Background

The following model was built to provide an insight into the cost-effectiveness of genetic testing for BRCA1/2 breast/ovarian cancer mutations in a UK setting. The model was made available to the GDG but was not discussed by the group or used to determine guideline recommendations. It must be stressed that the modelling presented here is a preliminary piece of work which is intended to highlight the important uncertainties that exist in this area but is not sufficiently well developed to be used for decision making.

Overview of model

The analysis compares a full gene genetic test to no testing for women at high risk of developing breast cancer due to familial history using a Markov decision tree model developed in DATA Professional¹. An overview of the model structure is shown in figure 1.

Figure 1

Genetic Testing Decision model.

If genetic screening is available then only a proportion of women eligible for the program will be carriers of the BRCA1/2 genetic mutation. For women that enter the program and undergo testing, the test provides either a positive or negative result. The model also allows for the possibility that women enter the program but decide not to undergo testing. Following testing a woman may decide to undergo prophylactic surgery (either mastectomy and/or oophorectomy) in order to reduce her risk of breast and ovarian cancer. Four Markov states are included in the model; no cancer, breast cancer, ovarian cancer and death. The model runs annually for a maximum of fifty cycles.

The analysis is undertaken from a health service perspective for women aged 25, 35, 45, and 65 years of age. Health benefits are expressed in terms of Quality Adjusted Life Years (QALYS). Costs are discounted at 6% per annum and benefits are discounted at 1.5% per annum. Costs are expressed in 2002 UK sterling.

Assumptions of the model:

• Breast cancer and ovarian cancer cannot be experienced simultaneously
• Prophylactic surgery, if undertaken, is immediate
• Prophylactic oophorectomy is only undertaken in conjunction with prophylactic mastectomy
• Women in a cancer state experience a constant decrement in Quality of life whilst in that state. There is distinction made between different stages of disease
• Women that undergo prophylactic surgery experience a constant decrement in quality of life.
• Reductions in quality of life from prophylactic surgery and cancer states are multiplicative
• Cancer states may be experienced for a maximum of 5 years. Women that have not progressed to death after 5 years return to the normal health state
• Progression from the normal health state to cancer states is not dependent on previous health states i.e. the model has no memory
• No account is taken of the possible gains that might accrue from individuals choosing not to undergo surgery given the information from a negative test who would have undergone surgery if no test were available
• No genetic testing takes place without counselling

Values used in the analysis

A table of base case values and sources is provided in Table A.

Table A

Base case results for Genetic Testing Model (£'s per QALY).

If genetic testing is available then women will be eligible for the program dependent on their level of risk. Clearly, the effectiveness and cost-effectiveness will differ according to the threshold at which this risk level is set. This threshold determines the probability that a woman is BRCA1/2 positive. The base case model uses a value of 0.15, which corresponds to the level of risk associated with women whose family history corresponds to the “high” risk definition used in the guideline, Sevilla et al. (2002).

For each woman entering the program, it is assumed that a preliminary counselling session is required. The cost of this session (£49.84) has been taken from Cohen et al. (in submission) In order to test for BRCA mutations an affected family member is tested to see if a family mutation can be found. In the base case analysis it is assumed that 100% of individuals entering the program have a living affected relative that is willing to be tested. This will depend on the criteria adopted for entry to the program. In the TRACE trial, for example, only 36 out of 48 women had such a relative (75%) and this value is used in the sensitivity analysis. It is recognised that the TRACE trial was not a trial of full gene testing. However, some of the cost elements are common to alternative testing programs.

The cost of this test is that for full sequencing in Manchester (Evans personal communication). Where a family mutation is detected in the affected woman then family members can be tested. Therefore, the cost per woman will depend both on the cost of the subsequent testing and the number of relatives that undergo testing per affected relative. We assume that a mean of two relatives will be tested per affected woman found positive. The unit cost for testing unaffected women is £28.84 for single batch cascade testing from the TRACE trial. Additionally, a positive test in the affected relative requires appropriate counselling to be provided both to that relative and to the unaffected relatives. This cost was estimated at £115.78 (Cohen et al. in submission - table 5 excluding patient travel costs) for affected women, plus £148.21 for pre and post test counselling in the unaffected relative.

