aOR indicates adjusted odds ratio; IVC, inferior vena cava.
DVT indicates deep vein thrombosis; IVC, inferior vena cava; and PE, pulmonary embolism.
Abbreviations: IVC, inferior vena cava; NA, not applicable.
Abbreviation: IVC, inferior vena cava.
aP = .004 for the difference between groups.
bP = .29 for the difference between groups.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Sarosiek S, Rybin D, Weinberg J, Burke PA, Kasotakis G, Sloan JM. Association Between Inferior Vena Cava Filter Insertion in Trauma Patients and In-Hospital and Overall Mortality. JAMA Surg. 2017;152(1):75–81. doi:10.1001/jamasurg.2016.3091
Does insertion of an inferior vena cava (IVC) filter in trauma patients affect mortality?
This cohort study demonstrated no statistically significant difference in overall survival between trauma patients who did and did not receive an IVC filter.
Inferior vena cava filters should not be placed in trauma patients in an attempt to improve overall survival.
Trauma patients admitted to the hospital are at increased risk of bleeding and thrombosis. The use of inferior vena cava (IVC) filters in this population has been increasing, despite a lack of high-quality evidence to demonstrate their efficacy.
To determine if IVC filter insertion in trauma patients affects overall mortality.
Design, Setting, and Participants
This retrospective cohort study used stratified 3:1 propensity matching to select a control population similar to patients who underwent IVC filter insertion at Boston Medical Center (a level I trauma center at Boston University School of Medicine) between August 1, 2003, and December 31, 2012. Among patients with an IVC filter and matched controls, age, sex, race/ethnicity, and Injury Severity Score were entered into a multivariable logistic regression model to calculate a propensity score. Matching was stratified by the date of injury.
Main Outcomes and Measures
Multivariable logistic regression was used to compare hospital mortality across both groups, adjusting for age, sex, race/ethnicity, Injury Severity Score, and brain injury severity using the head and neck Abbreviated Injury Score. To determine any significant difference in mortality, patient characteristics and mortality data from the National Death Index were analyzed in all patients and in those who survived 24, 48, and 72 hours after injury, as well as at hospital discharge.
Among 451 trauma patients with an IVC filter and 1343 matched controls without an IVC filter, the mean (SD) age was 47.4 (21.5) years. The median Injury Severity Score overall was 24 (range, 1-75). Based on a mean follow-up of 3.8 years (range, 0-9.4 years), there was no significant difference in overall mortality or cause of mortality in patients with vs without an IVC filter who survived more than 24 hours from the time of injury, independent of the presence or absence of deep vein thrombosis or pulmonary embolism at the time of IVC filter placement. Additional analyses at shorter intervals of 6 months and 1 year after discharge also showed no significant difference between the 2 groups of patients. Eight percent (38 of 451) of the IVC filters were removed at Boston Medical Center during the follow-up period.
Conclusions and Relevance
The research herein demonstrates no significant difference in survival in trauma patients with vs without placement of an IVC filter, whether in the presence or absence of venous thrombosis. The use of IVC filters in this population should be reexamined because filter removal rates are low and there is increased risk of morbidity in patients with filters that remain in place.
Venous thromboembolism (VTE) is a significant cause of morbidity and mortality in the United States. Approximately 900 000 patients per year have a clinically significant deep vein thrombosis (DVT) or pulmonary embolism (PE).1 Pharmacologic anticoagulation is typically used for VTE prophylaxis in hospitalized patients or for treatment of an acute venous thrombosis. Trauma patients often have a perceived contraindication to anticoagulation early in their postinjury course because of injury-specific and systemic factors.2,3 Placement of an inferior vena cava (IVC) filter is frequently considered in trauma patients because of the high risk of VTE associated with traumatic injury and the risks of anticoagulation in this population.4
Since the invention of the permanent percutaneous IVC filter in 1973 and the retrievable IVC filter in the 1990s, its use has become a standard part of treatment for select patients with acute lower-extremity venous thrombosis who cannot receive anticoagulation.5 The Prévention du Risque d’Embolie Pulmonaire par Interruption Cave (PREPIC) trial,6 prospectively studied patients with acute, proximal lower-extremity venous thrombosis after IVC filter placement. Patients randomized to filter placement and anticoagulation had a lower incidence of PE in the first 2 weeks of follow-up compared with patients receiving anticoagulation alone. Since that initial trial, a second PREPIC trial was published and demonstrated no decrease in symptomatic recurrent PEs in patients who undergo insertion of IVC filters while receiving anticoagulation.7 Based on retrospective reviews and smaller observational trials,8 the use of IVC filters has become commonplace in patients who have an acute VTE and are unable to tolerate anticoagulation. They are also often used as prophylactic therapy in patients at high risk of thrombosis.
