|Home | About | Journals | Submit | Contact Us | Français|
Van Ravesteyn NT, Heijnsdijk EAM, and de Koning HJ: Department of Public Health, Erasmus MC, P.O. Box 2040, 3000 CA Rotterdam, The Netherlands.
Miglioretti DL and Buist DSM: Group Health Research Institute, 1730 Minor Avenue, Suite. 1600 Seattle, WA 98101
Stout NK: Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, 133 Brookline Ave, 6th Floor, Boston, MA 02215
Lee SJ: Dana-Farber Cancer Institute, 3 Blackfan Circle, Boston, MA 02115.
Schechter CB: Dept. of Family & Social Medicine, Albert Einstein College of Medicine, 1300 Morris Park Ave - Mazer Bldg. 406, Bronx, NY 10461 USA.
Huang H:. Dana-Farber Cancer Institute, 44 Binney Street, Boston, MA 01720.
Trentham-Dietz A; University of Wisconsin-Madison, 610 Walnut St., WARF Room 307, Madison WI 53726.
Alagoz O: University of Wisconsin-Madison, 3242 Mechanical Engineering Building 1513 University Avenue, Madison, WI 53706
Near AM and Mandelblatt JS: Lombardi Comprehensive Cancer Center, 3300 Whitehaven Street, NW, suite 4100, Washington, DC 20007.
Kerlikowske K: University of California, San Francisco, 4150 Clement Street, VAMC (111A1), San Francisco, CA 94121
Nelson HD: Oregon Evidence-based Practice Center, Oregon, Health & Science University, Mailcode BICC, 3181 Southwest Sam Jackson Park Road, Portland, OR 97239-3098.
In the US biennial screening mammography is recommended for average-risk women aged 50–74 because the benefits outweigh the harms. For women with increased risk starting screening at age 40 may have a similar harm-benefit ratio.
Determine the threshold relative risk (RR) at which the harm-benefit ratio of screening women aged 40–49 equals that of biennial screening for women aged 50–74.
Comparative modeling study.
Surveillance, Epidemiology, and End Results, Breast Cancer Surveillance Consortium, medical literature.
A contemporary cohort of women eligible for routine screening.
Mammography screening starting at age 40 vs. 50 with different screening modalities (film, digital) and screening intervals (annual, biennial).
Benefits: life-years gained, breast cancer deaths averted; harms: false-positive mammography examinations; and harm-benefit ratios: false positives/life-year gained, false positives/death averted.
Screening average-risk women aged 50–74 biennially yields the same false positives/life-year gained as biennial screening with digital mammography starting at age 40 for women with a 2-fold increased risk above average (median threshold RR 1.9; range across models 1.5–4.4). The threshold RRs are higher for annual screening with digital mammography(median 4.3; range 3.3–10) and higher when false positives/death averted is used as outcome measure instead of false positives/life-year gained. The harm-benefit ratio for film mammography is more favorable than for digital, because film has a lower false-positive rate.
The threshold RRs changed slightly when a more comprehensive measure of harm was used and were relatively insensitive to lower adherence assumptions.
Risk was assumed to influence onset of disease without influencing screening performance.
Women aged 40–49 with a 2-fold increased risk have similar harm-benefit ratios for biennial screening mammography as average-risk women aged 50–74. Threshold RRs required for favorable harm-benefit ratios vary by screening modality, interval, and outcome measure.
National Cancer Institute.
Breast cancer is the most frequently diagnosed non-cutaneous cancer among women in the United States (US), where it is second only to lung cancer as a cause of cancer deaths. Mammography screening has been shown to reduce breast-cancer mortality rates in randomized trials (1–2) and nationwide screening programs (3).
