Inferior vena cava filter use and patient safety: legacy or science?

The vast majority of patients with venous thromboembolism (VTE) are treated with anticoagulants, which effectively stop the abnormal clotting process and thereby prevent progression of deep vein thrombosis (DVT) as well as pulmonary embolism (PE) and recurrent VTE.²  In addition, percutaneous insertion of a vena cava filter (VCF) has been used for more than 40 years in selected patients to trap large or moderate-sized emboli arising in the leg veins before they can travel to the pulmonary arteries.³ 

In this focused review, we address current knowledge related to the use of VCFs, evidence of their effectiveness, indications for filter insertion, and complications. We also provide an approach to the safe use of VCFs.

Currently, 3 types of VCFs are available. (1) Permanent VCFs were first developed in the late 1960s and were designed to be left in place permanently. For more than 20 years, permanent Greenfield filters were the most widely used VCF.⁴  (2) Retrievable VCFs (also called optional VCFs), first approved by the US Food and Drug Administration (FDA) in 2003, were developed with the objective of preventing PE in the short-term and with the ability to be removed once the patient’s high risk for PE had resolved or to be left in place permanently.^5,6  Filter retrieval was designed to reduce long-term complications associated with permanent filters. Over the past decade or so, most new VCF insertions have used retrievable filters, although most of these have been left in place indefinitely.⁷  (3) Tethered VCFs are attached to a central venous catheter. The catheter and filter are inserted and removed together, usually after a brief dwell period. The first of these catheter-filter combinations was recently approved but, to date, has not been widely used.⁸ 

In 2017, 11 different VCFs are approved for use in the United States (Table 1). Although VCFs differ with respect to their design, composition, size of the introducer, and supporting clinical evidence, the selection of a specific filter is primarily dependent on the duration of expected need, ease of insertion (and removal), experience of the inserter, local availability, and device-specific outcome data.

The use of VCFs has increased dramatically over the last 20 years.^9-11  Approximately 2000 filters were placed in the United States in 1997; by 2007, this number was ∼167 000, while in 2012, >250 000 VCFs were inserted.⁹  Possible reasons for the increased use of VCFs include expanded indications (especially for prophylactic indications), availability of retrievable filters, ease of insertion, belief that current VCFs are safer than older filters, greater number of VCF inserters, financial incentives for insertion, and practice influenced by medical legal concerns. At the same time, the FDA approval process for VCFs has been simplified, thereby eliminating the need for rigorous scientific studies of efficacy and safety for any of the currently available filters.

The use of VCFs varies widely from country to country, from region to region within the same country, and from center to center; this variation cannot be accounted for by differences in patient characteristics.^11-16  In 2012, there were 25 times more filters placed in the United States than in the combined 5 largest European countries with a similar total population.¹⁷  Among 33 American states, the annual rates of VCF use varies fivefold (13 to 68 per 100 000 residents) after controlling for demographic and clinical factors.¹³  Among 263 California hospitals, the use of VCFs in 130 643 patients with acute VTE varied from 0% to 39% between hospitals, even after adjusting for important clinical parameters.¹⁴  Similarly, in 14 000 cancer patients with acute VTE, VCFs were placed in 20% overall, with rates of insertion varying from 0% to 52% of patients among the 223 hospitals.¹⁵  Higher VCF use was seen in states with more litigious medical legal environments and in patients who had commercial or government health insurance.^13,18 

Despite several decades of VCF use, there remains a remarkable paucity of high-quality clinical trials assessing their effectiveness and safety. In fact, none of the indications for VCF insertion are based on high-quality evidence.^3,19,20  In a study of all 556 658 US Medicare patients hospitalized with PE from 1999 to 2010, VCFs were placed in 17% during their PE admission.¹¹  All-cause mortality at 30 days in the patients with and without a filter was 11.7% and 9.2%, respectively. Among 85 159 California patients with acute VTE, VCFs were placed in 11%.¹⁶  In the 80 697 patients with no contraindication to anticoagulation and the 1445 patients who underwent major surgery, VCF use did not reduce mortality at 30 days. Only in the 3017 patients with active bleeding was the use of filters associated with a reduced 30-day risk of death. In none of the 3 groups did VCF use reduce subsequent PE; however, subsequent DVT rates increased 135% among filter patients who had active bleeding (presumably related to decreased or delayed use of anticoagulation). Similarly, among the 14 000 cancer patients with VTE who had a VCF inserted, there was no reduction in 30-day mortality or PE, but there was a 60% increased risk of DVT compared with cancer patients who did not have a filter placed.²¹ 

