Total cost-effectiveness of mammography screening strategies
Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.
by Nicole Mittmann, Natasha K. Stout, Pablo Lee, Anna N.A. Tosteson, Amy Trentham-Dietz, Oguzhan Alagoz and Martin J. Yaffe
Implementation of screening programs can have significant budget implications, depending on the size of the population affected and the health care system resources involved. Recommendations for mammography screening are continually being updated and modified —the age range, frequency, effectiveness and cost-effectiveness of population-wide screening are ongoing topics of debate.Note 1Note 2Note 3 Decisions about whether to screen, who should be screened, what modalities to use, and how frequently to screen are best made when the trade-offs between improved health outcomes, potential harm, and the economic impact are understood.
Economic evaluation of mammography screening is particularly important in countries like Canada that have single-payer publicly funded health care systems. Studies have estimated the costs of breast cancer treatment from the Canadian perspective.Note 4Note 5Note 6Note 7Note 8 However, none have examined or incorporated the full cost of screening, even though breast cancer natural history models have been developed to project the impact of different mammography strategies.Note 9Note 10Note 11Note 12Note 13Note 14 The objective of this analysis was to evaluate the costs, outcomes and cost effectiveness of various mammography strategies, using a validated breast cancer simulation model.Note 15
Data and methods
The framework for this analysis is the Canadianized University of Wisconsin Breast Cancer Epidemiology Simulation Model, developed under the U.S. National Cancer Institute-funded Cancer Intervention and Surveillance Modeling Network (CISNET) programNote 15Note 16Note 17 (www.cisnet.cancer.gov/breast/). This study describes inputs specific to resource use and analyzes the cost-effectiveness of screening.Note 18 A total of 11 screening strategies are examined—annual, biennial, and triennial across different age groups (starting at age 40 or 50 and ending at age 69 or 74)—compared with No Screening. The results are based on calculations for a 1960 birth cohort of women in the general population. Clinical model inputs are described in the original modelNote 12; details of the Canadianized modified inputs are provided elsewhere.Note 18
Resources associated with screening, diagnosis, and cancer management were identified: mammography, clinic visits, physicians, diagnostic procedures, and treatment (surgery, radiation, medication). To account for the societal perspective, the costs of lost productivity related to screening, screening results and diagnosis, and the cost of premature death were included. Once identified, quantities of resources utilized were determined based on guidelines, reports, peer-reviewed literature, and expert opinion (Appendix Table A). The base case assumed that 100% of eligible women would be screened using digital mammography. All positive screens and all detection of suspicious findings outside of screening were assumed to incur non-invasive work-up costs; a subset of these positive screens and suspicious findings incurred invasive work-up costs. It was assumed that all women took time off work for the mammogram, and during the first year after a diagnosis. Productivity loss associated with premature death due to breast cancer was also counted.
All women diagnosed with breast cancer received either a mastectomy or a lumpectomy with or without radiation. Chemotherapy was assigned by stage of disease. Hormonal treatment was assigned by estrognen receptor status (ER+/ER-). Traztuzumab was assigned by human epidermal growth factor receptor 2 (HER2) status. Women with ductal carcinoma in situ did not receive trastuzumab (Appendix Table A).
Canadian unit costs ($2012CAN) were applied to each resource used and modelled ($1.01US = $1CAN, based on December 31, 2012).Note 19 Non-2012 costs were converted to 2012 values using the Consumer Price Index.Note 20 The sources consulted to determine costs were provincial drug formularies, national statistics programs, costing programs, health resource management, and the literature. Capital or institutional costs of the equipment were not included. Lost productivity costs were based on the literature and the average wage per woman (Appendix Table A).
Age-specific population health preference valuesNote 21 were derived from U.S. Medical Expenditure Panel Survey data applying the EuroQoL EQ-5D instrument using U.S. scoring.Note 22 For women with newly diagnosed breast cancer, decrements to quality of life based on stage lasting for one year post-diagnosis were assumed, after which the women would return to their appropriate age-specific value. For women diagnosed with Stage IV breast cancer, the decrement was applied to their remaining lifetime. For screening-related health states, disutilities of 0.006 for one week and 0.105 for five weeks were applied for a screening mammogram and a positive screening result, respectively.Note 23
To estimate lifetime costs of screening and management, the overall costs and for premature death for each strategy and for No Screening were calculated. The outcome was quality-adjusted life-years (QALY).
The costs and outcomes of each screening strategy were compared in an incremental stepwise cost-effectiveness analysis: a strategy was considered to be efficient if it was not “dominated,” that is, if no alternative strategy had improved outcomes for the same or lower cost, or no combination of two other strategies had improved outcomes for the same cost. Incremental cost-utility ratios (ICURs) were computed as the difference in cost divided by the difference in outcomes only for the efficient strategies. The average cost-effectiveness of each screening strategy relative to No Screening was also determined. All costs and health outcomes for the incremental analyses were discounted at a 5% rate. Univariate sensitivity analyses were conducted in which values for key parameters such as screening rates, screening sensitivity and specificity, health preference, treatment costs, cost of medications, and discount rate were varied (Appendix Table B).
