Cost-effectiveness of guidelines for insertion of inferior vena cava filters in high-risk trauma patients

From the American Venous Forum

Cost-effectiveness of guidelines for insertion of inferior vena cava filters in high-risk trauma patients Emily L. Spangler, MD,a Ellen D. Dillavou, MD,b and Kenneth J. Smith, MD,c Pittsburgh, Pa Background: Inferior vena cava filters (IVCFs) can prevent pulmonary embolism (PE); however, indications for use vary. The Eastern Association for the Surgery of Trauma (EAST) 2002 guidelines suggest prophylactic IVCF use in high-risk patients, but the American College of Chest Physicians (ACCP) 2008 guidelines do not. This analysis compares cost-effectiveness of prophylactic vs therapeutic retrievable IVCF placement in high-risk trauma patients. Methods: Markov modeling was used to determine incremental cost-effectiveness of these guidelines in dollars per quality-adjusted life-years (QALYs) during hospitalization and long-term follow-up. Our population was 46-year-old trauma patients at high risk for venous thromboembolism (VTE) by EAST criteria to whom either the EAST (prophylactic IVCF) or ACCP (no prophylactic IVCF) guidelines were applied. The analysis assumed the societal perspective over a lifetime. For base case and sensitivity analyses, probabilities and utilities were obtained from published literature and costs calculated from Centers for Medicare & Medicaid Services fee schedules, the Healthcare Cost & Utilization Project database, and Red Book wholesale drug prices for 2007. For data unavailable from the literature, similarities to other populations were used to make assumptions. Results: In base case analysis, prophylactic IVCFs were more costly ($37,700 vs $37,300) and less effective (by 0.139 QALYs) than therapeutic IVCFs. In sensitivity analysis, the EAST strategy of prophylactic filter placement would become the preferred strategy in individuals never having a filter, with either an annual probability of VTE of >9.6% (base case, 5.9%), or a very high annual probability of anticoagulation complications of >24.3% (base case, 2.5%). The EAST strategy would also be favored if the annual probability of venous insufficiency was <7.69% (base case, 13.9%) after filter removal or <1.90% with a retained filter (base case, 14.1%). In initial hospitalization only, EAST guidelines were more costly by $2988 and slightly more effective by .0008 QALY, resulting in an incremental cost-effectiveness ratio of $383,638/ QALY. Conclusions: Analysis suggests prophylactic IVC filters are not cost-effective in high-risk trauma patients. The magnitude of this result is primarily dependent on probabilities of long-term sequelae (venous thromboembolism, bleeding complications). Even in the initial hospitalization, however, prophylactic IVCF costs for the additional quality-adjusted life years gained did not justify use. ( J Vasc Surg 2010;52:1537-45.)

Prevention of venous thromboembolism (VTE) can be a difficult and complex issue in the trauma patient. Fatal pulmonary embolism (PE) may occur without warning, despite noninvasive venous screening and lack of clinical evidence of deep venous thrombosis (DVT). Immobilization devices may preclude mechanical prophylaxis, and ongoing bleeding or risk of bleeding may make anticoagulation contraindicated, at least early in the period, in up to 14% of trauma patients.1 Because of these factors, trauma surgeons increasingly have used inferior vena caval (IVC) From the University of Pittsburgh School of Medicine;a Department of Vascular Surgery, University of Pittsburgh Medical Center;b and Section of Decision Sciences, University of Pittsburgh School of Medicine.c Competition of interest: none. Oral presentation at the American Venous Forum, Amelia Island, Fla, Feb 13, 2010. Additional material for this article may be found online at www.jvascsurg.org. Correspondence: Emily Spangler, MD, Dartmouth-Hitchcock Medical Center Department of Vascular Surgery, 1 Medical Center Drive, Lebanon, NH 03756 (e-mail: [email protected]). The editors and reviewers of this article have no relevant financial relationships to disclose per the JVS policy that requires reviewers to decline review of any manuscript for which they may have a competition of interest. 0741-5214/$36.00 Copyright © 2010 by the Society for Vascular Surgery. doi:10.1016/j.jvs.2010.06.152

interruption as a form of PE prophylaxis. Patients at particularly high risk for PE immediately after a traumatic injury include those with spinal cord injury, pelvic fracture, lower extremity fracture, venous repair, injury severity score (ISS) ⬎9 or ⬎15, history of thromboembolism, coma (Glasgow Coma Scale [GCS] ⬍8), or age ⬎45 years with enforced bed rest for ⬎3 days.1-4 Two sets of guidelines provide conflicting recommendations for placement of IVC filters (IVCFs). The American College of Chest Physicians (ACCP) 2008 guidelines suggest IVCF insertion in patients with proven proximal DVT or PE who are unable to receive anticoagulation due to risk of bleeding, with anticoagulation added once the risk of bleeding resolves,5 but recommend against use of an IVCF as thromboprophylaxis in trauma patients.6 The Eastern Association for the Surgery of Trauma (EAST) 2002 guidelines, however, provide a level III recommendation that insertion of a prophylactic IVCF should be considered in very-high-risk trauma patients.7 These are patients unable to receive anticoagulation because of increased bleeding risk coupled with prolonged immobilization. Injuries fitting this pattern include severe closed head injury with GCS ⬍8, incomplete spinal cord injury with paraplegia or quadriplegia, complex pelvic fractures with 1537

1538 Spangler et al

Fig 1. Conceptual diagram of states modeled. Tree structure contains additional complexity for prophylactic and therapeutic anticoagulation as well as filter status (with/without during inpatient course and as an outpatient–never having a filter, successfully removed filter, or retained filter). VTE, Venous thromboembolism.

