|Breast cancer screening, outside the population-screening program, of women from breast cancer families without proven BRCA1/BRCA2 mutations: a simulation study
|Jacobi C E, Nagelkerke N J, Van Houwelingen J C, De Bock G H
This is a critical abstract of an economic evaluation that meets the criteria for inclusion on NHS EED. Each abstract contains a brief summary of the methods, the results and conclusions followed by a detailed critical assessment on the reliability of the study and the conclusions drawn.
The authors studied a programme of mammography screening for women under the age of 50 years from breast cancer families without proven BRCA1/BRCA2 mutations.
Economic study type
Cost-effectiveness analysis and cost-utility analysis.
The study population comprised a population of 1 million women aged 30 to 50 years, which were simulated using the Dutch demographic age structure. All simulated women had at least one first- or second-degree relative with breast cancer, but the authors excluded all women with family histories associated with BRCA1/BRCA2 mutations (i.e. with bilateral breast cancer, ovarian cancer, and/or male breast cancer).
The study setting appears to have been the community. The economic study was undertaken in the Netherlands.
Dates to which data relate
The effectiveness data were derived from studies published between 1996 and 2004. Resource use and costs were based on current costs in the Netherlands. The price year was not reported.
Source of effectiveness data
The effectiveness data were derived from a review and synthesis of published studies, supplemented with the authors' assumptions.
A simulation model for mammography screening was developed using information from the medical literature. For mammography screening under the age of 50, the authors considered four cohorts with 5-year starting ages (30 to 50, 35 to 50, 40 to 50, and 45 to 50 years) and 4 screening intervals (6, 12, 18 and 24 months). The authors estimated the lifetime risk of breast cancer, conditional on the absence of BRCA1/BRCA2 mutations, using the Jonker model 1 (Jonker et al. 2003, see 'Other Publications of Related Interest' below for bibliographic details). In the Jonker model, all non-BRCA1/BRCA2 mutations and other familial risk factors have been grouped together into a single hypothetical susceptibility gene. The women are then classified according to family history groups based on the number of affected relatives and the ages at diagnosis of these relatives. The screening model showed the number of tumours found and missed, and the relative gain in life-years, per 1,000 women screened.
Outcomes assessed in the review
The outcomes assessed from the literature were:
the sensitivity and specificity of mammography for women younger and older than 50 years;
the median tumour volume doubling time for women younger and older than 50 years;
the breast cancer induction risk due to mammography radiation for women younger and older than 50 years;
the relative risk of dying from breast cancer for women aged 30 to 50 years, compared with women above the age of 50; and
the remaining life expectancy for women whose breast cancer could be cured and for those with no cure.
Study designs and other criteria for inclusion in the review
To identify studies on sensitivity and specificity, the search strategy used terms such as breast cancer, age specific, sensitivity, specificity, mammography and human.
To identify studies on tumour growth, the search strategy used terms such as growth rate, age specific and breast cancer.
To identify studies on the effects of radiation due to mammography, the search strategy used terms such as breast neoplasms, BRCA, radiation, induction of breast cancer and mammography. This search was limited to articles published in English between 1993 and 2003.
To identify studies on prognosis and survival, the search strategy used terms like breast neoplasms and BRCA combined with a specific filter for prognostic studies and systematic reviews. It was limited to articles published in English between 1995 and 2001. A further search was done to identify similar original articles after the conduction of the reviews. The second search was conducted for articles published in 2000 and 2001. Articles in this search had to provide information on tumour size, lymph node involvement, or histologic grade of breast cancers, and provide point estimates and measures of variability or frequencies for at least one of the relevant variables.
Sources searched to identify primary studies
The authors undertook a systematic search of MEDLINE.
Criteria used to ensure the validity of primary studies
Methods used to judge relevance and validity, and for extracting data
The validity of the primary studies does not appear to have been assessed.
Number of primary studies included
Approximately 11 studies were included in the review.
Methods of combining primary studies
Investigation of differences between primary studies
The authors do not appear to have investigated differences between the primary studies.
Results of the review
The specificity of mammography was 0.967 to 0.996 for women younger than 50 and 0.970 to 0.997 for those older than 50.
The sensitivity of mammography was 64% (range: 22 to 82) for women younger than 50 and 85% (range: 56 to 94) for those older than 50.
