Predictors of Bleeding From Stable Pelvic Fractures

Major pelvic fractures after trauma are often associated with bleeding, intra-abdominal injuries, and death.¹ The overall mortality from pelvic fractures depends on the associated injuries and ranges from 5% to 14%.^2-4 Severe hemorrhage from the disrupted venous and arterial vessels near the fractured structures accounts for the early mortality, although a direct association between the pelvic bleeding and death is hard to draw if multiple injuries are present.^4,5 Several predictors, including fracture type, patient age, hemodynamic parameters, and laboratory values, have been described to identify those patients who may benefit from early angiography and embolization.^6,7

Stable pelvic fractures (SPFs) do not usually cause significant pelvic bleeding. Only case reports have been published on this issue.^8,9 The objective of this study was to identify the rate and predictors of bleeding among patients with SPFs. Our hypothesis is that these uncommon scenarios can be predicted on the basis of simple indicators, resulting in close monitoring or early intervention.

Patients with SPFs who were admitted to the level 1 academic trauma center at Massachusetts General Hospital, Boston, between January 1, 2002, and June 30, 2007, were retrospectively identified through our trauma registry. The SPFs were defined as fractures not requiring external or internal fixation. Significant bleeding (SigBleed) was defined as the need for blood transfusion and/or intervention for bleeding control within the first 24 hours after admission. The following categories of patients were excluded from the analysis: (1) patients with unstable pelvic fractures who needed operative stabilization but died before this was achieved; (2) patients without a computed tomographic (CT) scan of the pelvis because we were unable to characterize their fractures accurately and appreciate the extent of a pelvic hematoma; and (3) patients without pelvic or other sources of bleeding who received a blood transfusion for reasons clearly related to a chronic medical condition, such as anemia or chronic renal failure.

The outcome of the study was SigBleed caused by SPFs. We collected data on demographics; mechanism of injury; injury severity score; 6 abbreviated injury scores; admission systolic blood pressure (SBP); heart rate; Glasgow Coma Scale; admission hematocrit (Hct); units of packed red blood cells; fresh-frozen plasma transfused during the first 24 hours after admission; CT scan findings; angiography and embolization; hospital length of stay; intensive care unit (ICU) length of stay; and mortality. Because of the retrospective nature of our study, we could not evaluate the vectors of force causing the pelvic fracture and thus used the Tile classification, which focuses on pelvic stability, to categorize the fractures.^10,11 According to this classification, type A fractures are rotationally and vertically stable; type B fractures are vertically stable but rotationally unstable owing to an incomplete posterior arch disruption; and type C fractures are vertically and rotationally unstable owing to a complete posterior arch disruption. Three groups were defined: group A included patients without SigBleed; group B included patients with SigBleed of a nonpelvic cause (eg, splenic or liver laceration); and group C included patients with SigBleed caused by the SPF. The distinction between groups B and C was based on the presence of other injuries causing bleeding, such as a splenic or liver laceration (group B) or a large pelvic hematoma with extravasation on CT scans (group C).

Statistical analysis was performed with PASW for Windows, version 17.0 (SPSS Inc, Chicago, Illinois). Comparisons among the 3 groups were conducted with 1-way analysis of variance for continuous variables with normal distributions and with the Kruskal-Wallis test for continuous variables without normal distributions. Categorical variables were analyzed using the χ² test and the Fisher exact test. For all the continuous variables that could serve as risk factors, we created dichotomous variables using clinically relevant cutoff points. We then focused on the analysis of patients who did not have bleeding from a nonpelvic source. For this reason, we excluded group B patients from further analysis. Because of the small sample size of group C, we included the 3 most significant factors identified in the univariate analysis in a logistic regression model to identify independent predictors of SigBleed in group C compared with group A. The odds ratios (ORs) and 95% confidence intervals (CIs) of each independent predictor were calculated. P < .05 was considered statistically significant. The study protocol was reviewed and approved by the institutional review board of Massachusetts General Hospital.

