NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Segal JB, Brotman DJ, Emadi A, et al. Outcomes of Genetic Testing in Adults with a History of Venous Thromboembolism. Rockville (MD): Agency for Healthcare Research and Quality (US); 2009 Jun. (Evidence Reports/Technology Assessments, No. 180.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Outcomes of Genetic Testing in Adults with a History of Venous Thromboembolism

Outcomes of Genetic Testing in Adults with a History of Venous Thromboembolism.

Show details

4Discussion

Our literature review was designed to identify evidence to inform our Key Questions. We identified literature that was directly applicable to answering the questions about analytic and clinical validity (Key Questions 2 and 3), but these studies were only indirectly relevant to the overarching question (Key Question 1). The literature was also less directly applicable to addressing the question about the clinical utility of testing for these mutations (Key Question 4).

Key Question 1: Association of FVL Testing, Alone or in Combination With Prothrombin G20210A Testing, With Improved Clinical Outcomes in Adults With a Personal History of VTE or in Adult Family Members of Mutation-positive Individuals

We found no evidence that directly informed our overarching question (Key Question 1): Does FVL testing, alone or in combination with prothrombin G20210A testing, lead to improved clinical outcomes (e.g., avoidance of a recurrent VTE) in adults with a personal history of VTE or to improved clinical outcomes (e.g., avoidance of an initial VTE) in adult family members of mutation-positive individuals? In our discussion of Key Question 4, below, we address the implications of the absence of direct evidence, the value of indirect evidence, and the implications for future research.

One recent study that did not meet our inclusion criteria nevertheless provides some information to answer this overarching question.165 We could not include this study by Coppens et al. in our review because the patients had a variety of different thrombophilic conditions for which they were tested. Using a case-control design, the authors investigated whether thrombophilia testing reduced the rates of recurrent VTE among persons with a history of VTE. From a cohort of patients enrolled in the Dutch-based Multiple Environmental and Genetic Assessment (MEGA) trial, 197 patients with VTE recurrence and 324 without recurrence were studied. Physicians had ordered thrombophilia testing in 35 percent of cases and 30 percent of controls. The odds ratio for recurrence in the tested versus the non-tested patients was 1.2 (95 percent CI, 0.8–1.9) and changed little after adjusting for age, sex, and VTE risk factors. These results suggest that thrombophilia testing did not affect the rate of VTE recurrence.

Key Question 2: Analytic Validity of Tests to Identify FVL and Prothrombin G20210A Mutations

An assessment of the analytic validity of a genetic test refers not only to the test’s accurate identification of genotypes but also its reliability and robustness. Many of the studies we reviewed were preclinical (and pre-commercial) evaluations of assays. The majority of laboratories in the United States still use PCR-RFLP or invader chemistry technology, and less frequently the assays described in this report. The studies we reviewed demonstrated the high analytic validity of both the commercially available and pre-commercial tests. Most of the discordant results resolved with repetition of the respective test, suggesting that operator or administrative errors were responsible for the discordant results.

The majority of the studies reviewed used the well established and commonly used PCR-RFLP or AS-PCR as the reference (gold) standards. There is a theoretical concern that these non-sequence-based methods may not distinguish between the single nucleotide mutations of interest (e.g., FVL) and other nearby benign polymorphisms and could potentially yield false-positive results. For example, a rare silent A1692C Factor V polymorphism would be mistakenly genotyped as a FVL allele in an RFLP assay. We saw false-positives attributable to this mechanism only very infrequently in the studies we reviewed.

In the quality assurance studies, we found evidence that most laboratories are highly accurate, even perfectly accurate, when asked to classify a sample with a known mutation. The majority of errors came from a limited numbers of laboratories. As we described, among the responding laboratories in one study, 51 percent made at least 1 error; however, three of the 39 laboratories were responsible for 46 percent of all errors.63 We suspect that this situation illustrates a systematic quality assurance defect in these isolated laboratories. We did not review the evidence about how to identify laboratories that may perform poorly. We suggest that our findings underscore the need for ongoing internal and external quality control programs.

