Efficacy of inferior vena cava filters in anticoagulated patients.

Abstract

The benefit of an inferior vena cava (IVC) filter in addition to standard anticoagulation regimens is unknown.
We examined data for patients who received IVC filters with anticoagulation (AC-Filter) after an episode of venous thromboembolism (VTE) and compared them with data for those who received anticoagulation only (AC-Only). Outcome measures were new pulmonary embolism (PE), recurrent deep vein thrombosis (DVT), and mortality at 90 days and at 5 years. Demographic data included age, gender, and ethnicity/race, prior thromboembolic history, cancer, serum albumin, and time in therapeutic range. In addition, subsets matched for age, gender and race/ethnicity were examined in detail.
AC-Filter patients (n = 251), when compared to AC-Only patients (n = 1377), did not differ significantly with regard to gender or cancer status, but white males in general had better outcomes. AC-Filter patients were more likely to have had a previous history of PE or VTE (P < 0.001). In comparison to AC-Only patients, AC-Filter patients had lower mean serum albumin levels (3.1 +/- 0.8 vs. 3.6 +/-0.8 mg dL(-1), P < 0.001) and were older (65 +/- 16.1 years vs. 60 +/- 17.5 years, P < 0.001). After stratification according to previous history of PE or VTE prior to the index VTE event, no differences in the outcome measures of new PE, recurrent DVT or mortality were identified between groups, but patients with a prior history of PE from either group were more likely to have a new PE (hazard ratio 1.9, P < 0.001).
These data suggest that IVC filters may not provide any substantial additional benefit for patients who can tolerate anticoagulant therapy.

Full text

Page 1

ORIGINAL ARTICLE
Efficacy of inferior vena cava filters in anticoagulated patients
H. H. BILLETT, L. G. JACOBS, E. M. MADSEN, E. R. GIANNATTASIO, S. MAHESH and H. W. COHEN
Department of Medicine and Epidemiology, Montefiore Medical Center, Albert Einstein College of Medicine, Bronx, NY, USA
To cite this article: Billett HH, Jacobs LG, Madsen EM, Giannattasio ER, Mahesh S, Cohen HW. Efficacy of inferior vena cava filters in
anticoagulated patients. J Thromb Haemost 2007; 5: 1848–53.
Summary. Background and objectives: The benefit of an
inferior vena cava (IVC) filter in addition to standard
anticoagulationregimens is
examined data for patients who received IVC filters with
anticoagulation (AC-Filter) after an episode of venous throm-
boembolism (VTE) and compared them with data for those
who received anticoagulation only (AC-Only). Outcome
measures were new pulmonary embolism (PE), recurrent deep
vein thrombosis (DVT), and mortality at 90 days and at
5 years.Demographicdataincludedage,gender,andethnicity/
race, prior thromboembolic history, cancer, serum albumin,
andtimeintherapeuticrange.
matched for age, gender and race/ethnicity were examined in
detail. Results: AC-Filter patients(n = 251),whencompared
to AC-Only patients (n = 1377), did not differ significantly
with regard to gender or cancer status, but white males in
general had better outcomes. AC-Filter patients were more
likelytohavehadaprevioushistoryofPEorVTE(P < 0.001).
In comparison to AC-Only patients, AC-Filter patients had
lower mean serum albumin levels (3.1 ± 0.8 vs. 3.6 ±
0.8 mg dL–1, P < 0.001) and were older (65 ± 16.1 years vs.
