increased morbidity and mortality, including risk of fatal pulmonary embolism (PE) [¹²].

Anticoagulant prophylaxis to prevent DVT is recommended for trauma patients without intracranial or other serious hemorrhage [¹³]. This therapy is also effective in post-operative neurosurgical patients undergoing brain tumour excision [¹²]. Only one small quasi-randomized clinical trial has studied anticoagulant prophylaxis versus pneumatic compression devices in patients with intracranial hemorrhage following trauma. This trial found no benefit or harm associated with use of anticoagulant prophylaxis, but only had sufficient sample size to detect very large differences in rates of venous thromboembolism or intracranial hemorrhagic progression [¹⁴]. One observational study reported rates of hemorrhagic progression in 150 patients with intracranial hemorrhage that were treated with low molecular weight heparin prophylaxis 24 hours after the initial head injury [¹⁵]. One quarter already had radiological evidence of intracranial hemorrhagic progression at the time anticoagulant prophylaxis was started, but only six patients (4%) subsequently developed progression in the 24 hours following initiation of anticoagulation prophylaxis. The study lacked a control group limiting inferences regarding the safety and efficacy of this therapy.

Many physicians report routinely withholding anticoagulation prophylaxis in patients with traumatic intracranial hemorrhage because of concerns that it may increase the risk of potentially devastating intracranial hemorrhagic progression [¹⁶,¹⁷]. Observational studies have also documented variable practice patterns for anticoagulation prophylaxis in such patients. A small retrospective study of 88 patients with traumatic brain injury observed that only 42% ever received low-molecular weight heparin prophylaxis, and the mean time to initiation of therapy was 14 days [¹⁸]. Similarly, a multicentre retrospective study of trauma patients with ICU length of stay greater than seven days found that prophylaxis was initiated within two days in only 25% of patients, and another quarter did not receive prophylaxis until at least one week following injury [¹⁹]. Thus, the use and timing of anticoagulation prophylaxis in these patients remains uncertain and controversial, and most are treated with more conservative and less effective measures such as graduated compression stockings, intermittent pneumatic compression devices, and physiotherapy [¹³,¹⁶].

The lack of persuasive evidence to guide decisions about using anticoagulant prophylaxis in patients with traumatic intracranial hemorrhage implies that clinicians must make decisions based on their own assessments of the risks and benefits. Our objective was to explore the decision to use or withhold anticoagulant prophylaxis in a common clinical scenario characterized by competing risks of hemorrhagic versus thromboembolic complications. We used decision analysis modeling techniques to compare the risks of progression of intracranial hemorrhage following trauma versus the potential benefits of reducing venous thromboembolism with anticoagulant prophylaxis. Decision analysis involves identifying the most important available clinical choices and determining probabilities of all important potential outcomes following each of these choices [²⁰]. This methodology has been used previously to evaluate the risks and benefits of anticoagulant prophylaxis in neurosurgical patients undergoing craniotomy [²¹], but to our knowledge has not specifically been used to evaluate patients with traumatic intracranial hemorrhage.

The expected value associated with withholding anticoagulation prophylaxis (0.90) was similar to that associated with the LMWH strategy (0.89; Figure 4). A threshold value was reached within our range of estimates for the effectiveness of anticoagulant prophylaxis for preventing DVT (threshold 0.80, range of estimates 0.33 to 0.82; Figure 5), suggesting that providing LMWH would have a greater expected value than withholding LMWH only if its effectiveness for preventing DVT exceeded 80%. Similarly, the variable representing the effectiveness of withholding anticoagulant prophylaxis for reducing the risk of intracranial hemorrhagic progression also reached a threshold value within our range of estimates (threshold 0.05, range of estimates 0.001 to 0.990; Figure 6), suggesting LMWH would become the preferred strategy if it increased the risk of ICH progression by no more than 5% above the baseline risk.

We performed sensitivity analyses using the most frequently reported risk estimates obtained from our survey of Canadian neurosurgeons and neurointensivists. These analyses did not produce any additional threshold variables and the results did not qualitatively change our findings. We also considered sequential compression devices as a third strategy for preventing DVT. For this analysis, we considered the effectiveness of sequential compression devices for preventing DVT to be between that of LMWH and no anticoagulation prophylaxis (effectiveness 0.19, equivalent to point estimate of DVT incidence of 26%) and the risk of intracranial hemorrhagic progression to be the same as no anticoagulation prophylaxis. The expected value of the sequential compression device strategy was 0.90, similar to that of the other strategies (results not shown, but available upon request).

Our results suggest that the decision of whether or not to use anticoagulant thromboprophylaxis 24 hours after traumatic intracranial hemorrhage is a toss-up. Although the no prophylaxis strategy was associated with a slightly higher expected value (0.90 versus 0.89), this difference is unlikely to be clinically important. For example, the magnitude of this difference is equivalent to the disutility associated with administering a subcutaneous injection of anticoagulant prophylaxis.

