Public Health Aspects of Breast Cancer Gene Testing in Canada
Part 3: A Model of Potential Need and Demand

J Mark Elwood

Abstract

Centres offering expert counselling and genetic testing are already experiencing high levels of demand, and yet the potential demand is much greater. There have been few attempts to estimate the potential demand created by particular guidelines for referral or testing. A model of need and demand for genetic services is presented, and research questions are identified that should assist in better prediction of future requirements for genetic counselling and testing. The value of integrated routine data on referral criteria, demand and clinical service load is considerable. Attention needs to be paid to referral at primary care and general specialist levels as well as to expert centres.

Key words: Canada; genetic counselling; genetic screening

Introduction

This is the third of three related papers; the methods are described in the first.^1,2 The current literature on testing for genetic susceptibility to cancer, reviewed in parts 1 and 2,^1,2 is very limited with regard to the assessment of potential need and demand for services. This is partially because many genetic testing centres do not have a population perspective. Often, they were developed primarily from expertise in the laboratory aspects of testing, dealing with those subjects referred to them through varied and largely unrecorded referral processes. Although criteria for testing have been developed, there has been little work done to relate such selection criteria to the population base and to consider issues of demand, need and equity of access.

Current Services and Demand in Canada

In 1997, genetic testing for BRCA1 and BRCA2 and other major cancer-related genes was being offered in research centres in Montreal, Toronto, Vancouver, Victoria and Winnipeg (J Beauvais, Laboratory Centre for Disease Control, personal communication). Several of the centres had long waiting lists, and the demand for counselling exceeded available resources. The guidelines for referral for counselling and for gene testing vary among centres and, in general, represent criteria based on an informal assessment of the likelihood of mutation detection in subjects and families referred, using the available data (as reviewed in Part 2 of this series). The relation between the criteria used and the potential demand defined by those criteria has not been explored in any detail.

A Model of Need and Demand for Genetic Assessment

There appear to be no published data on demand for screening services in Canada; the potential demand is, however, very large. The BRCA1 carrier prevalence of 0.0012³ equates with 1200 mutation carriers per million population. Testing for breast and ovarian cancer susceptibility genes has tended to be concentrated in women aged 20S59. Given the age and sex structure of the Canadian population, there will be some 320 such women carriers per million total population. If (as shown in Part 2) genetic counselling and testing are warranted for women with a 10% risk of being a carrier, then some 3200 women per million population would be eligible for testing. To this needs to be added the numbers of men and women needing testing for other genes, such as those for colon cancer susceptibility, and men may also require testing for BRCA carrier status. Of course these are prevalence figures, whereas the demand on services relates to annual numbers of subjects coming for counselling.

At present, only a small fraction of eligible subjects are seeking testing; the concern is that if this changes the demand could increase greatly. At a later point saturation could be reached, when all carriers in a defined population have been identified. The period over which all mutation carriers in a population can be identified and the criteria (age, sex and any other) used to target assessment and testing are therefore critical in determining the demand on screening services. None of these issues has received much attention in the published literature; although a full assessment is impossible with the data available, a general model of the situation may be helpful.

To be identified as a gene carrier by a normal clinical service, three criteria must be met: the subjects must be aware of their increased risk, so that they seek advice about it; they must meet the criteria for testing used by the agency to whom they go for advice; and they must actually have a detectable genetic abnormality. To consider this issue further, it is useful to distinguish different subgroups that would apply in a total population or in a subpopulation defined by sex, age, ethnic group or other criteria.

1. Genetic abnormality (G): The number of subjects in the population with detectable relevant genetic mutations, that is, the prevalence of relevant gene alleles. At this point, the sensitivity of testing will be ignored and the prevalence taken as being that of detectable abnormalities.

2. Criteria for testing met (C): The number of subjects who are eligible for testing and expert counselling, based on the criteria of personal and family history used by the clinical service. This number will be larger than G, probably much larger. For example, many criteria suggest that testing is justified if the predictive value of the family history is 10%: that is, up to 10 subjects will be tested for each one found to have a carrier state (see Part 2 ).

3. Perception/action (P): The number of subjects who come forward to seek advice. This is likely to be only a small fraction of all subjects at increased risk since it depends on motivation, knowledge of the services available and access to them. Many people will be aware of their risk and may be concerned about it, but will not take any action because they either do not know of available services or find the services to be inaccessible, expensive, of unknown quality or of unknown value to them. However, many people will overestimate their risk, and many, perhaps most, in this category may not have a strong family history on objective criteria. The quantity P, unlike G and C, will change quickly as the services available vary and public perceptions change. Increasing the provision of services and knowledge about them is likely to increase the demand.

These three criteria—having a detectable genetic alteration (G), fitting objective criteria (C) and perceiving risk enough to lead to action (P)—can be represented in a Venn diagram, which shows eight sets of subjects in the population (Figure 1). P', C', and G' indicate those groups who do not meet these criteria.