Because of limited high-quality data supporting the use of IVC filters in the absence of an acute DVT, guidelines for their use vary by specialty. The American College of Chest Physicians9 recommends the use of an IVC filter for patients with an acute lower-extremity DVT or PE who cannot receive anticoagulation. The guidelines do not recommend the use of IVC filters for venous thrombosis prophylaxis. The Society of Interventional Radiology10 guidelines state that IVC filters can be used in patients with evidence of venous thrombosis or that they can be used prophylactically in patients without known thrombosis. The guidelines from the Eastern Association for the Surgery of Trauma8 recommend consideration of an IVC filter in high-risk trauma patients, even in the absence of a DVT or PE.
The data supporting the use of IVC filters in trauma patients are based on multiple, small matched control trials.8 These trials have been studied in systematic reviews and meta-analyses,11,12 which conclude that there may be an association between their use and a lower incidence of PE but that mortality is unchanged. The number needed to treat to prevent one additional PE is reported as 109 to 969 patients depending on a patient’s underlying risk.11,12 Despite a lack of high-quality evidence, there has been increasing use of IVC filters in the trauma population.13 In an effort to delineate the long-term mortality risk associated with IVC filter placement in trauma patients with and without a VTE, we performed a retrospective cohort study of trauma patients at Boston Medical Center, a level I trauma center at Boston University School of Medicine.
Approval by the Boston Medical Center and Boston University Medical Campus Institutional Review Board was obtained for a retrospective review of all trauma patients who were admitted to Boston Medical Center between August 1, 2003, and December 31, 2012. Because this was a retrospective study, patient consent was not obtained. Participant data were obtained by performing a search of Boston Medical Center’s trauma registry. This database is part of version 5 of the National Trauma Registry of the American College of Surgeons and includes all trauma patients admitted to Boston Medical Center who have a designated mechanism of injury and are diagnosed as having an injury that can be coded by the International Classification of Diseases, Ninth Revision, and the Abbreviated Injury Score (AIS).14 The AIS and the Injury Severity Score (ISS) are anatomical scoring systems used to categorize patients with multiple traumatic injuries.15 All identifying information was removed before use in data analysis and stored in a secure location.
There were 18 555 patients admitted to the Boston Medical Center trauma service during the specified time frame. Six hundred thirty-nine patients who were younger than 7 years or older than 95 years were excluded because of small selection samples in those age ranges. An additional 31 patients were excluded for incomplete ISS data, leaving 17 885 patients for analysis. Radiologic studies in a control population were reviewed to ensure that no IVC filter was present.
To identify comparison groups, the trauma database was searched for patients with an IVC filter placed during the index hospitalization. A medical record review was performed to confirm documented IVC filter placement. The final number of patients with an IVC filter was 451. The remaining patients did not have documentation of IVC filter placement.
Patient characteristics were extracted from the database, including the date of birth, sex, race/ethnicity, date of injury, type of injury, ISS, AIS, date of IVC filter placement, presence of VTE at the time of the filter placement, date of discharge, and date of death. To obtain mortality data for participants included in the data analysis, a query was submitted to the National Death Index (NDI) to determine if a patient death had been reported. Of those patients who had died, the primary cause of death was obtained from death certificate information supplied by the NDI to be used as a comparison between patients with vs without IVC filters.
Stratified 3:1 propensity matching was used to select a control population similar to patients with IVC filter insertion at Boston Medical Center.16 Three patients without an IVC filter were matched to each patient with an IVC filter. The following 4 variables were entered into a multivariable logistic regression model to calculate a propensity score: age, sex, race/ethnicity, and ISS. Matching was stratified by time to make the long-term follow-up comparable and to account for changing standards of care. Four study periods were used, including 2003 to 2005, 2006 to 2008, 2009 to 2010, and 2011 to 2012.
The matched sample was described in terms of counts and proportions for categorical variables. Continuous variables were described by means and SDs. Patient characteristics were compared across the groups using a t test for continuous variables and a χ2 test for categorical variables. The primary causes of death were compared between the groups using the Fisher exact test.
Multivariable logistic regression was used to compare hospital mortality across both groups, adjusting for age, sex, race/ethnicity, ISS, and brain injury severity using the head and neck AIS. This analysis was performed in all patients and in those who survived 24, 48, and 72 hours after injury, as well as at hospital discharge. Any associations were expressed with adjusted odds ratios and corresponding 95% CIs.
Survival in patients with IVC filter insertion and those without IVC filter insertion was compared using the log-rank test, and Kaplan-Meier plots were generated. Three events were selected as starting points for the survival analysis, including the date of injury, date of IVC filter placement, and date of discharge. Analyses were also performed at 6 months and 1 year after discharge. For patients without IVC filter insertion, a date of IVC filter placement was imputed from the corresponding matched patient with an IVC filter. Subset analyses were then performed for patients with IVC filters with and without VTE at the time of placement and their corresponding matches to determine if the indication for IVC filter placement affected patient outcomes.
