Introduction

Venous thromboembolism occurs in up to 40% of hospitalized medical and surgical patients in the absence of prophylactic anticoagulation [1, 2]. Even with prophylaxis, the risk of venous thromboembolism in critically ill patients approaches 10% and has serious consequences: untreated pulmonary embolism has a mortality rate approaching 25% [3–5]. Among critically ill patients, those developing venous thromboembolism have longer intensive care unit and hospital stays, longer duration of mechanical ventilation, and higher hospital mortality [6]. Consequently, venous thromboembolism not only is associated with serious morbidity and mortality, but also has major implications for healthcare resource utilization.

Appropriate use of prophylaxis to prevent venous thromboembolism in patients at risk has been identified as one of the most important patient-safety interventions for hospitals [7]. However, substantial variability is found in the use of such prophylaxis in practice. Prevention is most commonly achieved with anticoagulant drugs. Because important decisions about pharmacologic interventions are made with knowledge of their economic consequences, formal economic analyses are useful tools to guide clinicians and policy makers about the value of drug interventions and their consequences [8, 9]. However, many evaluations of new drugs do not occur prospectively with efficacy studies, and many are sponsored by the manufacturers, raising the possibility of bias.

We performed a systematic review of economic analyses of venous thromboembolism-prevention strategies in acutely ill hospitalized patients. Our objectives were to review and critically appraise the economic evaluations of a broad spectrum of strategies in diverse patient groups to help inform clinicians and policy makers about the cost-effectiveness of various approaches to venous thromboembolism prophylaxis.

Materials and methods

Date sources and searches

We searched MEDLINE, EMBASE, the Cochrane Database of Systematic Reviews, ACP Journal Club, Database of Abstracts of Reviews of Effects (DARE), and Cochrane Controlled Trials Register from 1946 to October 21, 2011, by using a combination of the following subject headings and text words: venous thrombosis, pulmonary embolism, low-molecular-weight heparin, LMWH, dalteparin, enoxaparin, nadroparin, tinzaparin, heparin, unfractionated heparin, UFH, anticoagulants, warfarin, aspirin, fondaparinux, rivaroxaban, dabigatran, intermittent pneumatic compression devices, compression stockings, vena cava filters, venous foot pump, economics, health care cost, cost-effectiveness analysis, cost-benefit analysis, and economic analysis (Additional file 1). No limits regarding publication type were initially applied. To identify additional potentially relevant studies, we checked the reference lists of identified systematic and narrative reviews and the personal files of the authors and collaborators. We also sent the full list of identified articles and inclusion criteria to venous thromboembolism experts in the field to identify additional published or relevant unpublished studies.

Study selection

From 5,180 potentially relevant citations, 4,816 were excluded based on title and abstract review (Figure 1). The full text versions of 89 manuscripts were retrieved for full evaluation. Two reviewers (ST, RF) independently assessed each of the articles and applied the following eligibility criteria: (a) the economic evaluation was based on data from randomized controlled trials or meta-analyses of randomized controlled trials; (b). the study described hospitalized patients; (c) the study compared at least two different venous thromboembolism prophylaxis strategies; (d) the study described drug-acquisition costs, the costs of providing prophylaxis, costs of complications (including venous thromboembolism treatment and prophylaxis failures); and (e) the study described the effect of prophylaxis with respect to the number of venous thromboembolism events prevented and diagnosed. We excluded evaluations based on the following study designs: 1. cohort studies or other observational studies; 2. studies on outpatient use of venous thromboembolism prophylaxis; 3. studies on the treatment of venous thromboembolism; 4. studies examining the efficacy of short-term versus long-term venous thromboembolism prophylaxis; 5. decision analytic models based on data from nonrandomized trials; 6. studies examining anticoagulants for conditions other than venous thromboembolism, and seven letters, editorials, or narrative reviews of economic issues in venous thromboembolism prophylaxis. We also excluded studies appraised as low to moderate quality, as defined later.

Figure 1
figure 1

Study eligibility diagram.

