Cumulative incidence of recurrent deep vein thrombosis. Kaplan-Meier analysis censoring for event, death, or date of last medical record documentation. IVC indicates inferior vena cava.
Cumulative incidence rate of subsequent/recurrent pulmonary embolism. Kaplan-Meier analysis censoring for event, death, or date of last medical record documentation. IVC indicates inferior vena cava.
Cumulate incidence rate of major bleeding. Kaplan-Meier analysis censoring for event, death, or date of last medical record documentation. IVC indicates inferior vena cava.
Cumulative incidence rate of all-cause mortality. Kaplan-Meier analysis censoring for death or date of last medical record documentation. IVC indicates inferior vena cava.
Spencer FA, Bates SM, Goldberg RJ, Lessard D, Emery C, Glushchenko A, Gore JM, White RH. A Population-Based Study of Inferior Vena Cava Filters in Patients With Acute Venous Thromboembolism. Arch Intern Med. 2010;170(16):1456–1462. doi:10.1001/archinternmed.2010.272
Our study objective was to describe the frequency, indications, and outcomes after inferior vena cava (IVC) filter placement in a population-based sample of residents of the Worcester, Massachusetts, metropolitan area who had been diagnosed as having acute venous thromboembolism (VTE) in 1999, 2001, and 2003.
A retrospective chart review of inpatient and outpatient medical records was conducted. Recorded indication(s) for IVC filter placement was determined among a subset of cases from 3 Worcester tertiary care hospitals. Three thrombosis specialists assessed the appropriateness of IVC filter placement.
Of 1547 greater Worcester residents with validated acute VTE and without a prior IVC filter, 203 (13.1%) had an IVC filter placed after acute VTE. Patients with an IVC filter were older, had more comorbidities, and had a higher mortality rate during 3 years of follow-up. There was unanimous agreement by panel members that the use of an IVC filter was appropriate in 51% of cases and inappropriate in 26% of cases, with no consensus in the remaining 23%.
In this community-based study, IVC filters were frequently used in the treatment of patients with acute VTE. Placement was deemed to be appropriate in approximately 50% of the patients but was not appropriate or debatable in the remaining cases. Given the increasing use of IVC filters, prospective studies are clearly needed to better define the indications for, and efficacy of, IVC filter placement.
Historically, insertion of an inferior vena cava (IVC) filter in patients with acute venous thromboembolism (VTE) is the last therapeutic option to prevent pulmonary embolism (PE). The 2001 American College of Chest Physician (ACCP) guidelines recommended placement of an IVC filter only if there was either a contraindication to or a complication of anticoagulant therapy or if recurrent thromboembolism occurred despite adequate anticoagulant therapy.1 Insertion of a filter has never been recommended as a primary treatment of VTE.
There are risks associated with the use of IVC filters, including bleeding, incorrect positioning or dislodgment, local thrombosis, and a 2-fold increase in the risk of recurrent lower-extremity deep vein thrombosis (DVT).2The most recent 2008 ACCP guidelines for the treatment of VTE have narrowed their recommendations for filter placement even further, advocating that an IVC filter be considered only in patients with acute proximal lower-extremity DVT in whom “anticoagulant therapy is not possible, because of the risk of bleeding.”3
Unfortunately, there are very few population-based data describing how frequently IVC filters are inserted in patients with acute VTE and how often these patients have a documented failure of, or contraindication to, anticoagulant therapy. The objectives of the present study were to determine the frequency of IVC filter use in the setting of acute VTE, to ascertain the indication(s) for filter placement, and to determine whether patients treated with an IVC filter met existing guidelines for filter placement.
The Worcester VTE study is a retrospective population-based surveillance study of VTE in Worcester in 1999, 2001, and 2003.4,5 The study years were chosen based on funding cycles, resource considerations, the desire to track events over time, and correlation with publication of updated ACCP guidelines for the management of VTE. Computerized printouts of all Worcester residents with health care encounters during calendar years 1999, 2001, and 2003 who were coded as having any of 34 International Classification of Diseases, Ninth Revision, diagnosis codes possibly consistent with VTE were obtained from each of the 12 hospitals serving the Worcester metropolitan area.4,5 These data queries not only were limited to discharge diagnoses but also encompassed all outpatient activities.
The medical records of all patients who met the geographic inclusion criteria were reviewed and validated by trained abstractors. Each case of VTE was classified as being definite, probable, possible, or not acute or negative based on a modification of the classification schema used by Silverstein et al6 (eAppendix). If the classification of VTE was not immediately clear using the specified criteria, the principal investigator (F.A.S.) reviewed the medical record. For purposes of this study, patients in whom an IVC filter was placed before the index VTE event were excluded from further analysis.