Of 465 patients with SPFs, 74 were excluded (64 patients did not have a pelvic CT scan; 6 patients were transfused for chronic diseases; and 4 patients with unstable pelvic fractures died before receiving their intended operation). Finally, 391 patients formed the study population and were included in the analysis. There were 280 patients (72%) in group A, 90 (23%) in group B, and 21 (5%) in group C (Figure 1). A total of 310 patients (79%) had a Tile type A fracture and 20 (5%) had a Tile type B. Also, 61 patients (16%) had an isolated acetabular fracture. By definition, there were no type C fractures. Six patients (29%) in group C underwent angiography, and 5 (24%) proceeded to embolization. The mean (SD) hospital stay for the entire population was 8.6 (8.6) days, and 16 patients died (4%). In 3 patients, the pelvic bleeding was the cause of, or a significant contributor to, their death. An 83-year-old patient with SPFs, lower-extremity fractures, and upper-extremity degloving injuries died 24 hours after receiving multiple blood transfusions. A 91-year-old patient with SPFs and a humerus fracture underwent angiography, but no extravasation was found, and she did not undergo embolization. On posttrauma day 8, she died of multiorgan failure related to the initial bleeding. Finally, a 75-year-old multitrauma patient (SPFs, head injury, and lower- and upper-extremity fractures) underwent massive resuscitation and bilateral internal iliac embolization twice. He developed abdominal compartment syndromes, requiring decompression. Despite successful control of the bleeding, he died on posttrauma day 10 for the same reasons as those of the previous patient.

As expected, compared with the patients without SigBleed, those in group C were older and had a lower admission SBP and Hct and a higher rate of coumadin use. More group A patients required ICU admission. They had longer ICU and hospital stays and a higher mortality (Table 1).

The 2 groups had comparable units of packed red blood cells and fresh-frozen plasma transfusions, type of fractures, length of hospital stay, and mortality. However, patients in group C were older, more frequently injured after a fall, and had a lower injury severity score and admission Hct. They were less likely to be admitted to the ICU (Table 1). The sources of hemorrhage in group B are described in Table 2.

The multivariable analysis of group C vs group A produced the following independent predictors of SigBleed from SPFs: an Hct of 30% or lower (OR, 43.93; 95% CI, 9.78-197.32; P < .001); the presence of pelvic hematoma on CT (OR, 39.37; 95% CI, 4.58-338.41; P < .001); and an SBP of 90 mm Hg or lower (OR, 18.352; 95% CI, 1.98-169.87; P = .01). When all 3 independent predictors were present, 100% of the patients had SigBleed; when all 3 were absent no one had SigBleed.

To our knowledge, this is the first study focusing on bleeding from SPFs. Although SigBleed occurred in one-third of the patients, the SPF was the cause of it in only 5% of them. The overall mortality rate of 0.8%, which is directly associated with the pelvic bleeding, is similar to the rate reported in other studies.^4,5 It is well known that among severe pelvic fractures there are specific patterns, such as vertical shear fractures, bilateral pubic rami (butterfly) fractures, and pubic symphysis widening of more than 2.5 cm, that are associated with a higher likelihood of bleeding.^6,12,13 According to our findings, the only 3 independent risk factors that predicted bleeding in patients with SPFs were an Hct of 30% or lower, the presence of a pelvic hematoma on CT, and an admission SBP of 90 mm Hg or lower. It is therefore reasonable to assume that hypotension and anemia in a patient with an SPF and no other sources of intra-abdominal hemorrhage indicate pelvic bleeding unless proved otherwise. These findings should translate to an increased level of vigilance for relevant patients. Such patients should either have an early intervention or be monitored closely in a higher-care unit. Since external fixation is not appropriate for an SPF, the only potential interventions would be angiographic embolization or preperitoneal pelvic packing, the former being a more preferable option than the latter.