60 ± 17.5 years, P < 0.001). After stratification according
to previous history of PE or VTE prior to the index VTE
event, no differences in the outcome measures of new PE,
recurrent DVT or mortality were identified between groups,
but patients with a prior history of PE from either group were
morelikely tohavea new
P < 0.001). Conclusions: These data suggest that IVC filters
may not provide any substantial additional benefit for patients
who can tolerate anticoagulant therapy.
unknown.Methods: We
Inaddition,subsets
PE(hazard ratio1.9,
Keywords: anticoagulation,
thrombosis.
inferiorvenacava filter,
Introduction
Thereareseveraldifferentpopulationswhoarenowconsidered
to be candidates for inferior vena caval (IVC) filters [1]. The
?failedtherapygroup?consistsofpatientswhohaveexperienced
a recurrent venous thromboembolism (VTE) on adequate
anticoagulation. Another group is composed of those who
have had a VTE but who also have a contraindication to
anticoagulation, such as a bleeding diathesis or history of
hemorrhage [2,3]. A third group is composed of those patients
who have not sustained a VTE, but in whom an IVC filter is
inserted prophylactically in connection with a procedure or
event that is associated with a high incidence of thrombosis,
such as trauma from motor vehicle accidents and bariatric
surgery [4,5]. A large amount of the published outcome data
regarding the use of IVC filters combines all of these types of
patients, prophylactic and therapeutic, on and off anticoagu-
lation, and with prothrombotic or hemorrhagic indications for
placement. Although there is reason to believe that the IVC
filter can provide benefit for patients who cannot be antico-
agulated and have no other alternative, few studies have
demonstrateda benefit of filterplacement inthesettingof VTE
in addition to continuation of anticoagulation [6–8]. To
determine the value of IVC filter placement in this common
clinical situation, we examined outcomes for patients with a
VTE who subsequently received an IVC filter in addition to
treatment with therapeutic levels of anticoagulation after filter
placement, and compared this cohort to patients with a VTE
who received therapeutic anticoagulation alone.
Methods
CLINICAL LOOKING GLASS (CLG) is an interactive software
application developed at Montefiore Medical Center to evalu-
ate health care quality, effectiveness, and efficiency. The system
integrates clinical and administrative datasets that allow the
production of epidemiologically cogent self-documenting
reports that, in addition to global assessment of care, can
identify specific patients in need of clinical remediation.
Replicate data derived from hospital records are produced in
a programmable format for statistical access as defined by
International Review Board (IRB)-approved protocols. Using
CLG, we retrospectively examined prospectively collected data
in the clinical electronic database of Montefiore Medical
Correspondence: Henny Billett, Montefiore Medical Center, Albert
Einstein College of Medicine, 3411 Wayne Ave, Bronx, NY 10467,
USA.
Tel.: +1 718 920 7631; fax: +1 718 882 8698; e-mail: hbillett@
montefiore.org and billett@aecom.yu.edu
Received 3 January 2007, accepted 18 June 2007
Journal of Thrombosis and Haemostasis, 5: 1848–1853
? 2007 International Society on Thrombosis and Haemostasis