The U.S. Preventive Services Task Force (USPSTF) recommends biennial breast cancer mammography screening for women between age 50 and 74 years based on a comprehensive evaluation of current research indicating a favorable balance of benefits and harms (4–5). Individual trials have not demonstrated significant breast-cancer mortality reductions from screening women in their 40s (6–7), but a meta-analysis of eight trials demonstrated a 15% mortality reduction (8). However, the absolute benefits (e.g. number of deaths prevented) are smaller than for older women, because of the lower incidence of breast cancer and lower sensitivity of mammography in women aged 40–49. At the same time, screening in this age group is accompanied by more harm (false-positive results and unnecessary biopsies) as a result of lower screening specificity. As a result, the USPSTF concluded that the ‘decision to start regular, biennial screening mammography before the age of 50 years should be an individual one and take into account patient context, including the patient’s values regarding specific benefits and harms’ (4, 9).
Technology that improves screening test performance might influence the balance of benefits and harms of screening in younger women. Digital mammography has rapidly replaced film mammography in most areas of the US (10). Younger women are more likely than older women to have dense breasts and screening regimens using digital mammography in women 40–49 might have a different balance of benefits and harms for them than strategies that use film mammography. Digital has a higher test sensitivity than film mammography in women below age 50 (11), has been found to detect more ductal carcinoma in situ and lead to more false positives (12). Thus, it is uncertain whether initiating screening at age 40 with digital mammography would yield a more favorable balance of benefits and harms. Another factor that changes the balance of benefits and harms is risk for developing breast cancer. Clearly, the absolute benefits of screening before age 50 will be larger for women with an increased risk of breast cancer than for average-risk women. A more risk-based screening approach might therefore be appropriate (13–17). To implement a risk-based screening approach, it is crucial to know the magnitude of the relative risk (RR) that would tip the balance of benefits and harms to recommend screening for women aged 40–49 (i.e., threshold RR) and which risk factors lead to that elevated risk.
This study was completed to determine the threshold risk at which the harm-benefit ratio of starting screening at age 40 equals the harm-benefit ratio of currently recommended biennial screening for average-risk women starting at age 50. We determined the impact of screening modality (film, digital) and screening interval (annual, biennial) on the threshold RR using four simulation models (18–21).
We used four micro-simulation models developed as part of the Cancer Intervention and Surveillance Modeling Network (CISNET), which is an international collaborative modeling consortium funded by the National Cancer Institute (22–23). The four models were: model D (Dana-Farber Cancer Institute, Boston, Massachusetts), model E (Erasmus University Medical Center, Rotterdam, the Netherlands), model G–E (Georgetown University Medical Center, Washington, DC, and Albert Einstein College of Medicine, Bronx, New York), and model W (University of Wisconsin, Madison, Wisconsin, and Harvard Medical School, Boston, Massachusetts). The models have been described in detail elsewhere (18–21) and information about the models can be found online (http://cisnet.cancer.gov/). Briefly, the models simulated life histories for individual women. After estimating breast-cancer incidence and mortality in the absence of screening and treatment, the models overlaid screening use and improvements in survival associated with treatment advances (22). A schematic representation of the influence of breast-cancer screening on (simulated) life histories is shown in Figure 1. The main model differences and assumptions are outlined in Appendix Table 1.
In the present study, the model inputs and assumptions were based on assumptions previously used in one of the supporting articles for the most recent USPSTF recommendation (23). We used a common set of age-specific variables for breast-cancer incidence, survival, and competing non-breast cancer causes of death (Appendix Table 2). A cohort of women born in 1960 was simulated and followed throughout their entire lifetime. We assumed 100% adherence with screening and adjuvant treatment guidelines.
The Breast Cancer Surveillance Consortium (BCSC) provided data on recent performance (between the years 2001 and 2007) of film and digital mammography(Table 1). The BCSC collects prospective data on breast imaging in community practice in five mammography registries and two affiliated sites in the US (http://breastscreening.cancer.gov). Each registry obtains annual approval from its institutional review board for consenting processes or a waiver of consent, enrollment of participants, and ongoing data linkages for research purposes. All registries have received Federal Certificates of Confidentiality that protect the identities of research subjects. The four models used BCSC data inputs for sensitivity, specificity, and stage distribution by age, screening modality (film, digital), and screening interval (annual, biennial) (http://breastscreening.cancer.gov/data/elements.html).