associated long-bone fractures, or multiple long-bone fractures.7 Given these conflicting recommendations regarding placement of IVCFs, we can examine the clinical and economic implications of each placement strategy and salient factors influencing the results by using Markov modeling. Here we examined the cost-effectiveness of prophylactic indications for IVCFs according to the EAST criteria vs traditional (therapeutic) indications for IVCFs according to AACP 2008 guidelines among high-risk trauma patients with acute contraindications to anticoagulation. METHODS Target population. Our hypothetic cohort was 46 years old and met EAST definitions for high-risk trauma patients: those who are unable to receive anticoagulation due to high risk for bleeding complications for 5 to 10 days after injury, including intracranial hemorrhage, ocular injury with associated hemorrhage, solid intra-abdominal organ injury, and/or pelvic or retroperitoneal hematoma requiring transfusion, with one or more of closed head injury with GCS ⬍8, spinal cord injury with paraplegia or quadriplegia, and complex pelvic fractures with associated long-bone fractures or multiple long-bone fractures.7,8 Model description. A Markov model with a 1-day cycle length was used to simulate hospital stay as well as long-term (post-discharge) complications over the lifetime of the patient. The target population began in an initial high-risk state for VTE while unable to be anticoagulated, and the model branched for either the EAST or ACCP guidelines for IVCF placement to be applied. A simplified schematic depiction of the inpatient portion of the model is shown in Fig 1, with the full tree structure available in Appendix A (online only). In the EAST branch, prophylactic filters were placed at a daily probability based on the literature’s average time to filter placement. During the inpatient stay, patients encountered daily probabilities of dying and of becoming symptomatic from a suspected DVT or PE (represented by the VTE state in the figure), leading to filter placement in the following cycle. Based on tree structure, all filters were placed after hospital day 1. The cohort cycled through the model, with or without a filter, either remaining asymptomatic or becoming symptomatic with a suspected PE or

JOURNAL OF VASCULAR SURGERY December 2010

DVT. Evaluation for suspected PE or DVT had a probability of positive imaging, which would result in anticoagulation (if not contraindicated), observation (if filter already placed and anticoagulation contraindicated), or placement of a filter (if one was not already inserted and anticoagulation remained contraindicated). As implemented, the ACCP strategy is identical to the EAST strategy except for having zero probability of prophylactic IVCF placement, such that IVCFs were placed only with a positive imaging study of PE or DVT when anticoagulation was contraindicated. Because the ACCP guidelines recommend instituting anticoagulation therapy once bleeding risk resolves,5,6 after a threshold number of days, the cohort could cycle through states in which individuals are able to receive prophylactic or therapeutic anticoagulation. While receiving prophylactic or therapeutic anticoagulation, individuals with and without filters again had probabilities of symptoms developing that required evaluation by imaging, or, if remaining asymptomatic, had a probability of hospital discharge. All patients with filters placed had a filter removal attempt before discharge. The daily probability of discharge, reflecting duration of hospital stay, was assumed to be primarily related to the injury mechanism and ability to safely anticoagulate the patient, rather than factors related to filter status; however, in the context of the model, one extra day of hospitalization was required to implant the filter, and one extra day was required for an attempt at filter retrieval. After discharge (not shown in Fig 1), individuals who never had a filter placed, had a successfully removed filter, or a retained filter, cycled through their own set of outpatient states in which daily age-specific mortality risks based on the 2004 U.S. life tables were encountered,9 modified by a slight survival advantage for those with retained filters. If surviving, an individual faced a daily probability of developing complications specific to never, currently, or formerly having had a filter. The initial mechanisms of injury putting individuals at very-high risk for VTE by EAST criteria7 were used to assume that all patients might continue to have sufficient mobility limitation to warrant an average of a 6-month course of anticoagulation after discharge. For individuals of each filter status, long-term complications encountered daily included VTE resulting in hospitalization for PE or DVT; if no VTE occurred within a cycle, a probability of a hemorrhage occurring during a 6-month post-hospital course of anticoagulation, as well as a probability of having venous insufficiency, were considered. Perspective, boundaries, time horizon. The societal perspective was used and included both direct and indirect costs, including lost wages in the long-term, but not acute, hospitalization states. The time horizon was the lifetime of the cohort, confined to a maximum age of 100 years. Probability data. Probability data were calculated from rates in the published vascular surgery and trauma literature. Because our modeling used a 1-day cycle length, probabilities and rates from the literature were converted to

JOURNAL OF VASCULAR SURGERY Volume 52, Number 6

1-day probabilities using the TreeAge (TreeAge Software Inc, Williamstown, Mass) commands ProbtoRate and RatetoProb to convert probabilities to rates and rates to probabilities over the appropriate time horizons which reflect the relation that p ⫽ 1-e–rt, where p represents probability, r, rate, and t, time. Using this relation, we also used median time to event data to determine daily probabilities. Table I reports detailed probabilities, parameters, and references. Costs and utilities. Costs were calculated from Centers for Medicare & Medicaid Services (CMS) fee schedules,41-43 the Healthcare Cost & Utilization Project (HCUP) database,40 and Red Book wholesale drug prices for 200744 (Table II). For data unavailable from the literature, assumptions were set forth based on similarities to other populations. Utilities, ranging from 0 (death) to 1 (perfect health), weight the time spent in a health state by the desirability of the health state to determine qualityadjusted life-years (QALY). Utilities were obtained from the published literature and the New England Medical Center Utility Search Database48 (Appendix B, online only). Future costs and utilities were discounted at 3%. Analysis. Base case cost-effectiveness analysis, determining the incremental cost-effectiveness ratio (ICER) of the EAST guidelines relative to the ACCP guidelines, was performed using TreeAge Pro 2008 software. One-way sensitivity analyses were also performed to test how robust the base case results were and determine influential variables. A conventional threshold for cost-effectiveness is an ICER of $50,000/QALY. We used the sensitivity analysis to determine which variables would change the preferred strategy from the base case, and at which threshold. The EAST strategy was considered the preferred filter placement strategy if one of the following occurred: ● ●

●

if the ICER of the EAST strategy relative to the ACCP strategy was ⬍$50,000/QALY; if the EAST strategy was less effective and less costly than the ACCP strategy, while the ICER of the ACCP strategy was ⬎$50,000/QALY; or if the ACCP strategy was dominated by the EAST strategy being less costly and more effective.