Full size image

Data extraction and quality assessment

We critically appraised each article by using established criteria [9]. Our goal was to include only those studies that adhered to a high methodologic quality. We assigned an ordinal score of quality based on the criteria set forth in the "User's guide to the medical literature: XIII. How to use an article on economic analysis of clinical practice" [9]. With a semiquantitative scale incorporating these characteristics, we assigned 1 point for each of the 12 categories. All studies were graded as high (≥ 9 points), medium (5 to 8 points), or low (0 to 4 points) quality by two authors (ST, RF). Disagreements about the inclusion of individual studies were resolved by consensus between two authors (ST, RF). Of the 84 articles selected for full review, 50 were appraised as low or moderate quality, and the remaining 39 studies were selected for data abstraction.

We identified 10 economic evaluations of mechanical prophylaxis, including intermittent pneumatic compression devices, compression stockings, or vena cava filters [10–17]. None of these articles met our previously stated eligibility criteria.

From each included study, we abstracted the following: the patient group, venous thromboembolism prophylaxis strategy, duration of prophylaxis, time frame of the study, source of the outcome data, source of the cost data, incremental costs and benefits of each strategy, results of any sensitivity analyses, the country in which the study was performed, and the declared source of funding for the economic analysis. We attempted to contact authors of studies for which no external support was declared to ensure that this was the case.

We abstracted the number of thrombotic events, costs, and complication rates of the prophylaxis and of the treatment of venous thromboembolism from each article. We then recorded or calculated the incremental cost-efficacy ratio for each venous thromboembolism prophylaxis strategy. If we were unable to calculate the incremental cost-efficacy ratio because of missing data, we attempted to contact the authors to obtain this ratio or original data. Costs were converted to 2009 US dollars and adjusted for country-specific temporal changes in gross domestic product [18–20]. We standardized the incremental effects we reported as "venous thromboembolism events avoided," "life-years or quality adjusted life-years gained' or 'deaths avoided per 1000 patients', as is commonly performed in venous thromboembolism literature. We chose in-hospital or near-term (< 90 days) events for the primary comparisons whenever possible, as the short and longer-term effects of inpatient thromboprophylaxis are often greatest during this time period.

Data synthesis and analysis

Heterogeneity of the interventions, perspectives, and time-horizons precluded meta-analytic techniques to combine incremental cost-efficacy ratios into a single summary statistic. We summarized cost-effectiveness ratios by graphic representation of point estimates on a cost-efficacy plane. Categoric variables and proportions were compared by using the χ2 or Exact tests as appropriate.

Results

Study comparisons, populations, and format

Among the 39 studies included in this review, the following comparisons were made: low-molecular-weight heparins versus placebo (five) [20–24]; unfractionated heparin versus low-molecular-weight heparins (12) [21, 23–33]; various low-molecular-weight heparins versus warfarin (eight) [34–41]; various low-molecular-weight heparins compared with one another or other agents (five) [31, 42–45]; fondaparinux versus enoxaparin (11) [46–56], rivaroxaban versus low-molecular-weight heparins or dabigatran [57], and dabigatran versus low-molecular-weight heparin [58] (Tables 1 and 2).

Table 1 Study characteristics
Full size table
Table 2 Study description and quality assessment
Full size table

Twenty-six evaluations were performed in orthopedic patients [25–27, 34–43, 46–58]; five in other surgical populations [28, 30–32, 44], and eight in medical patients [20–24, 29, 33, 45]. All 39 studies were either decision analytic models based on individual randomized controlled trials [20–22, 24, 26, 28, 31–49, 51–58] or meta-analysis [23, 27, 29, 30, 51].

Study perspectives, time horizon, and funding

The studies were conducted from a North American or European economic perspective: 18 of the studies were conducted in the United States [26, 29, 31–41, 45, 48, 51–53, 55], six in the United Kingdom [21, 23, 25, 46, 57, 58], four in Canada [23, 28, 34, 51], and three in Italy [27, 30, 44], two in Sweden [42, 47], and one each in Spain [43], Belgium [49], France [20], Norway [54], Switzerland [56], and Germany [24]. Seven studies were conducted from the perspective of the hospital [24, 41, 42, 44, 45, 49, 53], four from a societal perspective [20, 28, 30, 32], nine from the perspective of another specific payer [23, 31, 33, 40, 54–58], and the remaining 19 studies did not clearly specify which perspective was used.

