Inferior Vena Cava Filters and Mortality: Is It the Underlying Process, the Patient, or the Device?

The incidence of venous thromboembolism (VTE) is increasing with an aging and more medically complex patient population.¹ The Achilles heel of VTE management remains the lack of a medical therapeutic option that does not increase bleeding risk. In 1865, Armand Trousseau² suggested a physical barrier may decrease migration of emboli; in the 1930s, John Homans³ popularized femoral vein ligation to prevent pulmonary embolism (PE). It was not until 1970 that the percutaneous inferior vena cava (IVC) filter was developed,⁴ with the goal of lowering the risk of PE among patients at high risk for recurrence or who cannot tolerate anticoagulation. Although there are guideline recommendations and widespread use of IVC filters, there remain limited high-quality data supporting its use, particularly in patients with a contraindication to anticoagulation.^5,6 A 2017 meta-analysis that evaluated the association of IVC filter placement and risk of PE included only 11 studies, 6 of which were randomized clinical trials, and overall these were considered low to moderate quality.⁶ None of the randomized trials included patients with a VTE and a contraindication to anticoagulation. Despite a lack of high-quality data, use remains high, with a 2016 report demonstrating that 1 in 6 Medicare beneficiaries with a PE receive an IVC filter.¹

Turner et al⁷ explored whether IVC filter placement was associated with a difference in 30-day mortality among patients with VTE and a contraindication to anticoagulation. The study used all-payer administrative billing data for 3 states to identify adult patients with a deep vein thrombosis (DVT) and/or PE. A contraindication to anticoagulation was determined by the presence of 1 of over 850 selected International Classification of Disease, Ninth Revision diagnosis or procedure codes including, intracranial bleeding, other major bleeding, and thrombocytopenia. The analysis included 126 030patients categorized with contraindications to anticoagulation, with 45 771 (36.3%) receiving an IVC filter. Turner and colleagues⁷ found that IVC filter placement was associated with a hazard ratio of mortality (1.18; 95% CI, 1.13-1.22; P < .001), which was unchanged after additional adjustment by a propensity score for IVC filter placement.

Turner and colleagues⁷ should first be commended for exploring a topic with a scarcity of evidence. Their study addressed an often neglected and overlooked bias, the immortal time bias, by categorizing IVC filter status using a time-varying covariate approach. Immortal time bias can occur in observational comparisons when the inclusion into 1 treatment group, in this case the IVC filter group, is conditional on living to receive that treatment.⁸ In this analysis, if immortal time bias was not accounted for (as had not been done previously), patients in the IVC group would have to remain alive long enough after index hospitalization to receive an IVC filter, thus receiving a window of immortality. This bias would be the most influential if VTE were present at the time of admission, which may not have been true for all patients in this study.

It is also necessary to comment on the important limitations of this study. Turner and colleagues⁷ attempt to identify patients with a contraindication to anticoagulation by the International Classification of Disease, Ninth Revision billing codes, but the validity of the algorithm used is unknown. For example, claims data are unable to identify whether patients were using anticoagulation as a trial or tolerated anticoagulation, despite being classified as having a contraindication. A 2016 study of patients in California without cancer demonstrated decreased mortality for those who received an IVC filter with active bleeding.⁹ Nearly two-thirds of patients included in the study did not receive an IVC filter despite guideline recommendations to do so, which could reflect deviation of care from guidelines, but also the inclusion of patients without true contraindications for anticoagulation—particularly among patients not treated with IVC filters. In addition, Turner and colleagues⁷ are unable to differentiate whether VTE was present at admission or acquired during the hospitalization, both DVT and PE occurred together or sequentially during the hospitalization, or IVC filter placement preceded or followed DVT and/or PE.

The authors were unable to account for important unmeasured confounders between patient groups that are not captured in claims-based registries. Administrative claims databases are limited in their ability to elucidate the severity of VTE, clot burden, frailty of the patient, or overall clinical stability. For instance, it is possible that those with a more severe VTE presentation had an IVC filter placed, as placement of an IVC filter in those with limited cardiopulmonary reserve, regardless of candidacy for anticoagulation, is also guideline indicated.⁵ Turner and colleagues⁷ argue that sicker patients may be less likely to undergo a procedure; however, given the ease at which IVC filters can be placed, including at the bedside, this may not be the case. Based on the limitations of the data set, treatment selection bias is impossible to rule out, which could bias the results in either direction.

Thirty-day mortality may not be the only appropriate end point for the evaluation of the usefulness of IVC filters in the management of acute DVT and/or PE. The occurrence of PE among patients with DVT and recurrent PE among patients with PE are both associated with substantial morbidity and added health care costs. In addition, there is no clear causal explanation why an IVC filter may increase the risk of short-term death, as the failure rate of IVC filters to lower PE risk is limited,¹⁰ and early complications of IVC filters are uncommon.¹⁰

Turner and colleagues⁷should be commended for their attempt to add to the current dearth of data for IVC filter use. We believe the greatest value of the study is to call out how limited our current evidence base is to support such a commonly used device, and to challenge the clinical and research communities to demand higher-quality studies before practices become ingrained. Although randomized clinical trials are considered the gold standard of scientific investigation, there is an underlying conflict with randomization when strong clinical beliefs about the effectiveness of an intervention exist. Would clinicians feel comfortable randomizing a patient with a stable PE, a residual common iliofemoral DVT, and a contraindication to anticoagulation to not receive an IVC filter? In the absence of randomized evidence, what is the value of relying on observational data alone to guide this important treatment decision, with all of their limitations? These are the questions that remain at large for the cardiovascular community in determining how to best treat our patients with VTE with a contraindication to anticoagulation.

Accepted for Publication: May 30, 2018.

Published: July 13, 2018. doi:10.1001/jamanetworkopen.2018.0453

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2018 Secemsky EA et al. JAMA Network Open.

Corresponding Author: Robert W. Yeh, MD, MSc, Smith Center for Outcomes Research in Cardiology, Department of Medicine, Beth Israel Deaconess Medical Center, 375 Longwood Ave, Fourth Floor, Boston, MA 02215 (ryeh@bidmc.harvard.edu).

Conflict of Interest Disclosures: Dr Yeh reported grants and personal fees from Boston Scientific and personal fees from Medtronic outside the submitted work. No other disclosures were reported.