There are only 2 high-quality randomized trials of VCF use that have evaluated the consequences of filter placement.^22,23  The PREPIC study randomly assigned 400 patients with proximal DVT and increased risk of PE to therapeutic anticoagulation alone or anticoagulation plus 1 of 4 permanent inferior vena cava (IVC) filters.²²  At the 12-day follow-up, the rate of all PE (both symptomatic and those found on screening lung scans) was significantly lower in the IVC filter group (1.1% vs 4.8%; odds ratio, 0.22; P = .03). However, symptomatic PE and total mortality were not significantly different between groups at 12 days or 2 years. Furthermore, the incidence of recurrent DVT was greater in IVC filter patients at 2 years (21% vs 12%; odds ratio, 1.87; P = .02). An 8-year follow-up of patients in this trial also showed no survival advantage in patients with an IVC filter.²⁴  In summary, this study demonstrated an initial small reduction in symptomatic plus asymptomatic PE with a later increased risk of recurrent DVT and no difference in mortality at any time point.

PREPIC-2 was a subsequent randomized trial conducted in 17 French centers in 399 patients who had acute PE with leg vein thrombosis.²³  The investigators selected patients who were thought to be at risk for recurrent PE, and they used a retrievable filter to reduce long-term complications associated with permanent filters. Patients received standard anticoagulation alone or anticoagulation plus a retrievable VCF with the intent of removing the filter before 3 months. Successful retrieval was achieved in 93% of the patients in whom VCF removal was thought to be indicated. At 3 months, recurrent PE was seen in 3 out of 199 patients (1.5%, 2 fatal) in the control group and in 6 of 200 patients (3.0%, all fatal) who had a filter. All-cause mortality was not significantly different between groups. In the filter patients, access site hematoma, filter thrombosis, unsuccessful attempted retrieval, and nonretrieval rates were seen in 2.6%, 1.6%, 6.7%, and 21% of patients, respectively. These 2 randomized trials do not demonstrate a reduction in symptomatic PE or death with VCF use and, therefore, do not support the use of VCFs in patients with VTE who can be anticoagulated.

No randomized trials have compared permanent versus retrievable filters, and there are no studies comparing routine removal with leaving retrievable filters in place permanently.

The sole purpose of a VCF is the prevention of life-threatening and fatal PE from a lower extremity or pelvic vein DVT. Although VCFs have been inserted for many indications, in general, filters are placed in patients for 1 of 3 reasons: (1) DVT or PE and an absolute or relative contraindication to anticoagulation, (2) DVT or PE and assumed risk of major or fatal PE despite the use of anticoagulants, or (3) no VTE but the patient is felt to be at high risk of developing clinically significant PE.

Indications 1 and 2 are considered therapeutic, while indication 3 is considered prophylactic.^25,26  Although we believe the single appropriate indication or VCF use is the presence of a recent proximal DVT and a contraindication to therapeutic anticoagulation, there are no randomized trials of VCF use for this indication. Filters are also placed in patients who are thought to have an increased risk of PE despite being anticoagulated and in selected patient groups without VTE as primary thromboprophylaxis; this indication now accounts for the majority of filters inserted, at least in the United States.^7,10,25-27 Table 2 lists some of the common reasons for VCF insertion, along with our commentary as to why we do not agree with these indications.^3,28 

The American College of Chest Physicians (ACCP) guidelines recommend against use of IVC filters in patients with acute DVT or PE who are treated with anticoagulants, and they recommend against the use of IVC filters as thromboprophylaxis in trauma patients or those with spinal cord injuries.^2,35  However, physician compliance with guideline-recommended indications for VCFs is highly variable. In a population-based study from Worcester, Massachusetts, among patients who had a VCF placed in the setting of acute VTE, independent review of the indications revealed that only half of the filters inserted were consistent with rather liberal guideline recommendations; the indications in the other half were considered to be inappropriate or debatable.³⁶  Out of 952 filters inserted at another center from 2003 to 2011, only 14% were placed for indications for which there was consensus.²⁷ 

In some centers, the most frequent (but “off label” according to the FDA) reason for VCF insertion is in trauma patients who do not have VTE but who are thought to be at increased risk for PE and who are unable to receive appropriate anticoagulant thromboprophylaxis because of risk of bleeding or who are thought to be at increased risk of PE despite the use of usual thromboprophylaxis. However, a prospective study from a large trauma center showed no reduction in PE among patients who had a prophylactic VCF placed.³⁷  Furthermore, a study of 39 456 major trauma patients showed that patients who had a VCF did not have reduced mortality but were more likely to develop DVT.³²  Similarly, prophylactic VCFs have not been shown to reduce PE when placed in patients undergoing bariatric surgery.³³  A systematic review of prophylactic VCFs found no class I studies to support insertion of a filter in patients without VTE and concluded that “widespread use of prophylactic VCFs is not supported by evidence and should be discouraged.”³⁴ 

Complications associated with VCFs are common, although life-threatening complications are not, at least in the short-term.^7,38-42  Factors associated with increased rates of complications include retrievable vs permanent filters, certain specific filters, operator inexperience, and increased filter dwell time. Complications may occur at the time of insertion, in the early days to weeks after placement, at the time of a retrieval attempt, and later on related to filters that are not removed.