The ACCP strategy was preferable in the converse of the above cases. RESULTS Base case analysis. In base case analysis over the lifetime of a patient, the prophylactic IVCF (EAST) strategy was not cost-effective, being both more costly ($37,700 vs $37,300) and less effective by 0.139 QALYs (15.473 vs 15.612) than therapeutic IVCF placement. Sensitivity analysis. In one-way sensitivity analyses, the variables causing the biggest change in ICER magnitude were the probabilities of VTE after successful IVCF retrieval, and VTE with a retained IVCF. These were followed by the probabilities of anticoagulation-related bleeding events in individuals never having a filter or with a retained filter. The probability of inpatient discharge for

Spangler et al 1539

those with filters placed as well as the probability of VTE in individuals never having a filter placed also caused notable variation in the magnitude of the ICER (Fig 2). Although the above variables had a significant effect on the magnitude of the ICER, with the exception of the probability of inpatient discharge and probability of VTE in individuals never having a filter placed, the ACCP strategy of filter placement would remain preferred over the full range of clinically reasonable probabilities. The EAST strategy would be cost-effective, with an ICER of ⬍$50,000/ QALY at the probabilities of inpatient death or hospital discharge as listed in Table III. These sensitivity analysis results demonstrate that the EAST strategy of prophylactic filter placement would become cost-effective as the probability of inpatient death in patients without a filter increases, the probability of inpatient death in patients with a filter decreases, and as hospital stays become longer for those without a filter and shorter for those with a filter. Among the long-term probabilities of influence in the model, several could affect the preferred choice of strategy. In individuals never having a filter, an annual probability of VTE ⱖ9.6% (base case, 5.9%) would make the EAST strategy preferred, as would a very high annual probability of anticoagulation complications of ⱖ24.3% (base case, 2.5%). The EAST strategy would be favored among those with successfully removed filters if the annual probability of venous insufficiency were ⬍7.69% (base case, 13.9%) or ⬍1.90% among those with retained filters (base case, 14.1%). Owing to the influence of the probabilities of longterm VTE and other complications on the model, we also examined the cost-effectiveness of the EAST and ACCP guidelines over the more limited time horizon of only the hospitalization. In this scenario, the EAST guidelines were more costly ($16,518 vs $13,530) and more effective (by 0.007789 QALYs); however, the ICER of $383,638/ QALY greatly exceeded the commonly accepted threshold for practical cost-effectiveness of $50,000/QALY. By our above definitions, the ACCP guidelines would remain the preferred choice. It is also important to note several structural assumptions of our model that were shown not to affect the results in sensitivity analyses. In particular, the time to prophylactic filter placement may vary significantly by institution; however, even if all in the EAST cohort were required to receive a filter after day 1, the EAST strategy remains dominated, with a cost of $37,800 and effectiveness of 15.499 QALYs, whereas the ACCP strategy cost $37,300 with 15.612 QALYs. Similarly, our assumption of 6 months of anticoagulation therapy after the initial hospitalization for likely mobility limitation may vary by institution. We explored the role of extending the duration of anticoagulation by increasing the costs of anticoagulation (medication and surveillance testing) and increasing the duration of disutility from use of anticoagulation. Assuming these costs scale linearly, even increasing the duration of outpatient anticoagulation out to 5 years would not affect the model’s

JOURNAL OF VASCULAR SURGERY December 2010

1540 Spangler et al

Table 1. Probabilities and parameters Parameter Inpatient probabilities Weekly probability of death In initial high risk state Given no filter and receiving prophylactic anticoagulation Given no filter and receiving therapeutic anticoagulation Given filter placed With a filter while ineligible for anticoagulation With a filter receiving prophylactic anticoagulation With a filter receiving therapeutic anticoagulation Weekly probability of inpatient symptoms In initial high risk state Given no filter and receiving prophylactic anticoagulation During filter placement with complications At time of uncomplicated filter placement With a filter while ineligible for anticoagulation With a filter, receiving prophylactic anticoagulation With a filter, receiving therapeutic anticoagulation Daily probability of filter placement given no symptoms Daily probability eligible for prophylactic anticoagulation Threshold number of days for beginning anticoagulation Daily probability of discharge with filter Daily probability of hospital discharge given no filter

Base case (%)

Range (%)

0.24 1.47

0.15-1.47 0.00-5.00

10-12 11

0.06

0.06-1.47

11,13

0.84 0.07 1.47 1.47

0.0-5.00 0.06-1.32 0.0-5.00 0.0-5.00

14 15-17 11 Assumed same as prophylactic anticoagulation

2.41 1.58

0.0005-3.24 0.16-25.23

1.17 0.17

0.01-5.67 0.007-1.17

0.17 0.64

0.007-1.17 0.64-2.65

0.64

0.64-2.65

15.9 0 if days ⬍ threshold 1 if days ⱖ threshold 5

15.9-20.6

References

1,18,19 11,18,20 21-23 Assumed same as with filter while ineligible for anticoagulation 23-25 11 Assumed same as prophylactic anticoagulation 26,27

N/A 0-14

Assumption

4.8 5.6

3.6-7.4 4.0-9.4

Probability of positive imaging study in suspected PE/DVT Probability of a false negative on imaging study Probability of filter placement complication Probability of successful filter removal Outpatient probabilities Annual condition-specific decrease in mortality related to filter remaining in long-term [for those with filter] Annual probability of outpatient VTE Among those never having a filter Among those with successfully removed filters Among those with retained filters Conditional probability of PE given a VTE outpatient Among those never having a filter Among those with successfully removed filters Among those with retained filters Annual probability of bleed during 6 months of anticoagulation Among those never having a filter Among those with successfully removed filters