The overall costs (undiscounted) for annual screening of 1,000 women in the general population ranged from $11.3 million (ages 50 to 69) to $16.0 million (ages 40 to 74) (Table 1). Costs for biennial and triennial screening of 1,000 women ranged from $8.4 million to $11.2 million (ages 50 to 69 and 50 to 74), and from $7.6 million to $8.3 million (ages 50 to 60 and 50 to 74), respectively. The overall cost of No Screening for 1,000 women was $4.9 million ($4,875 per woman) over a lifetime.
Active screening itself was a cost-driver, making up a substantial portion of the overall cost. The ratio of the cost of screening to the overall cost was proportional to the aggressiveness of the strategy. Treatment costs were slightly higher for active strategies, compared with No Screening. Screening costs for more aggressive strategies were proportionally higher when compared with treatment and procedure costs.
Lowering the age of eligibility for screening from 50 to 40 added $2 to $3 million to the cost for 1,000 women ($2,000 to $3,000 lifetime cost per woman). Raising the upper age limit from 69 to 74 added approximately $1 million per 1,000 women ($1,000 per woman) to the total cost. If the cost of premature death was included in the cost, the overall cost increased by $7.4 to $10.8 million, depending on the strategy, with more aggressive screening being associated with fewer lives lost lives, and thus, lower costs due to premature deaths.
For the cost-effectiveness analysis, screening strategies for women aged 50 to 69 had lower incremental ratios than strategies in which screening continued to age 74 (Table 2). Annual screening at ages 40 to 69 and at ages 40 to 74 were considered the dominant strategies, compared with annual screening of narrower age ranges.
Compared with No Screening, all incremental ratios for the societal perspective were less than $155,000/QALY. The most favourable (lowest) ratios were for the least frequent screening (triennial) of the narrowest age group (50 to 69): $94,762/QALY. The most aggressive strategy (annual screening at ages 40 to 74) yielded the least favourable (highest) cost-effectiveness ratio: $154,187/QALY (Table 2, Figure 1).
A comparison of active screening strategies with No Screening revealed that the major cost drivers were no treatment (medication, surgery, or radiation) after a diagnosis and no subsequent screening (Table 3). The model was generally insensitive to changes in missed screening and no subsequent screening, but showed more favourable incremental ratios when the percentage receiving adjuvant chemotherapy was reduced, when screening costs were decreased, and when specificity and sensitivity were 100%. Not providing treatment, and thereby expediting disease progression, had the greatest impact on the ICUR. The model was sensitive to modifications in utility values, with more favourable (lower) incremental cost-utility ratios when the utility values were increased by 25%, thereby showing greater incremental benefits between the active screening strategies and No Screening. When the utility values were decreased by 25%, less favourable (higher) cost-utility ratios were modelled, owing to smaller differences in the benefit. Finally, changing the discount rate to 0% substantially reduced the ICURs.
Early diagnosis of breast cancer through screening mammography can save lives, but costs and possible harms may be associated with screening and subsequent treatment. This is the first analysis of the lifetime cost-utility of various mammography screening strategies for Canada using an accepted population health model.
The main cost driver of the active screening strategies was the frequency of screening. A significant reduction in the cost of premature death was associated with more frequent screening, with an inverse relationship between the cost of screening and the cost of premature death.
Narrower age groups and less frequent screening were optimal when cost-effectiveness rather than lifesaving was considered. The exception was annual screening at ages 40 to 74, which was not dominated, but had high incremental cost utility ratios.
All active screening strategies were more effective than not screening. Compared with No Screening, incremental ratios for active screening strategies generally fell below $150,000 per QALY. The average incremental ratios generated from the model are in line with other ICURs used in oncology decision-making ($/QALY).Note 24 Moreover, the incremental ratios are well below the willingness-to-pay threshold of $300,000/QALY, which is defined as “good value for money” by about half of Canadian and American oncologists, with another third using $100,000/LYG to define good value.Note 25 A recent editorial on incremental ratios questioned the wisdom of existing thresholds.Note 26
Annual screening strategies had higher average incremental ratios than did less frequent screening, but they were associated with greater benefits. For this analysis,assumptions about total costs were based on lost productivity—having a screening mammogram (1/2 day), a positive screen (5 weeks), and a diagnosis of invasive cancer (one year). Any change in these conservative time estimates would affect the overall cost. The more aggressive the screening strategy, the more recalls for no cancer, and the more cancers detected, both of which result in more lost productivity, higher incremental costs, and higher incremental ratios (that is, less favourable cost-effectiveness). Because QALYs and costs rise with the number of screens per woman, decisions about screening strategies are mainly related to willingness to pay and avoiding the recall of too many women for further examinations after positive screens. Thus, a screening tool that provides higher specificity would be beneficial.