Page 2

Center, a large diverse urban medical center in Bronx, NY for
all patients who had been discharged between October 1997
and September 2005 with a diagnosis of VTE and who had
documented adequate anticoagulation with warfarin [target
International Normalized Ratio (INR) 2.5, range: 2.0–3.0]
within the first 30 days. The date of discharge of this index
hospitalization was consideredtobe the index date. These were
grouped into two cohorts: VTE patients who had IVC filters
inserted during that index admission along with concomitant
anticoagulation (AC-Filter), and VTE patients who did not
receive IVC filters but who were also treated with adequate
anticoagulation (AC-Only). Data regarding demographic
characteristics and clinical history, including prior PE or
VTE, or cancer, as well as serum albumin, were also obtained.
The last serum albumin level prior to discharge was used as a
surrogate marker for morbidity. Cancers identified up to
5 yearspreviouslyaswellasthoseidentifiedaftertheindexdate
were assessed.
Outcome measures were recurrent deep vein thrombosis
(DVT), new pulmonary embolism (PE), and mortality at
90 days and at 5 years. The primary endpoint was new PE.
Mortality was assessed by use of both the hospital clinical
database and a social security record search. The potentially
confounding variables gender, self-declared race/ethnicity and
cancer status at the index discharge were compared by chi-
square;meanageandalbuminwerecomparedusingat-testfor
independent samples. Multivariable-adjusted hazard ratios
(HRs) were estimated with Cox models constructed for the
first 90 days from the date of discharge as well as a maximum
follow-up time of 5 years. Analyses were performed with SPSS
(SPSS Inc., Chicago, IL, USA). Cumulative event rates were
described using Kaplan–Meier curves, and group comparisons
of time-to-event were performed with log-rank tests and
assessment of a group indicator variable in Cox models. The
numberoftargeteventsachievedperperson-dayatriskandthe
incidence density per year of risk were obtained to determine
the relative risk and the risk difference.
In order to address indication bias, we stratified each group
intopatientswhohadhadeitherapreviousVTEorPEpriorto
the index thrombotic event (thus a minimum of two separate
events). We examined separately the rates of recurrence at
90 days and at 5 years in these subgroups.
As duration of anticoagulation might be different in patients
who have a filter vs. those who are anticoagulated without a
filter, and could therefore be another potential outcome
modifier, rates of recurrence were assessed at 5 years, whether
ornot anticoagulationhad continuedfor the full 5 years,anda
separate assessment was made for those patients who had had
continued anticoagulation. For the large cohorts, time in
therapeutic range (TTR) was calculated as the number of
laboratory tests within the therapeutic range (target INR 2.5,
range: 2.0–3.5 accepted).
To further identify and characterize clinical features that
might be associated with different outcomes, we selected
subsets of the two cohorts of patients for review in depth. The
smaller AC-Filter list was divided into different alphabetized
sections for each of the investigators involved, and the first 30
charts retrieved wereused. We then identified patients from the
AC-Only group who matched these for age, gender and race/
ethnicity, and used the most closely matched available. In all
but a few circumstances, there was only one match; when there
were two or more, proximity to actual birth date was the
deciding factor. Specific objectives for the analysis of these
matched cohorts included an evaluation of the incidence of
individual risk factors for thrombosis: cancer (subdivided into
breast, colon, lung, prostate, lymphoma, ?other?), immobility
(subdivided into acute or chronic), thrombophilia (subdivided
into factor V Leiden, prothrombin G20210A, protein C
deficiency, antiphospholipid antibody syndrome (APLS), pro-
tein S deficiency, antithrombin III deficiency, methylene
tetrahydrofolate reductase mutations (MTHFR) or hyper-
homocysteinemia, ?other?), body mass index (BMI) >40 or
weight >110 kg, hormonal therapy, recent surgery (divided
into total knee replacement, total hip replacement, hip fracture
surgery, abdominal surgery, genitourinary or gynecologic
surgery, cancer surgery, bypass surgery, ?other surgery?),
varicoseveins,diabetes,cigarettesmoking,presenceofvascular
access devices, or trauma. For the IVC filter group, indications
for filter placement, types of devices inserted and procedure-
related adverse events (operative complications, IVC throm-
bosis, filter fractures, migration, IVC penetration) were also
investigated. For analyses of these smaller matched subsets,
TTRs (with target INR of 2.5) were calculated from midpoint
to midpoint duration of INR values by the method of
Rosendaal [9], and were considered ?in-range? for values of
2.0–3.0 and ?high-in-range? for values of 3.1–3.5.
Results
ThetotalnumberofpatientswithadischargediagnosisofVTE
during the study period was 5263. Of these, 1726 patients had
documented INRs of at least 2.0 within 30 days of admission
and were included in the analysis. There were 1146 patients
who received an IVCfilter after an index VTE during this time:
303ofthesehaddocumentedINRofatleast2.0within30 days
of admission. Data were incomplete or missing in 31 (four AC-
Filter, 27 AC-Only). To minimize selection bias, both the 51
deathsontheindexadmission(45AC-Filter,sixAC-Only)and
the 16 recurrent VTEs while in hospital prior to discharge
(three AC-Filter and 13 AC-Only) were excluded. The
remaining 1628 formed the study sample, with 251 patients in
the AC-Filter group and 1377 in the AC-Only group. The
characteristics of the larger study sample, along with data from
the matched cohort groups, are shown in Table 1. The AC-
Filter cohort was significantly older, and fewer had normal
albumin levels when compared to the AC-Only cohort. Both
groups were predominantly female, reflecting the gender
differences of an older population, and the groups were similar
with regard to the presence of cancer. As expected, the group
who had IVC filter placement had an increased incidence of
both prior PE and prior DVT as compared with patients who
were treated with anticoagulation alone (9.2% vs. 2.3%, and
Inferior vena cava filters in anticoagulated patients 1849
? 2007 International Society on Thrombosis and Haemostasis