96.0

96.0-100.0

26,28,29 Computed from probability of discharge with filter 30

9.6 3.2 76.9

0.0-9.6 1.6-16.7 34.8-100

30 31-33 16,23,34

Among those with retained filters Annual probability of venous insufficiency Among those never having a filter Among those with a successfully removed filter Among those with retained filters Annual probability of ulceration given chronic venous insufficiency

0.7

0.005-0.7

Calculated from difference in death rates35

5.9 4.6 6.1

0.6-6.4 0.6-42.4 0.4-70.7

11,12, 36 23 14,25,32,35,37,38

33.7 15.4 12.6

7.5-83.4 1.1-24.8 0.5-15.4

11,12,36 23 14,25, 32,35,37,38

2.5 2.5

2.1-60.4 2.1-60.4

2.1

2.1-60.4

11,33 Assumed same as never having a filter 11

13.9 13.9

0.1-14.1 0.1-14.1

14.1 4.5

0.1-14.1 ...

35,37 Assumed same as never having a filter 32,35,37 39

JOURNAL OF VASCULAR SURGERY Volume 52, Number 6

Spangler et al 1541

Table 1. Continued Parameter

Base case (%)

Other parameters Time away from work during the 6 months of anticoagulation Time away from work per year for venous insufficiency Median length of stay in hospitalization for PE

Range (%)

References

2 hours

...

Assumption

1 hour

...

Assumption

5.0 days

5.0-6.3

Median length of stay in hospitalization for DVT

4.0 days

4.0-5.3

Median length of stay in hospitalization for GI bleed

3.0 days

2.0-3.5

40

(Upper age ranges only, ICD-9 code 415.19) 40 (Upper age ranges only, ICD-9 code 453.41) 40 (Upper age ranges only, DRG 175)

DRG, Diagnosis Related Group; DVT, deep vein thrombosis; GI, gastrointestinal; ICD, International Classification of Disease; PE, pulmonary embolism; VTE, venous thromboembolism.

Table II. Costs Parameter

Baseline ($)

Baseline cost of hospitalization

12,702.80

Daily cost of prophylactic anticoagulation Cost of initial day therapeutic anticoagulation Cost of subsequent days of therapeutic anticoagulation Cost of CT thorax Cost of filter placement Cost of filter placement complication Cost of filter removal Total cost of anticoagulation for 6 months Cost of hospitalization for PE

17.115

Range ($)

Reference

6979.00-23,653.00 40 0.78-36.37

44

169.14

4.89-439.64

42,44

143.51

7.10-513.16

42,44

230.56 1417.77 0

53.53-494.60 1110.08-1897.63 0-1417.77

41 41 Assumption

1612.08 219.14

1200.16-2295.42 106.15-465.99

41 44

11,004.77

7922.36-11,683.82 40,41

Cost of hospitalization for DVT

8300.91

5133.36-8900.61

40,41

Cost of hospitalization for GI bleed

6118.01

4489.66-6493.17

40,41

Daily cost of venous insufficiency Hourly private nonfarm worker wage Cost of death

0.87

0.29-2.02

17.50 5000.00

... ...

39,41,43,45

Description/Code(s) ICD-9 codes 344.1, 344.00, 432.9, 808.43 Average prophylactic cost of Lovenox or heparin use Heparin bolus, ½ day infusion, 1 day of testing PT/PTT q4hr, warfarin 1 day infusion, q24hr testing PT/PTT, warfarin HCPC 71260 HCPC 36010, 37620, 75940-26 Range: none to cost of placing filter again HCPC 37203,75961-26 9 venipunctures 9 PT checks, 180 days of warfarin 5 mg daily Evaluation and management, CT, PE hospitalization, anticoagulation Evaluation and management, CT, DVT hospitalization, anticoagulation Evaluation and management, EGD, GI bleed hospitalization, colonoscopy 2 new sets of compression stockings per year. 1 duplex every 5 years, annual office visit, annual cost of ulceration

46 Assumption 47

CT, Computed tomography; DVT, deep venous thrombosis; EGD, esophagogastroduodenoscopy; HCPC, Healthcare Common Procedure Coding; ICD, International Classification of Disease; GI, gastrointestinal; PE, pulmonary embolism; PT, prothrombin time; PTT, partial thromboplastin time.

preferred alternative, with the EAST guidelines ($39,900 for 15.431 QALYs) remaining dominated by the ACCP guidelines ($39,500 for 15.57 QALYs). Our proposed regimen had considered maintaining an international normalized ratio (INR) of 2.0 to 3.0 during anticoagulation because the probability of major bleeding events was derived from the Prevention du Risque d’Embolie Pulmonaire par Interruption Cave (PREPIC) trial,35 which had maintained this INR range; however, a lower-dose regimen could be considered and would primarily affect this probability. Considering the case of no warfarin use, the preferred outcome remained the ACCP strategy.

The risk of VTE in the trauma patient increases with age, but a numeric cutoff for the age at which the risk of VTE significantly increases has not been firmly established in epidemiologic analyses.7 The base case of age 46 was chosen to maintain some consistency with the cost-benefit analysis by Brasel et al8; however, in one-way sensitivity analyses varying the starting age, the ACCP guidelines remained the favored strategy throughout. For an 18-yearold, the EAST strategy would cost $43,838 with an effectiveness of 21.42 QALYs, whereas the ACCP guidelines would cost $44,275 with an effectiveness of 21.61 QALYs. The EAST strategy at this age is less costly and less effective,

JOURNAL OF VASCULAR SURGERY December 2010

1542 Spangler et al

Fig 2. A tornado diagram shows the magnitude of the incremental cost-effectiveness ratio over the range of reasonable literature values for the six variables displaying the greatest variation. The vertical yellow line represents the base case. ACCP, American College of Chest Physicians; EAST, Eastern Association for the Surgery of Trauma; GI, gastrointestinal; VTE, venous thromboembolism.