There are abundant technologies that can be used to detect the FVL and prothrombin G20210A mutations. Given the almost equal analytic validity of these methods, considerations such as shorter turn-around time, cost-effectiveness, high throughput, and availability of user-friendly software become important factors in selecting one method over another. The choice of methodology will likely be driven by considerations other than analytic validity.

Key Question 3: Clinical Validity of Testing for FVL and Prothrombin G20210A Mutations

Testing Probands

We found moderate-grade evidence that individuals with at least one prior thrombotic event who are homozygous or heterozygous for FVL have a higher risk of recurrent VTE than do those without the mutation. For heterozygous individuals, the odds ratio was 1.56 (95 percent CI, 1.14–2.12); for homozygous individuals, it was 2.65 (95 percent CI, 1.2–6).

There was moderate-grade evidence that prothrombin G20210A is not predictive of recurrent thrombosis. The odds ratio was 1.45 (95 percent CI, 0.96–2.2), similar to that for heterozygous FVL, but the confidence interval overlapped 1.0. There were too few pieces of data to allow us to refute or support the contention that homozygosity for prothrombin G20210A, a rare condition, is associated with recurrent thrombosis.

There was insufficient evidence that double heterozygosity (FVL plus prothrombin G20210A) is predictive of recurrent thrombosis. The pooled odds ratio was 4.8, but the confidence interval was wide, with very few studied individuals having double heterozygosity.

When we separately evaluated patients with idiopathic VTE as the index event, we found that the odds ratio associated with heterozygous FVL was close to one (1.17; 95 percent CI, 0.63–2.18), suggesting that that there may be little predictive value in knowing the mutation status in patients with idiopathic events.

There were limitations in this body of evidence, although the quality of the studies overall was fairly high. There was often insufficient description of potential confounders of the relationship between the mutation and the recurrent event. However, in those studies employing multivariate models that included other potential predictors of recurrent thrombosis, statistical adjustment did not result in any attenuation of the mutation-specific effect size,14, 111,116,120,127 suggesting that these other clinical variables were not driving the relationship between mutation status and recurrent thrombosis. Although not all of the studies were primarily designed to answer our question about risk among probands, the prospective design that we required for inclusion in this review is the optimal design for answering this question.

Testing Family Members

We found high-grade evidence that family members of probands who are homozygous for the FVL mutation have a substantially increased risk of VTE when compared to family members who do not have this mutation. Homozygous family members may have a thrombosis rate nearly 20-fold that of individuals without homoyzgosity, or an event rate of 2 percent per year. There was moderate evidence that the risk of an event for heterozygous individuals is increased roughly three-fold (odds ratio, 3.4; 95 percent CI, 2.4–4.9). The absolute increase in the rate of events, however, was low.

We saw that the annualized rate of venous thromboembolic events for family members without a mutation was approximately 0.1 percent per year. This value translates to an event rate in heterozygous family members of 0.3 percent per year, or an absolute increase of 0.2 percent per year (a change from an average of 1/1000 person-years to 3/1000 person-years).

There was insufficient evidence regarding family members who are heterozygous for prothrombin G20210A, with an odds ratio of 1.9 (95 percent CI, 0.35–10). Doubly heterozygous individuals, with one FVL and one prothrombin G20210A mutation, can be expected to have event rates that are higher than those for singly heterozygous individuals but lower than those of family members who are homozygous for FVL (odds ratio, 6.7; 95 percent CI, 2.9–15) This conclusion was supported by only low-grade evidence. There was little information about the rare condition of homozygosity for the prothrombin G20210A mutation.

