Thrombophilia testing in people with venous thromboembolism: systematic review and cost-effectiveness analysis

Thrombophilia

Thrombophilia is a heritable (genetic) or acquired defect in blood coagulation that leads to a predisposition towards thrombosis. A thrombus is a solid mass of blood constituents that can fragment and block vessels downstream (thromboembolism). Depending on the blood vessel occluded, venous thromboemboli can lead to pulmonary embolism (PE) or, rarely, stroke. Venous thrombosis often occurs in normal vessels, with the majority of venous thrombi forming in the deep veins of the leg (deep vein thrombosis, DVT).

Physiological blood coagulation is complex and is mediated through the interaction of numerous plasma proteins. These circulate in an inactive form to prevent unwanted clot formation, but when activated contribute to a potent cascade of interactions culminating in the generation of thrombin. The initiating event is interaction of factor VII/VIIa with tissue factor. Tissue factor is present in most cells and is made available as a result of injury. When factor VIIa binds to tissue factor this initiates the coagulation cascade, in which factors IX and X are activated. Through activated factors IX and X, in the presence of activated factors VIII and V, thrombin is generated from prothrombin (factor II). Thrombin converts factor I (fibrinogen) to insoluble fibrin, the principal component of thrombus. Thrombin also acts as a catalyst to its own formation by feedback activation of factors VIII and V. In addition, thrombin recruits platelets and promotes cross-linking of fibrin strands through activation of factor XIII.

There is a regulatory system to prevent uncontrolled coagulation. Tissue factor pathway inhibitor inhibits the early events. Antithrombin inhibits thrombin as well as activated factors IX, X and XI. When thrombin binds to thrombomodulin it is redirected to an anticoagulant role through activation of protein C. Activated protein C, together with the free form of protein S that acts as a cofactor for protein C, inactivates factors Va and VIIIa. Finally, a parallel system controls the generation of plasmin, the principal enzyme capable of lysis of fibrin. 1–6

Thrombophilia can be genetic, acquired or mixed (due to a mixture of genetic and environmental factors). Heritable (genetic) thrombophilia is caused most commonly by mutations in the genes for coagulation factors II and V. Acquired thrombophilia refers to conditions in which individuals without genetic defects in coagulation factors are at increased risk of thrombosis, for example those with lupus anticoagulant or anticardiolipin antibodies. Examples of mixed-type thrombophilias are elevation of factor VIII or homocysteine levels. Malignancy can lead to an increased risk of thrombosis. Transient risk factors for thrombosis are conditions in which individuals are temporarily at increased risk of thrombosis, for example pregnancy, oestrogen therapy from combined oral contraceptives or hormone replacement therapy, obesity, fractures and major surgery. There is an increased risk of thrombosis with increasing age. 7

Factor V Leiden (FVL) and the prothrombin G20210A mutation (PTG20210A) are genetic thrombophilias associated with increased procoagulant (promoting coagulation) activity. The FVL mutation is a point mutation in the gene for clotting factor V (1691G-A). Activated protein C (APC) is one of the major inhibitors of the coagulation system. An impairment in plasma anticoagulant response to APC is known as APC resistance. FVL is the most frequent cause, although not the only cause, of inheritable APC resistance. PTG20210A is a mutation of the prothrombin gene that is associated with elevated plasma prothrombin levels. In the general population the prevalence of FVL heterozygosity is 1–15%,8 of FVL homozygosity is 0.02–0.05%9 and of PTG20210A is 2–5%. 8 The prevalence of FVL and PTG20210A is higher in Caucasian populations than in African or Asian populations. 10 Compared with people without the mutation there is an increased risk of experiencing VTE of 3–8 times for heterozygous carriers of FVL, 80 times for homozygous carriers of FVL and 3 times for carriers of PTG20210A. 11

Antithrombin (AT), protein C (PC) and protein S (PS) are physiological anticoagulants and a deficiency of any of these can be heritable. AT inhibits thrombin and also some activated clotting factors. AT deficiency can be caused by either a reduction in the quantity of normal AT protein or production of abnormal protein. PC, when activated, is a major inhibitor of the coagulation system. PC deficiency can be caused by either a lower level of PC or less functional PC. PS is a cofactor for APC. PS deficiency can be caused by reduced production of PS, a defect in PS or reduced availability of PS. In the general population the prevalence of AT deficiency is 0.02–0.04%,9 of PC deficiency is 0.2–0.4%9 and of PS deficiency is 0.003–2%. 11,12 The low prevalence of these deficiencies makes it more difficult to accurately assess the relative risk (RR) of VTE, but estimates suggest an increased risk of first VTE compared with people without the deficiency of 19–50 times for individuals with AT deficiency,11,13 6.5–15 times for PC deficiency,11,12 and 5–10 times for PS deficiency,11,12 although this risk had been estimated to be 32 times as high by a retrospective study. 13

Antiphospholipid antibodies, that is anticardiolipin antibodies and lupus anticoagulant, are forms of acquired thrombophilia. The mix of genes and environmental factors can cause elevated levels of homocysteine, fibrinogen or clotting factors VIII, IX and XI. Sufficiently elevated levels of these can increase the risk of VTE. The prevalence of hyperhomocysteinaemia (levels > 18.5 µmol/l) in the general population is 5–7%,12 that of elevated factor VIII (> 150 IU/dl) is 11%,8 of elevated factor IX (> 129 IU/dl) is 3%8 and of elevated factor XI (> 120.8 IU/dl) is 10%. 8 Dysfibrinogenaemia is rare in the general population. 11

It is possible to have more than one type of thrombophilia. Most types of thrombophilia are considered neither necessary nor sufficient cause for thrombosis. Individuals can have thrombophilia without experiencing a thrombotic event. Thrombosis can occur in people without thrombophilia.

Venous thrombosis

The estimated annual incidence of VTE (not restricted to patients with thrombophilia) is 1 in 1000 individuals in the general population. 14 The incidence is higher in older age groups than in younger age groups. 7

VTE can be associated with, for example, pregnancy, oestrogen therapy, fractures and major surgery. When there is no known cause, VTE is referred to as idiopathic. Within a group of VTE patients it is estimated that approximately 30–50%10,15 will have a known form of heritable thrombophilia, depending on the population.

A higher prevalence of some types of thrombophilia has been found in patients with venous thrombosis than in the general population, although there is considerable variation in reported prevalence rates according to study/population. The prevalence rates of thrombophilia in unselected (including idiopathic and non-idiopathic) patients with VTE are shown in Table 1.

Venous thrombosis is an important cause of morbidity and mortality; approximately 90% of PEs are caused by dislodged fragments from asymptomatic DVTs. 19 PE can be fatal. Estimates of mortality rates vary widely, from 2.3%, based on patients enrolled in clinical studies, to 28%, based on a cohort study. 20 Post-thrombotic syndrome (PTS) is a long-term complication of DVT. Approximately 30–50%21,22 of DVT patients may suffer post-thrombotic symptoms in the long term. These symptoms include pain, swelling and venous ulceration of the affected leg. The risk of developing PTS has not been found to be affected by thrombophilia, as shown for FVL, PTG20210A or elevated factor VIII. 23

In the UK approximately 500,000 patients are being prescribed oral anticoagulants. 24 Anticoagulants suppress the synthesis of clotting factors in the blood and therefore prolong the time it takes for the blood to clot. Warfarin is the most commonly prescribed anticoagulant, but other oral anticoagulants include acenocoumarol and phenindione. 24 Warfarin interferes with the vitamin K-dependent synthesis of factors II, VII, IX and X, and also affects proteins C and S. 25 The dose of warfarin prescribed is determined by the international normalised ratio (INR), which is a measure of coagulability. Patients on oral anticoagulants need to be monitored regularly, with INR testing. The INR is derived from measurements of the time that it takes for a sample of the patient’s blood to clot, for example an INR of 2 means that the blood takes twice as long as normal to clot. Usual practice is to give a large initial loading dose of warfarin and adjust the daily dose according to the INR results from blood samples taken over the following days. Because there is a delay before the onset of the clinical effects of warfarin, in the initial stages of treatment, heparin is often given concomitantly as it has an immediate effect. Once the patient has achieved the target INR, the patient continues treatment with a maintenance dose of warfarin. The patient must undergo periodic blood tests to ensure that the target INR is maintained. Monitoring is time- consuming for patients and clinicians. The dose of warfarin is adjusted to maintain the INR within a target therapeutic range, which is determined by the indication for treatment. The recommended target INR for VTE is 2.5. 26

The benefits of anticoagulation in terms of a reduction in the risk of thromboembolic events must be balanced against the increased risk of haemorrhage. While taking oral anticoagulants long term there is an annual risk of haemorrhage of approximately 1–15%, with the risk increasing with higher INR. 24 The risk of haemorrhage is not affected by the presence of thrombophilia. 27,28

Objectives

The review addresses the following question: ‘Is thrombophilia testing following a venous thrombotic event clinically effective and cost-effective in the management of thrombosis compared with not testing for thrombophilia?’