Table III. Sensitivity analysis

Parameter Weekly probability of death In initial high-risk state Given no filter and receiving prophylactic anticoagulation Given no filter and receiving therapeutic anticoagulation With a filter receiving prophylactic anticoagulation Daily probability of discharge with filter Daily probability of hospital discharge given no filter

Probability at which Base ICER is ⬍$50,000/ case (%) QALY (%) 0.24

ⱖ22.9

1.47

ⱖ1.9

0.06

ⱖ12.5

1.47

ⱕ1.1

4.8

ⱖ6.2

5.6

ⱕ4.4

ICER, Incremental cost-effectiveness ratio; QALY, quality-adjusted lifeyears.

with the ACCP strategy having an ICER of $2305. The gap in costs narrows with age, and by age 35, the EAST strategy is dominated, being more costly and remaining less effective: $30,621 yielding 18.31 QALYs for EAST vs $40,610 and 18.47 QALYs for ACCP. At age ⱖ35, the EAST strategy remains dominated by the ACCP strategy. Considering only the duration of the initial hospitalization, the result is unaffected by the starting age because such a short time period is considered in the analysis, with EAST costing $16,518 and yielding 0.03 QALYs, whereas the ACCP strategy is less costly and less effective at $13,530 yielding 0.02 QALYs, giving the EAST strategy an ICER of $383,638.

Although the analysis assumed outpatient age-specific mortality risks equivalent to the general U.S. population life tables (modified by a slight survival advantage for retained filters as derived from the PREPIC results35), trauma patients may have decreased life expectancy. Considering cases as extreme as adding a 50% annual probability of dying did not affect the preferred result, though the utilities of both strategies decreased significantly due to the decrease in life span. DISCUSSION Prior examinations of cost-effectiveness related to IVCFs have been limited in applicability to the trauma population,49 in scope or perspective,50-52 or in providing only a cost-benefit analysis.8 The cost-effectiveness results presented suggest that prophylactic placement of IVCFs in high-risk trauma patients is not cost-effective from a societal perspective. These results are theoretic and constrained by both the assumptions and structure of the model used to derive them and the quality and extent of evidence in the literature, especially regarding long-term complications. Within these constraints, however, we found the ACCP strategy of therapeutic placement of IVCF in high-risk trauma patients is most appropriate. The importance of our analysis is that it highlights the conflict between these practice guidelines and should serve as a springboard for further comparative research, both clinical and cost-effective. Although this project models a theoretic cohort and is not intended for individual clinical decision making, it illustrates the costs to society and the importance of the probabilities of complications to the individual in which the consequences may outweigh the

JOURNAL OF VASCULAR SURGERY Volume 52, Number 6

benefits of temporary IVC filter placement. An important aspect of this type of modeling is its ability to identify thresholds of variables that could sway decisions. The thresholds from sensitivity analyses can be viewed either as confidence limits on the results where the literature is not strong or, potentially, as goals for device refinement, quality improvement, or risk stratification in order for a given filter utilization strategy to become cost-effective. The aim of our model was to adequately characterize both inpatient and outpatient long-term complications while avoiding unnecessary complexity. For example, the ACCP recommends ultrasound screening in patients who are at high risk for VTE and have received suboptimal or no thromboprophylaxis,6 but these schedules are not standardized. From a modeling perspective, it would be difficult to account for the proportion of positive asymptomatic examinations and the associated imaging or therapeutic costs incurred. In addition, the ACCP recommends against ultrasound screening for asymptomatic DVT in trauma patients as a whole.6 We thus omitted ultrasound screening from our model because it seemed clinically reasonable. Similarly, for modeling purposes, suspected PE and DVT were lumped together without consideration of progression of DVT to PE because confirmed VTE is an AACP indication for filter use. Although imaging evaluation for DVT might be reasonably limited to ultrasound, because our model considered symptomatic DVT and PE, cost of a computed tomography thorax scan represented a reasonable estimate of an imaging workup that might be undertaken. Use and costs of pneumatic compression devices were considered to be a component of the inpatient hospitalization. For outpatients, costs and utilities of anticoagulation complications were primarily limited to evaluation and management of nonfatal gastrointestinal bleeding. Probabilities were drawn from published literature to make the model as applicable to clinical practice as possible. For many probabilities, however, there was no perfect source of information. For example, the probability of death without an IVCF when anticoagulation is contraindicated was drawn from historic controls in which it is not documented how many trauma patients might have received some anticoagulation.10 Similarly, there is a relative paucity of literature regarding long-term consequences of retrievable IVCFs because the Food and Drug Administration first granted approval for use in July 2003. This restricted our ability to model long-term complications of a retained IVCF. We therefore assumed that these individuals had probabilities equivalent to the literature for permanent IVCFs, and that those with successfully removed filters were similar to individuals who had never had a filter. We did not restrict our filter data to one brand of IVCF in order to make the results broadly applicable, although within the literature, different makes of IVCF have different complication rates. For retained IVCFs, most of the probabilities were determined from the 8-year cumulative results of the PREPIC study, the only long-term randomized study of filter placement in the prevention of pulmonary embo-

