OUP user menu

Cost‐effectiveness of screening and extended anticoagulation for carriers of both factor V Leiden and prothrombin G20210A

M. Marchetti, S. Quaglini, G. Barosi
DOI: http://dx.doi.org/10.1093/qjmed/94.7.365 365-372 First published online: 1 July 2001


Carriers of a double thrombophilic mutation (factor V Leiden and prothrombin G20210A) are at high risk of a recurrent venous thromboembolism (VTE), and may benefit from a longer course of secondary prophylaxis. We examined the costs and health benefits of screening for both the mutations, provided that double heterozygotes undergo 2 years of anticoagulation as compared to the standard 6 months. We thus pooled the available evidence and calculated that the OR for recurrence in double heterozygotes was 5.9 (95% CI 2.65–13.20). A Markov model tracked patients’ health lifelong, and calculated that prolonged prophylaxis saved 26 quality‐adjusted days of life and $410 per double heterozygote treated. Screening all the patients with venous thromboembolism thus provided one additional day of life at the cost of 13624 $/QALY (95% CI 12 965–22 889). Screening was not cost‐effective in those cohorts with a low prevalence of the mutations, a high bleeding risk or in those where prophylaxis prevented <65% of recurrences. Screening for factor V Leiden and prothrombin G20210A, with prolonged prophylaxis of double carriers, is cost‐effective in most patients with VTE.


Deep vein thrombosis (DVT) is a chronic multifactorial disease with a high rate of recurrence that depends on the underlying inherited or acquired thrombophilic factors.1 Mutated factor V, called factor V Leiden, is the most prevalent genetic thrombogenic stimulus,2 but data on the role of factor V Leiden as an independent risk factor of recurrent DVT are conflicting.3,,4 The G20210A prothrombin mutation, the second most prevalent prothrombotic mutation, does not significantly increase the recurrence rate of DVT after a first episode.5,,6 Nevertheless, in two recent cohort studies, double heterozygotes for factor V Leiden and the G20210A prothrombin mutation were at increased risk of recurrent DVT.7,,8

In this decision analysis, we compared standard anticoagulant prophylaxis to the strategy of screening for factor V Leiden and prothrombin G20210A and extending anticoagulation only for double heterozygotes for these mutations. We evaluated the impact of these two strategies on both quality‐adjusted life expectancy and lifelong costs.



English‐written articles dealing with recurrent VTE were obtained from a computerized search of Medline 1966–2000 (July). In the MESH field, the following query was run: ‘Venous Thrombosis’ AND ‘Factor V*’ AND ‘Prothrombin*’ AND ‘Recurrence’. A full search of the journals dealing with thrombosis was also conducted: issues published after 1990 were reviewed in Thrombosis and Hemostasis, Blood Coagulation and Fibrinolysis, and Haemostasis. Indexed studies were eligible for meta‐analysis if they were English‐written papers, reported a complete 2×2 table, included more than 10 double heterozygotes and followed the patients retrospectively or prospectively for at least 2 years. The MEDLINE query and journals review identified 15 studies, but only two were eligible for meta‐analysis.7,,8

Under the assumption of a fixed‐effect model, test Q of homogeneity of outcomes among studies was performed. To combine study results, a common odds ratio (OR) was estimated for each outcome using Peto's method, since it provided wider confidence intervals. For calculation purposes we used Review Manager 4.1 (Cochrane Collaboration, freeware, downloadable at www.cochrane.nl). The results of the analysis can be found in Table 1.

View this table:
Table 1

Meta‐analysis of recurrent venous thromboembolism in carriers of factor V Leiden and prothrombin G20210A mutations: ORs calculated according to Peto's method

Decision model

We developed a computerized decision model (DATA software for Windows, TreeAge Software) that compared two intervention strategies for patients with a first episode of DVT. The first strategy provided patients with standard care, namely stopping anticoagulation 6 months after DVT. The second strategy consisted of screening for factor V Leiden and G20210A prothrombin mutation, followed by 2 years on warfarin for those carrying both the defects.9 All the patients incurring a recurrent thromboembolic episode underwent 2‐year‐long secondary prophylaxis with warfarin.