There was insufficient evidence to allow us to draw any conclusions about the risks associated with heterozygosity in FVL among pregnant family members. The point estimates were above one in both studies, but with very wide CIs. Because pregnancy and the postpartum period are high-risk times for VTE, the absolute increase in risk with pregnancy is likely to be much greater than the absolute risk increase from FVL. We suspect that the risk attributable to FVL during pregnancy is very small. There was only low-grade evidence that homozygosity for this mutation may increase the risk of venous thrombosis in pregnancy beyond that which is usually seen. The evidence concerning women who are doubly heterozygous for FVL and prothrombin G20210A was insufficient, and we could not draw any conclusions about the risk. We reviewed one additional informative study that could not be included because the pregnant women were either those with a personal history of thrombosis or a family history of thrombosis or a family history of thrombophilia. This study suggested that double heterozygosity does modestly raise the risk to women during their pregnancies.166

Key Question 4 Clinical Utility of Testing for FVL and Prothrombin G20210A Mutations

Effect of Testing on Clinicians’ Management

There was low-grade evidence that patient management by clinicians is altered on the basis of the results of FVL testing, although there was no evidence to indicate whether this action improves patient outcomes or not. No information was available on management practices resulting from testing for prothrombin G20210A. We identified a single study addressing this question, which assessed practitioners’ responses when presented with clinical scenarios.150 The evidence to address this question can be considered to be only indirect, as this study asked clinicians to respond to hypothetical cases; there was no observation of practice patterns. This study was performed in Canada, where clinicians’ management decisions may be influenced by the availability of resources and the health policy environment specific to that country. Also, all of the scenarios described pregnant women, and therefore the results may not be applicable to other patient populations.

We identified one additional study published after the end of our search period.167 In this study by Hindorff et al., 112 primary care physicians in Washington state (60 frequent and 52 infrequent prescribers of FVL testing) responded to a survey about their motivations for testing for FVL (response rate of 67 percent). Approximately 82 percent of the providers indicated they would order FVL in order to advise patients about VTE recurrence, while 67 percent would order the test to make decisions about VTE treatment or prevention. Fewer than 40 percent of the clinicians reported a high level of confidence in interpreting or communicating the results of FVL testing, or a high degree of confidence in determining when it is appropriate to order the test. As in the study by Rodger et al,150 this study did not provide direct evidence that clinicians manage patients differently based on FVL test results. However, the results suggest that physicians ordering this test may use results to inform management decisions. This study also highlighted a significant level of uncertainty regarding when to order FVL testing as well as how to interpret the results.

Effect of Testing and Resultant Management on VTE-related Outcomes

We conclude that there is no direct evidence that testing for FVL and prothrombin G20210A, and the resultant management, reduce VTE related-outcomes. In our search for indirect evidence to support or refute this hypothesis, we found high-grade evidence that anticoagulation reduces recurrent events in patients with FVL or prothrombin G20210A. However, there was only low-grade evidence that the magnitude of this relative reduction in outcomes is comparable to that seen in individuals without these mutations. Thus, the mutation status of the patient apparently does not, of necessity, play a role in the decision to extend anticoagulation in a patient with a history of VTE. This conclusion is based on four studies, none of which was specifically designed to directly answer the question about testing or treatment changes based on testing. The studies did not describe bleeding associated with anticoagulation that was stratified according to mutation.

The four studies we identified were heterogeneous in their designs and treatments. This body of evidence is limited by the relative lack of data on patients with prothrombin G20210A. Although Ridker et al. included individuals with prothrombin G20210A, these authors pooled them with the individuals having FVL for the analysis, so it was not possible to assess the effect of each mutation on the relationship between treatment and outcome.151 Also, some important safety outcomes were not analyzed according to the type of mutation of the participants. Of the three studies that addressed the risks associated with anticoagulation,122,123,151 only one reported these results as a function of mutation status.123

Effect of Testing and Results on Other Outcomes

We found moderate-grade evidence that little change in knowledge and behavior results from testing for FVL or prothrombin G20210A. Four studies directly answered this question; each study was designed to explore how FVL or prothrombin G20210A testing affected non-thromboembolic outcomes. The main themes from these studies were: (1) an individual’s understanding of the risk factors for VTE or the significance of the test results was not improved after testing, unless structured counseling as well as access to information before and after testing was provided; (2) daily life changes were uncommon, although some patients used the test results to make important medical decisions; (3) most individuals did not regard carrier status as a serious condition but tended to worry about the implications for their children and relatives. We conclude that there is moderate evidence that the process of testing for these mutations does not have serious adverse consequences but may possibly improve understanding of VTE risk factors.