Areas outside the scope of this appraisal

Screening of individuals exposed to conditions that are transient risk factors (e.g. major surgery and pregnancy) has been excluded from the scope of this appraisal, as have pregnancy outcomes. Case finding by testing of asymptomatic individuals with a family history of thrombophilia or thrombosis but no personal history of thrombosis is also outside the scope of this review. These are important issues but it was not feasible to appraise all of these within a single technology assessment report.

Identification of studies

A comprehensive search was undertaken to systematically identify clinical effectiveness literature concerning thrombophilia testing of patients with thrombosis and the resulting long-term anticoagulation management. The search strategy comprised searching of electronic databases, scrutiny of bibliographies of retrieved papers and contact with the project advisory group to identify key papers.

Searches of electronic databases were not restricted by language, publication date or publication type. Searches included arterial as well as venous thrombotic events as the review initially aimed to investigate these. Searches were conducted between September and November 2006. The MEDLINE search strategy is shown in Appendix 2 and was adapted for use on the other databases. The following electronic databases were searched from inception: MEDLINE (Ovid), CINAHL, EMBASE, PreMEDLINE, Cochrane Library including the Cochrane Database of Systematic Reviews, Cochrane Controlled Trials Register (CENTRAL), DARE, NHS EED and HTA databases, Science Citation Index (SCI), National Research Register (NRR), Current Controlled Trials, BIOSIS, Centre for Reviews and Dissemination (ongoing reviews database), Research Findings Register, Web of Science.

Inclusion criteria

Exclusion criteria

• Publications in languages other than English.

• Thrombophilia tests conducted while patient was taking warfarin.

• Thrombosis in pregnancy or pregnancy complications associated with thrombophilia.

• Thrombosis related to temporary risk factors, including major surgery or oestrogen therapy.

• Case finding by testing individuals with a family history of thrombosis or thrombophilia but no personal experience of thrombosis.

Based on the above inclusion/exclusion criteria, study selection was made by one reviewer.

Data extraction, critical appraisal and data synthesis

It had been planned for one reviewer to extract data using a standardised form, with no blinding to authors or journal, for the purpose of providing a narrative account of trial quality for the reader. Planned quality assessment was with criteria based on those proposed by the NHS Centre for Reviews and Dissemination for randomised controlled trials, or using the Downs and Black checklist for randomised and non-randomised studies if other study types had been accepted into the review.

The lack of data made any kind of data synthesis impossible. It had been planned that prespecified outcomes would be tabulated and discussed within a descriptive synthesis, or, if statistical synthesis had been appropriate, meta-analysis would have been conducted using fixed- and random-effect models.

Results

Following removal of duplicates the search yielded 10,341 citations. Of these, 38 were database citations of ongoing studies, all of which were excluded by title.

Of the remaining 10,303 published articles 10,185 were rejected from titles and abstracts. A total of 118 articles were accepted by title search, many of which did not have abstracts available. Of these retrieved papers none met the inclusion criteria for intervention and comparator for the review; that is there were no comparisons available of patients tested for thrombophilia with patients whose risk assessment was based on a personal and family history of thrombosis. This is illustrated in the flow diagram (Figure 1).

FIGURE 1.

Flow diagram of study selection.

Discussion

There were no studies available comparing patients tested for thrombophilia with patients whose risk assessment was based on a personal and family history of thrombosis. Thus, there were no studies meeting the inclusion criteria that could have provided data on any of the outcomes for which data were sought. Therefore clinical data for the cost-effectiveness model were not obtained from this clinical effectiveness systematic review. It may be that studies of this kind have not been conducted because thrombophilia testing is not routine after a first thrombotic event or because thrombophilia diagnosis does not alter anticoagulation management. A potential limitation of the search was that it excluded publications in languages other than English.

Prevalence

The prevalence of thrombophilias in patients with venous thrombosis was taken from the prevalence in unselected patients, as data restricted to idiopathic thrombosis were not identified. The prevalence rates of thrombophilia in unselected VTE patients (idiopathic and non-idiopathic VTE) are shown in Table 1.

Recurrence rate following VTE in patients without thrombophilia

For patients without prothrombotic abnormalities (FVL, PTG20210A, AT, PC or PS deficiency, hyperhomocysteinaemia, hyperfibrinogenaemia, or elevated factor VIII, IX or XI), the recurrence rate following first idiopathic VTE was found to be 32.4 per 1000 patient-years [95% confidence interval (CI) 19.2–51.2 per 1000 patient-years] with a mean follow-up of 7.3 years. 41

Relative risk of recurrence in thrombophilia

Following a venous thrombotic event, the risk of VTE recurrence after discontinuation of anticoagulants may be higher for patients with some types of thrombophilia than for those without. The relative risk (RR) of recurrence following idiopathic VTE was sought for each form of thrombophilia (Table 2) but was not available for all types of thrombophilia. The RR of recurrence following unselected (including idiopathic and non-idiopathic) VTE was sought to allow conversion of the RRs in unselected VTE to risks for idiopathic VTE (Table 3).

In idiopathic DVT, a retrospective cohort study found no increased risk of recurrence for heterozygous FVL based on 43 patients with recurrence (RR 1.0; 95% CI 0.5–2.0). 42 A recent meta-analysis looking at heterozygous FVL pooled 10 studies of VTE recurrence following first episode of VTE in 3104 patients. 47 Patients were unselected apart from the exclusion of malignancy. The meta-analysis found an increased odds of recurrence for patients with FVL [odds ratio (OR) 1.41; 95% CI 1.14–1.75], which resulted in a RR of 1.46. 47 A prospective study found that the odds of recurrence in homozygous FVL, with reference to non-carriers or heterozygous FVL, was 4.1 (95% CI 0.97–15.5), which resulted in a RR of 2.3. 48

In idiopathic DVT, a retrospective cohort study found that patients heterozygous for PTG20210A had no significant difference in hazard from patients without the abnormality, based on 41 patients with recurrence [hazard ratio (HR) 1.5; 95% CI 0.7–3.1; RR 1.5]. 43 A recent meta-analysis looking at PTG20210A pooled nine studies of VTE recurrence following first episode of VTE in 2903 patients. 47 Patients were unselected apart from the exclusion of malignancy. The meta-analysis found an increased odds of recurrence for patients heterozygous for PTG20210A (OR 1.72; 95% CI 1.27–2.31). 47

In idiopathic VTE, a retrospective cohort study found that patients heterozygous for both FVL and PTG20210A had an increased risk of recurrence (RR 5.4; 95% CI 2.0–14.1) with reference to patients heterozygous for FVL, based on 17 patients with recurrence. 42

In unselected (including idiopathic and non-idiopathic) VTE, the presence of AT deficiency, PC deficiency or PS deficiency or lupus-like anticoagulants was associated with a RR of recurrence of 1.44 (95% CI 1.02–2.01). 49 An anticoagulant study found that lupus anticoagulant was associated with an increased risk of recurrence following idiopathic VTE (RR 6.8; 95% CI 1.5–31), and that the RR for anticardiolipin antibodies did not reach significance (RR 2.3; 95% CI 0.5–11). 18

In idiopathic VTE, hyperhomocysteinaemia had an increased risk of recurrence (RR 2.7; 95% CI 1.3–5.8). 44 Data were not found that indicated the RR of recurrence for patients with dysfibrinogenaemia. A prospective study found that elevated factor VIII was associated with a RR of VTE recurrence of 5.43 (chromogenic factor VIII) or 6.21 (clotting factor VIII) following first idiopathic VTE. 45 This study found lower rates of recurrence and lower RRs, 2.62 for chromogenic factor VIII and 1.72 for clotting factor VIII, for patients with non-idiopathic VTE. 45 A prospective study in idiopathic VTE found an increased risk of recurrence (RR 3.06; 95% CI 1.29–7.28) in patients with elevated factor IX. 46 This study also found a RR of recurrence of 2.14 (95% CI 1.01–4.58) in patients with elevated factor XI. 46

There may be a higher risk of recurrence in men than in women. 50 This may be explained by women whose first VTE event was related to pregnancy or oestrogen therapy51 and so would not apply in idiopathic thrombosis. However, this explanation is not supported in all studies,52 and a higher risk of recurrence in men than women following idiopathic VTE has been found. 41

Methods

Although no literature was found concerning thrombophilia testing of patients with thrombosis and the resulting long-term anticoagulation management, enough data were found on the increased risk of VTE associated with each thrombophilia type, the prevalence of thrombophilia, the efficacy of warfarin, the risks of haemorrhage associated with warfarin, the outcomes of VTE and haemorrhages, and the costs and utilities to be able to provide cost-effectiveness analyses. No data have been found within a UK context on the benefits of treatment compared with the risks of warfarin coupled to costs and utility. Our analysis may provide clinicians with an evidence-based approach within which to evaluate the period of warfarin treatment to be provided.