We used a Markov simulation model10 in which a hypothetical cohort of patients moved from one health state to another at 6‐month intervals until all of them had died (Figure 1). Point probability estimates and plausible ranges for model variables were obtained by surveying published reports (Table 2). Both life expectancy and costs were discounted at a 3% yearly rate, according to common guidelines for health‐care economic evaluation.11

The baseline cohort consisted of males aged 60 years, i.e. the median age of patients with a first DVT. Age‐ and sex‐adjusted mortality was derived from Italian Life Tables (ISTAT, Statistiche della Sanità—Anno 1994).

Figure 1.

Simplified diagram of the Markov decision‐analysis model. The square at the left represents the decision to follow one of the treatment strategies. The M inside the circle indicates the Markov process, which leads to one of several health states. Plain circles represent chance events that may occur during each cycle and result in continuing the current state or shifting to a permanent disabling event or death. Natural death is not depicted. PTS, post‐thrombophlebitis syndrome. DVT, deep vein thrombosis.

View this table:
Table 2

Data used in the model

View this table:
Table 3.

Results of base‐case analysis

Clinical data

The recurrence rate of DVT was derived from the ‘Duration of Anticoagulation Study Group’ data.12,,13 Patients on warfarin have an annual risk of major bleeding that increases with age and ranges from 0.24% to 2.9%, with 1.35% being the risk at the age of 60 years.14,,15 Incidence rate for heparin‐related bleedings was from a review article.16 The rate of pulmonary embolism was calculated as a fixed portion of the overall DVT rate17 and case fatality was that reported by an international registry.18

Efficacy of prophylaxis was that reported by the East Anglian Thrombophilia Study Group in a retrospective cohort19 and by a recent randomized trial.9 We assumed that post‐thrombophlebitis syndrome (PTS) did not appear later than 5 years after an episode of DVT and used the time‐dependent incidence rate reported by a 8‐year‐long prospective study.20 All patients experiencing recurrent DVT were calculated to develop PTS within 3 years of recurrence.20

Quality of life estimates were those reported in the literature.21,,22 Quality of life for combined health states was obtained by multiplying the utilities of the single combined health states.

Economic data

We estimated a $60 cost of the sequential test for factor V Leiden (second‐generation APC‐resistance test) and G20210A prothrombin (PCR‐based molecular performed only in those individuals carrying factor V Leiden), according to local laboratory practice. The costs induced by the modelled health conditions were derived from Italian list prices and charges of the Italian Health System or economic analyses conducted in Italy. Since Italian estimates were lacking, we calculated the post‐discharge direct costs of patients after a haemorrhagic stroke from Spanish estimates by using the relative medical cost index.23,,24 Since a societal perspective was followed, in addition to direct costs, we also considered indirect costs calculated as lost income caused by absence from work.25 We assumed a total loss of productivity during hospital stay and for 1 month afterwards. Total impairment for 6 months after a haemorrhagic stroke and lifelong inability to work for one third of the patients were assumed. We did not account for loss of productivity as a result of premature death. Patients on warfarin also undergo chronic morbidity due to minor bleeding events, which occur at a 6% yearly rate.13 These events do not require hospitalization, yet each one causes an estimated one‐week home rest. The annual income used for calculations was the average Italian income of $21 000. Indirect costs of severe PTS were assumed to be $913 as estimated by the San Valentino Venous Disease Project.26

Cost and charges were converted to US dollars: the exchange rate used was L2000 to $1. Benchmark cost‐effectiveness was $50 000 per QALY saved.


Base‐case analysis of life expectancy and costs

Under the assumptions of our baseline analysis, 60‐year‐old patients with a first DVT and carrying both factor V Leiden and prothrombin mutation A20210G can expect a shorter life expectancy than controls. Prolonged anticoagulation saved 52% of the life potentially lost due to recurrent thromboembolism, that is 26 quality‐adjusted days (QALDs). The overall screened cohort therefore had its life expectancy prolonged of 0.003 QALYs, equivalent to 1 QALD.

