ARCHIVED - Organized Breast Cancer Screening Programs in Canada - Report on Program Performance in 2005 and 2006

Special Topic

Decision Analysis Models for Outcomes Related to Breast Cancer Screening

Introduction

Breast cancer is the second most frequent cancer among Canadian women, with a projected 23,200 new cases in 2010 alone.⁽¹⁾ At current rates, 11% of Canadian women are expected to develop breast cancer at some point in their lives, although only 2 to 3% of women between 50 and 90 years of age are expected to develop the disease in the next ten years^.(11) About 2/3 of cases are expected to survive, yet breast cancer is still the second-leading cause of cancer mortality among Canadian women, with a projection of 5,300 deaths in 2010.⁽¹⁾ The average 5-year survival rate is 87%.⁽¹¹⁾ Survival increases to 96 % (+/- 3%) if the cancer is found at Stage I, yet can be as low as 26% (+/- 10%) if found at Stage IV.⁽¹²⁾ Finding breast cancer early is critical.

Regular mammography screening can help detect breast cancers early, and consequently improves survival rates. Current Canadian guidelines recommend women to be screened biennially (every two years) in their 50s and 60s. Above the age of 70, early detection is deemed to have less benefit due to an increase in other competing risks for mortality. Below the age of 50, the risk for breast cancer is lower and it is preferable to avoid unnecessary exposure to radiation and unwanted anxiety. Yet almost half of all breast cancers occur in these age groups. Literature suggests there may be untapped screening benefits outside the 50 to 69 target screening population.^{(13,14,15,16)}

Decision support tools, often referred to as decision aids, are applications built based on data obtained through decision analysis. Decision support tools assist in informed decision making by providing important information regarding the risks and benefits of difficult health decisions, in this case, related to breast cancer screening mammography. A better understanding of the true outcomes related to breast cancer screening improve informed decision making and may increase the number of women returning for timely mammography (retention) which in turn increases the benefits from screening. Therefore a Markov decision analysis model for mammography screening among Canadian women was built to assess the potential long-term benefits and harms of screening and used to inform a decision support tool (www.publichealth.gc.ca/decisionaids).

ⁱ Incidence of non-melanoma skin cancer exceeds that of breast cancer in Canada, however, rates are typically not reported due to difficulty estimating true incidence.

Methods

A Markov decision analysis model was developed to estimate the outcomes related to three hypothetical cohorts of women and their screening experiences over a ten to 20 year period. The age cohorts and their comparators that were included were:

• 40 to 49 years screening annually for 10 years compared to no screening,
• 50 to 69 years screening biennially for 20 years compared to no screening, and
• 70 to 79 years screening biennially for 10 years compared to screening biennially for 20 years and stopping after age 69.

Detailed technical notes and literature references are available at www.publichealth.gc.ca/decisionaids.

The model was developed using TreeAge software and was based on an analogous Australian model.⁽¹⁷⁾ The model tracks a variety of outcomes including: number of abnormal and normal screen results, follow-up imaging and biopsy requirements, breast cancers detected, stage of disease at diagnosis, and deaths due to breast cancer or other causes. False negatives screens were estimated by the number of interval cancers within 12 months of the last screen among women in their forties, and within 24 months among women within their fifties. The model assumes time-constant transition probabilities, full compliance for return to screening and independent screening outcomes. Before applying the model to Canadian data, its structure was successfully validated against the Australian model outcomes generated by 1996-1998 BreastScreen Australia data.⁽¹⁷⁾

Screening data were acquired from the Canadian Breast Cancer Screening Database (2000, 2001, 2002, 2003, 2004), aggregate pan-Canadian data from provincial screening programs in practice. All provinces contributed some data and specific contributions are noted in Appendix A. Among the 40 to 49 year cohort, screening data were contributed predominately by British Columbia. Breast cancer mortality, breast cancer incidence, population counts, and all-cause mortality were obtained from Statistics Canada (2000, 2001, 2002, 2003, 2004). Canadian Community Health Survey cycles 1.1 and 2.1 were used to estimate national participation in screening. A literature review was performed to obtain relative risk reductions in breast cancer mortality due to screening.

A long-term scenario (10 to 20 years) was used to account for the mortality lag observed in a new screening population.⁽¹⁷⁾ Cancers observed in younger women are typically more aggressive therefore we applied the scenario of annual screening for women in their 40's but biennial screening for women 50 and above. For women between 40 and 69, we contrasted screening with full participation versus no screening to highlight the full range of differences between screeners and non-screeners. For women between 70 and 79, we contrasted screening with full participation starting at age 50 to 79 versus screening with full participation between age 50 and 69 only.