The effects of five screening scenarios were estimated per 1,000 average-risk women aged 40 followed over their lifetimes. These scenarios included biennial screening for women aged 50–74 extended with four screening scenarios for women aged 40–49 varying by screening interval (annual and biennial) and screening modality (film and digital):
The incremental effects of each scenario were determined by comparing it to the previous less intensive scenario. Thus, to determine the effects of adding biennial screening for women in their forties, we compared scenarios 2a and 2b with scenario 1 and for annual screening, scenarios 3a and 3b were compared with scenario 2a and 2b, respectively.
We estimated the impact of each screening strategy on the number of breast cancers detected, the number of breast-cancer deaths averted, the number of life-years gained, and the number of false-positive mammography screening tests. The time horizon for calculating effects was from age 40 until death of all simulated women.
First, the effects (benefits and harms) of biennial screening between the age of 50 and 74 years were determined. We defined harm as the number of false positives and benefits as the number of breast-cancer deaths averted and number of life-years gained. Then, the additional effects of screening between age 40 and 49 were assessed for two screening intervals (annual and biennial) and two screening modalities (film and digital). For each screening strategy, we determined the harm-benefit ratios by dividing the incremental harm by the incremental benefits.
We then implemented different relative risks (RRs) in the models. The higher risk was modeled over the entire lifetime of the simulated women. We calculated the incremental harms, benefits, and harm-benefit ratios of the five screening scenarios for women at increased risk. We used the harm-benefit ratio of biennial screening for average-risk women between age 50 and 74 (scenario 1) as a threshold value. Then, we determined for the four screening scenarios (scenarios 2a&b and 3a&b) that started at age 40 how high the RR needed to be in order to have the same harm-benefit ratio as the threshold value.
To evaluate how the harm-benefit ratios and threshold RRs were influenced by certain assumptions and parameter values we performed several sensitivity analyses. First, we explored the effect of reduced adherence (70%). Second, we considered alternative screening test characteristics of digital mammography (‘best-case scenario’) by using an improved sensitivity and specificity (using the upper limit of the 95% confidence interval). Third, we assessed the influence of a broader measure of harm on the threshold RRs by calculating quality-adjusted life-years (QALYs) lost. We applied quality-of-life decrements due to undergoing a mammogram and diagnostics (24), and life-years with breast cancer by stage of disease at diagnosis (25) (see Appendix Table 3). The number of ‘QALYs lost’ incorporates harm from undergoing a mammogram, having a false-positive screening test, and also harm from overdiagnosis, because with more overdiagnosis, more life-years are spent in disease stages, and there are, thus, more QALYs lost when there is a large amount of overdiagnosis. All four models incorporate overdiagnosis, because in all models there is a chance that breast cancer is detected in the presence of screening but would not have become symptomatic during a woman’s lifetime if no screening had taken place. Overdiagnosis estimates vary across models because of different underlying assumptions, e.g. on the progression rates of ductal carcinoma in situ to invasive disease and on the possibility of limited malignant potential of invasive disease. We used the harm-benefit ratio QALYs lost/life-year gained to re-estimate the threshold RRs.
This work was funded by a supplement from the National Cancer Institute. The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, and approval of the manuscript; or decision to submit this manuscript for publication.
In the absence of screening, the models estimate that 153 cases of breast cancer would be diagnosed (median; range across models 152–158) and 25 deaths from breast cancer would occur (range 19–35) among 1,000 40-year-old women followed over their lifetimes (data not shown).
If these women underwent biennial film mammography between the ages of 50 and 74, the models predict that 10,610 (range 10,529–10,660) mammograms would be performed over 12 screening rounds; 6.3 (range 5.3–6.9) breast-cancer deaths would be avoided; 109 (range 93–113) life-years would be gained; and 883 (range 795–939) mammography examinations would be read as false-positives. The harm-benefit ratios are estimated as 8.3 (range 7.8–8.7) false positives/life-year gained and 146 (range 128–151) false positives/death averted (Table 2).
In all models, the harm-benefit ratios for adding screening between ages 40 and 49 are less favorable than those for biennial screening starting at age 50 (Table 3). In all models, adding annual to biennial screening leads to slight increases in additional life-years gained and breast-cancer deaths averted, but at the expense of greater increases in incremental harm. Adding annual screening to biennial screening starting at age 40 therefore has a less favorable harm-benefit ratio than adding biennial screening starting at age 40 to biennial screening from age 50–74 in all four models (Table 3).