Spangler et al 1543

lism.11,35 However, the PREPIC participants differed from our cohort in several regards: they were not required to be trauma patients, only considered to be at high risk for PE and with evidence of an acute proximal DVT by venography with or without PE, and those with contraindications to anticoagulation were excluded. All patients in the PREPIC trial also received vitamin K antagonists for at least 3 months for a target INR of 2.0 to 3.0, although 35% of patients received the vitamin K antagonists during all 8 years of follow-up or until their death. Our model stipulated 6 months of warfarin therapy after hospital discharge for all individuals on the basis of the severity of the injuries classifying them at high risk by EAST criteria, which would lead to continued reductions in mobility in the initial months after discharge. A filter removal attempt was required within the structure of the model before discharge for any individual in whom a filter had been placed. The variable of success in this attempt in the model accounts for both the low rate of IVCF retrieval attempts in clinical practice as well as attempted but unsuccessful retrievals. Although this modeling strategy leaves a portion of each cohort with retained filters, it also may overcharge for filter removal. A review of the one-way sensitivity analysis indicates that this overestimation likely did not significantly alter the results, because the cost of filter removal was the 12th most important variable in the model, varying the ICER only $7515/ QALY over its range. Finally, after hospital discharge, patients were considered to have an age-adjusted mortality equivalent to that of the general population. A slight survival advantage (for PE-specific mortality) was given to those with retained filters based on the difference in death rates from the PREPIC study. However, we are aware that with the older population and increased number of cancer patients in the PREPIC study, this was another area in which assumptions were made to account for imperfect data sources. Sensitivity analysis of the survival advantage suggests a minimal effect on the outcome, varying the ICER by only $30/ QALY over its range. CONCLUSIONS Our analysis suggests prophylactic IVCFs are not costeffective. This result is influenced by probabilities of longterm sequelae (VTE, bleeding complications) that are poorly characterized in the literature due to the recent advent of retrievable filters and lack of long-term follow-up in patients with retained and removed filters. Consistent with the need to appropriately risk-stratify a potential recipient of a prophylactic IVCF, we found the EAST strategy could become preferred if the probability of inpatient death or longer hospital stays in those without a filter was sufficiently high, or the probability of these same events in those with a filter was sufficiently low. AUTHOR CONTRIBUTIONS Conception and design: ES, ED, KS Analysis and interpretation: ES, ED, KS

JOURNAL OF VASCULAR SURGERY December 2010

1544 Spangler et al

Data collection: ES Writing the article: ES, ED Critical revision of the article: ES, ED, KS Final approval of the article: ES, ED, KS Statistical analysis: ES, KS Obtained funding: Not applicable Overall responsibility: ES

20.

21.

22.

REFERENCES 1. Shackford SR, Davis JW, Hollingsworth-Fridlund P, Brewer NS, Hoyt DB, Mackersie RC. Venous thromboembolism in patients with major trauma. Am J Surg 1990;159:365-9. 2. Knudson MM, Lewis FR, Clinton A, Atkinson K, Megerman J. Prevention of venous thromboembolism in trauma patients. J Trauma 1994; 37:480-7. 3. Burns GA, Cohn SM, Frumento RJ, Degutis LC, Hammers L. Prospective ultrasound evaluation of venous thrombosis in high-risk trauma patients. J Trauma 1993;35:405-8. 4. Dennis JW, Menawat S, Von Thron J, Fallon Jr WF, Vinsant GO, Laneve LM, et al. Efficacy of deep venous thrombosis prophylaxis in trauma patients and identification of high-risk groups. J Trauma 1993; 35:132-8. 5. Kearon C, Kahn SR, Agnelli G, Goldhaber S, Raskob GE, Comerota AJ. Antithrombotic therapy for venous thromboembolic disease: American College of Chest Physicians evidence-based clinical practice guidelines (8th edition). Chest 2008;133(6 Suppl):454-545S. 6. Geerts WH, Bergqvist D, Pineo GF, Heit JA, Samama CM, Lassen MR, et al. Prevention of venous thromboembolism: American College of Chest Physicians evidence-based clinical practice guidelines (8th edition). Chest 2008;133(6 Suppl):381-453S. 7. Rogers FB, Cipolle MD, Velmahos G, Rozycki G, Luchette FA. Practice management guidelines for the prevention of venous thromboembolism in trauma patients: the EAST practice management guidelines work group. J Trauma 2002;53:142-64. 8. Brasel KJ, Borgstrom DC, Weigelt JA. Cost-effective prevention of pulmonary embolus in high-risk trauma patients. J Trauma 1997;42: 456-62. 9. Arias E. United States life tables, 2004. Nat Vital Stat Rep 2007;56. 10. Khansarinia S, Dennis JW, Veldenz HC, Butcher JL, Hartland L. Prophylactic Greenfield filter placement in selected high-risk trauma patients. J Vasc Surg 1995;22:231-6. 11. Decousus H, Leizorovicz A, Parent F, Page Y, Tardy B, Girard P, et al. A clinical trial of vena caval filters in the prevention of pulmonary embolism in patients with proximal deep-vein thrombosis. N Engl J Med 1998;338:409-16. 12. Spain DA, Richardson JD, Polk HCJ, Bergamini TM, Wilson MA, Miller FB. Venous thromboembolism in the high-risk trauma patient: do risks justify aggressive screening and prophylaxis? J Trauma 1997; 42:463-9. 13. Golob Jr JF Sando MJ, Kan JC, Yowler CJ, Malangoni MA, Claridge JA. Therapeutic anticoagulation in the trauma patient: is it safe? Surgery 2008;144:591-7. 14. Kinney TB. Update on inferior vena cava filters. J Vasc Interv Radiol 2003;14:425-40. 15. Rosenthal D, Wellons ED, Hancock SM, Burkett AB. Retrievability of the Günther Tulip vena cava filter after dwell times longer than 180 days in patients with multiple trauma. J Endovasc Ther 2007;14:406-10. 16. Kirilcuk NN, Herget EJ, Dicker RA, Spain DA, Hellinger JC, Brundage SI. Are temporary inferior vena cava filters really temporary? Am J Surg 2005;190:864-70. 17. Neuerburg J, Günther R, Vorwerk D, Dondelinger RF, Jäger H, Lackner KJ, et al. Results of a multicenter study of the retrievable Tulip vena cava filter: early clinical experience. Cardiovasc Intervent Radiol 1997;20:10-6. 18. Knudson MM, Collins JA, Goodman SB, McCrory DW. Thromboembolism following multiple trauma. J Trauma 1992;32:2-11. 19. Knudson MM, Ikossi DG, Khaw L, Morabito D, Speetzen LS. Thromboembolism after trauma: an analysis of 1602 episodes from the Amer-

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33. 34.