Those people who perceive themselves at risk and are sufficiently concerned to seek advice (group P) will come to the attention of a genetic service provided through clinics, phone contact or other means, and then the following subject groups should fall into place.

i. Group PCG will be correctly identified.

ii. Group PCG' will be tested, but the result will be negative. The balance between PCG and PCG' is determined by the criteria set for testing.

iii. Group PC'G' perceive themselves at risk, but do not meet the criteria for testing and do not have the genetic condition. They require good advice and counselling, without testing, in a cost-efficient way; the challenge is to correctly and successfully reassure this large group.

iv. Group PC'G perceive themselves at risk, but do not meet the criteria for testing; however, they do have the gene state. They are the false negatives of the testing criteria. This number will increase if the testing criteria are narrowed.

Other groups (P') will not come forward to a clinical service that depends on the subjects taking the initiative.

v. Group P'CG do not perceive themselves at risk, but meet the criteria for testing and carry the gene abnormality. This group represents the (probably large) number of gene carriers not identified because of the incomplete use of available services by those eligible for them, although some may be identified through another family member (by outreach investigation) or by a population-based screening program.

vi. Group P'CG' do not perceive themselves at risk, and, although they do meet the criteria for testing, they do not carry the gene abnormality. If referred, they would be unnecessarily tested. It could be said that they benefit from the incomplete access to testing.

vii. Group P'C'G do not perceive themselves at risk, and if surveyed do not meet the criteria for testing, although they do carry the gene abnormality; they will be identified only by a less selective screening system.

Finally, the rest of the population are P'C'G': they do not have the genetic state and correctly do not perceive their risk as high, nor do they meet the criteria for testing. However, many such people will be alerted by publicity on this issue and will need information to avoid being misled into thinking they may be at high risk.

In principle, the number of subjects in a defined population with a well-developed genetic service who fall into each of the different sets depicted in Figure 1 can be ascertained by a combination of good routine data collection and special surveys. Routine data would identify all subjects in groups P, PC, PCG and PCG'. A well-constructed population sample survey could supply a number for group C, and a calculation based on known gene frequency plus knowledge of the sensitivity of the test used could estimate group G.

The number in group PC'G could be calculated by a special survey of gene tests carried out on a sample of subjects who attend for counselling because of perceived high risk but do not fit the normal criteria. The subgroup P'CG could be ascertained by genetic testing of a sample of subjects identified in the community survey as fitting the normal criteria for gene testing, but who have not perceived this risk and taken action themselves.

From these steps, the numbers of subjects in all eight categories could be ascertained. In practice, such work would pose many difficulties, one being how to ensure that the responses to questions used in a population survey to assess criteria for testing were consistent with those given by the normally more intensive interview methods used in the clinical service. The results will be specific to the time and place they are generated, as the total P group (representing perception and action) will change quickly, whereas, in principle, the total G and C groups will be fairly constant.

The situation can be explored further, more realistically, by introducing more complexity. To avoid the two main simplifications in the scheme set out above, it is necessary to distinguish perception of risk (in response to some systematic inquiry) from perception that is strong enough to initiate action. The balance between perception and initiation of action will depend, of course, on the ease with which clinical services can be accessed. Thus we can consider the perception group as two concentric circles, the outer one relating to perception of risk and the smaller inner circle to perception of risk followed by initiation of action.

Similarly, the genetic circle can be considered as an outer circle representing the presence of a genetic abnormality and a concentric inner circle representing the presence of a detectable genetic abnormality, to emphasize the lack of total sensitivity of the genetic tests. This model ignores any possible false positives. Introducing these two situations makes the Venn diagram much more complex (Figure 2), since there are now 18 rather than 8 subgroups. Of these 18 groups, only one represents those subjects whose genetic susceptibility will be recognized by the clinical service (Òdetected groupÓ). As before, this is the PCG group, the intersection of perception plus action, fitting testing criteria and having a detectable genetic abnormality.

In estimating the demand and need for genetic testing, a number of approaches seem worth consideration.

1. Taking consistent standardized family histories from patients with newly diagnosed breast, ovarian, colon or other cancers, who are a representative or total sample from a defined population (perhaps limited by age), would give the numbers of families fitting various criteria of strength of family history, identified by a family member having an incident cancer in a given time period. These numbers are directly relevant to the guidelines for family interventions based on the strength of family history in newly incident cancer patients. This is a useful and fairly easily measured indicator of an annual number of families needing investigation. However, it identifies only a fraction, perhaps a small fraction, of all high-risk families in the population because of the limitation of requiring a newly incident case in a certain time period.

2. A population survey will give a direct estimate of the numbers of people who can identify themselves as having specified degrees of family history for specified cancers. Such a survey should be based on a representative series of subjects, limited or stratified by age and sex, and should inquire about relatives affected, type of cancer and the date of diagnosis of the cancer. It would be valuable to compare these results with the investigation of newly incident cancer patients in the same population, as that would show if the results of studying new cases (which would be easier to continue) could be used to estimate the population numbers derived by the survey. Results from self-completed questionnaires suggest that family history data, at least on first-degree relatives, can be collected with reasonable accuracy.⁴

Such a population survey should identify a much larger number of high-risk families than the patient- based method if the data are equally accurate, because there is no narrow time restriction on cancer incidence. However, because of the recent diagnosis, the patient-centred method may produce more detailed and more accurate data. The relation between the numbers of high-risk families identified by these two methods could be estimated from an analysis of extensive data on high-risk families from existing sources, in terms of the probability of having at least one member with a cancer incident in a given time.