Page 3

20.7% vs. 11.9%, respectively, both P < 0.001). To control
for this, groups were stratified for prior events in the outcome
analyses.
Bothgroupswereevaluatedfromtheindexdischargedateto
90 days and for a maximum period of 5 years. After adjust-
ment for age, male gender conferred protection for subsequent
PE at 90 days (HR 0.29, P = 0.003) and at 5 years (HR 0.58,
P = 0.02), and being Caucasian conferred a decreased risk at
5 years for recurrence of both PE (HR 0.43, P = 0.04) and
DVT (HR 0.64, P = 0.001) but at 90 days only significantly
for DVT (HR 0.61, P = 0.001).
For the total population of patients, there were 53 PEs at
90 daysand109PEsoverthe5-yearriskwindow,theincidence
of subsequent PE for the population being 6.7%. There was no
difference in the rate of subsequent PE for patients who were
anticoagulated, whether or not they had a filter. This was true
at 90 days (HR at 90 days 1.02, P = 0.95) and at 5 years (HR
0.96, P = 0.88; Fig. 1).
Major differences in the rate of new PE were seen in both
groups, however, for patients who had had a previous
admission for PE prior to the index admission vs. those who
had not had a prior PE. As seen in Table 2, 13% of patients
with a prior PE had new PEs as compared to 6.7% of patients
without a prior PE (P < 0.001). As the AC-Filter group had a
higherincidenceofrecurrentPEs,stratificationforthisvariable
wasnecessary.Afterstratification for prior PE, therewas again
no difference in the incidence of new PE for patients with or
without filters. This was true at 90 days (HR at 90 days 1.02,
P = 0.95) and at 5 years (HR 0.96, P = 0.88; Fig. 1). These
stratifiedriskdatasuggest that, althoughthere isa differencein
thetypesof patient whoreceivean IVC filter,in that these filter
patients are more likely to have had recurrent thrombotic
events, insertion of IVC filters in these high-risk patients does
not significantly benefit them.
There were 326 DVTs within 90 days and 603 DVTs over
thenext 5 years(37.0%).Forthetotalcohort,therewasanon-
significant trend (41.4% AC-Filter vs. 36.2% AC-Only,
P = 0.12) toward an increase in DVTs in patients who were
anticoagulated and had a filter. Unlike for PE, a history of
prior DVT did not predispose patients to a higher incidence of
either new PE or DVT.
TTRs were similar for both groups and are shown in Fig. 2;
theTTRforAC-Filterwasrelativelylowat52.3%,butwasnot
dissimilar to that for AC-Only at 54.3%. The duration of
anticoagulation, as depicted by the percentage discontinued at
eachdesignatedyear,isshowninFig. 3;61.5%ofpatientswith
AC-Filter and 59.2% of the AC-Only patients were anticoag-
Table 1 Population demographics. Significance was determined by two-
sided chi-squared tests for the categorical variables and by ANOVA for the
continuous variables
n
AC-Only AC-Filter
P
1377 251
Age (mean years) (SD)
% Female
% White
% Cancer
% Prior PE
% Prior DVT
Albumin (mg dL–1), mean (SD)
60 (±17.4)
66.3
25.3
26.7
2.3
11.9
3.6 (±0.8)
65 (±16.1)
62.5
29.1
30.7
9.2
20.7
3.1 (±0.8)
<0.001
NS
NS
NS
<0.001
<0.001
<0.001
PE, pulmonary embolism; DVT, deep vein thrombosis; NS, not
significant.
New PE
Days
Ac-only
Ac-filter
Cumulative percent with event
1.0
0.8
0.6
0.4
0.2
0.0
0.00500.001000.00 1500.002000.00
Fig. 1. Kaplan–Meier cumulative incidence rate for new pulmonary
embolism(PE).ThereisnosignificantdifferenceintherateofnewPEwith
or without filter insertion.
Table 2 Incidence of new event
TotalAC-OnlyAC-FilterP
All patients
n
New PE [% (n)]
New DVT [% (n)]
Deaths [% (n)]
Patients with a history of prior PE
n
New PE [% (n)]
New DVT [% (n)]
Deaths [% (n)]
1628
6.7 (109)
37.0 (603)
33 (538)
1377
6.8 (93)
36.2 (499)
30.9 (426)
251
6.4 (16)
41.4 (104)
44.6 (112)
0.83
0.12
<0.001
5431
12.9 (4)
45.2 (14)
38.7 (12)
23
13 (7)
44.4 (14)
40.7 (22)
13 (3)
43.5 (10)
43.5 (10)
>0.99
>0.99
0.78
PE, pulmonary embolism; DVT, deep vein thrombosis.
Time in therapeutic range
30%
25%
20%
15%
10%
5%
0%
1.0< INR <1.5 1.5< INR <2.0 2.0< INR <2.5 2.5< INR <3.0 3.0< INR <3.5 3.5< INR <4.0 4.0< INR <5.0 5.0< INR <10.0
INR >10.0
AC-only
AC-filter
Fig. 2. Time in therapeutic range (TTR) for both cohorts. There is no
difference in TTR for either group.
1850 H. H. Billett et al
? 2007 International Society on Thrombosis and Haemostasis