Despite the similar expected values associated with each strategy, anticoagulant thromboprophylaxis became the preferred approach only in situations where the incremental risk of hemorrhagic progression was very low, or when its effectiveness in preventing venous thromboembolism was very high. These situations reflect the limits of our plausible risk estimates, and therefore seem unlikely to apply to most clinical situations. Considering the uncertainty, routinely withholding anticoagulant prophylaxis seems an appropriate strategy based on our findings, especially in the early phase when risk of hemorrhagic progression is perceived to be highest [¹⁷].

Our results are consistent with the findings of our recent survey of Canadian practice [¹⁷]. Of the 160 intensivists and neurosurgeons surveyed, almost two-thirds (60%) of intensivists and neurosurgeons indicated they would use, at some time, anticoagulant thromboprophylaxis in patients with intracranial hemorrhage due to traumatic brain injury. However, only one-third (34%) of these respondents reported that they would start this thromboprophylaxis within two days of the surgery, reflecting their concerns about risk of hemorrhagic progression. However, slightly more than half (57%) would start anticoagulation within four days and most (80%) by one week, suggesting that the perceived risk of hemorrhagic progression decreases over time. We lacked sufficient data to evaluate the risks and benefits of initiating anticoagulant prophylaxis after more than 24 hours, but this could be the topic of future research. We only considered anticoagulant prophylaxis with LMWH rather than unfractionated heparin, reasoning that most available studies in neurosurgical patients used the former and that direct comparisons of these two types of prophylaxis have yielded similar results [⁴⁰].

Instead of considering anticoagulant prophylaxis, some clinicians may choose to routinely screen patients with ultrasound or other imaging modalities to identify patients with DVT who require treatment with vena cava filters or full-dose anticoagulation, but we did not consider this to be a strategy for DVT prophylaxis. Similarly, depending on available resources some clinicians may choose to use sequential compression devices to prevent DVT while withholding anticoagulation to avoid intracranial hemorrhagic progression. However, this strategy would have important cost considerations and did not yield an apparent benefit in our decision analysis or reduce DVT incidence in a recent large clinical trial in stroke patients [⁴¹].

A limitation of all decision analyses is that they rely on having accurate estimates of probabilities of outcomes following clinical choices, and such estimates often are lacking in the literature for the clinical scenario of interest. For example, a limitation of our decision analysis was the difficulty quantifying the incremental risk of intracranial hemorrhagic progression at 24 hours with anticoagulant prophylaxis. Our only estimate was derived from a single prospective study that lacked a control group, so we used a very wide confidence interval in sensitivity analysis. These analyses suggest that anticoagulant prophylaxis would still only become the most effective strategy if the true incremental risk of hemorrhagic progression were less than 5% above baseline. Furthermore, the expected values associated with clinical choices in a decision analysis will depend on the utilities that are assigned to each clinical outcome, and different patients and clinicians may weigh such outcomes differently in the real world.

Our model showed no clear advantage to providing or withholding anticoagulant prophylaxis for DVT/PE prevention at 24 hours after traumatic brain injury associated with intracranial hemorrhage. In the context of such a toss-up, and given that the most disastrous event (other than death) encountered in this model is the development of a disabling neurological deficit, we would recommend against the use of routine anticoagulant prophylaxis at 24 hours for these patients. However, considering that both strategies were associated with nearly equivalent expected values, our model would suggest that it should be ethical to conduct a randomized, controlled, clinical trial to evaluate the safety and efficacy of using anticoagulant prophylaxis in patients with traumatic brain injury and intracranial hemorrhage.

? Patients with traumatic brain injury associated with intracranial hemorrhage are at high risk for developing venous thromboembolism

? Anticoagulation prophylaxis is proven to decrease the risk of venous thromboembolism in other patient groups, but may increase the risk of progression of intracranial hemorrhage

? The decision to provide anticoagulation prophylaxis to these patients therefore represents a trade-off that can be examined using decision analysis techniques

? Our decision analysis model showed no clear advantage to providing or withholding anticoagulant prophylaxis for thromboembolic prophylaxis at 24 hours after traumatic brain injury associated with intracranial hemorrhage

? Randomized controlled trials of this study question are justifiable and needed to guide clinicians

CNS: central nervous system; DVT: deep vein thrombosis; ICH: intracranial hemorrhage; ICU: intensive care unit; LMWH: low-molecular weight heparin; PE: pulmonary embolism; SYS: systemic bleeding.

The authors declare that they have no competing interests.

DCS, JRC, JTG and ASD conceived of and designed the manuscript. DCS, JRC, VA and DW were responsible for data acquisition. DCS, JRC, DW and ASD were responsible for analysis and interpretation. DCS, JRC, DW, VA, JTG and ASD drafted and revised the manuscript. All authors read and approved the final manuscript.

DCS is supported by a New Investigator Award from the Canadian Institutes for Health Research.