3. Such a population survey could also be used to estimate the number of subjects who perceive themselves at high risk, and the number who want advice, counselling or testing. It would be necessary to measure this self-perception of risk independently from collecting data on actual family history, since the process of collecting family history data may change the perception. This could be achieved by a well-constructed telephone interview or by serial data collection methods. The link between perception of risk and the desire to seek help will vary with the services available and with the knowledge and perceptions of the services.

Referral Criteria and Counselling in Primary Care

In a community, the number of subjects whose family history is strong enough to warrant genetic testing is relatively small, but a larger number may benefit from expert counselling, even if they do not require gene testing. However, there are likely to be much larger numbers of subjects who have a concern, perhaps considerable anxiety, about their risk state, but who do not have a strong enough family or personal history to make referral for expert counselling appropriate. Consideration has to be given to how advice and support will be given to these people. It may be through impersonal media or voluntary organizations. In this case the development and testing of educational materials for direct use by the public or to assist relatively unskilled counsellors is important. The family physician is likely to be the first health professional to be consulted. The development of good management plans, materials, support and training for family physicians is also an important issue.

A triage approach developed in Australia⁵ for breast cancer risk defined three groups: 95% of women with no family history or only a weak history who could be advised by family physicians alone; up to 4% with a moderately increased risk who could be advised by family physicians using guidelines from expert centres, with consultation if required; and up to 1% with a strong history who required a referral for expert counselling and perhaps testing. Evaluation of this model is in progress.

Economic Issues

Major economic assessments of BRCA testing have yet to be published. In England, the cost per mutation detected has been estimated, but only in an approximate manner.⁶ An economic analysis of testing for HNPCC (hereditary non-polyposis colorectal cancer)⁷ shows that the major determinants of cost-effectiveness are the prevalence of the gene and the assumptions about the benefits of interventions for gene carriers. For BRCA genes, the benefits of interventions are far from established.

Discussion and Recommendations

Some recommendations for activities to co-ordinate and develop Canadian work on genetic testing for cancer susceptibility can be made. In Canada, as in other countries, the pace of development of laboratory expertise in genetic testing has been greater than that of expertise in counselling and studies of its effectiveness. There is only limited information on the current provision of genetic services, and there has been relatively little attention paid to population-based issues of demand and need. There is little information on the relation between the criteria used for referral and gene testing and the potential demand created if those criteria are applied on a population basis. The model of need and demand for services presented here and the research questions identified may assist in developing better estimates of future requirements for genetic counselling and testing. There is little information on costs and cost-benefit aspects of genetic testing.

Although there is good communication among different centres with similar expertise, for example, in laboratory techniques, there is less communication between groups from different disciplines and perspectives, for example, those with expertise in family medicine or in health economics. The workshops on cancer genetics organized by the Canadian Collaborative Group for Cancer Genetics (CCGCG) have shown the strengths of Canadian research in this area and the value of a forum for discussion. Further efforts to combine disciplinary strengths through networks, meetings or task forces would be valuable. A data monitoring system with adequate attention to issues of confidentiality would be useful in establishing a core data set to monitor counselling and genetic testing in different centres in Canada.

Acknowledgments

This paper is based on a report prepared initially for the Cancer Bureau, Laboratory Centre for Disease Control, Health Canada, under contract 502-8082 and project 502-0205.

Personal acknowledgments are noted in the first paper in this series of three.

References

1. Elwood JM. Public health aspects of breast cancer gene testing in Canada. Part 1: Risks and interventions. Chronic Dis Can 1999;20(1):3-13.

2. Elwood JM. Public health aspects of breast cancer gene testing in Canada. Part 2: Selection for and effects of testing. Chronic Dis Can 1999;20(1):14-20.

3. Ford D, Easton DF, Peto J. Estimates of the gene frequency of BRCA1 and its contribution to breast and ovarian cancer incidence. Am J Hum Genet 1995;57:1457-62.

4. Theis B, Boyd N, Lockwood G, Trichler D. Accuracy of family cancer history in breast cancer patients. Eur J Cancer Prev 1994;3:321-7.

5. Genetic Testing Working Group of the NHMRC National Breast Cancer Centre. Current best advice about familial aspects of breast cancer. A guide for general practitioners. Sydney: NHMRC National Breast Cancer Centre, 1996

6. Eccles DM, Englefield P, Soulby MA, Campbell IG. BRCA1 mutations in southern England. Br J Cancer 1998;77:2199-203

7. Brown ML, Kessler LG. The use of gene tests to detect hereditary predisposition to cancer: economic considerations. J Natl Cancer Inst 1995;87:1131-6

Author Reference
J Mark Elwood, Department of Preventive and Social Medicine, University of Otago, PO Box 913, Dunedin, New Zealand; Fax: 64-3-4797164;
E-mail: melwood@gandalf.otago.ac.nz

[Previous][Table of Contents] [Next]