The sensitivity and specificity of full gene sequencing is taken from myriad genetics (quoted in Tengs and Berry 2000). Women that receive a positive test result are more likely to undergo risk reducing surgery (either mastectomy, or mastectomy and oophorectomy) with these probabilities taken from Evans (personal communication). Expected health outcomes are therefore dependent on age dependent cancer risks and other cause mortality (ONS, Stratton et al 1998, Grann et al 2002), adjusted according to true BRCA1/2 status and whether risk reducing surgery has been taken. Quality of life values for all cancer and risk reducing surgery states are taken from Grann et al. (2002) and Tengs et al (1998).

Costs for cancer states and prophylactic surgery are taken from NHS reference costs (2002). Cancer costs accrue are assumed to accrue for each year that a person remains in a cancer state.

Results

Base case results, shown in Table A, show that genetic testing women at a very young age is dominated compared to a no testing alternative. The reason for this is that the model assumes that those women that undertake risk reducing surgery do so immediately. At a young age the risk of breast or ovarian cancer is relatively low compared to the reduction in quality of life suffered from risk reducing surgery. The benefits of risk reducing surgery are experienced to a greater extent in later years and are consequently not valued particularly highly due to discounting.

The base case results for testing at all other ages (except 65 years and over) indicate that health benefits are generated at a relatively low additional cost. The cost per QALY is relatively high (£55k) for women aged 65 years. The reason for this is that whilst the costs of providing testing and surgery are immediate (financial and quality of life reduction for women), the benefits occur in later years (reduced incidence of disease). However, all cause mortality in older women is obviously higher and therefore the benefits accrued in future years are limited.

Sensitivity analyses

Only simple, one-way sensitivity analyses have been performed with the aim of showing those key estimates that have substantial impact on the model. Results are shown in Table B.

Table B

Sensitivity Analyses for genetic Testing Model (£’s per QALY).

Firstly, it can be seen that the results are relatively insensitive to the probability that a woman entering the scheme has an affected relative and therefore receives the test.

The prior probability of women entering the program having genetic mutations has a substantial impact. If the true proportion is 20% instead of 15% then, except in women aged 25 and below, then testing becomes a relatively cost effective option and in some case (women aged 45 and 55 years) dominates the no test option. Table B also shows that the results are sensitive to the QALY values used for the relevant health states, to the proportion of women that undertake risk reducing surgery that would not have done so in the absence of genetic testing and to the cost of the test. For the latter value, the Myriad quoted cost was used although the result here may not be a completely accurate reflection of such testing arrangements since the myriad system would not require the involvement of affected relatives.

Discussion and Limitations

There is no combined ovarian and breast cancer health state in the model.

It is assumed that no women undergo risk reducing oophorectomy without mastectomy.

There is little evidence that genetic testing alters the likelihood that women undergo risk reducing surgery (or any other type of risk reducing behaviour that is not accounted for in the model). These estimates are therefore extremely uncertain.

No comparison of different genetic testing programs has been undertaken. Sevilla et al (2002) suggest that the incremental cost effectiveness of some full gene testing programs is extremely high compared to less sensitive/specific alternatives. Such an analysis should be undertaken from an NHS perspective.

Only health benefits are included in the model. There may be an inherent value of information and there is also likely to be utility in waiting less time for results. If this wider perspective were adopted then it would be important to compare different types of genetic testing, including the private provision of testing by Myriad genetics (for example, comprehensive BRCA testing results can be received in 10 days for a unit cost of $4,140).

The costs and benefits associated with breast and ovarian cancers are relatively crude.

The results for women aged 25 years should not be interpreted as an indication that genetic testing is not cost-effective at this age but that surgery undertaken in the light of additional information may be best delayed (as many women choose currently).

Footnotes

1

1998–2003 Treeage Software Inc.