Results

The results indicated that 3,000 women, aged 40 to 49 years, would need to be screened annually in order to prevent 1 death from breast cancer. Comparatively, 250 women aged 50 to 69 and 400 women aged 70-79, would need to be screened biennially to prevent 1 death from breast cancer.

Most women who undertake mammography screening are given peace of mind from knowing they are breast-cancer free. Among women aged 40 to 49 years, 631 of the 10,000 mammograms will be abnormal resulting in 549 women being recalled at least once. After further testing, 533 abnormal screens will result in no breast cancer and 16 cancers will be detected. Three post-screen cancers will be diagnosed between rounds of screening (Figure 10. pg39 and Table 9. below ).

Similarly, among women aged 50 to 69 years, 717 of the 10,000 mammograms will be abnormal resulting in 574 women being recalled at least once. After further testing, 529 abnormal screens will result in no breast cancer and 45 cancers will be detected. Fifteen post-screen cancers will be diagnosed between rounds of screening (Figure 11. pg41 and Table 10. pg41 ).

Lastly, among women 70 to 79 years, 270 of the 5,000 mammograms performed will be abnormal, resulting in 244 women being recalled at least once. After further testing, 213 abnormal screens will result in no breast cancer and 31 cancers will be detected. Eight post-screen cancers will be diagnosed between rounds of screening (Figure 12. below and Table 11. pg43 ).

Discussion

The results of this modelling exercise suggest important variation in outcomes related to mammography screening among women of different age groups, highlighting the need for informed decision making by women considering attendance. In addition, the benefits and limitations of a modelling approach to screening outcomes are important to understand, as is the method in which the results are presented to women.

A high number of abnormal screening mammograms were predicted for all age groups (Table 9-11. pg40-43 ). This seems like disturbing news; a large proportion of screened women will expect to receive further testing only to confirm they have no cancer. The impact of this anxiety should not be underestimated because women who experience false positive mammograms are less likely to return for regular screening in the future. However, these abnormal results represent a small fraction of the total mammograms performed. Because we assumed independent outcomes, screening recall rates were applied at random to the population, resulting in a high cumulative chance of being recalled at least once over the 10 to 20 year screening period. In practice, some women (those with dense breasts or prior false positive screens) are more likely to receive a false positive than others; the model did not reflect such individual factors. Further, technological changes, such as the introduction of digital mammography, will change outcomes likely altering numbers of abnormal screening mammograms and patterns of cancer diagnoses, and ultimately will require changes to the decision support tool.

The primary benefit of screening is early detection, which typically results in simpler treatment, a lower chance of recurrence and a greater chance of survival. All screened cohorts had a fewer number of cancers found at a late stage (III or IV) and a greater number of cancers found at stage I than the unscreened cohorts (Table 9-11. pg40-43 ). Yet there were also more cancers found overall. Some may assume that screening increases breast cancer incidence through exposure to radiation; however, research suggests that benefit from screening outweighs the risk attributed to radiation exposure.(18,19,20) The likely cause of increased breast cancers among women who were screened is due to the detection of a large number of asymptomatic ductal carcinoma in situ (DCIS) that would not have been found without screening. DCIS is considered to be stage 0 breast cancer that can be easily treated, and does not pose an immediate threat to the woman. Overdetection, the detection of cancers that never would have presented clinically during the patient's lifetime, is also an issue in screening mammography.(21) Randomized controlled clinical trials need to be conducted to accurately estimate overdetection in Canada.

Prevention of death through screening is a strong motivation for women to undertake regular mammography. The greatest mortality benefit falls within the target screening group of 50 to 69 years, followed by the 70 to 79 year group. Women in their 40's received a lesser mortality benefit. The benefits of screening by women in different age groups may vary depending which outcomes are considered. Years of life saved may be an important measure among younger women where fewer deaths are expected but they occur at an earlier age compared to older women.

It should be noted that the model uses "average" population-level results to give relative estimates of likelihoods in the screening process. Individual risks and experiences will differ, depending on risk factors such as the BRCA mutation, age, and family history. It is important to take such factors into consideration when trying to assess an individual's risks. The decision support tool is accompanied by risk factor information to facilitate this (www.publichealth.gc.ca/decisionaids).

Models can never fully and completely predict future outcomes but can add evidence to the bigger picture to inform decision-making. These outcomes are included in the decision support tool (www.publichealth.gc.ca/decisionaids) to help women make informed decisions about participation in mammography screening. The decision aids contain general information about mammography and specific sections targeted at three age groups: 40 to 49, 50 to 69, and 70 to 79. For each age group, outcomes derived from Figures 1 through 3 are used to provide a general picture of each outcome. Mortality benefits are also shown. Women are then given an opportunity to rank their feelings about each benefit and limitation of the screening experience. The goal is to use a combination of these personal feelings, individual risk factors, and provided simulated outcome data to make a more informed decision about mammography screening.