The main limitation of these studies was that they were all performed outside the United States. The political and cultural factors in these countries, as well as the healthcare environment, may have affected how the patients experienced genetic testing, interpreted the results, or behaved in response to those results. Another limitation of this body of evidence is that all of the studies involved self-selected participants who were interested in the research question, a situation that could have skewed the study results. Only one study reported how outcomes differed based on the test result. It may be useful for clinicians to know how patients are affected by the results of the test, and not just by the process of testing.

Two other important limitations were the paucity of studies addressing outcomes for patients undergoing prothrombin G20210A testing and the fact that all of the results were exclusively based on patient perceptions and behaviors. No study (except for the one by Wahlander et al.123) addressed clinical outcomes arising from changes in management as a consequence of testing (such as bleeding, mortality, or hospitalization rates) or quantifiable nonclinical outcomes, such as cost to patients or utilization of healthcare services.

Cost-effectiveness of FVL and prothrombin G20210A testing in the care of probands and their relatives

The cost-effectiveness studies all used decision analytic models. These models can point to the need for further testing of the utility of an intervention if the assumptions in the models are compatible with actual practice. The data ranges explored in the sensitivity analyses demonstrate the variables to which the cost-effectiveness of the interventions are most sensitive.

The five studies that modeled the experience of probands following a VTE all suggested that testing for FVL alone or in combination with testing for prothrombin G20210A could be cost-effective in certain patients.159,161–164 The strategy of testing for FVL alone161,163 or in combination with prothrombin G20210A,159,162 with extended anticoagulation for 2 or 3 years for identified carriers, was either the dominant or the most cost-effective option in patient populations having a high prevalence of the mutations (e.g., 5 percent or more), a high risk of recurrence (e.g., 10 percent or more within 2 years), and a low risk of bleeding. The models were not robust. They were extremely sensitive to the parameters chosen for the input, and we challenge some of the assumptions that were used for the base case analyses.

The models’ results were most sensitive to the rate of venous thrombosis recurrence, the prevalence of FVL and prothrombin G20210A mutations, the risk of adverse bleeding events, and anticoagulant efficacy. We compared the model input to the summary results from our meta-analyses. In our analyses, the odds ratio for recurrence for individuals heterozygous for FVL relative to non-carriers was 1.56. This odds ratio is above the threshold suggested by Auerbach et al. at which a test for the mutation should be included in a hypercoagulability panel.159 Also, our pooled odds ratio was slightly higher than the value used in the model by Marchetti et al.; using our data for input would have more strongly favored FVL testing and extended anticoagulation.163

However, there are examples in these models of the use of parameters that seem flawed. One model made the assumption that there is an average risk of recurrence of 16 percent per year for the first 3 years, followed by 0 percent in subsequent years.161 These percentages were based on a single study with a rate of recurrence that we found to be an outlier.120 Similarly, one model assumed that prophylaxis against recurrent VTE with enoxaparin would be 50 percent effective.160 This assumption was based on a study of women with antiphospholipid syndrome and may not be applicable to women with FVL during pregnancy.

Two studies considered the use of universal or selective testing for FVL160 with or without prothrombin G20210A158 in cohorts at high risk for thrombosis recurrence. The absence of quality-of-life weighting makes interpretation challenging. Without a way to standardize outcomes, the models’ results can only be meaningfully compared within a study, as in Wu et al,158 or across similar studies, as in the pregnancy cohorts of Wu et al 158 and Clark et al.160 Although these models shared many assumptions, the incremental cost-effectiveness ratios differed, presumably because of the different costs included in the models. However, selective screening based on personal or family history of VTE was preferred in both studies, although the authors of both studies acknowledged that it may be challenging to obtain an accurate family history.

While an incremental cost-effectiveness ratio must be interpreted carefully in the context of the study from which it is derived, the values obtained in these four studies, which ranged from $11,100 to $13,624/QALY, were well within the range considered to be cost-effective in the economic evaluation literature. Lifelong anticoagulation was modeled in two studies159,161 but was the preferred option only in scenarios of high recurrent VTE risk and low risk of anticoagulant-induced bleeding.