Published rates of VCF complications vary considerably and depend on the specific filters involved, the duration of follow-up, definitions of complications used in the reports, and whether routine imaging was performed. Society of Interventional Radiology (SIR) Standards of Practice guidelines reported the following complication rates associated with VCF use: insertion problems (5% to 23%), access site thrombosis (3% to 10%), IVC occlusion (2% to 30%), IVC penetration (0% to 41%), and filter fracture (2% to 10%).³⁸  Less common complications included deployment outside of the target area (1% to 9%), filter embolization (0.1%), recurrent PE (0.5% to 6%), and death (0.12%).

Among 688 retrievable filters inserted at Massachusetts General Hospital, 4% had insertion-related complications (malposition and marked angulation), 19% had complications during the indwell phase (DVT, 11%; PE, 4%; and IVC thrombosis, 4%), 3% had a major complication during retrieval attempts, and 67% of these retrievable filters were not removed.⁴³  Among patients who had an abdominal contrast computed tomography (CT) scan at least 4 years after filter placement, the overall complication rates were as follows: caval performation in 48%, filter fracture in 14%, and IVC thrombosis in 13%.⁴¹  A systematic review of 37 studies reporting outcomes with 6834 retrievable filters found DVT in 5.4% of patients, filter migration to the lungs or heart in 1.3%, and IVC thrombosis or stenosis in 2.8%.⁷ 

Retrievable filters (particularly those that are not removed) are associated with a greater complication rate than permanent filters.^39-41  A 6-year study reported outcomes in patients who had 785 permanent and 449 retrievable filters.⁴⁰  Device-related complications were much more common with retrievable filters than with permanent filters (7.5% vs 0.9%; P < .0001). Another study reported that 87% of 1606 IVC filter–related complications submitted to the voluntary FDA MAUDE database from 2009 to 2012 were related to retrievable filters, while only 13% were related to permanent filters.³⁹  For each type of adverse event, the frequency was significantly higher with retrievable filters. Complication rates associated with retrievable filters also increase with time after filter insertion.^44,45 

While the majority of IVC filters penetrate at least 3 mm outside the wall of the IVC, most of these penetrations are asymptomatic.^7,41,46-49  However, filter struts can penetrate adjacent structures including the aorta, duodenum, and vertebral body. Rates of penetration are more common with retrievable filters than with permanent filters and tend to be progressive over time, which increases penetration into adjacent organs and makes retrieval more challenging and less likely to be successful.^46,47,50  When the results of abdominal CT scans were analyzed in 465 patients with a VCF, the rates of IVC perforation were 47% in those with a retrievable filter and only 2% of those with a Greenfield filter; perforation into an adjacent organ was seen in 34% and 0%, respectively.⁴⁷  In another study, among 262 patients who had a post–filter insertion abdominal CT scan, penetration of the filter outside the cava was documented in 46%, with 14% of cases having at least 1 strut penetrating a retroperitoneal organ.⁴⁸  Penetration was seen in 5% of permanent filters and 49% of retrievable filters (P < .0001), and increased with time after insertion (18% <30 days vs 57% ≥30 days; P < .0001). A systematic review of 9002 patients from 88 IVC filter studies reported caval penetration in 19% of patients, with 5% experiencing a major complication.⁴⁹ 

Potentially fatal migration or embolization of a filter to the right atrium, tricuspid valve, right ventricle, or pulmonary arteries has been described multiple times.^38,51  These aberrant filters require removal by endovascular methods or cardiopulmonary surgery.