The paper and electronic medical records of each patient's index, as well as previous hospitalizations and/or outpatient visits at participating Worcester hospitals, were reviewed to determine whether the index VTE event represented an incident (initial) or a recurrent case and to obtain additional information about patient comorbidities and risk factors for VTE. Follow-up hospital record reviews to ascertain the specified outcomes of recurrent VTE and bleeding complications were conducted at a minimum of 1 year and a maximum of 3 years after the index VTE event. State and national mortality records were reviewed on an annual basis (last reviewed in 2007) for the purposes of assessing each patient's vital status; follow-up of vital status was obtained in more than 99% of all patients.
Information abstracted from medical records included demographic and clinical characteristics, results of all tests for VTE, and hospital treatment and management, including the date of IVC filter insertion. Medical history variables defined as recent were those occurring or active in the 3 months before the index diagnosis of VTE. Medical records for each individual at the other Worcester area hospitals were also screened in case the patients sought treatment for their initial VTE event or subsequent complications at more than 1 area hospital.
Major bleeding was defined as any episode of bleeding that required transfusion or hospitalization or was life-threatening (resulted in myocardial infarction, stroke, or death). The nurse abstractors were instructed to include only those cases of bleeding that met 1 or more of these criteria as outcome events. For the 1999, 2001, and 2003 cohorts, major bleeding events were not independently reviewed by the lead investigator.
Potential recurrent VTE events were classified using criteria identical to those used for incident cases, with the exception that a definite recurrence of DVT or PE required the documentation of thrombosis in a previously uninvolved venous or pulmonary arterial segment, respectively, by diagnostic imaging (compression ultrasonography, ventilation perfusion imaging, or computed tomographic scan). The lead investigator reviewed the medical record and diagnostic imaging reports in each case of potential VTE recurrence; only definite or probable recurrences were included in this study.
In the subset of patients (n = 165) seeking care at 1 of the 3 participating major tertiary care centers in Worcester, indication for IVC filter insertion was ascertained by an additional careful retrospective review of the medical record. Admission, progress, and discharge notes written by the primary care physician or by any consultant involved in the placement of an IVC filter were reviewed to identify the documented indication for IVC filter use or the attending physician's rationale for insertion of the IVC filter.
Using these abstracted data, the lead investigator classified the documented rationale for IVC filter placement into the following 9 groups: (1) perception of increased bleeding risk with anticoagulant therapy; (2) bleeding within 30 days before VTE; (3) major bleeding after initiation of anticoagulant therapy; (4) minor bleeding after initiation of anticoagulant therapy; (5) suspected PE occurring after initiation of anticoagulant therapy; (6) DVT occurrence or extension despite anticoagulant therapy; (7) perceived high risk for subsequent PE despite anticoagulant therapy; (8) planned interruption of anticoagulant therapy for future operations or procedures; and (9) other. These categories were created to simplify analysis and reporting of our observations but were not strictly defined or necessarily exclusive.
Three experts in the field of thromboembolism (F.A.S., S.M.B., and R.H.W.) independently reviewed the data pertaining to rationale or indication for filter placement for each of the patients in whom an IVC filter was placed; each of these cases was categorized as meeting or not meeting 2001 ACCP guidelines. The criteria for appropriate filter placement were (1) documented recurrent VTE during adequate anticoagulant therapy; (2) presence of a contradiction or complication of anticoagulation in a patient with, or at high risk of, a proximal lower-extremity DVT; (3) chronic recurrent embolism with pulmonary hypertension; and (4) concurrent performance of surgical pulmonary embolectomy or pulmonary thromboendarterectomy. Agreement was measured as the proportion of cases with complete agreement for placement of an IVC filter, the proportion with complete agreement for not placing a filter, the proportion with 2 raters favoring the use of a filter, and the proportion with 1 rater favoring the use of a filter. The results were analyzed using the Fleiss κ nominal scale agreement among many raters.7
The prevalence of prespecified baseline comorbidities in patients with VTE in whom a filter was placed was compared with that in patients in whom this device was not placed using χ2 tests of statistical significance for categorical variables and t tests for continuous variables. The incidence rates of our principal study outcomes among patients who received and did not receive an IVC filter were compared using Kaplan-Meier survival curves. The patients were censored at time of event, death, or date of last medical chart documentation, whichever came first.