The median tumour volume doubling time was 80 days (95% confidence interval, CI: 44 to 147) for women younger than 50 and 157 days (95% CI: 121 to 204) for those older than 50.
The breast cancer induction risk due to mammography radiation was 0.0000165 for women younger than 50 and 0.0000114 for those older than 50.
The average remaining life expectancy for women who could not be cured was 5.1 years. For those cured, the remaining life expectancy for women aged 30 to 50 years was (80.52 minus the detection age) multiplied by 0.966.
For women aged 30 to 50 years, the relative risk of dying from breast cancer is linearly, negatively, associated with age.
Methods used to derive estimates of effectiveness
The authors supplemented the results of the review with their own assumptions, which were supported with findings from the literature.
Estimates of effectiveness and key assumptions
The authors made the following assumptions.
Screening caused a shift to finding tumours at early stages. This led to a 50% better survival rate among women with screen-detected breast cancer compared with clinically detected tumours. Women with a familial predisposition for breast cancer were at a higher risk of radiation-induced cancer than women without such a predisposition. Hereditary and familial tumour had an underlying mechanism that caused unfavourable prognosis. The authors included a worse prognosis of up to 25% due to larger-sized and higher-graded tumours, based on the extent of family history.
Measure of benefits used in the economic analysis
The measures of benefits were the life-years and quality-adjusted life-years (QALYs) gained. To adjust life-years by quality of life, the authors combined the changes in life expectancy with the expected changes in morbidity, according to a published Dutch study. The authors used a 10% reduction in quality of life for women who could be cured from breast cancer, and a 50% quality-of-life reduction for those not cured. The authors assumed no effect of the screening process on quality of life if no cancers were detected.
The authors did not report explicitly whose direct costs were included in the analysis. However, only the costs associated with the screening programme appear to have been included in the analysis. Such costs included the costs of a mammogram, the interpretation of the mammogram, a related visit to the general practitioner, and the costs of biopsies. The costs were based on current costs in the Netherlands, and were used independently of age. Since the costs appear to have been incurred during a short time period, discounting was not relevant and was thus not performed. The costs of the actual programmes were not reported. The price year was not reported.
Statistical analysis of costs
The costs were treated as point estimates (i.e. the data were deterministic).
The indirect costs were not included in the analysis.
The authors performed three sensitivity analyses to test the effect of assumptions in the model.
In one scenario, the authors omitted the correction factor regarding unfavourable prognosis due to larger-sized and worse-graded familial cancers.
In another scenario, the authors omitted the two correction factors regarding unfavourable prognosis: the one due to larger-sized and worse-graded familial cancers and the one on young age at diagnosis.
In the final scenario, the authors tested the effect of increased sensitivity to radiation-induced breast cancer among women with a familial predisposition to cancer based on the assumption that women with BRCA1/BRCA2 mutations were at higher risk of radiation-induced breast cancer than women without such mutations.
Estimated benefits used in the economic analysis
Although the authors used QALYs and life-years as an outcome measure, these were not reported in the results.
The authors found that the number of breast cancers detected by screening depended strongly on the familial predisposition of the woman and the screening cohort. It was found that if screening started at older ages (i.e. 45 years instead of 30 years), the detection rate per screening round increased. The authors also reported that with a screening interval of 12 months, approximately 70% of all the tumours were detected. However, with a biannual screening programme, approximately 90% of cancers would be identified, but at the expense of twice the dose of ionising radiation per woman per year. Screening intervals higher than 12 months would result in a lower ionising radiation, but would substantially reduce the number of cancers detected by screening (<50%).
The authors therefore focused on the 12-month interval as the optimal screening interval for women under the age of 50 years.
The authors did not report the mean or total costs of each programme.
Synthesis of costs and benefits
The authors defined screening under the age of 50 as cost-effective if the costs of screening per life-year gained, per 1,000 screened women from a certain family history group, were equal or less than the screening costs per life-year gained among women from the general population of 50 to 52 years.
The authors reported that annual screening with mammography was found to be cost-effective in women with:
a lifetime risk of breast cancer of 27% and higher from age 45;
a lifetime risk of breast cancer of 32% and higher from age 40;
a lifetime risk of breast cancer of 38% and higher from age 35; and
a lifetime risk of breast cancer of 47% and higher from age 30.
The authors reported that the outcomes would be similar if the life-years gained were not corrected for quality of life.
The results of the sensitivity analysis showed the following.