The fracture rate associated with retrievable filters increases over time after implantation.^38,41,52  Patients who have fracture of 1 or more components of a VCF commonly have embolization of the fractured component to the heart or lung. Among 80 patients who had a retrievable VCF, fracture of at least 1 strut was detected in 16%; 3 patients had a life-threatening arrhythmia or pericardial tamponade, and 1 had resulting sudden death.⁵² 

The presence of a VCF does not reduce the risk of DVT; in fact, there is evidence of increased risk of DVT after filter placement.^7,16,22,27  In the PREPIC trial, patients who had a permanent filter were much more likely to develop a DVT than those without a filter (DVT rates at 2 years were 21% vs 12% in the filter and nonfilter groups, respectively).²²  In a population-based study, recurrent DVT occurred in 21% of patients with an IVC filter and in 15% of those without a filter (P < .01).³⁶  Venous thromboembolic events occurred in 8% of 952 patients who had a VCF placed at Boston Medical Center; 48% of these were in patients who did not have VTE at the time of filter insertion.²⁷  These DVTs require long-term anticoagulation with its attendant risk of bleeding, and some result in postthrombotic syndrome with chronic leg swelling, discomfort, and reduced quality of life.

VCFs are thrombogenic. Macroscopic evidence of thrombi on temporary filters was shown in 75% of patients after a mean duration of insertion of only 5 days.⁵³  Filter-induced thrombosis begins early, another reason for starting (or restarting) anticoagulant as soon as it is safe to do so after VCF placement. In the PREPIC trial, symptomatic IVC thrombosis was seen in 13% of patients at 8 years.²⁴  These thrombi arise either from thrombosis at the site of filter implantation or are related to thrombus propagation from emboli that are trapped by the filter. Many VCF thrombi are asymptomatic, but others lead to acute, massive leg swelling, chronic leg pain, and skin changes, including venous leg ulcers. Vena caval thrombosis is less frequent when patients are anticoagulated after filter placement.⁵⁴ 

PE and occasional fatal PE, the conditions that IVC filters are designed to prevent, still occur despite the presence of a filter in 0.4% to 13% of patients.^{5,7,23,24,27,38,42,43,45}  In the first PREPIC study, 6% of patients with a permanent IVC filter developed new PE with a filter in place, whereas in PREPIC-2, the PE rate at 3 months was 3% in patients who had a VCF.^22,23 

While almost all VCFs have a temporary indication and >90% of retrievable filters can technically be removed, in practice, only 0.5% to 60% of filters are actually retrieved.^{7,27,42,43,48}  There are now centers that have developed advanced techniques to retrieve almost all difficult filters.^50,55  Therefore, the vast majority of failures to retrieve filters are due to making no attempt to remove them. Furthermore, the success rate of removal decreases as the duration of placement increases.⁷  In the systematic review of 6834 retrievable VCFs, only 34% were removed.⁷  Some of the reasons for failure to remove a retrievable IVC filter are listed in Table 3. Because retrievable filters have only been used since 2003, the long-term risks of these devices, if they are not removed, are largely unknown. For all of these reasons, in 2010 and again in 2014, the FDA issued Safety Alerts about VCFs and recommended that they be removed “as soon as protection from pulmonary embolism is no longer needed,” generally between 29 and 54 days after insertion.⁵⁶ 

Based on our review of the literature and experience, we summarize our suggestions for the safe use of VCFs in Table 4.

Despite the absence of direct evidence that VCFs save lives or reduce major or life-threatening PE in any studied population or for any indication, their use has proliferated in both the treatment and prophylaxis of VTE. Most of the increase in filter use has been the placement of retrievable filters, and many of these have been placed for prophylactic reasons. The most widely accepted indication for VCF use is in patients with a recent proximal DVT and an absolute contraindication to therapeutic doses of anticoagulant. Filters cannot be justified in addition to anticoagulation or as primary prophylaxis. Patients who have recurrent VTE despite therapeutic anticoagulation require a change in the anticoagulant option or in its dose rather than insertion of a VCF. A high proportion of retrievable filters are not removed, and these are associated with substantial rates of complications and considerable patient anxiety without providing long-term benefit.

Clinicians should consider having a filter placed only after careful consideration of the benefits and risks for each patient. They must carefully select patients with an appropriate indication, implement a standardized monitoring and follow-up plan, and arrange for filter removal when they are no longer required. If a VCF is justified, a retrievable VCF is preferred, since the vast majority can be removed when the contraindication to anticoagulation resolves. A flurry of recent multidistrict litigations against VCF manufacturers may ultimately lead to improvement in the FDA approval process and the quality of informed consent for patients.⁵⁷  Hematologists can take a leadership role in supporting the appropriate use of VCFs in their organizations and in ensuring that indicated filters are retrieved in a timely manner.

William Geerts, Thromboembolism Program, Sunnybrook Health Sciences Centre, Room D674, 2075 Bayview Ave, Toronto, ON M4N 3M5, Canada; e-mail: william.geerts@sunnybrook.ca.