The prophylaxis time horizons considered were variable: either for 5 days [33, 51]; 7 days [25, 28, 34, 38, 44, 46–50, 53–56]; 6 to 14 days [23]; 10 days [31, 32, 37, 40, 41, 52], 8 to 12 days [43]; 7 to 14 days [26]; 6 to 14 days [57], 10 to 14 days [20, 22–24, 27, 30, 35, 36, 39, 42–46]; 30 days [29], 8 to 33 days [59], and another for 8 weeks [45]. Twenty-five studies received some sponsorship by pharmaceutical companies [20, 22–25, 28, 30, 33–38, 40, 41, 43, 44, 46, 47, 50, 54–56, 58].

Study quality

All 39 studies that were of high quality were included in this systematic review (Table 2). Eighteen of these showed complete cost data but did not present complete effectiveness data [25–27, 29, 31, 37–45, 51, 54–56]. The remaining 17 studies had complete cost and effectiveness data. Six of the studies obtained effectiveness data from meta-analysis [23, 25, 27, 31, 50, 57], whereas the remainder obtained effectiveness data from one or more randomized controlled trials. Six of these studies obtained effectiveness data from a single trial [20, 22–24, 33, 41]. All studies, with the exception of one [23], obtained cost data from multiple sources, including actual and estimated healthcare system costs, randomized controlled trials, literature reviews, and other national government sources. All performed sensitivity analyses of some description.

Cost and effect estimates

Low Molecular Weight Heparins versus Placebo

Among the five studies comparing low-molecular-weight heparins with placebo in medical patients, enoxaparin was the most economically attractive strategy in all five studies and dominant in two [20, 22], with incremental cost-efficacy ratios ranging from $83 to $1,711 per venous thromboembolism event avoided in three others [21, 23, 24]; cost per life-year or quality-adjusted life-year gained were not investigated (Table 3). Sensitivity analysis did not alter these findings. Four of the five studies were sponsored by the manufacturer of enoxaparin [20–22, 24].

Table 3 Incremental costs, effects, and cost-efficacy ratios for the different modes of venous thromboembolism prophylaxis
Full size table

Unfractionated Heparin versus Low Molecular Weight Heparins

Among the 12 studies comparing low-molecular-weight heparins with unfractionated heparin among medical and surgical patients, 11 found that low-molecular-weight heparins were more effective (Table 3 and Figure 2). Eight of the 12 studies comparing low-molecular-weight heparins with unfractionated heparin found low-molecular-weight heparins to be the dominant strategy [21, 23–25, 27, 29, 30, 32, 33]. Two studies reported an incremental cost-efficacy ratio of $1,180 and $1,445 per venous thromboembolism event avoided when using enoxaparin [26, 29], and two studies found incremental cost-effectiveness ratios of $10,360 per death avoided and $20,337 per quality-adjusted life-year gained with low-molecular-weight heparins. A single study of venous thromboembolism prevention among patients undergoing colorectal cancer surgery found no difference in efficacy, yet costs of low-molecular-weight heparins were greater [28]. Three studies reported increased bleeding risk with low-molecular-weight heparins, and three studies reported lower risk. Sensitivity analyses did not change these results. Of these 12 studies, eight received financial support from the manufacturer of the low-molecular-weight heparins [23–26, 28–30, 33].

Figure 2
figure 2

Comparison of the incremental effects and costs of low-molecular-weight heparin versus unfractionated heparin for venous thromboembolism prophylaxis.

Full size image

Warfarin versus Low-Molecular-Weight Heparins

Low-molecular-weight heparins were reported to be economically more attractive than warfarin in all eight studies among surgical patients, with incremental cost-efficacy ratios of $874 to $26,711 per venous thromboembolism event avoided in five of the comparisons [34, 35, 37, 38, 40], the dominant strategy in three comparisons [36, 39, 41]. Long-term outcomes varied widely, with $16,200 to $334,055 per death avoided, $32,158 per life-year and $4,340 per quality-adjusted life-year gained (Table 3 and Figure 3) [34, 38]. Sensitivity analyses did not change the results in individual studies. Of these eight studies, seven received pharmaceutical sponsorship [34–38, 40, 41].

Figure 3
figure 3

Comparison of the incremental effects and costs of low-molecular-weight heparin versus warfarin for venous thromboembolism prophylaxis.