The study sample consisted of 1547 male and female residents of the Worcester metropolitan area with confirmed acute DVT (66%), PE (19%), or both (15%). Of these, 203 patients (13.1%) had an IVC filter placed during the index hospital stay after diagnosis of acute VTE. The median duration of follow-up was 926 days. During 1999, 2001, and 2003, IVC filters were inserted in 63 of 472 patients (13%), 75 of 547 patients (14%), and 65 of 528 patients (12%), respectively.
Patients who received an IVC filter were more likely to be older, male, diagnosed as having VTE after hospital admission for a different principal medical condition, have a medical or surgical hospitalization within the prior 3 months, require care in an intensive care unit before VTE diagnosis, or have other illnesses or comorbidities (Table 1). Patients who received an IVC filter were also more likely to have a platelet count lower than 100 × 103/μL (to convert to × 109/L, multiply by 1.0) at the time of the VTE diagnosis, to have been diagnosed as having a recurrent VTE, to have had a recent episode of bleeding, and to have been diagnosed as having DVT together with PE (rather than isolated DVT or PE). Seven patients who underwent IVC filter placement had an isolated calf DVT.
Of the patients who were discharged from the hospital, those with an IVC filter were much less likely to be discharged on a regimen of anticoagulant therapy (low-molecular-weight heparin sodium or warfarin sodium) (Table 1). In-hospital mortality was significantly higher among patients who received an IVC filter compared with those who did not receive a filter (9.9% vs 5.1%; P = .01). Cumulative incidence rates of recurrent DVT, new PE, major bleeding, and death are shown in Figures 1, 2, 3, and 4. The 90-day incidence of objectively confirmed PE was similar among the patients treated (1.7%) and not treated (1.4%) with an IVC filter. However, after 90 days, no further PE events were diagnosed in the patients with an IVC filter, whereas the cumulative incidence of PE had increased to 5.3% among the patients without a filter (P = .18). Three years after the index event, the incidence of recurrent DVT was 21.0% among the filter-treated patients and 14.9% among those without a filter (P = .009). The incidence of major bleeding (including bleeding only after initiation of anticoagulant therapy in the nonfilter-treated group and after placement of the IVC filter in the filter-treated group) was not significantly (P = .17) different between the 2 groups. All-cause mortality was significantly higher (P < .001) in patients in whom an IVC filter was placed compared with patients who did not receive an IVC filter.
The indications for filter placement were determined in 160 of 165 patients who had an IVC filter placed in 1 of 3 area hospitals. In 5 patients, some or all of the chart information required was not available. Based on medical record review, 57 patients were deemed to be at high risk for bleeding (but had no recent bleeding), 39 patients had a history of recent bleeding, 26 patients had major bleeding after starting anticoagulant therapy, and 9 patients had minor bleeding after starting anticoagulant therapy. Seven patients had a presumed PE after starting anticoagulant treatment (1 to 44 days) after the index VTE event. In 1 patient, warfarin therapy had been stopped for unspecified reasons for 6 days before a PE that occurred on day 44. In 2 other patients, the initial partial thromboplastin time was less than 55 seconds and PE developed on days 1 and 2, respectively. An additional 7 patients received an IVC filter after the index VTE because it represented “failure” of anticoagulant treatment (for prior VTE [n = 5], atrial fibrillation [n = 1], or stroke [n = 1]). Two of the 7 patients had a subtherapeutic international normalized ratio at the time of admission. Eight patients who were treated with full-dose anticoagulant therapy also received a filter to prevent PE; 3 patients had a filter placed because of a planned procedure; and 4 patients had another “indication.”
Overall, the 3 reviewers agreed that there was an indication for IVC filter placement in 81 of 160 patients (51%), but they also agreed unanimously that an IVC filter was not indicated in 41 patients (26%) (overall agreement, 122 of 160 patients, or 77%). In 20 cases, 2 of the reviewers thought that an IVC filter was indicated, but 1 reviewer disagreed. The Fleiss κ score was consistent, with substantial agreement between multiple raters (κ = 0.65).
Table 2 shows the agreement and disagreement between the reviewers about whether the indication for IVC filter placement was consistent with the ACCP 2001 guidelines. All 3 reviewers agreed that filter placement was appropriate in 34 of the 39 patients (87%) with a recent history of bleeding. Similarly, there was agreement that use of a filter was indicated in 23 of 26 patients (88%) who had major bleeding after starting anticoagulant therapy. There was also unanimous agreement among the reviewers that an IVC filter was indicated in 21 of the 57 patients (37%) who were simply judged to be at increased risk for bleeding and that an IVC filter was not indicated in 21 of the 57 patients (37%). Regarding the appropriateness of IVC filter placement in the remaining 15 patients (26%), there was disagreement among the reviewers.