Page 4

ulated for only 1 year. When the duration of anticoagulation
was examined specifically for those who died in the first year,
72.5% of AC-Filter patients and 73.8% of AC-Only patients
wereanticoagulatedfor>75%ofthetimefromdischargeuntil
death (Fig. 3, inset). Although the duration of anticoagulation
was very similar in both groups, a separate assessment was
made for those who continued anticoagulation. At 3 years,
40.8% of the AC-Filter patients and 46.8% of the AC-Only
patients had a recurrent VTE (P = 0.78); for PE, it was 7.7%
for the AC-Filter patients and 6.3% for the AC-Only patients
(P = 0.84). This was not further stratified for events prior to
theindexevent,norweretheseassessedat5 years,owingtothe
smaller numbers involved.
The overall mean follow-up time of 2.57 ± 1.8 years from
index date reflects the substantial death rate of 12.4% at
90 days and 33% at 5 years. Prior to risk factor adjustment,
patients with IVC filters had increased mortality (HR 1.4, 95%
CI: 1.14–1.71, P < 0.002). This greater HR seen for patients
with AC-Filter was not observed when age, cancer or albumin
levels were adjusted for in a multivariate Cox regression model
(HR 0.98, CI: 0.79–1.2, P = 0.85). At 5 years, the HR for
albumin levels<3.5 mg dL–1
P < 0.001), whereas for age and cancer the HRs were 1.028
(CI: 1.022–1.034, P < 0.001) and 2.22 (CI: 1.87–2.64,
P < 0.001) respectively. At all stages, the surrogate marker
for health status, serum albumin, had a strong association with
mortality and recurrent DVT (Table 3).
was4.58(CI:3.7–5.6,
In addition to stratification, we addressed indication bias
using the matched cohorts. Results from these patients were
similar to those from the larger cohorts from whom they were
derived, with no difference being identified in the incidence of
cancer, BMI, varicose veins, immobility, or recent surgery
status. As in the larger cohorts, patients in the AC-Filter
matched subcohort were more likely to have had a DVT or PE
prior to the index event (P = 0.04 and P = 0.003 respect-
ively). Known ?thrombophilia? risk factors for VTE were
positively identified in 16.7% of the AC-Filter patients vs. 10%
of AC-Only patients. For the matched cohort group, the type
of IVC filter was recorded. Vena-Tech filters were inserted in
58%, Gunther-Tulip Filters in 17% and Opt-Ease/Trap-Ease
Filters in 11%. Temporary filters were less common and, when
used, were rarely retrieved. The major indications for filter
placement in the AC-Filter matched cohort were massive PE
(36.7%) and recurrent VTE while on adequate anticoagulation
(36.7%), and the third most common indication was for the set
of patients labeled as having ?a poor prognosis? (13.3%). Other
common reasons for filter insertion were non-compliance, fall
risk, and dementia. TTRs were somewhat lower than for the
full cohorts (40.8% for the AC-Only patients vs. 38.2% for the
AC-Filter patients) as assessed by mid-duration range to mid-
duration INR [9]. There was only one PE in these matched
cohorts, in a patient with a filter who had not been a ?recurrent
VTE? or a prior ?massive PE?, so no outcome data on PE could
be determined. As in the larger group, there were no significant
advantages or disadvantages to filter insertion for recurrent
VTE for patients already on anticoagulation.
Discussion
More than a century after interruption of the IVC as a means
of preventing PE was first suggested by Trousseau (1868) and
performed by Bottini (1893) [10], IVC filter insertion has
become an increasingly utilized treatment modality, with 2000
procedures being performed in 1979 and 49 000 in 1999 [11].
This increased use appears to be more location-driven than
data-driven, with more IVC filters being placed in the
northeastern USA than in the rest of the country, and fewer
in some European countries than in the USA [12]. Filters,
traditionally used in situations where systemic anticoagulation
is deemed contraindicated or excessively risky, have also
become an increasingly popular method of preventing recur-
rent PE in patients who are anticoagulated but still thought to
be at high risk, yet few data exist to support their efficacy and
widespread usage.
In most studies, rates of PE and DVT with IVC filters have
been generally comparable to rates without an IVC filter: PE,
2.6–3.8%, DVT, 5.8–32%; and IVC thrombosis, 3.6–11.2%
[13]. However, it is generally unclear in these studies which
patients were on anticoagulation, what the duration and/or
efficacy of the anticoagulation was, and how these groups
matched up against a comparable population without a filter.
One recent multicenter study [1] noted the difficulties of
gathering data from these disparate groups: in this study, 14%
60%
50%
40%
30%
20%
10%
0%
0%
% of population 80%
Time anticoagulated from diagnosis
until death - first year
70%
60%
50%
40%
30%
20%
10%
1
AC only
AC filter
0-25%26-50% 51-75% 76-100%
2
Years
3
5 or more
Fig. 3. Duration of anticoagulation for each cohort as determined by last
International Normalized Ratio recorded (inset: subset who died within
the first year). The duration of anticoagulation is similar for both the AC-
Only group and the AC-Filter group in each case.
Table 3 Hazard ratios (HRs) and 95% CIs for abnormal albumin
(<3.5 mg dL–1) estimated with Cox models adjusting for age, sex, race,
cancer and prior event
90-Day follow-up5-Year follow-up
HR (95% CI)P HR (95% CI)P
PE
DVT
Mortality
0.80 (0.44, 1.44)
1.93 (1.53, 2.43)
12.9 (8.1, 20.7)
0.45 0.88 (0.58, 1.33)
1.30 (1.09, 1.54)
4.58 (3.77, 5.57)
0.55
0.003
<0.001
<0.001
<0.001
PE, pulmonary embolism; DVT, deep venous thrombosis.
The presence of an IVC filter did not significantly influence these
endpoints (not shown).
Inferior vena cava filters in anticoagulated patients 1851
? 2007 International Society on Thrombosis and Haemostasis