Lifelong treatment cost of the entire unscreened cohort was $3000. Eighty‐six percent of the costs were directly related to recurrent thromboembolism, 2% to oral anticoagulation therapy, 2% to PTS and 10% to major bleeding. Indirect costs represented 29% of overall costs (Table 2 ). The screening strategy, on the contrary, cost $3040 per patient, $40 more than standard therapy.

From a utilitarian perspective, screening of factor V Leiden and prothrombin G20210A and planning 2‐year selective prophylaxis for a cohort of 1000 individuals prevented 6 thromboembolic events and 1.1 pulmonary embolisms, and induced 0.9 major bleedings due to prolonged warfarin therapy.

Further, quality of life issues were also relevant. Screened patients incurred on average a 21 days longer exposure to oral anticoagulants than unscreened ones, but avoided 15 days of oral anticoagulants following thromboembolic recurrences. Thus screened patients lost 0.09 QALDs (7 days of exposure×0.013 disutility of taking oral anticoagulants) due to exposure to oral anticoagulants. Screened patients, however, avoided 54 days with severe post‐thrombophilic syndrome and consequently gained further 0.27 QALDs, with a neat gain of 0.18 QALDs.

Sensitivity analyses

We tested the model across a wide spectrum of rates to identify a threshold at which effectiveness and cost‐effectiveness remained favourable to screening. One‐way sensitivity analysis showed the four critical variables of the decision model: prevalence of double heterozygotes, bleeding rate and efficacy of prophylaxis with warfarin, mortality of pulmonary embolism. (Figure 2). Threshold analysis further calculated that screening cost more than $50 000/QALY in patients with expected bleeding rates higher than 1.6% in 6 months or an expected prevalence lower than 1.4%. Thus, screening was not indicated for those patients with VTE who are at high bleeding risk while on warfarin, or who have a very low prevalence of the double heterozygote state for factor V Leiden and prothrombin G20210A. At low mortality rates of pulmonary embolism (i.e.<10%) screening and prolonged prophylaxis cost more than $50 000 per QALY saved since the health benefit of averted VTE was greatly reduced.

In those patients who are poorly compliant to anticoagulation or cannot be monitored by a specialized center, prophylaxis is expected to have a low efficacy, but at efficacy rates <65%, screening for double heterozygotes was no longer cost‐effective. However, the cost of screening tests can be decreased by a high number of exams, new techniques or market policies and this could be important: had the double screening test cost $43 or lower, the whole preventive strategy would become cost‐saving.

Two‐way sensitivity analysis identified the classes of patients that might be targeted for cost‐effective screening (Figure 3). Individuals at high risk of bleeding but also a high prevalence of factor V Leiden and prothrombin G20210A mutation could still benefit from screening, and prolonged prophylaxis at a reasonable cost. In the same way, screening was cost‐effective in high‐prevalence populations even when efficacy rates of prophylaxis were expected to be low (Figure 4). On the other hand, populations at moderate and low prevalence need to be provided screening only if a high efficacy is forecasted from anticoagulation.

Figure 2.

One‐way sensitivity analysis: tornado diagram. Cost‐effectiveness of screening over non‐screening was >50 000 $/QALY for patients at high risk of bleeding (six‐month rate >1.6%) and those populations with a low prevalence of double heterozygotes (prevalence <1.4% in individuals with VTE).

Figure 3.

Two‐way sensitivity analysis. The variables analysed were: 6‐month rate of major bleeding in patients on warfarin; and the prevalence of double heterozygotes for factor V Leiden and prothrombin G20210A. Grey area indicates those cohorts for whom screening and prolonged prophylaxis was cost‐effective, i.e. cost <50 000 $/QALY. Dotted lines indicate isocontours for cost‐effectiveness thresholds of 20 000 and 100 000 $/QALY.

Figure 4.