Regarding the 14 patients who had recurrent DVT or new PE, there was agreement among the reviewers that a filter was not indicated in 6 patients, whereas there was no agreement as to filter placement in the remaining 8 patients.
There was universal agreement that filter placement was not indicated in the 8 patients in whom the rationale for filter placement was to prevent PE (in addition to anticoagulant therapy).
Our study suggests that approximately 1 in every 8 patients with a documented episode of VTE will receive an IVC filter in their early treatment. As far as we are aware, no other population-based study has reported on the frequency of IVC filter use as an acute management strategy for VTE. However, a number of studies have suggested that the utilization rates of IVC filters overall (for VTE prophylaxis and/or treatment) are growing rapidly. In a retrospective study using the National Hospital Discharge Survey database, the overall use of IVC filters was estimated to have increased from approximately 2000 in 1979 to more than 49 000 in 1999.8 In a more recent single-center study, placement of IVC filters in patients with VTE approximately doubled between 2004 and 2007.9
While the rate of IVC filter use as an acute management strategy for VTE in our study was higher than we had anticipated, our data suggest that in most patients this treatment strategy was appropriate. Patients in whom IVC filters were placed were significantly older, were much more likely to have had recent bleeding, and had an increased prevalence of a number of comorbidities (eg, recent surgery or intensive care unit stay, active malignancy, prior cerebrovascular accident) that may have predisposed them to an increased risk for serious bleeding during anticoagulant therapy.
Given the complexity of these patients, and to gain a better understanding of specific indications for IVC filter placement, we performed an additional chart review in a subset of patients. Approximately 40% of patients had bleeding after their index episode of VTE or had bleeding in the prior 30 days. Our review panel agreed unanimously with IVC filter placement in most of these cases, as such placement was clearly consistent with prior ACCP guidelines.1
The ACCP guidelines also indicate that IVC filter placement is appropriate in patients deemed to be at high risk for bleeding or who have a contraindication to anticoagulant therapy but do not define what constitutes high risk. Indeed, being at increased risk for anticoagulant-related bleeding was the listed indication in 35% of our reviewed cases. Interestingly, all of the experts agreed with filter placement in only 1 of 3 of these cases; they unanimously disagreed with filter placement in an additional 1 of the 3 cases; and they failed to reach a consensus in the remaining cases. This finding highlights the fact that guidelines that use implicit phrases such as “at risk for bleeding” or “have a contraindication to anticoagulants” are subjective and that interpretation of these terms may vary widely. Although the reviewers were cognizant of these limitations, in one-third of the cases they unanimously agreed that a trial of anticoagulant therapy was justified before resorting to placement of an IVC filter. Included in this group were patients who were deemed to be at high risk for falling, patients with a remote history of bleeding, and patients who underwent a major operation more than 2 weeks before being diagnosed as having VTE.
Admittedly, predicting which patients will have major bleeding during anticoagulant therapy after VTE remains difficult. Decision tools to better define the risk of anticoagulant-associated bleeding have been developed.10- 13 To find out whether such models can be used to help clinicians decide on whether anticoagulant therapy is contraindicated (and an IVC filter is appropriate) will require further prospective studies in various at-risk groups.
Prevention of PE (in addition to anticoagulant therapy) or “anticoagulation failure” was the indication for IVC filter placement in 22 patients (14%). The expert reviewers unanimously agreed that an IVC filter was not indicated in 14 of these 22 patients (64%). Given the efficacy of available anticoagulation strategies, all reviewers thought that placement of an IVC filter to “protect” a patient from PE (in addition to anticoagulant therapy) was not an appropriate indication for use. Among the patients who received a filter because of “failure of anticoagulant treatment,” the reviewers thought that this rationale was not supported in 6 cases because anticoagulation was subtherapeutic at the time of the recurrent event (n = 5) or because the recurrent event was considered to be presumptive (n = 1) .
Despite the relatively frequent use of IVC filters, clinical data about their efficacy and safety in patients with acute VTE are sorely lacking. The relative lack of such data has led to somewhat vague guidelines regarding indications for filter use and the potential for overuse. As far as we are aware, no clinical trials have evaluated the efficacy of IVC filters alone as an acute VTE treatment strategy. Use of an IVC filter as an adjunct to standard anticoagulation was compared with anticoagulation alone in the PREPIC (Prévention du Risque d’Embolie Pulmonaire par Interruption Cave) study, a randomized clinical trial of 372 patients with documented lower-extremity DVT.2 At the 2-year follow-up visit, use of an IVC filter was associated with an increased risk of recurrent lower-extremity DVT without any effect on survival.