Page 5

of all patients with a VTE received the filters as therapy with or
without anticoagulation. The filter patients were more likely to
have had a history of PE, VTE, or multiple system organ
failure,makingcomparisonsof thevalueoffilterswith/without
anticoagulation difficult. As the majority of patients with IVC
filters are not anticoagulated (indeed, our selected patient
population represented only 20% of the total filters inserted),
clinicians have interpreted ?equivalency? to mean that IVC
filters confer a benefit to a more at-risk group. We, too,
identified an increased ?at-risk? group; patients who received an
IVC filter had higher rates of both prior VTE and PE and had
lower serum albumin levels. We therefore thought that it was
important to assess these factors in our risk stratification
scheme. We found that there was no benefit to the incidence of
new PE and that the incidence was significantly higher in the
group who had prior PEs, irrespective of whether or not they
had a filter inserted. As the biggest difference between the
groups was the higher incidence of prior PE in the AC-Filter
group, it could be argued that the reason why the AC-Filter
group was ?equivalent? was that it was indeed protecting the
higher-risk group, and the filter succeeded in decreasing these
patients? PE incidence to that of patients without prior PE. We
therefore did a sensitivity analysis in which the sample was
restricted to those with prior PE in both groups. As we note in
Results, after stratification, there was no difference in subse-
quent PE between the two study groups.
It is of interest that the duration of anticoagulation was very
similar in both groups, even in the subgroup that died. When a
separate analysis of only those with continued anticoagulation
was performed, presumably reflecting those considered to be at
higher risk, the total event rate was slightly higher, but the AC-
Filter and AC-Only groups were not significantly different.
Many believe that thrombophilia risk factors are risk factors
for recurrence. In addition to stratification, in order to
minimize indication bias, we selected matched cohorts. In the
matched cohort, as with the whole study sample, the AC-Filter
patients were more likely than AC-Only patients to have had
PE or a previous DVT prior to the index VTE. We found that
the matched cohort AC-Filter patients were only slightly more
likely to have other thrombophilia risk factors (16.7% vs.
10%), attesting to the greater importance given to clinical
events than to laboratory evidence of a thrombophilia.
Poor prognosis was another indication cited for filter
placement. Low albumin is commonly used as a surrogate
marker for poor prognosis; we therefore assessed this as a
potential contributor to indication bias by adjusting for
albumin in the multivariable models. The AC-Filter group
had a slightly higher proportion with existing cancer at the
index date (30.7% vs. 26.7%), and so we also adjusted for
cancer – either before, subsequent to, or ?ever?. Thus, while it is
possible that indication bias could influence our results, we
took care to employ all available reasonable measures to
account or adjust for it.
The recurrence risk was analyzed for all patients at 90 days
and at 5 years, as we felt that the decision to stop therapy is
part of the therapeutic process, with or without a filter.
Interestingly, a similar proportion of those alive at 1 year in
each group (61.5% vs. 59.2%) were on anticoagulation at the
end of that year. This leaves about 40% of patients without
anticoagulation for some of that time, and one can argue that
this could tend to create a bias toward a benefit for IVC filters,
but such a benefit was not seen. In addition, despite our INR
target of 2.5, about 25% of our anticoagulated population had
INR levels of about 1.5–2.0. This inadequate level of antico-
agulation would again tend to create bias against the AC-Only
group, as this group would have little to protect itself from
recurrence, whereas, it could be argued, the filter group at least
had the inserted filter.
The only prospective randomized but not blinded trial to
date, the Prepic [14,15] trial, first published in 1998 and then
updated in 2005, examined 400 patients with VTE who were
anticoagulated for at least 3 months. Initially, at 12 days, these
investigators found a lower incidence of PE in patients with a
filter but no difference in rates of death. At 2 years, no
significant difference in PE was found, but at the end of
8 years, the incidence of PE was reduced in patients with IVC
filters (6.2% vs. 15.1%, P = 0.008). Our study, with a larger
number of patients, albeit with a shorter follow-up time of up
to 5 years, demonstrated no significant difference. The differ-
ence may lie in the study population. The Prepic trial was a
controlled trial withpurely prospective data. In the Prepic trial,
patients were excluded if they had demonstrated failure of
anticoagulant therapy, short life-expectancy, hereditary throm-
bophilias, or indications for thrombolysis. These indications,
especially failure of anticoagulant therapy, are among the most
oftencitedoftherecommendedindicationsforfilterplacement.
Therefore, by excluding these groups, the trial did not provide
an answer to the question of whether IVC filter placement is
beneficial for patients with these common, difficult manage-
ment problems.
Another limitation of this study is that only symptomatic
events requiring admission were evaluated, and therefore
asymptomatic cases of VTE or PE may have been missed,
and it is dependent on good data entry from hospital billing/
coding personnel. This problem should have been present for
both groups. Given that this study seeks to answer the
question of whether IVC filters confer any additional benefit
over anticoagulation alone, only patients with IVC filters
with early and adequate anticoagulation were investigated,
rather than all filter patients. One of our requirements was
that both sets of patients had to have had a documented
INR of at least 2.0 within 1 month of anticoagulation. This
helped ensure that these data, especially the 90-day data,
would be valid. It also ensured and restricted us to a
population that was followed by our hospital system (which
traditionally serves the majority of Bronx residents requiring
health care). Although our TTR was somewhat low, it was
similar in both groups, and comparable to those in other
studies [16,17]. In addition, the duration of anticoagulation
was comparable in both the AC-Filter and AC-Only groups,
and similar to that seen in the Pepic trial [15], with the
majority maintaining anticoagulation.
1852 H. H. Billett et al
? 2007 International Society on Thrombosis and Haemostasis