Testing for FVL or for FVL and prothrombin G20210A, followed by extended anticoagulation for 2 to 3 years in a carrier following a VTE could prove to be a cost-effective strategy, although these models have demonstrated that the level of cost-effectiveness depends heavily on the prevalence of the mutations, the risk of recurrent VTE, the risk of bleeding, and the effectiveness of anticoagulation. Lifelong anticoagulation after testing may prove to be cost-effective for individuals with a very high risk of recurrent VTE and low risk of anticoagulant-induced bleeding events. The studies offer a robust model structure to assess the potential effectiveness and cost-effectiveness of various interventions.

The models are a useful starting point in the evaluation of cost-effectiveness, but since the literature did not strongly support the effectiveness of testing, the results of cost-effectiveness models are challenging to interpret and apply to patient care.

Summary of the Evidence

In Tables 1315 (see also Appendix G, Evidence Table 28), we summarize the evidence to answer our Key Questions. In brief, we found no direct evidence to indicate whether testing for FVL and prothrombin G20210A improves outcomes for probands or family members.

Table 13. Strength of the evidence regarding Key Questions 1 and 2: overarching question and analytic validity.

Table 13

Strength of the evidence regarding Key Questions 1 and 2: overarching question and analytic validity.

Table 14. Strength of the evidence regarding Key Question 3: clinical validity.

Table 14

Strength of the evidence regarding Key Question 3: clinical validity.

Table 15. Strength of the evidence regarding Key Question 4: clinical utility.

Table 15

Strength of the evidence regarding Key Question 4: clinical utility.

There was high-grade evidence to support the conclusion that existing laboratory assays accurately detect these mutations, and most laboratories do an adequate job of detecting these mutations in a clinical setting. There was moderate-grade evidence that homozygosity or heterozygosity for FVL is predictive of recurrent thromboembolism among probands, and high-grade evidence that homozygosity is predictive of VTE events in family members of probands. There was moderate-grade evidence that heterozygosity for FVL is predictive of thromboembolic events in family members and that heterozygosity for prothrombin G20210A is not predictive of VTE in probands; there was insufficient evidence as to whether double heterozygosity is predictive.

There was low-grade evidence that double heterozygosity is predictive of VTE in family members and insufficient information to allow us to draw any conclusions about the predictive value of homozygosity for prothrombin G20210A. There was low-grade evidence that clinicians might change their practice based on testing results. There was high-grade evidence that anticoagulation can reduce VTE in individuals with these mutations, but only low-grade evidence that the relative risk reduction is comparable to that in individuals without mutations. There was moderate-grade evidence that there are neither harms nor benefits associated with these genetic tests. The modeling studies suggest that testing followed by treatment (for 2 to 3 years) for carriers of a mutation could be cost-effective, although the level of this evidence was only low-grade.

Limitations of This Report

In addition to the reported deficits in the literature, there are limitations to this report. In our assessment of clinical validity, we used pooled odds ratios rather than time-dependent measures of recurrence (such as hazard ratios or incident rate ratios). This approach necessarily excluded some studies from the pooled estimates. We recognize that odds ratios may be biased if the follow-up duration varied systematically between individuals with and without the mutations. However, there is little reason to suspect that follow-up duration varied according to mutation status in these prospective studies. In the studies that reported time-dependent analyses, the results of these analyses were generally similar in direction and magnitude to the unadjusted odds ratios that we calculated from the raw event data.113 Odds ratios are often misinterpreted as being highly clinically significant when the absolute difference in the rates of events is very low. We did not calculate pooled rates of events, since we expected the rates of events in the probands to be very dependent on the timing of the study relative to the index event and change over time.26,110,111,125,138

The odds ratios should approximate the relative rates of events in most studies, since these were relatively rare outcomes. We pooled the results using the DerSimonian and Laird random effects methods, a conservative method that often results in wide confidence intervals. In our sensitivity analyses, we also repeated the pooling using several fixed effect methods. Given the near-absence of heterogeneity among the studies in our comparisons, the results were very similar. We opted to report the results from our pooling incorporating random effects because we think this more accurately represents the truth (the odds ratios for the individual studies coming from a distribution of the odds ratios).