In a large population-based case-control study, White et al14 compared outcomes after acute VTE in 3632 patients who had a filter implanted and 64 333 patients who did not receive an IVC filter. Similar to the results of the current study, the patients who received an IVC filter were older and had more comorbidities. Nevertheless, the incidence rates of recurrent PE at 1 year did not differ significantly between the patients who received an IVC filter and those who did not. In the current study, however, we observed a trend toward a higher incidence of recurrent PE in the patients who were not treated with a filter over longer-term follow-up. The IVC filter–treated patients had a lower cumulative incidence of PE at 3 years (1.7%) than patients who did not receive a filter (5.3%). A similar finding was also reported by the PREPIC study group after 8 years of follow-up.15 Any firm conclusions about the efficacy and utility of IVC filters alone as a therapy in patients who develop DVT will require a randomized trial of patients who meet agreed-on criteria for bleeding risk associated with the use of anticoagulant therapy.
Like any observational study, the present investigation has several limitations. Although we conducted a broad screening for all possible cases of VTE in the greater Worcester population, we cannot claim complete case ascertainment of index VTE events, episodes of VTE recurrence, or episodes of major bleeding. As in any retrospective study based on medical record review, the quality of data abstracted with respect to other medical conditions is limited by the quality of the medical documentation itself. Another limitation of this study is that we did not analyze any patients diagnosed as having acute VTE after 2003. During our study period, retrievable filters were not commonly used in the greater Worcester community; we suspect but cannot yet verify that the introduction of the retrievable IVC filter has resulted in even more liberal use of this treatment modality. Interestingly, in several studies, rates of removal of temporary IVC filters have been quite low (5%-20%), suggesting that most of these devices are in fact permanent.9,16
Finally, it should be noted that our results may not be generalizable to other communities because it is likely that IVC filter use rates in a given community are likely affected by local physician practice. Such local variation suggests the need for health care systems to evaluate practice within their own communities.
In conclusion, the results of this observational community-based study document that an IVC filter is frequently inserted as part of an acute management strategy for acute VTE. In approximately 50% of all cases, placement of a filter appears to be appropriate and consistent with contemporary guidelines, whereas in approximately 25% of all cases, the use of a filter was deemed to be inappropriate. Guidelines for IVC filter use could be improved by developing explicit criteria for contraindications to anticoagulant use and identification of patients who might benefit from the receipt of IVC filters.
Spencer et al have highlighted an important area of “Less Is More”: the inappropriate use of IVC filters. In a retrospective analysis of a large cohort, only half of all IVC filters placed for prevention of acute VTE were found to be placed for appropriate indications by professional society guidelines. In a recent online article in the Archives, Nicholson et al1 found a high rate of fracture with serious adverse consequences associated with vena caval filters. The combination of invasive devices being placed in persons with no expectation of benefit and having clear risks leads to this “Less Is More” classification.
1. Nicholson W, Nicholson WJ, Tolerico P, et al. Prevalence of fracture and fragment embolization of bard retrievable vena cava filters and clinical implications including cardiac perforation and tamponade [published online August 9, 2010]. Arch Intern Med. doi:10.1001/archinternmed.2010.316.
Correspondence: Frederick A. Spencer, MD, Department of Medicine, Divisions of Cardiology and Hematology/Thrombosis, Faculty of Health Sciences, McMaster University, 1200 Main St W, Hamilton, ON L8N 3Z5, Canada (email@example.com).
Accepted for Publication: February 8, 2010.
Author Contributions: Dr Spencer had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Spencer, Goldberg, and Gore. Acquisition of data: Spencer, Bates, and Emery. Analysis and interpretation of data: Spencer, Bates, Goldberg, Lessard, Glushchenko, Gore, and White. Drafting of the manuscript: Spencer, Goldberg, Emery, and Glushchenko. Critical revision of the manuscript for important intellectual content: Spencer, Bates, Goldberg, Lessard, Gore, and White. Statistical analysis: Goldberg, Lessard, Glushchenko, and White. Obtained funding: Spencer. Administrative, technical, and material support: Spencer, Bates, and Emery. Study supervision: Emery and White.
Financial Disclosure: None reported.
Funding/Support: This study was supported by a grant from the National Heart, Lung, and Blood Institute (R01-HL70283). Dr Spencer also has a Career Investigator Award from the Heart and Stroke Foundation of Canada.
Role of the Sponsors: The sponsors had no role in the design and conduct of the study; in the collection, management, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.