With digital mammography screening, more life-years are gained and more breast-cancer deaths averted than with plain film mammography, but because of the lower specificity of digital mammography for women in their 40s, it also yields more false-positives. In all four models, there is greater harm relative to benefit from digital than from film mammography in women age 40–49 (Table 3).
In all models, screening women with increased risk for breast cancer leads to a more favorable harm-benefit ratio. Screening women with increased risk results in more life-years gained and more breast-cancer deaths averted with approximately the same number of false-positives (see Appendix Table 4 and 5 for RR 2). The models predict that annual screening with digital mammography for women aged 40–49 years with a 4-fold increased risk above average (median threshold RR 4.3; range across models 3.3–10) would yield similar false positives/life-year gained estimates as biennial screening for average-risk women aged 50–74 years. To find similar false positives/life-year gained estimates for biennial screening with digital mammography for women aged 40–49 years, the threshold RRs are lower in all models (median threshold RR 1.9; range across models 1.5–4.4).
When screening with film mammography, the threshold RRs are predicted to be somewhat lower in all models than those estimated for digital mammography (median threshold RR for biennial screening 1.6; range 1.5–3.7; Table 4). When deaths averted was considered as an outcome measure instead of life-years gained, all models estimated higher threshold RRs of 3.3 (range 2.3–5.7) for biennial screening and even higher threshold RRs for annual screening (Table 4). The incremental changes in the benefits and harms of starting screening at age 40 instead of at age 50 per 1,000 women for the threshold RRs are listed in Appendix Table 6.
When 70% adherence was assumed instead of 100% the harms diminished by 30%, while the benefits diminished less (by 26% for biennial and 5% for annual screening). Therefore, more favorable harm-benefit ratios were found for biennial screening (e.g., for film mammography in model E: 12.2 false positives/life-year gained vs. 12.7 when assuming 100% attendance) and for annual screening (in model E: 14.9 false positives/life-year gained vs. 20.4 when assuming 100% attendance). However, adding annual screening to biennial screening starting at age 40 still had a less favorable harm-benefit ratio than adding biennial screening.
Changing screening test characteristics of digital mammography had only a very small influence on the benefit-harm ratios. The ‘best-case scenario’ resulted in somewhat more benefits and less harms, but the differences in harm-benefit ratios were very small (less than 8%).
Use of the more comprehensive measure of harm (QALYs lost) led to somewhat higher predicted median threshold RRs, e.g. of 2.1 (range 1.4–2.9) for biennial screening with digital mammography (vs. 1.9 when using false-positives).
For women with approximately 2-fold increased risks for breast cancer, the balance of benefits and harms (life-year gained vs. false positives) of starting biennial screening at age 40 approximates that of biennial screening for average-risk women starting at age 50. The models consistently showed that the additional benefits of adding annual screening are small and that there is greater harm relative to benefit from digital than from film mammography in women age 40–49. To obtain similar harm-benefit ratios as the one that results from the currently recommended screening, the false-positive rates for biennial screening with digital mammography would have to decrease substantially among women in their 40s.
The model results on the difference between annual and biennial screening are largely in line with previous work. A retrospective study found that women screened annually vs. biennially had similar distributions of prognostic factors (e.g., tumor size, lymph node status, and histological grade) (26). However, another study found that among women age 40–49 years specifically, those undergoing biennial screening were more likely to have late-stage disease at diagnosis than those undergoing annual screening (27). It has been suggested that for younger women annual screening would be more beneficial than biennial screening because of the faster tumor growth rates in this age group (28–30). This is reflected in the model outcomes showing that adding annual screening to biennial screening in the 40–49 age group is somewhat more beneficial than in older age groups. For example, a previous study showed that 72–89% of the mortality benefit is maintained in these four models when women aged 50–74 move from annual to biennial screening scenarios (23). The present study shows that in the 40–49 age group this percentage of mortality benefit maintained is somewhat lower (66–77%).