Full size image

Low-Molecular-Weight Heparins versus One another, and Other Comparisons

Within the studies comparing low-molecular-weight heparins with one another and with other anticoagulants among surgical patients, bemiparin and dermatan sulfate were the dominant prophylaxis over enoxaparin [43, 44]. Desirudin had an incremental cost-effectiveness ratio of $3,794 per life-year gained, whereas enoxaparin was favored over tinzaparin but was more expensive (Table 3) [45]. Dalteparin, 5,000 units once daily, was more efficacious than dalteparin, 2,500 units, with an incremental cost-effectiveness ratio of $24,357 per quality-adjusted life-year gained [30].

Fondaparinux versus Low Molecular Weight Heparins

Among the 11 studies comparing fondaparinux with enoxaparin, all were conducted in orthopedic surgery patients, and all concluded that fondaparinux was economically attractive. In six, fondaparinux was dominant [47, 51–53, 55, 56], and in one, enoxaparin [48] (Table 3 and Figure 4). In four studies, incremental cost-effectiveness ratios of fondaparinux over enoxaparin were $158 to $1,077 per venous thromboembolism event avoided, $104 to $6,782 per death avoided, and $32,144 per quality-adjusted life-year gained [46, 49, 53, 54]. In eight of 11 studies, fondaparinux was associated with increased bleeding risk. Sensitivity analyses of the various costs did not alter the findings. The manufacturer of fondaparinux provided sponsorship for six of the 11 studies.

Figure 4
figure 4

Comparison of the incremental effects and costs of low-molecular-weight heparin versus fondaparinux for venous thromboembolism prophylaxis.

Full size image

Dabigatran and Rivaroxaban versus Low Molecular Weight Heparins

Among orthopedic patients, dabigatran, in comparison with enoxaparin, was dominant, with cost savings of $103,050 and $8,162 and six and seven quality-adjusted life-years gained per 1,000 patients with a total hip replacement and total knee replacement, respectively [58]. Comparing rivaroxaban, dabigatran, and enoxaparin among orthopedic surgery patients, rivaroxaban was dominant, with cost savings of $24,104 and $213,452 and 7 life-years gained per 1,000 patients with a total hip replacement and total knee replacement, respectively [57].

Sponsorship and Economic Comparisons

When comparing different populations and the different modes of venous thromboembolism prophylaxis, we observed several interesting trends. All studies comparing fondaparinux, rivaroxaban, or dabigatran with enoxaparin were performed in orthopedic patients, and the remainder of the studies in this patient population examined various low-molecular-weight heparins or warfarin. Sixteen of the 25 studies among orthopedic patients were sponsored in some manner by the pharmaceutical industry: six studies favored fondaparinux [46, 47, 50, 54–56]; one, rivaroxaban [57]; one, dabigatran [58]; and the remainder favored low-molecular-weight heparins [25, 34–38, 40, 41, 43]. In comparison, five of the eight studies conducted in medical patients compared enoxaparin with placebo [20–24]; two compared unfractionated heparin with enoxaparin [29, 33]; and the final one compared enoxaparin with tinzaparin [45]. Five of these eight studies were sponsored by pharmaceutical companies, and all studies favored enoxaparin [20, 22–24]. Of the five studies in other surgical populations, three were sponsored by the pharmaceutical industry, and these studies favored unfractionated heparin [28], nadroparin [30], and dermatan sulfate [44].

Of the economic evaluations included in this review, 25 (64%) received funding by manufacturers of one of the comparators. The "new" agent within the comparison was deemed to have a favorable (dominant or attractive incremental cost-effectiveness ratio) outcome in 38 (97.4%) of the 39 economic evaluations (95% CI, 86.5% to 99.9%). Among the 25 studies funded by a pharmaceutical company, the sponsored medication was reported to be economically attractive in 24 (96.0%) (95% CI, 80.0% to 99.9%).