35.

36.

37.

38. 39.

40.

ican College of Surgeons National Trauma Data Bank. Ann Surg 2004;240:490-8. Geerts WH, Jay RM, Code KI, Chen E, Szalai JP, Saibil EA, et al. A Comparison of low-dose heparin with low-molecular-weight heparin as prophylaxis against venous thromboembolism after major trauma. N Engl J Med 1996;335:701-7. Linsenmaier U, Rieger J, Schenk F, Rock C, Mangel E, Pfeifer K. Indications, management, and complications of temporary inferior vena cava filters. Cardiovasc Intervent Radiol 1998;21:464-9. Rogers FB, Strindberg G, Shackford SR, Osler TM, Morris CS, Ricci MA, et al. Five-year follow-up of prophylactic vena cava filters in high-risk trauma patients. Arch Surg 1998;133:406-12. Meier C, Keller IS, Pfiffner R, Labler L, Trentz O, Pfammatter T. Early experience with the retrievable OptEase vena cava filter in high-risk trauma patients. Eur J Vasc Endovasc Surg 2006;32:589-95. Patton JHJ, Fabian TC, Croce MA, Minard G, Pritchard FE, Kudsk KA. Prophylactic Greenfield filters: acute complications and long-term follow-up. J Trauma 1996;41:231-7. Winchell RJ, Hoyt DB, Walsh JC, Simons RK, Eastman A. Risk factors associated with pulmonary embolism despite routine prophylaxis: implications for improved protection. J Trauma 1994;37:600-6. Offner PJ, Hawkes A, Madayag R, Seale F, Maines C. The role of temporary inferior vena cava filters in critically ill surgical patients. Arch Surg 2003;138:591-5. Hoff WS, Hoey BA, Wainwright GA, Reed JF, Ball DS, Ringold M, et al. Early experience with retrievable inferior vena cava filters in high-risk trauma patients. J Am Coll Surg 2004;199:869-74. Morris CS, Rogers FB, Najarian KE, Bhave AD, Shackford SR. Current trends in vena caval filtration with the introduction of a retrievable filter at a level I trauma center. J Trauma 2004;57:32-6. Millward SF, Oliva VL, Bell SD, Valenti DA, Rasuli P, Asch M, et al. Gunther Tulip retrievable vena cava filter: results from the registry of the Canadian Interventional Radiology Association. J Vasc Intervent Radiol 2001;12:1053-8. Schultz DJ, Brasel KJ, Washington L, Goodman LR, Quickel RR, Lipchick RJ, et al. Incidence of asymptomatic pulmonary embolism in moderately to severely injured trauma patients. J Trauma 2004;56: 727-33. Rosenthal D, Wellons ED, Levitt AB, Shuler FW, O’Conner RE, Henderson VJ. Role of prophylactic temporary inferior vena cava filters placed at the ICU bedside under intravascular ultrasound guidance in patients with multiple trauma. J Vasc Surg 2004;40:958-64. Langan III EM, Miller RS, Casey III WJ, Carsten III CG, Graham RM, Taylor SM. Prophylactic inferior vena cava filters in trauma patients at high risk: follow-up examination and risk/benefit assessment. J Vasc Surg 1999;30:484-90. Streiff MB. Vena caval filters: a comprehensive review. Blood 2000;95: 3669-77. Lam RC, Bush RL, Lin PH, Lumsden AB. Early technical and clinical results with retrievable inferior vena caval filters. Vascular 2004;12: 233-7. The 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 Interruption Cave) randomized study. Circulation 2005;112:416-22. McKinley WO, Jackson AB, Cardenas DD, DeVivo MJ. Long-term medical complications after traumatic spinal cord injury: a regional model systems analysis. Arch Phys Med Rehabil 1999;80:1402-10. McMurtry AL, Owings JT, Anderson JT, Battistella FD, Gosselin R. Increased use of prophylactic vena cava filters in trauma patients failed to decrease overall incidence of pulmonary embolism. J Am Coll Surg 1999;189:314-20. Greenfield LJ, Proctor MC. Recurrent thromboembolism in patients with vena cava filters. J Vasc Surg 2001;33:510-4. Labropoulos N, Waggoner T, Sammis W, Samali S, Pappas PJ. The effect of venous thrombus location and extent on the development of post-thrombotic signs and symptoms. J Vasc Surg 2008;48:407-12. Agency for Healthcare Research and Quality. HCUPnet: national and regional estimates on hospital use for all patients from the HCUP

JOURNAL OF VASCULAR SURGERY Volume 52, Number 6

41.

42.

43.

44. 45.

46.

47.

Nationwide Inpatient Sample (NIS) 2007; http://hcupnet.ahrq.gov/. Accessed: Jun 25, 2009. Centers for Medicare & Medicaid Services. Physician fee schedule look-up 2007; http://www.cms.hhs.gov/PfsLookup/. Accessed: Jun 30, 2009. Centers for Medicare & Medicaid Services 2007 Clinical diagnostic laboratory fee schedule 2007. http://www.cms.hhs.gov/Clinical LabFeeSched/, http://www.cms.hhs.gov/apps/ama/license.asp?file⫽/ ClinicalLabFeeSched/downloads/07clab.zip. Accessed: Jun 30, 2009. Centers for Medicare & Medicaid Services. Durable medical equipment, prosthetics/orthotics & supplies fee schedules 2007; http:// www.cms.hhs.gov/DMEPOSFeeSched/downloads/d07_jan.zip. Accessed: Jun 30, 2009. Murray L. Red Book 2007: Pharmacy’s fundamental reference (red book drug topics). Montvale, NJ: Thomson Healthcare; 2007. Olin JW, Beusterien KM, Childs MB, Seavey C, McHugh L, Griffiths RI. Medical costs of treating venous stasis ulcers: evidence from a retrospective cohort study. Vasc Med 1999;4:1-7. Bureau of Labor Statistics. Real earnings in December 2007 2008. http://www.bls.gov/news.release/History/realer.txt. Accessed: Jan 29, 2010. Aujesky D, Smith KJ, Cornuz J, Roberts MS. Cost-effectiveness of low-molecular-weight heparin for secondary prophylaxis of cancerrelated venous thromboembolism. Thromb Haemost 2005;93:592-9.