All four models found only small differences between film and digital mammography with regard to the benefits of screening, which is in line with a study that found that improvements in sensitivity did not have a marked effect on breast cancer mortality (31). However, digital mammography did result in substantially more false-positive results than film. This translated into greater harm relative to benefits for digital than for film mammography in younger women. Therefore, it seems unlikely that data on the performance of digital mammography, if it had been available, would have led the USPSTF to recommend screening women starting at age 40. A recent study in the Netherlands found that referral and false-positive rates first increased after the implementation of digital mammography, but then these rates started to decrease over time and stabilized at a somewhat higher level than film mammography after a little more than one year (32). However, in a recent US study comparing the screening performance of digital and film mammography, excluding the first year after the transition to digital mammographydid not influence results (33). Another recent study found that the availability of a comparison mammogram halved the false-positive recall probability (34). It remains to be investigated whether false-positive rates in the US can be reduced without also decreasing sensitivityor detection rates.
The results of the models are consistent regarding differences between outcome measures, predicting considerably higher threshold RRs when breast-cancer deaths averted is used instead of life-years gained, because in the 40–49 age group there are more life-years to gain by averting a death than in older age groups. Life-years gained may be considered preferable, because as a summary measure, it incorporates both the number of lives saved and the number of life-years gained per life saved. Our results indicate that the outcome measure used is a main determinant of the screening strategy that will be chosen for women aged 40–49, highlighting the importance of taking into account preferences of individual women about specific benefits and harms.
Several limitations in this study should be taken into account. Importantly, we calculated the harm-benefit ratios for women aged 50–74 screened biennially and used these as threshold values for younger women. However, it might be that younger women have different concerns and preferences than older women and that these preferences vary between individual women.
Additionally, in calculating the harm-benefit ratio we only included false-positive mammography screening examinations as the harm. Ideally, all harms and all benefits are taken into account when determining optimal screening scenarios. Besides false positives, harms of screening mammography include false-positive biopsies, radiation exposure, false reassurance, pain related to the procedure, overdiagnosis (the detection of lesions that would not have become clinically apparent without screening), overtreatment, and the burden of performing medical tests on healthy individuals. Several studies have shown that the risk of radiation is minimal (35–36), and false reassurance has been found to play only a minor role in breast cancer screening (37). Although many women experience pain during the procedure (range 1% to 77%), very few consider this a deterrent from future screening (8, 38). Estimates of overdiagnosis vary widely, ranging up to 54% (39). Although a proportion of invasive cancer diagnosed by mammography may never have presented clinically, the proportion is likely to be small for women aged 40–49, ranging up to 7% (39). For ductal carcinoma in situ this proportion might be larger, but is surrounded by uncertainty. For these reasons we chose to focus on false-positive examinations as the main harm for women in their forties. We did, however, perform a sensitivity analysis in which we considered a more comprehensive measure of harm, QALYs lost. Although this measure is more comprehensive, capturing disutility of false positives and the impact of overdiagnosis, it is less transparent than the number of false-positives, and there is the possibility that the preferences of individual women diverge from the assumed societal utilities.
Also, the models differed for some outcomes. For biennial film screening, three models (E, G, and W) found comparable threshold RRs (1.5–1.6), whereas one model (D) estimated a threshold RR of 3.7. This discrepancy relates to differences in the estimated benefits, reflecting differences in model structures. In model D (19) the stage distribution data are directly incorporated in constructing breast-cancer specific survival. Thus small incremental changes in stage shifts between annual/biennial or between film/digital led to smaller incremental benefits. The other models used a combination of sensitivity values and stage distribution to calibrate parameters and show larger benefits. Additionally, the models differed regarding the predicted incremental benefit of adding annual to biennial screening (range life-years gained 7–21), because the models make different assumptions for unobservable variables, such as sojourn time, which is the duration of the preclinical, screen-detectable phase of the tumor. No randomized controlled trials have directly compared annual and biennial screening. The range in model outcomes, thus, reflects uncertainties in current knowledge of the incremental benefits of screening women aged 40–49 and about shortening the screening interval.