All analyses were performed using statistical software (SAS, version 9.3; SAS Institute Inc). P < .05 was considered statistically significant.
The total number of patients included for the final analysis was 1794 (451 with IVC filters placed and 1343 matched controls). Their characteristics are compared in Table 1. There were no significant differences between the groups in the variables used in the propensity score matching, including age, sex, race/ethnicity, ISS, and study period.
Most trauma patients at Boston Medical Center were male, with 1266 (70.6%) male and 528 (29.4%) female in the present sample. Overall, 1311 patients (73.1%) were of white race/ethnicity, 278 (15.5%) were of black race/ethnicity, 95 (5.3%) were of Hispanic race/ethnicity, and 110 (6.1%) were of other races/ethnicities. The mean (SD) age of all patients included in the analysis was 47.4 (21.5) years. The median ISS score overall was 24 (range, 1-75). The ISS is further divided into the AIS for each patient to characterize anatomical areas of injury.14 The AIS was not used for matching, but the head and neck AIS was calculated. The mean (SD) score was 2.3 (2) overall, with mean (SD) scores of 2 (2) in the patients with IVC filters and 2.5 (2) in the control group (P < .001) (Table 1). The extremities AIS indicated that 309 patients (68.5%) in the IVC filter group and 611 (45.5%) in the control group had an extremity injury, with mean (SD) scores of 1.9 (1.4) and 1.2 (1.3), respectively (P < .001). Patient injuries were categorized by blunt or penetrating trauma (Table 2). In the group without an IVC filter, 1154 patients (85.9%) had blunt trauma and 189 (14.1%) had penetrating trauma. In the group with an IVC filter, 416 patients (92.2%) had blunt trauma, and 35 (7.8%) had penetrating trauma.
In those who received an IVC filter, most were placed in patients without thrombosis or PE, with only 69 patients (15.3%) having evidence of venous thrombosis at the time of IVC filter placement. In the control group, 10 patients (0.7%) had a DVT during the index hospitalization and 11 (0.8%) had a PE. The median time from injury until IVC filter placement was 3 days (range, 0-66 days). Eight percent (38 of 451) of the IVC filters were removed at Boston Medical Center during the follow-up period.
There was a significantly lower rate of hospital deaths among the patients who received an IVC filter compared with those who did not (5.5% [25 of 451] vs 22% [296 of 1343], P < .001), and most hospital deaths were associated with early complications after trauma (Table 3). The higher overall in-hospital mortality (adjusted odds ratio, 6.5; 95% CI, 3.9-10.6) in patients not receiving IVC filters remained in the multivariable-adjusted model. In those who did not receive an IVC filter, most hospital deaths occurred before the opportunity for IVC filter placement, with 14.5% (195 patients) dying less than 1 day from the date of injury, while 0.2% (1 patient) in the IVC filter group died in the first 24 hours. This difference was not statistically significant if patients survived the initial 24 hours (P = .07; adjusted odds ratio, 1.62; 95% CI, 0.97-2.72) or 48 hours after injury (P = .28; adjusted odds ratio, 1.34; 95% CI, 0.79-2.28). It remained nonsignificant if patients survived 72 hours after injury, which is the median time of IVC filter placement (adjusted odds ratio, 1.13; 95% CI, 0.65-1.97) (Figure 1).
During this study, mortality data from the NDI were available through December 31, 2012. Participants herein had a mean follow-up at Boston Medical Center of 3.8 years (range, 0-9.4 years) (Table 1). When measured from the date of discharge or the date of IVC filter placement, there was no significant difference in overall mortality between the 2 groups (Figure 2A). Additional analyses at shorter intervals of 6 months and 1 year after discharge also showed no significant difference between the 2 groups of patients. These survival trends did not change when analyzing subgroups of patients with and without VTE at the time of IVC filter placement (Figure 2B and C).
The primary cause of death was also obtained from the NDI and analyzed in all participants. The most common cause of death overall and in both subgroups before discharge from the hospital was assault, injury, or accident, including 40.7% (35 of 86) of deaths in the IVC filter group and 61.4% (296 of 482) of deaths in the control group (Table 3). There were no deaths during the index hospitalization associated with a VTE. After discharge, the most frequent cause of death in patients with an IVC filter was circulatory or cardiac disorder. In those without an IVC filter, mental disease or central nervous system disorder was the most common cause of death, followed by trauma and circulatory or cardiac disorder. Death certificate information from the NDI did not indicate PE or DVT as a cause of death for any patient, but it is not known if an underlying thrombosis may have contributed to death in some patients. While deaths due to trauma were statistically more frequent in patients without IVC filters during hospitalization, there was no statistically significant difference in cause of death between the 2 patient groups after hospital discharge (Table 3).