Discussion

In this systematic review of economic analyses of venous thromboembolism-prevention strategies in hospitalized patients, we found that all of the high-quality studies focused on pharmacologic prophylaxis. Low-molecular-weight heparins were the most commonly studied "new" class of medication and were generally reported to be the dominant or economically attractive strategy in comparison with placebo, unfractionated heparin, or warfarin, among medical and surgical patients. However, among orthopedic patients, fondaparinux was favored over low-molecular-weight heparins. The two newest agents, dabigatran and rivaroxaban, are favored in the two most recent orthopedic surgery evaluations included in this review. Among the various strategies compared, the new agent had a favorable cost-efficacy ratio in 97% of the studies, and most of these studies were sponsored by the manufacturers of the new agent.

Few systematic reviews of economic analyses comparing different modes of venous-thromboembolism prophylaxis in hospitalized patients have been conducted. In 1994, one such review of cost-efficacy analyses of low-molecular-weight heparins, unfractionated heparin, and warfarin in the prevention and treatment of venous thromboembolism in surgical patients, concluded that low-molecular-weight heparin was more efficacious and cost-effective after total hip arthroplasty [34]. Most recently, a trial comparing low-molecular-weight heparin with unfractionated heparin in hospitalized patients found low-molecular-weight heparin to be cost saving compared with unfractionated heparin and that low-molecular-weight heparin was associated with a lower venous thromboembolism readmission rate at 30 and 90 days [59]. Our review includes 37 subsequently published analyses, focusing on both medical and surgical patient populations, and comparing newer pharmacologic agents for venous thromboembolism prophylaxis, such as fondaparinux, dabigatran, and rivaroxaban.

In this review, low-molecular-weight heparins appeared to offer superior prophylaxis efficacy compared with warfarin, unfractionated heparin, and placebo for orthopedic, general surgical, and medical patients. Fondaparinux was found to be economically more attractive for venous thromboembolism prevention compared with heparins because of greater efficacy in surgical and orthopedic patients, but may also be associated with increased bleeding. Among 11 economic analyses comparing enoxaparin with fondaparinux, all found that fondaparinux was economically attractive. More than half of these studies were either directly sponsored by the manufacturers of fondaparinux, or were based on original randomized controlled trials funded by the manufacturer.

Historically, many economic evaluations of new drugs have been sponsored by the drug manufacturer. However, this introduces the potential for bias in model construction and interpretation of the results. In a retrospective analysis of 107 trials in five leading medical journals with regard to outcome and sources of funding, studies sponsored by pharmaceutical companies were much less likely to favor traditional therapy over new drug treatment [60, 61]. It is not surprising that new agents are incrementally efficacious; this is the nature of progress in medicine. However, new agents, typically still under patent protection, are virtually always substantially more expensive than comparator drugs. In our systematic review, 25 of the 39 studies were funded by pharmaceutical companies, and, with the exception of a single study [28], each of these found the sponsored drug more economically attractive than the comparator drug. Such consistency in incremental cost-effectiveness among more-expensive drugs is striking. Importantly, we could not detect a consistent bias in outcome between sponsored and nonsponsored evaluations; however, only a minority of evaluations did not receive sponsorship.

Strengths of our review include adherence to rigorous systematic review methods, which consisted of a comprehensive search strategy, broad eligibility criteria, and study selection by two independent adjudicators using a priori criteria to minimize selection bias. Economic analyses are susceptible to investigator bias, often due to retrospective decision-model generation and retrospective acquisition of cost-and-effect data. To reduce this risk, we included only economic evaluations that incorporated outcome data from prospective randomized controlled trials. We conducted data abstraction and critical appraisal in duplicate, by using established criteria for assessing economic evaluations. We also addressed the relation of recency to market and for-profit sponsorship in influencing the reporting of economic evaluations. This review also has limitations. Many of the analyses within studies that we included come from a limited number of trials and cost-comparison models. For example, five studies used outcome data from the MEDENOX trial [20, 22–24, 33]. If most of the data are derived from a limited number of efficacy trials and cost models, similar results are likely to be found across economic evaluations. Although the studies included in this review received high ratings of internal validity, studies varied widely with respect to patient population, time-horizon of therapy, and payer perspective, making generalizability to other health care difficult. In addition, many evaluations rely on radiologic as opposed to clinical venous thromboembolism detection, which may overestimate the real-life clinical consequences of venous thromboembolism. Side effects of thromboprophylaxis may be underestimated, as randomized controlled trials often exclude patients at higher risk of bleeding. Furthermore, trials are generally underpowered to detect differences in rare drug-specific complications such as heparin-induced thrombocytopenia. This may lead to an overestimation of cost-efficacy, as reported in the economic evaluations in this review. Finally, our review included a predominance of orthopedic, general surgery, and medical patients, and thus, our findings may not generalizable to other patient populations.