A comparison of the active screening strategies with No Screening showed a relatively tight range of incremental ratios: within $40,000 to $60,000 of each other. Raising the age of eligibility from 69 to 74 marginally increased the incremental ratios because of additional screening costs, but it also improved outcomes. Lowering the age of eligibility from 50 to 40 increased incremental ratios, again because of higher screening costs, but this, too, yielded more QALYs.
When examined by age group (50 to 69 and 50 to 74), the cost-effectiveness ratios for annual, biennial and triennial screening compared with No Screening showed similar values. The choice of a screening strategy based on cost-effectiveness should also consider the improvement in QALY associated with more frequent screening.
For the univariate sensitivity analysis, based on the analyses for the societal perspective (discount = 5%), where the QALYs of different screening strategies were compared with No Screening, the model was generally robust. The incremental ratio was generally insensitive to changes in missed screening, no subsequent screening, and reductions in the percentage of women receiving adjuvant chemotherapy. The model, however, was sensitive to decreased screening costs and improved specificity and sensitivity, which led to more favourable (lower) ICURs. Not providing treatment (medication, radiation, surgery), and thus expediting disease progression, had the greatest impact on the ICUR. Modifying the same parameters for the cost-utility analysis produced results similar to the cost-effectiveness analysis, but also showed that the model was sensitive to modifications in utility values—more favourable(lower) ICURs were predicted when the utility values were increased by 25%, thereby showing greater incremental benefits between active screening and No Screening. By contrast, when the utility values were decreased by 25%, less favourable (higher) ICURs were generated as a result of smaller differences in the benefit, which led to higher ratios. Most notably, incremental ratio values fell substantially when the no discount rate was applied; discounting significantly affected long-term effectiveness outcomes over the model time horizon, whereas costs were up front.
According to an analysis published in 2014 of population-based mammography screening from a Canadian health system perspective using screening diagnostic, treatment costs and utility values, the most cost-effective strategies were biennial screening of women aged 50 to 69 and 40 to 69.Note 27 That model used different costs per treatment, did not consider triennial screening strategies, and did not include the cost of lost productivity or consider the societal perspective.
Strengths and limitations
The Wisconsin model allowed simulation of the growth of a distribution of breast cancers within a cohort of women and consideration of the individual effects of various detection strategies and treatment regimens on mortality and other outcomes. The strength of the Wisconsin model is that it has been validated against empirical U.S. data. Modified for use in the Canadian context, the model performed well in predicting breast cancer incidence in the absence of screening.Note 18 In addition, the model used empirical data on the sensitivity and specificity of screening mammography versus age and breast density to describe the screening process. Canadian data on the use of therapies and on costs were employed, and no assumptions about the mortality reduction associated with screening were applied explicitly in the model.
Most studies that have evaluated the cost-effectiveness of screening strategies have been conducted from a U.S. health system perspective and focused on different risk factors such as early and late age and genetic profile.Note 9Note 11Note 28Note 29 Unlike the present analysis, no studies have examined the impact of lost productivity from the societal perspective in a general population of women, or for Canada.
The results of this analysis should be interpreted in the context of several limitations. First, the model itself has shortcomings that have been outlined in previous work.Note 18
Secondly, the model assumed that 100% of eligible women would be screened, whereas compliance is markedly lower. In Ontario, the 2010/2011 screening rate through an organized program was 61% for women aged 50 to 74.Note 30 When the screening rate was lowered to 50% in the present model, incremental ratios remained similar to those of the base case, mainly because along with a decrease in the costs of screening, the number of invasive cancers detected, which affects QALY, also decreased.
Finally, to avoid double-counting, the incremental evaluations did not include the cost of premature death. However, more frequent screening might be expected to reduce costs associated with premature deaths from breast cancer.
The results of this analysis may be used to help determine appropriate breast screening strategies for a population. The single greatest cost contributor in a screening program is the mammography itself, which exceeds the costs of therapy. Because lives saved and costs both rise with the number of screens per woman, decisions about screening strategies are mainly related to willingness to pay and avoiding recalling too many women for further examination with no cancer detected. Future models will consider the impact of different screening technologies and populations on both costs and outcomes.
Acknowledgments and funding
The work on which this report was based was supported by a contract from The Canadian Breast Cancer Foundation. The University of Wisconsin breast cancer simulation model was supported by grant number U01 CA152958 from the National Cancer Institute through the Cancer Intervention and Surveillance Modeling Network (CISNET). Model input data on the performance of screening mammography were provided by the National-Cancer-Institute-funded Breast Cancer Surveillance Consortium (BCSC), grant number UC2CA148577 and contract number HHSN261201100031C.The content of this study does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health. The collection of BCSC cancer data was supported by several state public health departments and cancer registries throughout the United States. A full description of these sources is available at: http://www.breastscreening.cancer.gov/work/acknowledgement.html. The authors thank the participating women, mammography facilities, and radiologists for the data they provided. A list of the BCSC investigators and procedures for requesting BCSC data for research purposes is provided at: http://breastscreening.cancer.gov/