Modelling structure

An individual patient-based discrete event simulation (DES) model was constructed in Simul8© (Simul8 Corporation). The rationale for this approach is that it provides more flexibility than a cohort model as the history of the patient can be incorporated. This allows risks that are dependent on time since an event, such as the risk of a haemorrhage being greater in the initial months of warfarin treatment or the risks of VTE being highest immediately after a VTE, to be considered. Thus, a more accurate determination of the costs and quality-adjusted life-years (QALYs) associated with each treatment option is possible. Additionally, the cohort approach relies on an arbitrary definition of cycle duration, which is not needed within the DES model. Note that the only intervention evaluated to prevent recurrence of a VTE is warfarin.

The model simulates the experiences of hypothetical patients who have just suffered an index DVT or index non-fatal PE. The events that can occur are an additional VTE, a warfarin-induced haemorrhage or death due to a non-VTE-related cause (which can be reached from any non-absorbing state). Each outcome is associated with a cost and utility impact that, because of the individual patient modelling structure, can be incorporated into future time periods. At the resolution of a non-fatal event the time to the next event is simulated. This continues until all patients are in an absorbing state (fatal PE, fatal haemorrhage or death through non-VTE-related causes). A representation of the model is given in Figure 2.

FIGURE 2.

A flow diagram representation of the mathematical model. Shaded cells represent absorbing states. The state of mortality unrelated to a VTE or haemorrhage event can be reached from any non-absorbing state. The arrows are omitted for clarity.

In accordance with clinical advice, lifelong warfarin therapy would be prescribed following a subsequent VTE event. If a patient sustained a haemorrhage, warfarin treatment would be discontinued but would be restarted following a subsequent VTE if the patient had not bled intracranially.

Because of the differential rate of recurrence between men and women, the differential ratio of PE and DVT in those whose initial VTE was a DVT and in those in whom it was a PE, and the increased rates of haemorrhage as a patient becomes older, a large number of analyses were conducted.

Population of the model

The literature retrieved from the cost-effectiveness review was used to identify sources for the economic model that were not covered within the review of clinical effectiveness or the increased risk of recurrence associated with thrombophilia. The source used for populating the parameters within the model and the rationale for using this value are provided in the accompanying text.

Population start age and life expectancy

Hypothetical cohorts of men aged 30, 40, 50, 60 and 70 years were simulated. Life expectancies have been obtained from interim life tables published by the Office for National Statistics. 53 We have assumed that a previous DVT or non-fatal PE will not affect life expectancy except through VTE or a haemorrhage, events that are explicitly modelled. The life expectancy and starting utility associated with age are given in Table 4. The underlying utility for patients aged 30 and 40 years is not provided and has been assumed to increase in a fairly linear manner with respect to the values for 50 and 70 years.

TABLE 4 - The life expectancy and starting utility associated with age

Author-estimated values.

Age (years)	Life expectancy (years)53	Starting utility54
30	47.75	0.950a
40	38.24	0.900a
50	29.04	0.850
60	20.49	0.829
70	13.09	0.727

Risk of recurrent VTE

To accurately assess the implications of increased risks associated with thrombophilia, the recurrence rates of VTE associated with patients without thrombophilia must be known. The most appropriate data found come from Christiansen et al. ,41 who followed up an untreated cohort of patients without FVL mutation, PTG20210A, anticoagulation deficiency, elevated levels of factors VII, IX and XI, hyperfibrinogenaemia or hyperhomocysteinaemia. The Christiansen et al. data may include patients with lupus anticoagulant and anticardiolipin antibodies and could thus overestimate the risk for non-thrombophilic patients. The rate of recurrence of VTE for these patients was 3.24 per 100 patient-years, which we have assumed to be applicable regardless of whether the patient had previously sustained a DVT or a PE. Although this paper discusses correcting results for the age of the patient, no data were presented that showed how rates change according to age of the patient, and we have assumed that the probability of recurrence is independent of age, although data7 show that the risk of an initial VTE increases as a person ages. The likely effect of an increased risk of recurrence as a patient ages has been included in the discussion.

It is reported that men have an increased HR of 2.7 (95% CI 1.8–4.2). 41 This value has been combined with the percentage of males within the study to calculate that the risk of recurrence for males would be 5.1 per 100 patient-years, with the corresponding value for females 1.9 per 100 patient-years.

The data from Christiansen et al. 41 show a non-significant decrease in risk with time since VTE, which may then increase beyond 6 years of follow-up. Smaller studies by Prins et al. ,55 Kearon et al. 18 and Simioni et al. 56 all suggest that the risk might be increased in the initial 3 months following a short course of warfarin treatment. The data available are insufficient and too heterogeneous to conclusively determine whether the risks are higher in this period, and we do not dismiss the notion that 6 months of warfarin treatment may be as appropriate as 3 months of treatment. This decision is, however, largely independent of any known thrombophilic status and has been excluded from our analyses. We have assumed that 3 months of treatment is the standard course of treatment but note that the decision on whether to provide lifelong treatment or not will be equally valid if the initial course of treatment is for 6 months.

Data were not found related to the increased probability of subsequent VTE in the high-risk period immediately following a VTE in patients who did not receive warfarin, either because the VTE was undetected or because of a previous intracranial haemorrhage. In the absence of better data we have assumed that for a patient who does not receive warfarin treatment following a VTE there is a 20% probability of a recurrent VTE in the next 6 months, which is independent of gender. This risk is double that which has been estimated to occur in patients who have taken a course of warfarin following a VTE,57 with the magnitude of the increase tempered in the expectation that the VTEs that remain undetected will be less severe and less likely to recur. This value is altered in the sensitivity analyses. Beyond this time period the risks of recurrent VTE in patients not receiving warfarin taken from Christiansen et al. have been used. 41

The increased risk of VTE due to thrombophilia

Some forms of thrombophilia may increase the risk of recurrent VTE, with the RR being dependent upon the cause of the condition. Because of the scarcity of data received regarding the marginal costs of individual tests, we have assumed the cost characteristics of the thrombophilia tests as detailed in Wu et al. ,30 i.e. tests for FVL, PTG20210A, AT deficiency, PC deficiency, PS deficiency, lupus anticoagulants and anticardiolipin antibodies. Thus, only these types of thrombophilia are included within our modelling. Ideally we require data on the increased risks of recurrent idiopathic VTE, and these data, where available, are summarised in Table 5; however, where data are available only for unselected (including idiopathic and non-idiopathic) VTE, these are given in Table 6. These tables also provide estimates of the prevalence of the thrombophilia in patients with VTE. It is assumed that these prevalences are independent of age, sex and previous VTE event; however, no data were found to support or refute this assumption.

TABLE 5 - The relative risk of recurrence following idiopathic VTE for thrombophilic people compared with people without thrombophilia

With reference to heterozygous FVL.

Type of thrombophilia	Relative risk (95% CI)	Prevalence (%)
Heterozygous for FVL	1.0 (0.5–2.0)42	10–508
Heterozygous for PTG20210A	1.48 (0.84–2.62)47	5–189
Heterozygous for both FVL and PTG20210Aa	5.4 (2.0–14.1)42	3.458,9
Lupus anticoagulant	6.8 (1.5–31)18	318
Anticardiolipin antibody	2.3 (0.5–11)18	318

TABLE 6 - The relative risk of unselected VTE for thrombophilic people compared with people without thrombophilia

With reference to non-carriers or heterozygous FVL.

Type of thrombophilia	Relative risk (95% CI)	Prevalence (%)
Homozygous for FVLa	2.26 (0.93–5.46)48	1.59
AT deficiency, PC deficiency or PS deficiency or lupus-like anticoagulants	1.44 (1.02–2.01)49	1349

We have to convert the RR in unselected VTE to risks for idiopathic VTE. To do this we need to use a thrombophilia that has data for both idiopathic and unselected VTE to estimate a relationship between the two groups. These data are available for heterozygous FVL, which has a RR of 1.0 in idiopathic and 1.46 in unselected VTE. These values are 1.48 and 1.73, respectively, for heterozygous PTG20210A. As unselected VTE includes provoked VTEs, such as those that occur in surgical patients, which are less likely to spontaneously recur, the RRs are higher in patients with unselected VTE than in those with idiopathic VTE. We have arbitrarily assumed that the RR of 2.6 for unselected VTE in patients homozygous for FVL compared with non-carriers or patients heterozygous for FVL would be reduced to 2.0 for idiopathic VTE in patients homozygous for FVL compared with non-carriers or patients heterozygous for FVL. Similarly we have assumed that the RR of idiopathic VTE for patients with AT deficiency, PC deficiency or PS deficiency or lupus-like anticoagulants is 1.25, reduced from 1.44 when compared with unselected VTE. The risks for patients with AT deficiency, PC deficiency or PS deficiency may be overestimated if the lupus-like anticoagulants have a similar RR to lupus anticoagulant, and the RR may indeed be 1. We have modelled using the value of 1.25 but have commented on the difference in treatment between patients with AT deficiency, PC deficiency or PS deficiency and those who are FVL heterozygous, which has an estimated RR of 1.

Efficacy of warfarin treatment in preventing VTE

Warfarin reduces the risk of subsequent VTE. We do not make comment on the quantity of warfarin prescribed to a patient and have assumed that this decision is made by the clinician. The exact relative risk reduction (RRR) in recurrent VTE is not known, with estimates ranging from 90% to 95%. These are given in Table 7, with a decision made to use 95% as our mean estimate. It is assumed that this RRR is applicable throughout the duration of warfarin treatment and that this effect is instantly removed following the cessation of warfarin treatment.

TABLE 7 - The efficacy of warfarin in preventing VTE

Publication	RRR of warfarin in preventing VTE (95% CI or reported range)
Prins et al.55	0.900
Marchetti et al.58	0.950 (0.65–1.00)
Kearon et al.18	0.950 (0.63–0.99)

The outcomes following a recurrent VTE

The outcome following a VTE is dependent on whether or not the person receives warfarin treatment on the detection of the VTE. 59 Because of scarce data we have assumed that the effect of warfarin on the VTE is independent of whether the patient has a course of warfarin prescribed following the VTE or remains on lifelong warfarin. A person with a VTE will not receive warfarin if the VTE remains undetected or if there has been a history of intracranial haemorrhage. We have assumed that the sensitivity of clinical tests in detecting a VTE is 95% as this value is representative of current detection methods for DVT and it is assumed that patients with a previous VTE will receive more sensitive tests before a decision to prescribe lifelong warfarin is taken. 60 We have assumed that this value is also applicable for the detection of PE. It was further assumed that all patients with a non-fatal VTE will see a clinician and be referred for testing.

The outcome of a VTE is additionally dependent on whether the previous VTE manifested in a DVT or a PE. 61 Among patients with a previous PE, 81.1% of VTE resulted in PE, whereas for patients with a previous DVT 78.6% of VTE resulted in DVT. The fatality rates were also markedly different, with 26.4% of VTE resulting in death after a previous PE, compared with 7.6% after a DVT. 61

The probability of a fatal, or non-fatal, PE following a recurrent VTE for a patient on warfarin after a VTE has been taken from Douketis et al. 61 Following a subsequent DVT there were 19 fatalities in 250 VTE events (7.6%), and there were 19 fatalities in 72 events (26.4%) following a subsequent PE. The probability of a VTE that is a DVT resulting in PTS for patients receiving warfarin has been taken from Goodacre et al. 59

Data on patients with a VTE who do not receive warfarin are by definition scarce, as these are patients in whom the VTE has been unidentified or who have sustained an intracranial haemorrhage. The probability of a fatal, or non-fatal, PE or PTS in patients who remain untreated following a DVT has been taken from Goodacre et al. 59 Data on patients with a PE who do not receive warfarin treatment were not found and had to be estimated by the authors. Data from Goodacre et al. 59 showed that the probability of a VTE progressing to a PE was 10 percentage points higher in those who do not receive warfarin treatment (68.6% DVT and 31.4% PE) than in those who receive warfarin treatment (78.6% DVT and 21.4% PE). We have assumed that this increase in PE without treatment is also applicable following a PE. We further assume that two-thirds of PEs are fatal compared with the one-third observed in treated patients. These assumptions are tested in sensitivity analyses.

For all outcomes following a VTE it is assumed that any cost and disutility associated with the event occur immediately.

These assumptions result in the risks provided in Tables 8 and 9.

TABLE 8 - The outcome of a recurrent VTE in patients who have had a previous DVT but not a previous PE related to whether the person receives warfarin treatment

For sources see accompanying text.

	Treated (%)	Untreated (%)
Fatal PE	7.60	20.93
Non-fatal PE	13.80	10.47
PTS	4.22	29.21
Resolved VTE	74.38	39.39

TABLE 9 - The outcome of a recurrent VTE in patients who have had a previous PE related to whether the person receives warfarin treatment

For sources see accompanying text.

	Treated (%)	Untreated (%)
Fatal PE	26.40	60.73
Non-fatal PE	54.70	30.37
PTS	1.01	3.31
Resolved VTE	17.89	5.59

Risk of haemorrhage due to warfarin use

The rate of haemorrhage is greater in the initial months of anticoagulation therapy,62,63 which may be due to factors that predispose the patient to haemorrhage,63 a patient initially being prescribed too great a dose of anticoagulant whilst the optimal level is determined,63 or co-prescribed medication. 64 During the initiation period the INR of the patient is closely monitored,64 resulting in higher resource use. The rates of haemorrhage adopted in the model within the initial 3-month period have been calculated from Linkins et al. 62 and are presented in Table 10. It is assumed that there is no risk of haemorrhage following the withdrawal of warfarin therapy.

TABLE 10 - The rates of haemorrhage associated with the initial 3 months of warfarin treatment62

Haemorrhage location/type	Absolute rate in the initial 3 months of treatment with warfarin
Fatal haemorrhage	0.0034
Non-fatal intracranial haemorrhage	0.0009
Non-fatal non-intracranial haemorrhage	0.0175

For longer-term risks of haemorrhage associated with warfarin use we have used data from Prins et al. 55 as this study has related the risks of haemorrhage to age, with the rates increasing as a patient ages. The risks used beyond the initial 3-month period by age are presented in Table 11. The proportion of haemorrhages that are fatal, non-fatal and intracranial, and non-fatal and non-intracranial have been set to the proportions of long-term events reported in Linkins et al. 62

TABLE 11 - The annual rates of haemorrhage associated with warfarin treatment after the initial 3-month period

For sources see accompanying text.

Haemorrhage location/type	Patient age (years)
	Less than 40	40–49	50–59	60–69	Over 69
All haemorrhages	0.0060	0.0100	0.0150	0.0220	0.0320
Fatal haemorrhage	0.0014	0.0009	0.0014	0.0051	0.0074
Non-fatal intracranial haemorrhage	0.0008	0.0011	0.0017	0.0028	0.0041
Non-fatal non-intracranial haemorrhage	0.0038	0.0080	0.0119	0.0141	0.0205

Warfarin treatment is withdrawn from any patient experiencing a haemorrhage requiring medical intervention. On clinical advice we have assumed that, when warfarin has been discontinued because of a haemorrhage and the patient experiences a subsequent VTE, warfarin would be reinstated provided that the haemorrhage was not intracranial.

Sensitivity and specificity of thrombophilia testing

Auerbach et al. 65 assumed that the sensitivity and specificity of thrombophilia testing were 99%, figures that are similar to those provided by a manufacturer for the accuracy of a FVL test (Roche Diagnostic, 30 January 2007, personal communication). To analyse specificity the following assumptions were made: that 50% of the idiopathic VTE population were non-thrombophilic (as may be the case from Tables 10 and 11); that, of those misdiagnosed, the proportion of false positives by thrombophilia type would be equal to their corresponding proportion in a population with an idiopathic VTE. Results are shown assuming 100% sensitivity and specificity and with the base-case values of 99% so that the magnitude of the change in results when sensitivity and specificity are changed can be gauged. Such high values are likely in DNA-based tests such as those for FVL or PTG20210A66 but are likely to be overestimates for tests for other types of thrombophilia; however, values were not found for individual thrombophilia types. It is noted that, where thrombophilia testing has been suggested to be cost-effective, these results are driven by those patients with lupus anticoagulant or who are heterozygous for both FVL and PTG20210A. Although the estimate of 99% for both sensitivity and specificity is likely to be relatively accurate for FVL and PTG20210A, if the true sensitivity and specificity of the tests for detecting patients with lupus anticoagulant were markedly lower than 99%, the cost-effectiveness ratios for thrombophilia testing produced in this report will be favourable to thrombophilia testing.

Costs of VTE-related events

Costs have been extracted from standard literature sources67–69 where available. When these costs are not readily available for a specific event we have used data from the literature updating to 2005–6 prices using the inflation indices in Curtis and Netten. 68 The summarised costs used in the model are presented in Table 12, with detailed calculations given in Appendix 5.

TABLE 12 - Cost of VTE-related events (all costs are in 2005–6 prices)

Description of variable	Mean value (£)	Source
Treatment of a resolving deep vein thrombosis	183.46	See Appendix 5
Warfarin treatment
Cost of warfarin, first quarter	538.48	See Appendix 5
Cost of warfarin, subsequent quarters	211.38	See Appendix 5
Cost of treating post-thrombotic syndrome	3284.70	See Appendix 5
Cost of treating a fatal pulmonary embolism	1803.86	NHS reference costs69
Cost of treating a non-fatal pulmonary embolism	1390.54	NHS reference costs69
Cost of treating a fatal haemorrhage	6792.65	Sandercock et al.70
Cost of treating a non-fatal intracranial haemorrhage
Initial one-off cost	5774.78	See Appendix 5
Ongoing cost per year	4798.19	See Appendix 5
Cost of treating a non-fatal non-intracranial haemorrhage	736.93	NHS reference costs69
Cost of thrombophilia test that detects FVL, PTG20210A, AT deficiency, PC deficiency, PS deficiency, lupus anticoagulants and anticardiolipin antibodies	70.60	Wu et al.30

Utility

The utility multipliers for the VTE-related health states are presented in Table 13.

TABLE 13 - Utility multipliers for VTE-related heath states

Health state	Utility multiplier	Chosen sources
Post DVT receiving warfarin	0.987	Gage et al.71
Post DVT not receiving warfarin	1.000	Assumption
PTS	0.977	O’Meara et al.72
Non-fatal PE	0.940	Goodacre et al.59
Non-fatal intracranial haemorrhage	0.290	O’Meara et al.72
Non-fatal non-intracranial haemorrhage	0.997	Goodacreet al.59

Utility scores are combined multiplicatively within the model so that a patient receiving warfarin after sustaining a non-fatal PE would have a utility multiplier of 0.931 (0.990 × 0.940). This value is then multiplied by the average utility for patients at the specified age as reported by Kind et al. 54 and presented in Table 4.

The utility for a patient who has recovered from a DVT without any adverse effect and is no longer receiving warfarin could not be determined from the literature. The assumption made is that this patient has recovered to the utility associated with people of that age within the general population as reported by Kind et al. 54

Another reported utility multiplier for PTS was 0.995. 58 However, we chose to use the value of O’Meara et al. 72 because it was the value assumed to be most appropriate by the authors in the recent review of diagnostic tests for detecting DVT59 and because the effect assumed in the former paper was so negligible.

Discount rates

The discount rates for both the costs and QALYs were set at 3.5% per annum in accordance with published guidelines. 73

Results

Probabilistic sensitivity analyses

PSA was conducted on the parameters included in Appendix 6 using a Monte Carlo methodology as detailed in Claxton et al. 74 This approach samples once from the probability distribution for each variable contained within the model to produce a parameter configuration. This process is repeated until a predetermined number of parameter configurations have been sampled. The model is then run using the parameter configuration to generate the prespecified number of estimates of cost-effectiveness. This is recognized by NICE and is a requirement in the reference case. 73 In our analyses, 100 parameter configurations were generated for each combination of age, sex, previous VTE and thrombophilia type.

For each parameter configuration the model was run simulating 5000 hypothetical patients. The results produced by each parameter configuration were then ranked in order of cost-effectiveness to produce a cost-effectiveness acceptability curve (CEAC),75 which shows the likelihood of a given intervention having a cost per QALY below a given MAICER. The PSA results shown in Tables 14–21 have been based on the assumption that thrombophilia testing was both 100% specific and 100% sensitive. The changes in the cost per QALY were the tests assumed to have both sensitivity and specificity of 99% are also shown, and it is noted that the cost per QALY does not markedly change.

The rationale for the groups of patients on which PSAs were conducted was based on those with cost per QALY ratios that are most borderline cost-effective with reference to published guidelines. 73 Thus, we initially analysed women aged 50 years with a previous idiopathic DVT, as the cost per QALY of conducting thrombophilia testing on such patients was approximately £20,000. When it was determined that the PSAs were systematically producing cost per QALY values that were less favourable to thrombophilia testing than the mid-point analyses, we analysed subgroups of patients with cost per QALY ratios below £20,000 until the cost per QALY of thrombophilia testing remained below £20,000 in the PSA.

Women aged 50 years with a previous idiopathic DVT

The CEAC for thrombophilia testing women with a previous idiopathic DVT who are aged 50 years is given in Figure 3. This assumes that those who were found to have lupus anticoagulant or who were heterozygous for both FVL and PTG20210A would be treated with the course of warfarin treatment indicated in our mid-point results (10 years for both). It is seen that the median value of cost-effectiveness is dominated, with the mean value calculated as £37,671. This value is markedly higher than the £20,746 estimated in the mid-point analyses, showing that the combination of sampled values produces a non-linear model. The cause of this is the non-normal distributions assigned to key parameters such as the increased risk of a recurrent VTE associated with each type of thrombophilia, the disutility associated with warfarin use and the efficacy of warfarin in preventing VTE.

FIGURE 3.

The cost-effectiveness acceptability curve (CEAC) for women aged 50 years with a previous idiopathic DVT, and who have lupus anticoagulant or are heterozygous for both FVL and PTG20210A and who are receiving treatment.

To explain these results the distributions for the increased risks of recurrence associated with each thrombophilia type must be noted. Figure 4 shows this for lupus anticoagulant, and a wide uncertainty in the true RR compared with patients without thrombophilia is seen. For women aged 50 years with a previous DVT, with an increased risk of 5.4, as seen for patients who are heterozygous for both FVL and PTG2021A, and analysing the mid-point results, a cost per QALY of £18,034 is seen (Table 41), whilst with an increased risk of 2.3, as seen for patients with anticardiolipin antibody, extended warfarin treatment is dominated by the standard treatment period for warfarin (Table 42). A value of increased risk of approximately 5.2 (authors’ estimation) would be needed for treatment to have a cost per QALY of £20,000. The mode and median values associated with the distribution for the increased risk associated with lupus anticoagulant (Figure 4) are 2.05 and 4.55, respectively, resulting in the majority of simulations having a cost per QALY of more than £20,000 and a sizeable proportion being dominated by the standard course of warfarin. The additional costs associated with thrombophilia testing will also increase the incremental cost-effectiveness of testing at this age.

FIGURE 4.

The distribution of increased relative risk associated with lupus anticoagulant.

The effect of the non-normal distributions for the increased risk of thrombophilia becomes less influential as the difference between the mean of the log-normal distribution for lupus anticoagulant and the increased risk required for treatment to be deemed cost-effective becomes greater. This is seen in the additional examples provided.

Further analyses were undertaken assuming that only those women with lupus anticoagulant would receive treatment. This reduced the cost per QALY of thrombophilia testing to a mean of £39,525, with a median value of approximately £75,000, as shown in Figure 5. The mean is slightly higher than when patients who are heterozygous for both FVL and PTG20210A are additionally treated. This is because the number of thrombophilia tests undertaken remains relatively constant and no QALYs are gained or costs accrued from treatment of patients who are heterozygous for both FVL and PTG20210A.

FIGURE 5.

The cost-effectiveness acceptability curve (CEAC) for women aged 50 years with a previous idiopathic DVT, who have lupus anticoagulant and who are receiving treatment.

Men aged 70 years with a previous idiopathic DVT

The cost per QALY of thrombophilia testing had a mean value of £27,006, with a median value of approximately £30,000, as shown in Figure 6. For reference the mean cost per QALY from the mid-point analysis was £16,641 (Table 14).

FIGURE 6.

The cost-effectiveness acceptability curve (CEAC) for men aged 70 years with a previous idiopathic DVT, who have lupus anticoagulant or who are heterozygous for both FVL and PTG20210A and who are receiving treatment.

Women aged 40 years with a previous idiopathic DVT

The cost per QALY of thrombophilia testing had a mean value of £18,689, with a median value of approximately £18,000, as shown in Figure 7. For reference the mean cost per QALY from the mid-point analysis was £14,384 (Table 15).

FIGURE 7.

The cost-effectiveness acceptability curve (CEAC) for women aged 40 years with a previous idiopathic DVT, who have lupus anticoagulant or who are heterozygous for both FVL and PTG20210A and who are receiving treatment.

Men aged 60 years with a previous idiopathic DVT

The cost per QALY of thrombophilia testing had a mean value of £13,516 and a median value of approximately £16,000, as shown in Figure 8. For reference the mean cost per QALY from the mid-point analysis was £10,239 (Table 14).

FIGURE 8.

The cost-effectiveness acceptability curve (CEAC) for men aged 60 years with a previous idiopathic DVT, who have lupus anticoagulant or who are heterozygous for both FVL and PTG20210A and who are receiving treatment.

Women aged 70 years with a previous idiopathic PE

The cost per QALY of thrombophilia testing had a mean value of £14,692 and a median value of approximately £15,000, as shown in Figure 9. For reference the mean cost per QALY from the mid-point analysis was £10,782 (Table 17).

FIGURE 9.

The cost-effectiveness acceptability curve (CEAC) for women aged 70 years with a previous idiopathic PE, who have lupus anticoagulant or who are heterozygous for both FVL and PTG20210A and who are receiving treatment.

Given the PSA results presented above it was assumed that all remaining combinations of age, sex and previous idiopathic VTE that had cost per QALY ratios of below £20,000 in the mid-point analyses would also remain below £20,000 when PSA was conducted.

Calculating the cost per QALY ratios from PSA for the thrombophilia type that had the lowest increased risk of recurrence whilst having a cost per QALY of treatment of below £20,000 in the mid-point analyses and for which global thrombophilia testing had a cost per QALY of below £20,000 in the mid-point analyses

Having determined that the cost per QALY ratios provided in the mid-point analyses were systematically lower than the cost per QALY values generated in the PSA, analyses were undertaken on the thrombophilia type that had the lowest increased risk of recurrence whilst having a cost per QALY of treatment of below £20,000 for each age, sex and type of previous idiopathic VTE event, denoted in Tables 14–21 as the lightest shaded box at each age. Thus, for example, for men aged 50 years with a previous DVT event, being homozygous for FVL had the lowest risk of recurrence whilst remaining under a cost per QALY of £20,000 for treatment (Table 14), and PSA was undertaken assuming the treatment durations provided in Table 14 to calculate a cost per QALY ratio for thrombophilia testing. Only the cost of the repeat thrombophilia test to determine the specific thrombophilia is included in these calculations as it is assumed that the costs of global testing are borne by thrombophilia with greater mean chances of VTE recurrence.

All other thrombophilia types that can be successfully treated at this age, sex and previous VTE combination are assumed to have cost per QALY values that are lower than this figure. These analyses were conducted for all age, sex and previous VTE combinations that were not included in the previous sections on PSA. The cost per QALY values from these analyses are presented in Tables 22 and 23 for a previous DVT and Tables 24 and 25 for a previous PE.

TABLE 22 - The mean cost per QALY ratios from PSA for the thrombophilia type that had the lowest increased risk of recurrence whilst having a cost per QALY of treatment below £20,000 in the mid-point analyses and for which global thrombophilia testing had a cost per QALY of below £20,000 in the mid-point analyses: men with a previous idiopathic DVT event

Age (years)	Thrombophilia type	Treatment duration (years)	Mean cost per QALY of treatment (£)
30	FVL heterozygous	10	22,385
40	PTG20210A	10	25,026
50	FVL homozygous	10	23,502
60	Heterozygous for both FVL and PTG20210A	20	13,516
70	Heterozygous for both FVL and PTG20210A	10	20,506

TABLE 23 - The mean cost per QALY ratios from PSA for the thrombophilia type that had the lowest increased risk of recurrence whilst having a cost per QALY of treatment below £20,000 in the mid-point analyses and for which global thrombophilia testing had a cost per QALY of below £20,000 in the mid-point analyses: women with a previous idiopathic DVT event

Note that the thrombophilia for women aged 50 years with a previous DVT has been omitted because of a cost per QALY of greater than £40,000. See section, Women aged 50 years with a previous idiopathic DVT, p.29, for further explanation.

Age (years)	Thrombophilia type	Treatment duration (years)	Mean cost per QALY of treatment (£)
30	Heterozygous for both FVL and PTG20210A	20	11,626
40	Heterozygous for both FVL and PTG20210A	20	15,144

TABLE 24 - The mean cost per QALY ratios from PSA for the thrombophilia type that had the lowest increased risk of recurrence whilst having a cost per QALY of treatment below £20,000 in the mid-point analyses and for which global thrombophilia testing had a cost per QALY of below £20,000 in the mid-point analyses: men with a previous idiopathic PE event

Age (years)	Thrombophilia type	Treatment duration (years)	Mean cost per QALY of treatment (£)
30	FVL heterozygous	Lifelong	8396
40	FVL heterozygous	Lifelong	10,898
50	FVL heterozygous	20	12,613
60	FVL heterozygous	10	17,482
70	PTG20210A	10	16,784

TABLE 25 - The mean cost per QALY ratios from PSA for the thrombophilia type that had the lowest increased risk of recurrence whilst having a cost per QALY of treatment below £20,000 in the mid-point analyses and for which global thrombophilia testing had a cost per QALY of below £20,000 in the mid-point analyses: women with a previous idiopathic PE event

Note that the thrombophilia with the next lowest increased risk of recurrence was heterozygous for both FVL and PTG20210A, with a cost per QALY of £6898.

Note that the thrombophilia with the next lowest increased risk of recurrence was heterozygous for both FVL and PTG20210A, with a cost per QALY of £13,855.

Age (years)	Thrombophilia type	Treatment duration (years)	Mean cost per QALY of treatment (£)
30	FVL heterozygous	10	19,106
40	Deficiency in either AT, PC or PS	10	19,427
50	FVL homozygous	10	14,440
60a	Anticardiolipin antibody	10	38,010
70b	Anticardiolipin antibody	10	230,335

Discussion

Our work has enabled an evidence-based assessment of the most cost-effective duration (3 months, 10 years, 20 years or lifelong) of warfarin treatment based upon age, sex and previous VTE type, which may help clinicians decide on the most appropriate treatment length by indicating the scenarios in which our results suggest that thrombophilia testing is cost-effective. However, as depicted in the CEACs presented, there is a great deal of uncertainty in the cost-effectiveness of thrombophilia testing. As the prevalence of each thrombophilia type was not altered within the model the uncertainty around the cost-effectiveness of thrombophilia testing is likely to be greater than that shown. Figure 7 shows that, even in circumstances in which the expected cost per QALY ratio is below £20,000, there is a sizeable probability (20%) that the cost per QALY exceeds £100,000. This uncertainty is driven primarily by the wide confidence intervals associated with the increased risk of recurrence for each thrombophilia, particularly those with the higher mean RRs, such as lupus anticoagulant and being heterozygous for both FVL and PTG20210A.

The sensitivity and specificity of tests for each thrombophilia type could not be obtained. In the model we used a sensitivity and specificity of 99%, which are the values reported for DNA tests such as those for FVL or PTG20210A; however, this is likely to overestimate the accuracy of other tests. This will be favourable to thrombophilia testing, particularly if the sensitivity and specificity of the tests for identifying lupus anticoagulant are markedly lower than 99%. This casts further uncertainty around the robustness of the results produced.

We have not considered the extended use of warfarin after an initial VTE event without the diagnosis of thrombophilia. The results for men aged less than 39 years with a previous PE suggest that, using a MAICER of £20,000, patients who are heterozygous for FVL would benefit from extended warfarin treatment (Table 27). As the mean RR for these patients compared with patients without thrombophilia is 1, any conclusion on the use of extended warfarin for patients who are heterozygous for FVL would also apply to those patients without thrombophilia. If a MAICER of £30,000 per QALY were adopted then this age limit is estimated to increase to men aged 49 years or less. Further research would need to be conducted to estimate the numbers of people that such a policy would affect.

Our results have been predicated on the assumption that 3 months of treatment is the standard duration of warfarin treatment. Were the standard treatment under certain circumstances to be lifelong, for example in young men with a PE, then our estimate of the cost-effectiveness of thrombophilia testing could markedly change. Such analyses were not conducted as there was no clear guidance on when alternative treatment strategies would be employed, and it was assumed that clinicians would be risk averse when prescribing medications with potentially fatal side effects.

Not all types of thrombophilia have been evaluated in this report, primarily because no information was found on the marginal cost of performing tests for them. If this marginal cost is small then these tests should be considered, with an approximation of the duration of warfarin treatment taken from a thrombophilia with a similar increased risk of recurrence. For instance, factor VIII has not been considered in this report; however, the increased risk of recurrence is very similar to that for patients who are heterozygous for both FVL and PTG20210A (Table 2) and the duration of warfarin treatment associated with factor VIII would be assumed to be equal to that in patients who are heterozygous for both FVL and PTG20210A. Were it feasible to include additional tests at a relatively small marginal cost then it is likely that the cost-effectiveness results produced in this report have been unfavourable to thrombophilia testing.

We assumed that the risk of recurrence of VTE was constant with respect to age. There were limited data that indicated that the risk may increase with age. Although we have not explicitly modelled this relationship, it will have the effect of increasing the cost-effectiveness of extended warfarin treatment in the elderly and thus the cost-effectiveness of thrombophilia testing in patients of this age. At the age of 70 years or older only those patients who present with an idiopathic PE are recommended for thrombophilia testing; if there is an increased risk of recurrence associated with increasing age it is possible that thrombophilia testing in men aged 70 years or older who present with a DVT could also become cost-effective. Conversely, if the risk of recurrence is age related, the cost-effectiveness of thrombophilia testing in younger patients would become less favourable.

We have also not investigated the costs saved by omitting tests from the panel of tests in cases in which it is shown that even if the patient were to have a particular thrombophilia then management would not change. For example, analysis of men aged 70 years with a DVT (Table 18) suggests that only tests for lupus anticoagulant and FVL and PTG20210A need to be undertaken. We have left the logistics of removing redundant tests, if it is deemed appropriate, for those in the field.

The average population utilities for people aged 30 and 40 years were estimated by the authors, as these were not reported in the source used for utilities. It is possible that these utilities may slightly overestimate the true utilities. If this is the case, the cost–utility ratios predicted for thrombophilia testing at these ages would be slightly favourable to thrombophilia testing as the QALYs lost because of an adverse event or mortality would be lower. It is not expected that this would alter the conclusions produced by the modelling as the cost per QALY ratios at these ages were not bordering on the £20,000 threshold.

Only warfarin has been evaluated as an intervention aimed at reducing recurrent VTE. The cost-effectiveness of other potential interventions is an area for future research.

Strengths

There was a comprehensive literature search, which was unlikely to have missed any relevant articles. A mathematical model was constructed that allowed the risks of recurrence to vary by age, sex, previous VTE event and type of thrombophilia. The individual patient approach allowed the increased risks in the period immediately following a VTE to be considered throughout the lifetime of a patient. To our knowledge this is the first model that incorporates these features. If the estimated results from our model are correct then the most cost-effective period of warfarin treatment from those considered can easily be interpreted from two tables, one for patients presenting with an idiopathic DVT and one for those presenting with an idiopathic PE. However, these results come with the caveat that there is a great deal of uncertainty in key parameters, which has resulted in wide confidence intervals for the cost per QALY of thrombophilia testing.

Limitations

The question of whether alterations in anticoagulation management result from thrombophilia testing is flawed as, at the time of writing this report, a change in anticoagulation management is not currently undertaken according to a diagnosis of heritable thrombophilia. 17,26 The data for the model population are limited with wide confidence intervals around many key parameters, which limits the robustness of the cost per QALY ratios. Only warfarin has been evaluated as an intervention to prevent recurrent VTE. We did not undertake systematic reviews for all of the parameters in the model, relying on previous economic evaluations and non-systematic reviews when necessary. It is possible that some relevant data were missed. A paper published after the search dates for this review indicated that there would be little advantage in increasing the duration of oral anticoagulant therapy from 3 to 6 months;77 however, the paper is not of direct relevance to this review because it excluded patients with known thrombophilia, was not restricted to first events and was not restricted to idiopathic events.

Clinical effectiveness

1. clinical trial.pt. (225210)

2. meta$.pt. (11138)

3. review.pt. (683286)

4. exp review literature/(2443)

5. exp clinical trials/(87937)

6. meta-analysis/(4893)

7. exp guidelines/(47776)

8. health planning guidelines/(916)

9. or/1–8 (992360)

10. randomized controlled trials/(36135)

11. controlled clinical trial.pt. (26436)

12. randomized controlled trials/(36135)

13. random allocation/(21037)

14. double blind method/(42876)

15. single blind method/(7621)

16. or/10–15 (128447)

17. clinical trial.pt. (225210)

18. exp clinical trials/(87937)

19. (clin$adj25 trial$).tw. (79059)

20. ((singl$or doubl$or trebl$or tripl$) adj25 (blind$or mask$)).tw. (41776)

21. placebos/(7536)

22. placebo$.tw. (52195)

23. random$.tw. (220009)

24. research design/(23074)

25. or/17–24 (475260)

26. Comparative Study/(560069)

27. exp evaluation studies/(279253)

28. follow-up studies/(157398)

29. prospective studies/(131914)

30. (control$or prospectiv$or volunteer$).tw. (894853)

31. or/26–30 (1600442)

32. 16 or 25 or 31 (1758984)

33. animal/(1390937)

34. human/(3698913)

35. 33 not 34 (936064)

36. 32 not 35 (1425207)

37. exp “Sensitivity and Specificity”/(167664)

38. sensitivity.tw. (151370)

39. specificity.tw. (97754)

40. ((pre-test or pretest) adj probability).tw. (454)

41. post-test probability.tw. (131)

42. predictive value$.tw. (22089)

43. likelihood ratio$.tw. (2584)

44. or/37–43 (331092)

45. exp case-control studies/(222729)

46. case control stud$.mp. (70352)

47. exp cohort studies/(323926)

48. cohort analysis.mp. (796)

49. exp longitudinal studies/(288791)

50. exp prospective studies/(131914)

51. exp follow-up studies/(157398)

52. cohort$.tw. (71183)

53. or/45–52 (523656)

54. meta-analysis/(4893)

55. meta analy$.tw. (12804)

56. metaanaly$.tw. (469)

57. meta analysis.pt. (11138)

58. (systematic adj (review$1 or overview$1)).tw. (8999)

59. exp review literature/(2443)

60. or/54–59 (27802)

61. cochrane.ab. (6521)

62. embase.ab. (4721)

63. (psychlit or psyclit).ab. (655)

64. (psychinfo or psycinfo).ab. (864)

65. (cinahl or cinal).ab. (1773)

66. science citation index.ab. (556)

67. bids.ab. (168)

68. cancerlit.ab. (312)

69. or/61–68 (9349)

70. reference list$.ab. (2752)

71. bibliograph$.ab. (4046)

72. hand-search$.ab. (1317)

73. relevant journals.ab. (222)

74. manual search$.ab. (664)

75. or/70–74 (7994)

76. selection criteria.ab. (6672)

77. data extraction.ab. (2971)

78. 76 or 77 (9084)

79. review.pt. (683286)

80. 78 and 79 (6373)

81. comment.pt. (214279)

82. letter.pt. (259780)

83. editorial.pt. (112306)

84. animal/(1390937)

85. human/(3698913)

86. 84 not (84 and 85) (936064)

87. or/81–83,86 (1328208)

88. 60 or 69 or 75 or 78 (38808)

89. 88 not 87 (35917)

90. 9 or 36 or 44 or 45 or 46 or 53 or 89 (2257794)

91. factor v Leiden.mp. (2486)

92. activated protein c resistance.mp. or exp activated protein c resistance/(1007)

93. apc resistance.mp. (468)

94. exp protein c deficiency/(565)

95. protein c deficienc$.tw. (410)

96. exp protein s deficiency/(591)

97. protein s deficienc$.tw. (549)

98. exp antithrombin III deficiency/(259)

99. anti thrombin deficienc$.tw. (0)

100. antithrombin deficienc$.tw. (200)

101. antiphospholipid antibod$.tw. (2560)

102. antiphospholipid antibodies.mp. or exp Antibodies, Antiphospholipid/(4074)

103. lupus anticoagulant.mp. or exp lupus coagulation inhibitor/(1582)

104. anticardiolipin antibodies.mp. or exp Antibodies, Anticardiolipin/(2015)

105. homocysteine.mp. or exp Homocysteine/(8131)

106. dysfibrinogenaemia.mp. (13)

107. factor VIII.mp. or exp factor VIII/(5055)

108. factor 8.mp. (1036)

109. d-dimer.mp. (2315)

110. factor IX.mp. or exp Factor IX/(1523)

111. factor 9.mp. (323)

112. factor XI.mp. or exp Factor XI/(450)

113. factor 11.mp. (39)

114. dilute russell viper venom time.mp. (33)

115. prothrombin G20210A.mp. (385)

116. MTHFR C677T.mp. (435)

117. kaolin clotting time.mp. (38)

118. or/91–117 (26120)

119. thrombophilia.mp. or exp Thrombophilia/(7295)

120. mass screening.mp. or exp Mass Screening/(43753)

121. screen$.mp. (159255)

122. test$.mp. (765004)

123. exp “diagnostic techniques and procedures”/or diagnostic tests, routine/(1024968)

124. (diagnostic test$and procedure$).mp. (1579)

125. or/120–124 (1629523)

126. 119 and 125 (2335)