In this retrospective cohort study, 451 trauma patients with IVC filters placed after their initial injury were compared with 1343 matched trauma patients without IVC filters. There was higher initial mortality in patients who did not receive an IVC filter, which can be attributed to death secondary to early trauma complications or high injury burden. Most of these deaths occurred in the first 72 hours, before the median time of IVC filter placement, which was an expected phenomenon: patients with early trauma-related mortality did not survive long enough to undergo IVC filter placement, resulting in the appearance of improved early survival in the IVC filter group.
Once patients survived the initial injury burden (24 hours), there was no significant difference in mortality between those who received an IVC filter and those who did not. Likewise, mortality did not differ in patients who survived until the time of IVC filter placement or hospital discharge, with 3.8 years of mean follow-up. As seen in these data, there is no demonstrable benefit or decrement in long-term mortality for trauma patients with IVC filter insertion, independent of the presence or absence of DVT or PE. Most of these filters were placed in the absence of a VTE, and 91.6% (413 of 451) were not removed, indicating that long-term IVC filter placement did not increase survival through prevention of future PE or decrease survival because of IVC filter complications (eg, embolism, ruptured viscera, and filter fracture).
This information, coupled with a lack of good-quality data regarding improvement in short-term mortality, should cause clinicians to consider the significant risks and expenses17 associated with the insertion of IVC filters. There are known risks of IVC filter placement,18 but morbidity associated with IVC filters that remain in place is a significant concern. The literature has described morbidity associated with filters that are not retrieved, including IVC filter fracture, filter thrombosis, filter protrusion outside the IVC, and lower-extremity venous thrombosis.19-23 However, the present study captured mortality only, and any nonlethal morbidity attributed to filter insertion would not be evident in these data.
Risk of morbidity should be considered before placing an IVC filter, especially in younger trauma patients because complications are known to increase with time. Retrievable filters that are not removed carry a higher risk profile than the previously used permanent filters,18 and many retrievable IVC filters are not removed when the initial indication for placement resolves.17 The IVC filter retrieval rates are as low as 1.2% in some populations,24 so the likelihood of filter removal should be considered before placement. At specific institutions with focused efforts to improve IVC filter retrieval rates,25-28 significant improvements have been made to reduce morbidity associated with retained IVC filters.
Patient-specific indications for filter placement not captured in our study may influence the interpretation of these data. In addition, the ISS was used for matching: although this scoring system is routinely used to classify severity of trauma, it is not an exact indicator of prognosis, and residual confounding may exist. Severity of trauma may also be affected by certain anatomical subsets of the ISS, such as head and neck injury, or by the mechanism of injury. Patients with IVC filters and matched controls in this study had a significantly different head and neck AIS and extremity AIS, although the mortality risk was not significantly affected based on multivariable analysis. In addition, the mortality data in this study were obtained from the NDI and therefore rely on the accurate reporting of a patient’s death. Mortality data were available only through the end of 2012, and it is possible that more deaths have occurred since that time. In addition, the rate of IVC filter complications is known to increase with time, and this study did not capture morbidity associated with IVC filter insertion or survival differences that may arise with longer follow-up.
Overall, these data indicate that IVC filters should not be placed in trauma patients in an effort to decrease all-cause mortality. Among patients who survived until the time of IVC filter placement, no survival benefit from the filter was observed. There was also no improvement or decrement in long-term mortality with placement of an IVC filter in trauma patients, whether in the presence or absence of VTE. Given the expected morbidity of long-term IVC filter use, filters should be removed as soon as a patient’s contraindication to anticoagulation resolves.
Corresponding Author: Shayna Sarosiek, MD, Department of Hematology, Boston University School of Medicine, 820 Harrison Ave, Boston, MA 02118 (email@example.com).
Accepted for Publication: June 1, 2016.
Published Online: September 28, 2016. doi:10.1001/jamasurg.2016.3091
Author Contributions: Drs Sarosiek and Sloan had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Sarosiek, Burke, Sloan.
Acquisition, analysis, or interpretation of data: Sarosiek, Rybin, Weinberg, Burke, Kasotakis.
Drafting of the manuscript: All authors.
Critical revision of the manuscript for important intellectual content: Sarosiek, Weinberg, Burke, Kasotakis.
Statistical analysis: Rybin, Weinberg.
Administrative, technical, or material support: Kasotakis.
Study supervision: Sarosiek, Kasotakis, Sloan.
Conflict of Interest Disclosures: None reported.
Create a personal account or sign in to: