Professor Ian A Greer, Division of Developmental Medicine, University of Glasgow, Glasgow Royal Infirmary, 10 Alexandra Parade, Glasgow G31 2ER, UK. E-mail: email@example.com
Laboratory testing for the identification of heritable thrombophilia in high-risk patient groups have become common practice; however, indiscriminate testing of all patients is unjustified. The objective of this study was to evaluate the cost-effectiveness of universal and selective history-based thrombophilia screening relative to no screening, from the perspective of the UK National Health Service, in women prior to prescribing combined oral contraceptives and hormone replacement therapy, women during pregnancy and patients prior to major orthopaedic surgery. A decision analysis model was developed, and data from meta-analysis, the literature and two Delphi studies were incorporated in the model. Incremental cost-effectiveness ratios (ICERs) for screening compared with no screening was calculated for each patient group. Of all the patient groups evaluated, universal screening of women prior to prescribing hormone replacement therapy was the most cost-effective (ICER £6824). In contrast, universal screening of women prior to prescribing combined oral contraceptives was the least cost-effective strategy (ICER £202 402). Selective thrombophilia screening based on previous personal and/or family history of venous thromboembolism was more cost-effective than universal screening in all the patient groups evaluated.
Thrombophilia has been defined as ‘disorders of the haemostatic mechanisms which are likely to predispose to thrombosis’ (Walker et al, 2001). Heritable thrombophilias include deficiencies of antithrombin, protein C and protein S, and common genetic mutations, such as factor V Leiden (FVL), the prothrombin G20210A mutation, and the thermolabile variant of the methylene tetrahydrofolate reductase (MTHFR) gene (Walker et al, 2001). Other relatively common thrombophilias with a combination of heritable and acquired components include elevated plasma factor VIIIc (Koster et al, 1995), hyperhomocysteinaemia (den Heijer et al, 1996) and acquired activated protein C resistance (Clark & Walker, 2001).
In recent years, there has been a rapid increase in the understanding of the contribution of these thrombophilic defects to the occurrence of venous thromboembolism (VTE). This had resulted in growing pressure to initiate laboratory tests on an increasing number of patients. Performance of a comprehensive laboratory screen for thrombophilia has become common practice in individuals presenting with deep vein thrombosis or pulmonary embolism. Detection of a heritable prothrombotic states may also lead to screening of family members. It has been estimated that approximately 25 000 tests for activated protein C resistance/FVL are performed each year in the UK alone (Greaves & Baglin, 2000).
Few studies in the literature have attempted to examine the cost-effectiveness of screening for some thrombophilias in different patient groups (Creinin et al, 1999; Palareti et al, 1999; Marchetti et al, 2001; Clark et al, 2002; Eckman et al, 2002). The cost of screening for thrombophilia in women prior to prescribing oral contraceptives has been shown to range from US$433 to detect one case of increased activated protein C resistance to US$7795 for protein S deficiency (Palareti et al, 1999). In another study, Creinin et al (1999) estimated that over 92 000 FVL carriers would need to be identified, at costs in excess of US$300 million to prevent one venous thromboembolic death attributable to the use of oral contraceptives.
Another study has evaluated the cost-effectiveness of FVL screening in a hypothetical female population who had prior VTE events (Eckman et al, 2002). The study examined three hypothetical cohorts: (i) all patients receiving standard anticoagulation therapy for 6 months without testing, (ii) testing followed by all FVL positive patients receiving 3 years of anticoagulant therapy and (iii) testing followed by all FVL positive patients receiving life-long anticoagulant therapy. The study concluded that of the three scenarios examined, testing and treating with 3 years of anticoagulation therapy was the preferred screening strategy. However, this was based on a very small margin of relative cost-effectiveness [$279·33 per quality-adjusted life year (QALY) with testing followed by three years of treatment compared with $299·39 per QALY with no screening]; therefore, screening for FVL is unlikely to be cost-effective. More recently, a cost-effectiveness analysis evaluated the testing of thrombophilia followed by anticoagulation for 6–36 months in patients with idiopathic deep vein thrombosis (Auerbach et al, 2004). This study showed that testing for thrombophilia followed by 2 years of anticoagulation was more cost-effective ($54 820; 23·76 QALYs) than the standard practice of 6 months anticoagulation without thrombophilia testing ($55 260; 23·72 QALYs).
Only one UK study has assessed the cost-effectiveness of thrombophilia screening and concluded that neither universal nor selective screening based on prior history of VTE was cost-effective in pregnancy (Clark et al, 2002). Based on data from a prospective cohort (n = 967), this study reported an additional management cost of £7535 with selective history-based screening and £13 281 with universal screening, when compared with no screening for FVL to prevent one vascular event.
As with most screening strategies, indiscriminate thrombophilia screening is inappropriate and unjustified (Walker et al, 2001; Machin, 2003). In contrast, a targeted screening approach based on specific high risk patient groups or additional risk factors may be of some value (Martinelli, 2003). However, which high-risk patient group benefits the most, clinically and economically from thrombophilia screening is not clear. The objective of this study was to assess the relative cost-effectiveness of a complete thrombophilia screen from the perspective of the UK National Health Service (NHS), in women prior to prescribing oral oestrogen preparations including combined oral contraceptives and hormone replacement therapy, in women at the onset of pregnancy, and in patients prior to major orthopaedic surgery, compared with no screening.
An incremental cost-effectiveness analysis was conducted, from the perspective of the NHS in the UK, to determine the relative cost-effectiveness of universal and selective screening based on a personal and/or family history of VTE compared with no screening for thrombophilia. A decision analytical model was developed to analyse the range of possible clinical events associated with screening and no screening in hypothetical populations (Fig 1). Four thrombophilia screening scenarios were evaluated:
1Thrombophilia screening in a hypothetical population of 10 000 women prior to prescribing combined oral contraceptives. Those tested positive would be perceived as at increased risk of VTE and would not be prescribed combined oral contraceptives, so avoiding the risk of VTE.
2Thrombophilia screening in a hypothetical population of 10 000 women at the onset of pregnancy (at week 6 of gestation). Those tested positive would be perceived as at increased risk of VTE and adverse pregnancy outcomes. These women would be prescribed prophylaxis to prevent VTE and early pregnancy loss.
3Thrombophilia screening in a hypothetical population of 10 000 women prior to prescribing hormone replacement therapy. Those tested positive would be perceived as having an increased risk of VTE and would not be prescribed hormone replacement therapy, so avoiding the risk of VTE.
4Thrombophilia screening in a hypothetical population of 10 000 patients prior to major elective orthopaedic surgery. These patients would already be given standard thromboprophylaxis, which is common clinical practice. Those tested positive would be perceived as having an increased risk of VTE and would be given extended thromboprophylaxis to prevent VTE events.
In the universal screening model, all the patients in each of the four groups would be tested for thrombophilia. However, in the selective screening model, only those who have a previous personal and/or family history of VTE would be tested for thrombophilia.
Major assumptions and model inputs
For the purpose of modelling, it was assumed that thrombophilia screening comprised of testing for FVL, prothrombin G20210A, deficiencies of antithrombin, protein C and protein S, lupus anticoagulants and anticardiolipin antibodies. There are limited data in the literature on the sensitivity and specificity of the individual tests for the various thrombophilias. Therefore, based on the existing data on the accuracy of laboratory diagnosis of familial thrombophilia (Preston et al, 2003), the overall test sensitivity and specificity of the thrombophilia screening tests were assumed to be 80%. This is a conservative estimate and was tested in the sensitivity analysis.
Women who were tested positive in the combined oral contraceptives and hormone replacement therapy groups were not given oral oestrogen preparations or thromboprophylaxis. In the pregnancy arm of the model, those tested positive would be given antenatal and 6 weeks postnatal prophylactic low molecular weight heparin (Nelson-Piercy, 2004). It was assumed that this would result in a 50% reduction in the occurrence of vascular complications including VTE and early pregnancy loss (Clark et al, 2002). It is routine practice to prescribe thromboprophylaxis for the prevention of VTE in patients undergoing major elective orthopaedic surgery; therefore, those tested positive for thrombophilia would be given additional extended prophylaxis for 4 weeks. It was assumed that this would result in a 50% reduction in the occurrence of VTE (SIGN, 2002).
The proportion of patients estimated to have had a prior personal and/or family history of VTE were based on data from the literature and expert opinion. Based on the data from a large prospective cohort study, the proportion of women who have had a prior personal and/or family history of VTE was estimated to be 12% (Clark et al, 2002). It is recognised that the risk of VTE is associated with age (Stein et al, 2004). Women in the pregnancy group were likely to be of similar age to those in the combined oral contraceptives group. Therefore, the same proportion of prior personal and/or family history of VTE (12%) was also applied to the combined oral contraceptives group. Similar data on prior VTE history data relating to women taking hormone replacement therapy or patients undergoing major elective orthopaedic surgery were not found in the literature. Through discussions with experts in vascular medicine and orthopaedics and taking into account the age factor, proportions of patients with prior VTE history in the hormone replacement therapy group and the orthopaedic surgery group were assumed to be 15% and 20% respectively. These assumptions were tested in the sensitivity analysis.
Major clinical outcomes were defined as VTE including deep vein thrombosis and pulmonary embolism, and in the pregnancy arm of the model, adverse pregnancy outcomes including early pregnancy loss, late pregnancy loss, preeclampsia (defined as mild and severe), placental abruption and intrauterine growth restriction (IUGR), were also incorporated in the model. The respective baseline probabilities and thrombophilia prevalences used in the model were based on data from published literature (Table I). The risks of VTE in thrombophilic patients while taking combined oral contraceptives, during pregnancy, while taking hormone replacement therapy and during major elective orthopaedic surgery were determined by an extensive systematic review and meta-analysis (Robertson et al, 2004; Wu et al, 2005a,b). Similarly, the risks associated with individual pregnancy complications in thrombophilic patients who were pregnant were also calculated.
Table I. Probability of events in the model base-case.
An extensive search on all major electronic databases was carried out, supplemented by manual searching of references cited in the relevant studies. Both prospective and retrospective studies of VTE events and thrombophilia in women taking combined oral contraceptives, women taking hormone replacement therapy and patients undergoing major elective orthopaedic surgery; and studies of VTE events and adverse obstetric complications in women with thrombophilia during pregnancy were assessed according to predefined inclusion and exclusion criteria. Relevant data were extracted from all the studies that met the inclusion criteria and their methodological quality were assessed. Odds ratios associated with individual clinical outcomes, stratified by thrombophilia type were calculated for each patient group. Where appropriate, meta-analysis was conducted based on the random effects model and testing of heterogeneity was carried out by standard chi-squared test. Subsequently, the estimated odds ratios for VTE and adverse pregnancy complications associated with individual thrombophilic defects in each patient group were converted into probabilities, taking into account the background rate of events in patients with no additional risks (Table I).
The clinical management strategy and healthcare resource use associated with all major adverse clinical complications were obtained from two Delphi studies (Jones & Hunter, 1995). Two questionnaires requiring quantitative and qualitative answers, regarding the clinical management of VTE and pregnancy complications in women and VTE in orthopaedic patients were designed and pre-piloted among a small of group of consultants in obstetrics and orthopaedics. Following feedback from the pilot group, appropriate revisions were made to the questionnaires.
The two questionnaires regarding clinical management of VTE and pregnancy complications in women and VTE in orthopaedic patients, were sent to all consultants of obstetrics (n = 108) and orthopaedics (n = 115) in Scotland by post and by email. Respondents were asked to indicate the routine diagnostic and treatment strategies used in patients with DVT, pulmonary embolism and various adverse pregnancy outcomes. In addition, they were also asked to estimate, if any, the average length of hospital stay associated with these clinical complications.
A two-round Delphi study was originally intended, where the results from the first round would be summarised and fed back to the respondents through a second questionnaire. However, the results from the first round of the study showed high level of convergence among the responses and average management strategies to the various clinical complications were recorded. Therefore, a second round was not conducted. There was divergence in the estimated length of hospital stay associated with various clinical complications, but due to the nature of the modelling, an absolute agreed length of stay is not necessary and, indeed, unlikely in clinical practice. As a result, the median length of stay was used in the base-case scenario, while the range obtained from the Delphi study was used in the sensitivity analysis.
Only direct health service costs were measured. The costs associated with thrombophilia screening and managing associated adverse clinical complications were calculated. The cost of thrombophilia screening consisted of the purchasing and processing cost of the diagnostic tests, staff time and the cost of prophylaxis and extended prophylaxis in the pregnancy and orthopaedics arm respectively. The costs associated with managing adverse clinical complications included costs of all diagnostic investigations, hospitalisations, outpatient consultations, counselling and drug treatments.
Unit costs for all healthcare resources used were obtained from routinely collected data and the literature. These were combined with the quantity of resource use, which were determined by the results of the Delphi study reflecting real clinical practice, to obtain a net cost per patient associated with various major clinical complications (Table II). All costs were calculated at 2002 values (UK £).
Table II. Indicative resource use and unit costs in the base-case analysis.
Average resource use
Unit costs (2002 UK£)
Sources of unit costs
*One test per person screened; in addition, those tested positive would receive a repeat test to confirm results. The most commonly prescribed combined oral contraceptive†, hormone replacement therapy‡ and transdermal hormone replacement therapy§ in Scotland 2003 (Information and Statistics Division, 2001). The second most commonly prescribed combined oral contraceptives‡, hormone replacement therapy¶ and transdermal hormone replacement therapy** in Scotland 2003 (Information and Statistics Division, 2001).
LMWH, low molecular weight heparin; BNF, British National Formulary.
Thrombophilia Screen* Testing for factor V Leiden, prothrombin G20210A, antithrombin deficiencies, protein C deficiencies, protein S deficiencies, lupus anticoagulants and anticardiolipin antibodies
Clinical Services Division, Laboratory Directorate, North Glasgow University Hospitals NHS Trust
Cost-effectiveness is measured as a ratio of cost to effectiveness. The effectiveness of screening was measured by the number of major clinical outcomes averted. The incremental cost-effectiveness ratio (ICER) is an estimate of the cost per unit of effectiveness of one strategy in preference to another. Therefore, the lower the ICER, the more cost-effective the strategy. In this study, ICERs presented as net costs per major clinical outcome averted, comparing universal and selective screening with no screening, were calculated for each individual patient group. ICERs were calculated by dividing the difference in cost (in this case, costs associated with screening and treating the major clinical outcomes that the particular strategy failed to prevent) by the difference in effectiveness (the number of major clinical outcomes prevented by the particular strategy) in the comparison groups.
Univariate sensitivity analysis was carried out to test the robustness of the base-case analysis to changes in key parameters. The parameters that were the least confident and those that were expected to have significant impact on the estimated cost-effectiveness of thrombophilia screening were tested. These include thrombophilia test sensitivity and specificity (sensitivity range 50–100% was explored), the effectiveness of the prophylaxis in preventing major clinical outcomes (sensitivity range 20–80% explored), the impact of varying unit costs data (inflated and reduced by 20%) and model input probabilities (the extreme values of the 95% confidence intervals associated with the calculated odds ratios were used).
Scenario analysis was also conducted to test other assumptions made in the model. The most commonly prescribed combined oral contraceptive (Microgynon 30®; Schering Health, West Sussex, UK) and hormone replacement therapy (Premique®; Wyeth, Berks, UK), based on national prescribing data in Scotland (Information and Statistics Division, 2001), were selected for the respective screening arms. This was tested using the second most commonly prescribed oral oestrogen preparations (Cilest®; Janssen-Cilag, Bucks, UK and Premarin®; Wyeth, Berks, UK, respectively) in the sensitivity analysis.
In the case of hormone replacement therapy, the literature suggested that transdermal preparations did not incur similar risks to oral preparations (Daly et al, 1996; Perez Gutthann et al, 1997; Scarabin et al, 2003). Therefore, it has been suggested that transdermal hormone replacement therapy should be considered in women who are at increased risk of VTE (SIGN, 2002). A scenario analysis was carried out to investigate the cost-effectiveness of prescribing transdermal preparations to women who were tested positive for thrombophilia.
In the absence of the knowledge of thrombophilia status, it is common in some clinical practice to prescribe prophylaxis to all patients with prior personal and/or family history of VTE without testing for thrombophilia. Therefore, scenario analysis was also carried out to examine the cost-effectiveness of no thrombophilia testing, but prescribing prophylaxis and extended prophylaxis based on the patients’ VTE history in the pregnancy and orthopaedic surgery groups respectively.
The estimated event probabilities determined from the literature and meta-analysis, and the indicative resource use determined from the Delphi studies are detailed in Tables I and II.
Based on a hypothetical model of 10 000 patients in each screening scenario, in the absence of thrombophilia screening, adverse clinical complications would be found in approximately seven women on combined oral contraceptives, 2921 pregnant women, 104 women on hormone replacement therapy and 1265 patients undergoing major orthopaedic surgery, at costs of £119 147, £513 592, £1 185 428 and £1 217 935 respectively (Table III). From a pure cost perspective, in this cohort, thrombophilia screening in women prior to prescribing combined oral contraceptives and restricting prescribing to those tested negative for thrombophilia was the least costly strategy to implement (approximately £709 640); while screening women at the onset of pregnancy and prescribing prophylaxis to those tested positive for thrombophilia was the most expensive (£5 384 320).
Table III. Clinical complications averted, costs and incremental cost-effectiveness ratios (ICERs) by screening strategies (n = 10 000).
Clinical complications prevented
All figures are rounded to two decimal places.
*ICER = Δcosts/Δbenefits; i.e. costs (screening − no screening)/clinical complications prevented (screening − no screening).
Combined oral oestrogen
5 384 319·69
Hormone replacement therapy
1 185 427·59
1 469 464·38
1 217 935·42
2 466 342·94
Combined oral oestrogen
1 081 232·90
Hormone replacement therapy
1 185 427·59
1 220 316·35
1 217 935·42
1 459 103·05
However, when taking effectiveness of screening into account, universal screening of patients prior to prescribing hormone replacement therapy and restricting prescribing to those tested negative for thrombophilia would prevent 42 VTE events in this hypothetical population and was the most cost-effective screening strategy (ICER £6824). In contrast, screening women prior to prescribing combined oral contraceptives would only prevent three VTE events and was the least cost-effective strategy (ICER £200 402).
Overall, selective screening based on the presence of previous personal and/or family history of VTE prevented fewer cases of adverse clinical complications but was more cost-effective than universal screening in all four screening scenarios. The most significant improvement in cost-effectiveness was observed with the hormone replacement therapy and the combined oral contraceptives groups, when the ICERs for selective history-based screening was reduced by approximately 60% (from £6824 to £2447) and 64% (from £200 402 to £79 085) respectively.
One-way, univariate sensitivity analysis showed that the results of the model were relatively robust (Fig 2). The model was most sensitive to test sensitivity and specificity, but changes in the key parameters did not alter the overall results. Screening women prior to prescribing hormone replacement therapy remained the most cost-effective strategy when test sensitivity and specificity, effectiveness of prophylaxis, unit costs and probabilities of developing adverse clinical complications were varied individually.
Scenario analysis was conducted to test the scenario of prescribing transdermal hormone replacement therapy in place of withholding therapy for those tested positive for thrombophilia. The risk of VTE associated with transdermal preparations are much lower (OR 1·27) than that observed in oral preparations (OR 3·16). In this hypothetical population of 10 000, the prescription of transdermal preparations to those tested positive for thrombophilia would incur additional costs of approximately £491 434, resulting in a total cost of £1 676 862 and an ICER of approximately £12 404 for this strategy.
The purchasing cost of the second most commonly prescribed combined oral contraceptives were greater than the most commonly prescribed preparations (Table II). However, this was not the case with oral hormone replacement therapy. Marginal improvement on cost-effectiveness (ICER = £186 905) was observed with combined oral contraceptives, however, substituting with the second most commonly prescribed hormone replacement therapy was less cost-effective than the base-case – the costs per event prevented were greater with oral hormone replacement therapy (ICER = £11 440). However, the relative cost-effectiveness between the groups remain unchanged.
Scenario analysis on no thrombophilia testing and prescribing prophylaxis to those with a VTE history resulted in ICERs of £192 728 and £15 317, for the pregnancy and the orthopaedic surgery group respectively. The cost of prescribing prophylaxis to all patients who had a prior personal and/or family history of VTE without thrombophilia testing was less cost-effective than screening followed by prescribing prophylaxis to those tested positive.
The total cost of screening for thrombophilia in a hypothetical population of 10 000 ranged from £709 640 (combined oral contraceptives group) to £5 384 320 (pregnancy group). In comparison with no screening, universal screening of women prior to prescribing hormone replacement therapy was the most cost-effective strategy, at a net cost of £6824 per adverse clinical complication prevented. Selective VTE history-based screening was more cost-effective than universal screening in all the patient groups examined in this study. Subsequently, screening women with a personal or family history prior to prescribing hormone replacement therapy was shown to be the most cost-effective at a net cost of £2446 per adverse clinical complication prevented.
Thrombophilia is associated with a substantial increase in relative risk of VTE, in particular, in patient groups such as women on combined oral contraceptives and hormone replacement therapy, the odds ratios for the combined risk of FVL and taking oral oestrogen preparation was 15·62 and 13·16 respectively. However, in view of the prevalence of thrombophilia, the absolute risk remained low. Therefore, the absolute numbers of expected events and the estimated number of prevented events in these groups were low. This was particularly apparent with the combined oral contraceptives group, when only three VTE events would be prevented in the hypothetical population of 10 000, and subsequently resulting in a large ICER.
Selective screening based on prior personal or family history of VTE has been recommended (Walker et al, 2001). The results of this study support such recommendations and showed selective history-based screening to be more cost-effective than universal screening in all four clinical situations. Nonetheless, the effectiveness of history-based screening is highly dependent on the reliability of the data source and the sensitivity of family history and the accuracy of such data has been reported to be as low as 49% (Schambeck et al, 1997). The findings of this study further highlight the importance and value in accurate recording of patient history.
This is the first study that attempted to evaluate the relative cost-effectiveness of a complete thrombophilia screen in various patient groups in a White population. The prevalence for thrombophilia varies in different populations (Gregg et al, 1997). Irrespective of individual thrombophilic defects, White people remained the most prevalent group, in which screening is relevant. Clark et al (2002) evaluated the cost-effectiveness of universal and selective screening for FVL in pregnancy, and gave comparable results to our pregnancy arm of the cost-effectiveness analysis.
This study has potential limitations that are inherent to all cost-effectiveness analysis. Based on a decision analysis, this study used estimates from several sources, such as probabilities of clinical events reported in the medical literature and expert opinion on management of events. In an attempt to overcome the potential bias, a systematic review and meta-analyses were conducted to estimate probabilities of clinical events and a Delphi study was conducted to determine the average treatment strategy for all the adverse clinical complications, which is believed to reflect current clinical practice. In addition, extensive sensitivity analysis was carried out to examine the effect that variations of model inputs would have on the results. The results of the sensitivity analysis showed that the overall results were robust.
In this analysis, cost-effectiveness was expressed as ‘cost per major clinical outcome prevented’. In the oral oestrogen preparation and the orthopaedic surgery groups, the adverse clinical complications referred to VTE events. However, in the pregnancy, adverse pregnancy outcomes were also considered, therefore, the ‘major clinical outcome’ in this group referred to an aggregation of VTE and adverse pregnancy events. Different clinical complications are of different significance to the NHS and to patients. Although such aggregated measure of outcomes is not ideal, it allows standardised comparison across the patient groups and offers some prioritisation order. In order to take into account the different value of the different clinical events to the NHS and to patients, the method of calculating a generic outcome measure, such as QALYs, may be used. However, in the case of pregnancy, such measure is potentially problematic as the QALYs associated with the fetus need also to be taken into account.
This cost-effectiveness study was taken from the perspective of the NHS. Indirect costs such as loss of production and quality of life impairment associated with VTE and adverse pregnancy outcomes, were not taken into account. This model was designed to investigate the most cost-effective strategy for thrombophilia screening, based on the assumption that a decision has been made to undertake screening and did not consider the relative cost-effectiveness of screening compared with other uses of scarce NHS resources. In order to determine the cost and the relative value of a thrombophilia screening programme to other healthcare programmes, alternative forms of economic evaluation, such as cost-benefit and cost-utility analysis, is required respectively. Currently, there is insufficient data in the literature to allow us to do that. In addition, other important issues, such as acceptability, psychological consequences deriving from the diagnosis of thrombophilia, potential consequences of false-positive and false-negative results, all need to be taken into account.