Spangler et al 1545

48. Center for the Evaluation of Value and Risk in Health. The costeffectiveness analysis registry 2009. https://research.tufts-nemc.org/ cear/default.aspx. 49. Chau Q, Cantor SB, Caramel E, Hicks M, Kurtin D, Grover T, et al. Cost-effectiveness of the bird’s nest filter for preventing pulmonary embolism among patients with malignant brain tumors and deep venous thrombosis of the lower extremities. Support Care Cancer 2003; 11:795-9. 50. d’Othée BJ, Faintuch S, Reedy AW, Nickerson CF, Rosen MP. Retrievable versus permanent caval filter procedures: when are they costeffective for interventional radiology? J Vasc Intervent Radiol 2008;19: 384-92. 51. Gonzalez RP, Cohen M, Bosarge P, Ryan J, Rodning C. Prophylactic inferior vena cava filter insertion for trauma: intensive care unit versus operating room. Am Surg 2006;72:213-6. 52. Nunn CR, Neuzil D, Naslund T, Bass JG, Jenkins JM, Pierce R, et al. Cost-effective method for bedside insertion of vena caval filters in trauma patients. J Trauma 1997;43:752-8.

Submitted Mar 11, 2010; accepted Jun 7, 2010.

Additional material for this article may be found online at www.jvascsurg.org.

1545.e1 Spangler et al

JOURNAL OF VASCULAR SURGERY December 2010

Appendix A (online only). Conceptual diagram of states modeled. Tree structure contains additional complexity for prophylactic and therapeutic anticoagulation as well as filter status (with/without during inpatient course and as an outpatient—never having a filter, successfully removed filter, or retained filter).

JOURNAL OF VASCULAR SURGERY Volume 52, Number 6

Spangler et al 1545.e2

Appendix B (online only). Utilities State Initial high risk state (including hospitalization, trauma) Not eligible for prophylactic anticoagulation, asymptomatic, with filter in place DVT or PE symptoms Anticoagulation with warfarin Hospital prophylactic or therapeutic anticoagulation Disutility of filter placement Filter placement complication (insertion site thrombosis) Disutility of filter removal Post-hospital discharge Age-based utility table Hospitalization for PE Hospitalization DVT Hospitalization for (GI) bleed while on anticoagulation Venous insufficiency

Baseline

Range

0.43

0.27-0.80

0.43

0.27-0.80

0.80

0.60-0.93

0.99 0.99

0.92-1.00 0.94-1.00

⫺0.02 0.84

— 0.84-0.931

⫺0.02

—

0.96

—

References Median utility for paraplegia (0.43)1,2 Assumption Average of PE utility (.76) and DVT utility (.84)3-5 6,7 6

0.76 0.84 0.73

0.60-0.89 0.84-0.931 0.47-0.92

8 Assume utility of insertion site thrombosis similar to that of a DVT 3,4 Assume equal to disutility of filter placement8 Utility of trauma, 3 yrs after treatment in ICU 2 Medians of males and females averaged by age group1 3-5 3,4 1

0.93

0.63-0.98

1,3,9,10

table

—

DVT, deep venous thrombosis; ICU, intensive care unit; GI, gastrointestinal; PE, pulmonary embolism.

APPENDIX B REFERENCES 1. Gold MR, Franks P, McCoy KI, Fryback DG. Toward consistency in cost-utility analyses: using national measures to create conditionspecific values. Med Care 1998;36:778-92. 2. Tengs TO, Wallace A. One thousand health-related quality-of-life estimates. Med Care 2000;38:583-637. 3. Haentjens P, De Groote K, Annemans L. Prolonged enoxaparin therapy to prevent venous thromboembolism after primary hip or knee replacement. A cost-utility analysis. Arch Orthop Trauma Surg 2004;124:507-17. 4. Duriseti RS, Shachter RD, Brandeau ML. Value of quantitative D-dimer assays in identifying pulmonary embolism: implications from a sequential decision model. Acad Emerg Med 2006;13: 755-66. 5. Perlroth DJ, Sanders GD, Gould MK. Effectiveness and cost-effectiveness of thrombolysis in submassive pulmonary embolism. Arch Intern Med 2007;167:74-80. 6. Aujesky D, Smith KJ, Cornuz J, Roberts MS. Cost-effectiveness of low-molecular-weight heparin for secondary prophylaxis of cancer-

7.

8.

9.

10.

related venous thromboembolism. Thromb Haemost 2005;93: 592-9. Sarasin FP, Eckman MH. Management and prevention of thromboembolic events in patients with cancer-related hypercoagulable states: a risky business. J Gen Intern Med 1993;8:476-86. Aujesky D, Smith KJ, Cornuz J, Roberts MS. Cost-effectiveness of low-molecular-weight heparin for treatment of pulmonary embolism. Chest 2005;128:1601-10. Botteman MF, Caprini J, Stephens JM, Nadipelli V, Bell CF, Pashos CL, et al. Results of an economic model to assess the cost-effectiveness of enoxaparin, a low-molecular-weight heparin, versus warfarin for the prophylaxis of deep vein thrombosis and associated long-term complications in total hip replacement surgery in the United States. Clin Ther 2002;24:1960-86. Gould MK, Dembitzer AD, Sanders GD, Garber AM. Low-molecularweight heparins compared with unfractionated heparin for treatment of acute deep venous thrombosis: a cost-effectiveness analysis. Ann Intern Med 1999;130:789-99.