Finally, model outcomes largely depended on the inputs and assumptions. One assumption was that the higher risk influenced only the incidence (onset of disease) and not the screening performance (sensitivity, specificity) or natural history of disease (e.g., the tumor growth rate, breast-cancer survival). However, at least some risk factors, including breast density and family history, have been found to influence both breast-cancer risk and screening performance (40–41). If this is taken into account, the harm-benefit ratio could change for women with risk factors that influence performance. The psychological impact of false positives might also differ by risk group. For example, the amount of anxiety or distress might be higher for younger women and for women with a family history of breast cancer than for average-risk women (42).
Our finding that women with increased risks for breast cancer have similar harm-benefit ratios from starting biennial screening mammography at age 40 is in line with studies finding that breast cancer risk or detection for women with a first-degree relative is similar to that for women a decade older without such a history (40). Several other countries have risk-based screening guidelines. For example, guidelines in the Netherlands state that women with a moderately increased risk, defined as RR 2–3, should be offered annual screening starting at age 40. Similarly, in Australia, guidelines specify that a starting age below 50 or more frequent examinations should be considered individually for women with moderately increased risk, defined as RR 1.5–3.
A systematic review and meta-analysis of risk factors and their prevalence rates in women age 40–49 years in the U.S. was conducted jointly with the present study (43). Two risk factors were associated with a 2-fold or higher RR: having a first-degree relative with breast cancer (9% of women in the U.S.) and extremely dense breasts on mammography (13% of women with Breast Imaging Reporting and Data System category 4 breast density). Results of these two studies imply that women with these characteristics could benefit from biennial screening starting at age 40, and that for them the balance of benefits and harms of screening would be similar to the balance of benefits and harms for average-risk women starting screening at age 50. In addition to these single risk factors, combinations of risk factors could potentially reach this risk threshold (16–17). A potential difficulty with including breast density in screening recommendations is that breast density is not uniformly reported and requires baseline mammography examinations to determine breast density, introducing additional potential screening harms.
Our results provide important information towards more individualized, risk-based screening, suggesting that starting biennial screening at age 40 for women with an increased risk for breast cancer (RR ≥1.9) has a similar balance of benefits and harms as biennial screening for average-risk women aged 50–74 years. For women below this level of risk, the harm-benefit ratio of starting screening at age 40 is less favorable than that of biennial screening between ages 50–74. Reducing the false-positive rate is crucial to improving the balance of benefits and harms for screening regimens for women of all ages.
This research was supported by a National Cancer Institute Activities to Promote Research Collaboration supplement (U01CA086076-10S1), and, in part, by National Cancer Institute Grant U01CA88283, U01CA152958, and KO5CA96940 (JM). Data collection for this work was supported by the Breast Cancer Surveillance Consortium funded by the National Cancer Institute Cooperative agreement U01CA63740, U01CA86076, U01CA86082, U01CA63736, U01CA70013, U01CA69976, U01CA63731, U01CA70040.
The authors would like to thank Drs. Kathleen Cronin and Brian Sprague for their valuable advice and consultation on this project. We thank the BCSC investigators, participating women, mammography facilities, and radiologists for the data they have provided for this study. A list of the BCSC investigators and procedures for requesting BCSC data for research purposes are provided at: http://breastscreening.cancer.gov/. The collection of breast cancer data used in this study was supported in part by several state public health departments and cancer registries throughout the US. For a full description of these sources, please see: http://www.breastscreening.cancer.gov/work/acknowledgement.html. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health.
This is the prepublication, author-produced version of a manuscript accepted for publication in Annals of Internal Medicine. This version does not include post-acceptance editing and formatting. The American College of Physicians, the publisher of Annals of Internal Medicine, is not responsible for the content or presentation of the author-produced accepted version of the manuscript or any version that a third party derives from it. Readers who wish to access the definitive published version of this manuscript and any ancillary material related to this manuscript (e.g., correspondence, corrections, editorials, linked articles) should go to www.annals.org or to the print issue in which the article appears. Those who cite this manuscript should cite the published version, as it is the official version of record