The indications for angiographic embolization after pelvic fractures are controversial. It has been suggested that SBP, Hct, heart rate, age, blood transfusion, a blush on CT, and fracture pattern are predictors of active bleeding on angiography.^7,14,15 Six of our group C patients underwent angiography, and 5 proceeded to embolization within 24 hours after presentation to the emergency department (Figure 2). Three patients underwent embolization of an internal iliac artery. Two patients underwent subselective embolization of the obturator and the superior gluteal arteries, respectively, and 1 patient underwent bilateral internal iliac artery embolization, as previously described in the literature.¹⁶ Of those 5 patients who underwent embolization, all demonstrated evidence of hemodynamic instability during the initial hours of their hospitalization, and 1 eventually died.

Preperitoneal pelvic packing has been described as a method to control acute pelvic bleeding if (1) angiography is not readily available, (2) the patient is in extremis, or (3) an emergent operation is needed before angiography.¹⁷ In our population, no preperitoneal pelvic packings were performed. In our institution, the interventional radiology team provides an around-the-clock service, which allows expeditious transfer of even severely injured patients to the angiography table, with an intensive care team providing continuous care. For this reason, preperitoneal pelvic packing for control of pelvic bleeding is infrequently practiced.

We therefore suggest the following clinical pathway for patients with SPFs and no other clinically significant injuries: If none of the 3 risk factors is present, the patient can be safely treated in a regular hospital bed or potentially discharged. If 1 or 2 of the risk factors exist, the patient should be monitored in an intensive care environment, and a low threshold for intervention should be maintained. If all of the risk factors exist, the patient should undergo emergent angiography or preperitoneal pelvic packing according to the circumstances.

Except for Hct and SBP, the multivariate analysis did not identify other independent predictors of bleeding. In accordance with other studies emphasizing age,¹⁸ our univariate analysis showed that the rate of patients older than 65 years was more than double in group C compared with group A (86% vs 42%). However, age was not identified as an independent predictor of bleeding. Similarly, other factors, such as coumadin use (P = .79) and associated injuries (injury severity score) (P = .13), which were expected to make a difference, did not achieve significance in the multivariable analysis. The limited sample size of this retrospective study may affect the results of the analysis and may be the main reason for preventing seemingly important clinical factors from becoming statistically significant. However, it is unlikely that large numbers of bleeding patients with SPFs will be described in single-institution studies. These patients are infrequent and will remain so. For the same reason (limited sample size), the 95% CI intervals of the selected independent risk factors were very wide. Obviously, the small sample size decreases the confidence in the true effect size of these factors from a statistical point of view. Clinically, it makes sense that 1 radiographic and 2 hemodynamic parameters predict SigBleed. Another limitation of our study is that 1 of our outcomes (blood transfusion) was not standardized and therefore was subjected to the treating physicians' personal preferences. Finally, the decision to perform angiography was left to the discretion of the trauma surgeon, although our team operates under a written protocol that provides guidelines for the management of pelvic fractures.

An SPF is often dismissed as a source of SigBleed, possibly resulting in delays of care. Our study shows that SigBleed from SPFs is indeed an infrequent event and happens in only 5% of patients with SPFs. However, it can be predicted by 3 simple parameters: an Hct of 30% or lower, the presence of a pelvic hematoma in CT scan, and an admission SBP of 90 mm Hg or lower. For those few patients with SPFs and a high likelihood of SigBleed, early intervention or close monitoring is essential.

Correspondence: George C. Velmahos, MD, PhD, Division of Trauma, Emergency Surgery, and Surgical Critical Care, Massachusetts General Hospital, 165 Cambridge St, Ste 810, Boston, MA 02114 (gvelmahos@partners.org).

Accepted for Publication: September 10, 2010.

Published Online: December 20, 2010. doi:10.1001/archsurg.2010.277

Author Contributions:Study concept and design: Bramos, Velmahos, Butt, Fikry, and Smith. Acquisition of data: Bramos and Butt. Analysis and interpretation of data: Bramos, Velmahos, and Chang. Drafting of the manuscript: Bramos, Velmahos, and Butt. Critical revision of the manuscript for important intellectual content: Fikry, Smith, and Chang. Statistical analysis: Bramos, Butt, Fikry, and Chang. Study supervision: Bramos, Velmahos, and Smith.