Page 6

There was a high mortality rate in patients with VTE, which
was not necessarily because of a recurrent thrombotic event.
Interestingly, over 70% of patients who died within the first
year were anticoagulated. In this study, we did not examine
those patients with VTE who were not undergoing anticoag-
ulation, and thus could not assess whether the addition of
anticoagulationtoafilterdecreasedmortalityascomparedtoa
filter alone.
Prior to risk adjustment, an initial increase in the number of
deaths in the filter group was noted. Because massive PE may
be considered an indication of filter placement and therefore
mightprejudiceoutcomeagainstthefiltergroup[18],thegroup
that died in hospital was excluded. The serum albumin level
appeared to be the most important prognostic indicator on all
levels. This yardstick of general wellbeing, more than any other
parameter, differentiated those who would have a recurrent
DVTandthosewhowoulddie.Indeed,whenwecontrolledfor
albumin, we saw no difference in the rate of recurrent VTE in
cancer patients. This is in contrast to studies [15,19] that have
shown a higher rate of recurrent VTE in cancer patients with a
filter vs. those without (32% vs. 17% overall [19]), but it is
unclear whether these patients were anticoagulated at the time
ofrecurrenceorwhetherthestudywascontrolledformorbidity
assessments [15]. Post-filter PEs have also been reported in 1.3–
4.3% of cancer patients [20,21], but these studies were
performed without a comparable non-malignant group.
Our study could not demonstrate any advantage in any
population for insertion of an IVC filter in patients who had
successful anticoagulation. We hope that this initial analysis
will help clarify questions that need to be addressed to
determine whether there is any clinical benefit of IVC filter
insertion.
Acknowledgements
We gratefully acknowledge the help of H. Chan with the data
input for the analyses of the matched cohorts.
Disclosure of Conflict of Interests
The authors state that they have no conflict of interest.
References
1 JaffMR,GoldhaberSZ,TapsonVF.Highutilizationrateofvenacava
filters in deep vein thrombosis. Thromb Haemost 2005; 93: 1117–9.
2 Buller HR, Agnelli G, Hull RD, Hyers TM, Prins MH, Raskob GE.
Antithrombotic therapy for venous thromboembolic disease. Chest
2004; 126: 401S–28S.
3 Schutzer R, Ascher E, Hingorani A, Jacob T, Kallakuri S. Preliminary
resultsofthenew6 FTrapeaseinferiorvenacavafilter.AnnVascSurg
2003; 17: 103–6.
4 Keeling WB, Haines K, Stones PA, Armstrong PA, Murr MM,
Shames ML. Current indications for preoperative IVC filter insertion
in patients undergoing surgery for morbid obesity. Obes Surg 2005; 15:
1009–12.
5 Ku GH, Billett HH. Long lives, short indications: the case for
removable inferior cava filters. Thromb Haemost 2005; 93: 17–22.