127. 118 or 126 (27217)

128. deep vein thrombosis.mp. or exp Venous Thrombosis/(14230)

129. dvt.mp. (2271)

130. pulmonary embolism.mp. or exp Pulmonary Embolism/(9702)

131. pe.mp. (7031)

132. venous thromboembolism.mp. (3551)

133. vte.mp. (1159)

134. stroke.mp. or exp Cerebrovascular Accident/(58898)

135. cva.mp. (563)

136. peripheral vascular disease$.mp. or exp Peripheral Vascular Diseases/(5229)

137. pvd.mp. (511)

138. myocardial infarction.mp. or exp Myocardial Infarction/(50751)

139. mi.mp. (9201)

140. coronary heart disease.mp. or exp Coronary Disease/(62297)

141. chd.mp. (5642)

142. exp Lateral Sinus Thrombosis/or exp Hepatic Vein Thrombosis/or exp Sagittal Sinus Thrombosis/or exp Thrombosis/or exp Coronary Thrombosis/or exp Sinus Thrombosis, Intracranial/or exp Cavernous Sinus Thrombosis/or exp “Intracranial Embolism and Thrombosis”/or exp Carotid Artery Thrombosis/or exp Venous Thrombosis/or exp Intracranial Thrombosis/(37174)

143. exp Embolism/(20440)

144. exp Thromboembolism/(10415)

145. thrombo$.mp. (94723)

146. thromboembolism$.mp. (10235)

147. embol$.mp. (32214)

148. occlu$.mp. (57340)

149. or/128–148 (292284)

150. exp Anticoagulants/(43413)

151. anticoag$.mp. (28066)

152. warfarin.mp. or exp Warfarin/(6252)

153. blood coagulation test$.mp. or exp Blood Coagulation Tests/(6685)

154. or/150–153 (55019)

155. 90 and 127 and 149 and 154 (3411)

156. *Pregnancy Complications/(10779)

157. *Pregnancy Outcome/(5261)

158. *Abortion, Spontaneous/(1519)

159. *Contraceptives, Oral/(1495)

160. *Hormone Replacement Therapy/(2483)

161. *Estrogen Replacement Therapy/(5519)

162. or/156–161 (25911)

163. pregnancy complication$.ti. (141)

164. pregnancy outcome$.ti. (1336)

165. pregnancy loss$.ti. (437)

166. miscarriage$.ti. (644)

167. foet$.ti. (1044)

168. puerperium$.ti. (257)

169. oral contraceptive$.ti. (1781)

170. oral contraception.ti. (125)

171. hormone replacment therap$.ti. (0)

172. oestrogen therap$.ti. (26)

173. estrogen therap$.ti. (209)

174. oestrogen replacement.ti. (60)

175. estrogen replacement.ti. (606)

176. or/163–175 (6620)

177. 162 or 176 (28991)

178. 155 not 177 (4908)

Cost-effectiveness

1. factor v leiden.mp.

2. activated protein c resistance.mp. or exp activated protein c resistance/

3. apc resistance.mp.

4. exp protein c deficiency/

5. protein c deficienc$.tw.

6. exp protein s deficiency/

7. protein s deficienc$.tw.

8. exp antithrombin III deficiency/

9. anti thrombin deficienc$.tw.

10. antithrombin deficienc$.tw.

11. antiphospholipid antibod$.tw.

12. antiphospholipid antibodies.mp. or exp Antibodies, Antiphospholipid/

13. lupus anticoagulant.mp. or exp lupus coagulation inhibitor/

14. anticardiolipin antibodies.mp. or exp Antibodies, Anticardiolipin/

15. homocysteine.mp. or exp Homocysteine/

16. dysfibrinogenaemia.mp.

17. factor VIII.mp. or exp factor VIII/

18. factor 8.mp.

19. d-dimer.mp.

20. factor IX.mp. or exp Factor IX/

21. factor 9.mp.

22. factor XI.mp. or exp Factor XI/

23. factor 11.mp.

24. dilute russell viper venom time.mp.

25. prothrombin G20210A.mp.

26. MTHFR C677T.mp.

27. kaolin clotting time.mp.

28. or/1–27

29. thrombophilia.mp. or exp Thrombophilia/

30. mass screening.mp. or exp Mass Screening/

31. screen$.mp.

32. test$.mp.

33. exp “diagnostic techniques and procedures”/or diagnostic tests, routine/

34. (diagnostic test$and procedure$).mp.

35. or/30–34

36. 29 and 35

37. 28 or 36

38. deep vein thrombosis.mp. or exp Venous Thrombosis/

39. dvt.mp.

40. pulmonary embolism.mp. or exp Pulmonary Embolism/

41. pe.mp.

42. venous thromboembolism.mp.

43. vte.mp.

44. stroke.mp. or exp Cerebrovascular Accident/

45. cva.mp.

46. peripheral vascular disease$.mp. or exp Peripheral Vascular Diseases/

47. pvd.mp.

48. myocardial infarction.mp. or exp Myocardial Infarction/

49. mi.mp.

50. coronary heart disease.mp. or exp Coronary Disease/

51. chd.mp.

52. exp Lateral Sinus Thrombosis/or exp Hepatic Vein Thrombosis/or exp Sagittal Sinus Thrombosis/or exp Thrombosis/or exp Coronary Thrombosis/or exp Sinus Thrombosis, Intracranial/or exp Cavernous Sinus Thrombosis/or exp “Intracranial Embolism and Thrombosis”/or exp Carotid Artery Thrombosis/or exp Venous Thrombosis/or exp Intracranial Thrombosis/

53. exp Embolism/

54. exp Thromboembolism/

55. thrombo$.mp.

56. thromboembolism$.mp.

57. embol$.mp.

58. occlu$.mp.

59. or/38–58

60. exp Anticoagulants/

61. anticoag$.mp.

62. warfarin.mp. or exp Warfarin/

63. blood coagulation test$.mp. or exp Blood Coagulation Tests/

64. or/60–63

65. *Pregnancy Complications/

66. *Pregnancy Outcome/

67. *Abortion, Spontaneous/

68. *Contraceptives, Oral/

69. *Hormone Replacement Therapy/

70. *Estrogen Replacement Therapy/

71. or/65–70

72. pregnancy complication$.ti.

73. pregnancy outcome$.ti.

74. pregnancy loss$.ti.

75. miscarriage$.ti.

76. foet$.ti.

77. puerperium$.ti.

78. oral contraceptive$.ti.

79. oral contraception.ti.

80. hormone replacment therap$.ti.

81. oestrogen therap$.ti.

82. estrogen therap$.ti.

83. oestrogen replacement.ti.

84. estrogen replacement.ti.

85. or/72–84

86. 71 or 85

87. Economics/

88. exp “Costs and Cost Analysis”/

89. economic value of life/

90. exp economics hospital/

91. exp economics medical/

92. economics nursing/

93. exp models economic/

94. Economics, Pharmaceutical/

95. exp “Fees and Charges”/

96. exp budgets/

97. ec.fs.

98. (cost or costs or costed or costly or costing$).tw.

99. (economic$or pharmacoecomomic$or price$or pricing$).tw.

100. quality adjusted life years/

101. (qaly or qaly$).af.

102. or/87–101

103. 37 and 59 and 64 and 102

104. 103 not 86

Men with a previous DVT

The cost per QALY of alternative treatment periods was compared with a standard 3-month treatment period assuming that the thrombophilic status of the patient is known without cost.

The most cost-effective strategy at each age assuming a MAICER of £20,000 is shaded and was established by undertaking incremental analyses (data not shown). No shading denotes that the standard 3-month treatment period is most cost-effective.

Women with a previous DVT

The cost per QALY of alternative treatment periods was compared with a standard 3-month treatment period assuming that the thrombophilic status of the patient is known without cost.

The most cost-effective strategy at each age assuming a MAICER of £20,000 is shaded and was established by undertaking incremental analyses (data not shown). No shading denotes that the standard 3-month treatment period is most cost-effective.

Men with a previous PE

Cost per QALY of alternative treatment periods wascompared with a standard 3-month treatment period assuming that the thrombophilic status of the patient is known without cost.

The most cost-effective strategy at each age assuming a MAICER of £20,000 is shaded and was established by undertaking incremental analyses (data not shown). No shading denotes that the standard 3-month treatment period is most cost-effective.

Women with a previous PE

The cost per QALY of alternative treatment periods was compared with a standard 3-month treatment period assuming that the thrombophilic status of the patient is known without cost.

The most cost-effective strategy at each age assuming a MAICER of £20,000 is shaded and was established by undertaking incremental analyses (data not shown). No shading denotes that the standard 3-month treatment period is most cost-effective.