Among economic analyses in this review, incremental cost-effectiveness ratios were commonly expressed in costs per venous thromboembolism events avoided, and they ranged from $500 to $8,000 per venous thromboembolism event avoided. These ranges are difficult to interpret, as no firmly established willingness-to-pay benchmarks exist for venous thromboembolism prevention. Costs per life-year or quality-adjusted life-year gained were less commonly reported, making economic comparisons of venous thromboembolism-prevention strategies and other interventions in healthcare similarly challenging. Comparing and combining ICERs performed with country-specific costs is challenging, as patient, disease, provider, and health-care system factors may influence transferability. We have not adjusted costs based on country-specific purchasing power parity but have adjusted based on changes in gross domestic product over time, and country-specific exchange rates [62].

An informative economic analysis should include both benefits and harms of interventions and the full associated costs over a relevant time horizon. Full costs of venous thromboembolism prophylaxis were not included in some studies. The common complications of venous thromboembolism prophylaxis include prophylaxis failure, leading to thrombotic events, bleeding, and heparin-induced thrombocytopenia [1]. All 39 studies evaluated in this systematic review accounted for breakthrough thrombotic events; however, only half included bleeding complications (Table 3), and none fully accounted for heparin-induced thrombocytopenia. The omission of these potentially serious complications may considerably affect the cost-comparison data of the individual studies. Most studies ascertained costs retrospectively and from the literature. This is often less complete or less accurate compared with prospective determination alongside a randomized controlled trial. Finally, none of the studies included in this review was prospectively designed before results of the randomized controlled trials were published; accordingly, they may be at risk for subjective decision-tree construction and interpretation.

Few studies have evaluated the cost-effectiveness of mechanical venous thromboembolism-prevention strategies, and none of the existing studies met our eligibility criteria. The UK National Institute for Health and Clinical Excellence have recommended considering graduated compression stockings in most patients, although economic evaluations of mechanical venous thromboembolism prophylaxis have generally been of low quality [63]. The paucity of rigorous evidence about the cost effectiveness of mechanical prophylaxis is striking. However, this may be explained by the fact that manufacturers of mechanical devices are often not required to furnish either effectiveness or cost data to regulatory bodies before their introduction and marketing.

Conclusion

In this systematic review of economic analyses of venous thromboembolism-prevention strategies in hospitalized patients, we found that low-molecular-weight heparins appear to be the most economically attractive strategy for venous thromboembolism prevention among the majority of medical and surgical patients, whereas fondaparinux is more economically attractive for orthopedic patients. The studies, however, may be at risk of overestimating efficacy and underestimating side effects such as bleeding. Approximately two thirds of all evaluations were directly funded by the manufacturer of the new drug, and such drugs were more likely to be found economically attractive in comparison to other strategies. Limited opportunity for peer-reviewed and independent funding for economic evaluations unfortunately leads to reliance on industry sponsorship in this field. In the future, we recommend that high-quality, prospective, cost-effectiveness analysis be planned alongside the intervention trials and that these be designed, conducted, analyzed, and reported independent of industry sponsors.

Key messages

  • Low-molecular-weight heparins appear to be the most economically attractive strategy for venous thromboembolism prevention among the majority of medical and surgical patients, whereas fondaparinux is more economically attractive for orthopedic patients.

  • However, approximately two thirds of all evaluations were directly funded by the manufacturer of the new drug.

  • Such drugs were more likely to be found economically attractive in comparison to other strategies.

  • Limited opportunity for peer-reviewed and independent funding for economic evaluations may lead to reliance on industry sponsorship and bias in this field.

Authors' information

Dr. Fowler is a Career Scientist of the Ontario Ministry of Health and Long-term Care and an incoming Clinician Scientist of Heart and Stroke Foundation of Canada. Dr. Cook holds a Canada Research Chair from the Canadian Institutes of Health Research.