We opted not to pool time-dependent outcomes, including the rates of VTE. The reporting of time-dependent outcomes has inherent limitations in a study of thrombosis recurrence, since recurrence rates are highest in the months following anticoagulation cessation. This situation renders absolute event rates (such as the number of events per 100-patient years) challenging to interpret, because a longer duration of follow-up after termination of anticoagulation will tend to bias the results in the direction of lower annualized incidence rates. For this reason, we did not primarily compare incidence rates across studies when the follow-up or anticoagulation duration varied or was unstated.

Another potential source of bias was that anticoagulation practices are not independent of mutation status (e.g., the longer-duration anticoagulation or more aggressive preventive strategies in those with mutations). Most studies mitigated this potential difficulty by excluding patients who were chronically anticoagulated14,110,113,114,118,119,121,123–125 or by using a pre-defined anticoagulation approach.116,119,123–125

In those studies that reported the duration of anticoagulation after the index event in mutation-positive and -negative subgroups, there was no obvious discordance in the anticoagulation duration between the two groups.111,112,120,121,127 If any bias was introduced by changes in clinical management based on knowledge of mutation status, it would tend to reduce the association between the mutation(s) and recurrent events, presuming that mutation positivity led to more intensive anticoagulation.

There was substantial heterogeneity in the composition of the control groups across studies, a matter of concern in that the rates of events in the control groups could have differed substantially. However, all studies were internally consistent, in that those that included other prothrombotic defects in the control group included those same defects in the group with mutations. When data were presented on the prevalence of other defects in groups with and without mutations, there was no evidence that the prevalence of other thrombophilic defects differed across groups.108,112,113 We cannot exclude, however, an interaction between other thrombophilic defects and our mutations of interest and thrombosis recurrence. In these cohort studies, ascertainment bias is possible. In the studies of probands, the individuals were not blinded to their mutation status. It is possible that patients with mutations were more likely to seek medical attention for symptoms consistent with deep vein thrombosis or pulmonary embolism and might have been over-diagnosed with recurrence (due to false-positive testing), or that those without mutations were under-diagnosed (because they did not seek medical attention for a thrombotic event that ultimately resolved without therapy). Ascertainment bias would tend to augment the association between the mutations and recurrent thrombosis. None of the studies we included had scheduled periodic radiographic testing to limit the potential for ascertainment bias.

The majority of the observational studies concerning family members were retrospective, with some notable exceptions.129–133 Retrospective studies are prone to important biases, including recall bias. Although this potential source of bias can be mitigated by interviewing participants before they have knowledge of their mutation status, this process was variably described in these studies.

The limitations that are specific to Key Question 4 – the clinical utility question – are described above in the discussion of that question.

Implications for Future Research

Studies to directly address our overarching question (Key Question 1) would ideally be designed as trials in which participants with venous thrombosis, and/or their family members, would be randomized to a test arm or a no-test arm. Individuals would be managed by their physicians on the basis of the test results (with evidence-based recommendations). Sufficient follow-up time would be included in the study design so that VTE events could be witnessed and compared between the tested and untested groups. An alternative approach would be a well-designed prospective cohort study. The population of interest would be defined as individuals with recent VTE. The subgroup of these individuals who had testing for the mutation would be considered the “exposed” group. A comparison group of closely comparable individuals who were not “exposed” to testing would also be defined, and the outcomes in the groups would be compared, with careful attention to complete follow-up in both groups.

Analytic Validity

Although the mutation detection methods were found to have high analytic validity, a small minority of laboratories accounted for a disproportionate percentage of the errors in the performance of these tests. This result suggests an ongoing need for participation of molecular diagnostic laboratories in external quality assurance programs to assure consistent provision of high-quality genetic testing services. In the United States, laboratories doing molecular testing on human samples are required to follow the guidelines set by the Clinical Laboratory Improvement Amendments (CLIA) 1988 that include requirements for proficiency testing. In addition, a large majority of the laboratories performing these tests have additional accreditation from bodies such as from the College of American Pathologists, which also requires semi-annual proficiency testing to ensure accurate and precise testing.