The survival of women with breast cancer increased when adjustment for unfavourable prognosis was omitted. Therefore, it would be more cost-effective to screen women younger than 50 who are at lower risk of breast cancer.
The survival of women with breast cancer increased when both adjustments of worse prognosis were omitted.
More gain in life-years was needed to compensate for the years lost due to induced tumours among women with a familial predisposition.
Annual breast cancer screening with mammography for women under the age of 50 would appear to be cost-effective in women with strong family histories of breast cancer, even when no BRCA1/BRCA2 mutation was found in affected family members.
CRD COMMENTARY - Selection of comparators
A justification was given for using a screening programme for women aged older than 50 years. It represented current practice in the Netherlands. You should decide if this intervention represents current practice in your own setting.
Validity of estimate of measure of effectiveness
The authors reported that a systematic review of the literature published in MEDLINE was undertaken to identify relevant research. For each variable used in the model, the authors appropriately reported the literature search they undertook, and the methods used. For example, the authors reported the search strategy used, the dates to which the search related, the number of studies included, and any other inclusion criteria. The authors appropriately reported the assumptions, and the underlying rationale, used to derive effectiveness estimates. However, very limited sensitivity analyses were performed to test the effects of varying the parameter estimates of these assumptions.
Validity of estimate of measure of benefit
The authors reported that life-years and QALYs gained were used as the measures of benefit. However, these were not reported. The authors reported only the risk of breast cancer up to 80 years of age, starting ages of mammography screening, and the number of tumours found per screening episode.
Validity of estimate of costs
Although the perspective from which the costing was carried out was not specified, it appears to have been that of the health care provider. However, the authors reported very limited information on their costing study. Consequently, it is not possible to determine whether all the categories of cost relative to the perspective adopted were included in the analysis, or if all major relevant costs were included. In addition, the resource use quantities and the prices were not reported separately. These facts will limit the generalisability of the authors' results. The costs were derived from Dutch sources, but no sensitivity analysis of the costs was performed. As with the benefits, the costs were not reported for each of the programmes under study. Further, it was unclear if cost-effectiveness was determined using average cost-effectiveness/utility ratios, rather than using an incremental cost-utility ratio. Again, such lack of details will limit the generalisability of the authors' results and their internal validity.
The authors did not compare their findings with those from other studies. The issue of generalisability to other settings was not addressed. The authors do not appear to have presented their results selectively. However, although the authors reported in detail the methods used in their study and compared many different screening programmes (i.e. age of women to be screened, presence of breast cancer relatives, screening intervals), the results were not presented satisfactorily. For example, the life-years or QALYs gained with each programme were not reported or summarised, nor were the costs. Further, incremental cost-utility/effectiveness ratios were not reported.
The authors reported a number of further limitations to their study. First, the presented study was a simulation study which included values presented in the medical literature. Second, the authors used a model that had not been validated. Finally, the authors did not include other predisposing factors for breast cancer such as early menarche, null parity and late menopause.
Implications of the study
The authors recommended further research to determine the impact of screening on different sub-groups of women.
Source of funding
Supported by the National Health Insurance Board of the Netherlands.
Jacobi C E, Nagelkerke N J, Van Houwelingen J C, De Bock G H. Breast cancer screening, outside the population-screening program, of women from breast cancer families without proven BRCA1/BRCA2 mutations: a simulation study. Cancer Epidemiology, Biomarkers and Prevention 2006; 15(3): 429-436
Other publications of related interest
Jacobi CE, Jonker MA, Nagelkerke NJ, et al. The prevalence of family histories of breast cancer in the general population and the incidence of related health care visits. J Med Genet 2003;40:e-83.
Jonker MA, Jacobi CF, Hoogendoorn WE, et al. Modeling familial clustered breast cancer using published data. Cancer Epidemiol Biomarkers Prev 2003;12:1479-85.
Subject indexing assigned by NLM
Adult; Age Factors; Breast Neoplasms /epidemiology /genetics; Cost-Benefit Analysis; Female; Genes, BRCA1; Genes, BRCA2; Humans; Mass Screening /economics /methods; Middle Aged; Models, Theoretical; Mutation; Netherlands /epidemiology; Patient Simulation; Pedigree; Prevalence; Research Support, Non-U.S. Gov't; Risk Assessment; Survival Rate
Date bibliographic record published
Date abstract record published