Two‐way sensitivity analysis. The variables analysed were: efficacy of warfarin in preventing recurrences; and the prevalence of double heterozygotes for factor V Leiden and prothrombin G20210A. Grey area indicates those cohorts for whom screening and prolonged prophylaxis was be cost‐effective, i.e. cost <50 000 $/QALY. Dotted lines indicate isocontours for cost‐effectiveness thresholds of 20 000 and 100 000 $/QALY.


DVT has an estimated annual incidence of 16 per 10 000 white adults27 and carries major medical, social and economic consequences. Current therapy practice involves treating with anticoagulants for a period of 6 months after a thromboembolic episode. However, recent intervention trials documented extended anticoagulation beyond the standard 6 months is effective in preventing recurrent venous thromboembolism9 and this might apply also to double heterozygotes for factor V Leiden and G20210A prothrombin mutations.

In the present study, a cost‐effectiveness methodology was thus used to assess screening for factor V Leiden and prothrombin G20210A followed by prolonged prophylaxis as compared with the current practice of 6 months of anticoagulation. A Markov tree showing paths to a variety of possible outcomes was implemented using data from the literature and a meta‐analysis. The result of base‐case analysis showed that screening for factor V Leiden and prothrombin G20210A was effective, in that the beneficial effect of reducing the risk of DVT recurrences overwhelmed the negative aspect of increasing the risk of bleeding due to anticoagulant therapy. Although screening saved an average of 1 QALD with respect to standard treatment, this was similar to some accepted preventive interventions.28 Furthermore, implementing this strategy was cost‐effective because of averted thromboembolic events and the low neat number of induced bleedings. In particular, screening prevented as many as six thromboembolic events and induced 0.7 major bleedings per 1000 patients.

As compared to an extensive prophylaxis of overall carriers of factor V Leiden identified at screening,29 the strategy of double screening and selective prophylaxis had a similar cost‐effectiveness since it targeted rarer patients with a higher risk of recurrence and the screening procedure had similar costs. However, at variations of the odds ratio for recurrent VTE within its confidence interval, the cost‐effectiveness of double screening ranged from 12 965 to 22 889 $/QALY, while the cost‐effectiveness of single screening for factor V Leiden changed up to 72 750 $/QALY. Thus screening for both factor V Leiden and prothrombin G20210A could be more attractive from the economic point of view, despite the fact that a cohort of 1000 individuals with VTE might get 2 days of life from prolonged prophylaxis of all the carriers of factor V Leiden, as compared to 1 day for a prophylaxis restricted to double heterozygotes. Whether to support the former or the latter preventive strategy is thus both a policy matter, dealing with budget and local costs, and a research matter, dealing with uncertain estimates of recurrence risk. The two decision models, however, are available from the authors for institutional or personal use whenever a policy need to be redesigned on the basis of precise computations (i.e. local costs for screening tests or the direct costs related to acute VTE). In fact, this study demonstrated that detailed models of the natural history of diseases allow a proper evaluation of long‐term and chained consequences of health interventions. In a short time horizon evaluation, where long‐term benefits such as avoided events and morbidity would have not been considered, screening would have saved only 8 h of life at a cost of nearly 40 000 $/QALY.

The current analysis has some drawbacks, since the efficacy of long‐term anticoagulation in thrombophilic patients is still under evaluation, and costs were specific to the Italian Health Care System. However, the model assessed the uncertainty related to input data and assumptions, and provided a framework for generalization of results and updating. Well‐designed prospective studies are needed to definitely set the risk of recurrence in double‐mutation heterozygotes; however, current evidence provided by the two large retrospective studies7,,8 is consistent with the hypothesis of a substantial higher risk.

In conclusion, prolonged prophylaxis of double heterozygotes for factor V Leiden and G20210A prothrombin mutation is warranted after a first episode of DVT, and screening for the two mutations is both effective and cost‐effective. Local policies and patient‐specific recommendations still need to be implemented to apply the results of the current analysis to clinical practice.


  • Address correspondence to Dr M. Marchetti, Laboratory of Medical Informatics, IRCCS Policlinico S.Matteo, viale Golgi 19, 27100 Pavia, Italy. e‐mail: marchettimsmatteo.pv.it


View Abstract