Clinical Validity

Future studies should report event rates over time (and relative rates of recurrence between specified groups), rather than just the number of events. Studies should consistently differentiate between heterozygous and homozygous individuals, since there is a different rate of recurrence in these two groups. Future studies should continue to use objectively measured thrombosis (radiographically proven) as a criterion for the index and recurrent thromboses but should provide more detail about both the index events and the recurrence, including whether the events had other precipitants (e.g., peri-procedural, idiopathic, associated with hospitalization, or cancer-associated). Such data were presented in an inconsistent fashion in the studies we reviewed, and when reported, they were generally given for the entire study cohort, rather than separately for the groups of interest. By examining specific subsets of patients, it may be possible to clarify whether there are any interactions between mutation status and clinical variables in terms of predicting recurrence. With regard to the prothrombin G20210A mutation (alone or in conjunction with FVL), additional studies are needed to more precisely quantify the effect size.

There remains uncertainty about the estimates of risk for family members, given the very wide confidence intervals surrounding the odds ratios and the rarity with which the studies reported actual rates of events (rather than counts). Also, the studies that we included were exclusively studies of European populations. It is well known that the mutation frequency varies markedly across populations (and is particularly low in African-derived populations), but it is still unclear whether the risk attributable to the mutation differs in other populations having different genetic or environmental contributors to VTE risk. Future research would be appropriate in Caucasian populations outside of Europe or in other populations with appreciable frequencies of mutations. Also, future research could better explore the age-mutation interaction.

Clinical Utility

Future studies should directly address whethe r clinicians change their recommendations in response to the results of FVL and prothrombin G20210A testing. Rather than surveys based on hypothetical situations, we suggest that chart reviews or analyses of utilization data (such as tracking prescriptions or the number of referrals) based on actual patients referred for testing would more directly and cogently answer this question.

To assist clinicians in the management of patients with VTE, future studies should be powered sufficiently to evaluate the risks associated with prolonged anticoagulation, as they relate to patients with specific thrombophilic mutations. Future studies addressing this question might move away from primarily focusing on the effect of treatment on absolute recurrence rate toward whether management decisions based on testing results affect the rates of recurrence in carriers of each of these mutations. Even though the evidence suggested that neither FVL nor prothrombin G20210A attenuates the prevention of recurrence during ongoing anticoagulation in probands, future studies in both probands and family members might focus on whether management decisions (duration of therapy, use of thromboprophylaxis) affect rates of VTE, particularly during times of heightened thromboembolic risk.

Future studies should ensure an adequate representation of patients with FVL and prothrombin G20210A. Studies based in the United States may give a clear understanding of how patients here might respond to the testing process and results. Larger sample sizes should also be used to increase the ability to detect rarer events, such as stigmatization and discrimination by insurers. Efforts should be made to recruit representative patient populations, and relevant comparison groups should be included (e.g., carriers and non-carriers) to increase the practical applicability of the study findings. Quantitative studies may be preferable, involving the use of standardized, validated questionnaires to evaluate patients’ experiences.

The cost-effectiveness analyses should be updated when there are additional data to support the assumptions of the models and the factors on which the results most depend, including the magnitude and duration of VTE recurrence risk, anticoagulant efficacy in preventing recurrent VTE, and anticoagulant-induced bleeding risk. To more definitively determine the cost-effectiveness of testing for these mutations, clinical trials could include an assessment of the costs associated with a testing strategy, as compared to care without testing.

Our literature review included articles through December 2008. We do not anticipate any important secular changes in the event rate that would markedly change the event rates in the upcoming years. We also do not expect major changes in the coming years in terms of the methods used to detect mutations. The most anticipated change would be an increase in options to reduce risk as new drugs become available. Future research will need to include an evaluation of the risks and benefits associated with use of new anticoagulant drugs in probands and family members at high risk of events.