6 Girard P, Stern JB, Parent F. Medical literature and vena cava filters:
so far so weak. Chest 2002; 122: 963–7.
7 Ansell J. Vena cava filters: do we know all that we need to know.
Circulation 2005; 112: 298–9.
8 Brender E. Use of emboli-blocking filters increases, but rigorous data
are lacking. JAMA 2006; 295: 989–90.
9 Azer AJ, Deckers JW, Rosendaal FR, van Bergen PF, van der Meer
FJ, Jonker JJ, Briet E. Assessment of therapeutic quality control in a
long-term anticoagulant trial in post-myocardial infarction patients.
Thromb Haemost 1994; 72: 347–51.
10 Becker DM, Philbrick JT, Selby JB. Inferior vena cava filters: indica-
tions, safety, effectiveness. Arch Intern Med 1992; 152: 1985–94.
11 Stein PD, Kayali F, Olson RE. Twenty-one-year trends in the use of
inferior vena cava filters. Arch Intern Med 2004; 164: 1541–5.
12 Emanuelli G, Farina A, Segramora V, Pecora N. 16-year follow-up of
Vena Cava Filters Group. Minerva Chir 2005; 60: 119–35.
13 StreiffMB.Venacavalfilters:acomprehensivereview.Blood.2000;95:
3669–77.
14 Decousus H, Leizorovicz A, Parent F, Page Y, Tardy B, Girard P,
Laporte S, Faivre R, Charbonnier B, Barral FG, Huet Y, Simonneau
G. A clinical trial of vena caval filters in the prevention of pulmonary
embolism in patients with proximal deep vein thrombosis. Prevention
du Risque d?Embolie Pulmonaire par Interruption Cave Study Group.
N Engl J Med 1998; 338: 409–15.
15 Prepic Study Group. Eight-year follow-up of patients with permanent
vena cava filters in the prevention of pulmonary embolism: the
PREPIC (Prevention du Risque d?Embolie Pulmonaire par Interrup-
tion Cave) randomized study. Circulation 2005; 112: 416–22.
16 Sarawate C, Sikirica MV, Willey VJ, Buliano MF, Hauch O. Monit-
oring anticoagulation in atrial fibrillation. J Thromb Thrombolys 2006;
21: 191–8.
17 Beyth RJ, Quinn LM, Landefeld CS. A multicomponent intervention
to prevent major bleeding complications in older patients receiving
warfarin. Ann Intern Med 2000; 133: 687–95.
18 Deshpande KS, Hatem C, Karwa M, Ulrich H, Aldrich TK, Kvetan
V. The use of inferior vena cava filter as a treatment modality for
massive pulmonary embolism. Respir Med 2002; 96: 984–9.
19 Elting LS, Escalante CP, Cooksley C, Avritscher EB, Kurtin D,
Hamblin L, Khosla SG, Rivera E. Outcomes and cost of deep venous
thrombosis among patients with cancer. Arch Intern Med 2004; 164:
1653–61.
20 Wallace MJ, Jean JL, Gupta S, Eapen GA, Johnson MM, Ahrar K,
Madoff DC, Morello FA, Murthy R, Hicks ME. Use of IVC filters
and survival in patients with malignancy. Cancer 2004; 101: 1902–7.
21 Jarrett BP, Dougherty MJ, Calligaro KD. Inferior vena cava filters in
malignant disease. J Vasc Surg 2002; 36: 704–7.
Inferior vena cava filters in anticoagulated patients 1853
? 2007 International Society on Thrombosis and Haemostasis