Clinical Review & Education
Special Communication
Quantifying the Benefits and Harms of Screening Mammography H. Gilbert Welch, MD, MPH; Honor J. Passow, PhD Related article page 417
Like all early detection strategies, screening mammography involves trade-offs. If women are to truly participate in the decision of whether or not to be screened, they need some quantification of its benefits and harms. Providing such information is a challenging task, however, given the uncertainty—and underlying professional disagreement—about the data. In this article, we attempt to bound this uncertainty by providing a range of estimates—optimistic and pessimistic—on the absolute frequency of 3 outcomes important to the mammography decision: breast cancer deaths avoided, false alarms, and overdiagnosis. Among 1000 US women aged 50 years who are screened annually for a decade, 0.3 to 3.2 will avoid a breast cancer death, 490 to 670 will have at least 1 false alarm, and 3 to 14 will be overdiagnosed and treated needlessly. We hope that these ranges help women to make a decision: either to feel comfortable about their decision to pursue screening or to feel equally comfortable about their decision not to pursue screening. For the remainder, we hope it helps start a conversation about where additional precision is most needed. JAMA Intern Med. 2014;174(3):448-453. doi:10.1001/jamainternmed.2013.13635 Published online December 30, 2013.
C
ancer screening involves trade-offs. Screening offers the potential benefit of avoiding advanced cancer and subsequent cancer death. It also produces the harms of false alarms, overdiagnosis, and unnecessary treatment. Because different individuals value these benefits and harms differently, there is no single calculation to answer the question of what to do. Instead, each of us needs information about both the benefits and harms to arrive at our own decision. Simply knowing that there are benefits and harms to screening is not sufficient to make the decision; information about their relative magnitude is essential. If 100 people benefit by avoiding a cancer death at the expense of the harms of 50 false alarms and 10 overdiagnoses with the ensuing unnecessary treatments, then the decision is easy. However, if for the same harms, the benefit is only 1 person avoiding a cancer death, the decision may be considerably more difficult. In this article we quantify the benefit-harm trade-off for screening mammography.
Methodological Approach: General Principles Our goal was to produce simple tables that convey the relative magnitude of the benefits and harms of screening mammography. We focus on quantifying 3 outcomes related to screening mammography for women aged 40, 50, and 60 years: reduction in breast cancer death, false-positive results (including subsequent biopsies), and overdiagnosis. We do not estimate the benefit of avoiding metastatic disease or the harm of unnecessary treatment. Fortunately, these 2 outcomes closely mirror those we do quantify: more than 80% of women diagnosed as having metastatic disease die from 448
Supplemental content at jamainternalmedicine.com
Author Affiliations: The Dartmouth Institute for Health Policy and Clinical Practice, The Geisel School of Medicine at Dartmouth, Hanover, New Hampshire. Corresponding Author: H. Gilbert Welch, MD, MPH, The Dartmouth Institute for Health Policy and Clinical Practice, 35 Centerra Pkwy, Lebanon, NH 03766 (H.Gilbert.Welch @dartmouth.edu).
breast cancer, and more than 90% of women diagnosed as having breast cancer are treated.1 We quantify these outcomes in terms of 1000 women screened annually for 10 years, which is long enough for benefit to accrue and short enough to be contemplated by an individual. At the outset, we acknowledge there is no single “right” number to describe the magnitude of either the benefits or harms of screening mammography. There are many sources of variability, such as statistical uncertainty as well as heterogeneity of the populations studied, the mammography intervention itself (eg, frequency of screening and the radiologist’s diagnostic threshold), and the methods and assumptions investigators use to assess the effects of screening. While it is tempting to provide a “best” estimate, we believe that doing so would convey a false sense of certainty and thus be misleading. Our approach is to convey uncertainty using a range, involving both a lower and upper bound. We sought published data to identify the extreme values for each of the 3 outcomes. We strive to make our method simple and transparent (for our 4 Tables and Figure presented herein, interactive spreadsheet versions [Table Calculators] are available in Supplement 1) so that readers who believe they know the single right number (eg, for mortality reduction or rate of overdiagnosis) can quickly recalculate the benefit-harm trade-off. Because our interest is to convey the order of magnitude, not a precise estimate, numbers have been rounded (down for lower bound and up for upper bound). Finally, our focus is on annual screening mammography, since this remains the most common practice in the United States.2,3 Because the risk of false-positive results is considerably lower in other countries4 and with biennial screening,5,6 these data should not be generalized elsewhere.
JAMA Internal Medicine March 2014 Volume 174, Number 3
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a Ndsu Library Periodicals User on 06/02/2015
jamainternalmedicine.com
Quantifying the Benefits and Harms of Mammography
Special Communication Clinical Review & Education
Table 1. Upper- and Lower-Bound Estimates for the Number of Breast Cancer Deaths Avoided Because of a 10-Year Course of Annual Screening Mammogramsa Lower Bound (5% Reduction) Data and Estimates
Notation and Calculation
SEER 15-y risk of dying from breast cancer per 1000 (2007-2009)1
Upper Bound (36% Reduction)
Age 40 y
Age 50 y
Age 60 y
Age 40 y
Age 50 y
Age 60 y
a
3.27
6.45
9.87
3.27
6.45
9.87
Relative mortality reduction attributable to screening, %
b
5
5
5
36
36
36
Proportion screened (NHIS 2008), %
c
61
73
75
61
73
75
15-y Risk without screening per 1000
d = a/[c × (1 − b) + (1 − c)]
3.37
6.70
10.25
4.19
8.76
13.52
15-y Risk with screening per 1000
e = (1 − b) × d
3.20
6.36
9.74
2.68
5.61
8.66
Calculated
f=d−e
0.17
0.33
0.51
1.51
3.16
4.87
Rounded
f (Rounded down for lower bound and up for upper bound)
0.1
0.3
0.5
1.6
3.2
4.9
10-y Absolute mortality reduction per 1000
Abbreviations: NHIS, National Health Interview Survey; SEER, Surveillance Epidemiology and End Results. a
An interactive spreadsheet version is available in Supplement 1.
How Many Breast Cancer Deaths Are Avoided by Screening Mammography? Upper-Bound Estimate Of the 9 randomized trials of screening mammography, the Swedish Two-County Trial is typically viewed as the most optimistic.7,8 Our high estimate for the number of breast cancer deaths avoided thus comes from the recent 30-year follow-up of this trial, in which 45 geographic clusters of women were randomized to screening vs control.9 The most favorable relative risk reduction in breast cancer mortality over the period was 31%, with a screening adherence rate of 85%. Therefore the estimated breast cancer mortality reduction for those attending screening was 36% (≈0.31/0.85).10
Lower-Bound Estimate At the other extreme, the Canadian trials show no effect on mortality,11,12 suggesting that the lower bound should be no mortality reduction. Whether any of the randomized trials are still relevant, however, is in question, given substantial improvements in breast cancer treatment.13 Epidemiologic evidence demonstrates that the timing of introduction of screening in geographically similar regions has little bearing on ongoing trends of declining breast cancer mortality.14,15 These epidemiologic data may better reflect the effectiveness of current treatment and suggest that screening itself has little or no effect on breast cancer mortality. While we believe that a lower-bound estimate of no mortality reduction could be justified, we find it implausible that no woman is helped by screening. Thus we opt for a more optimistic—and admittedly arbitrary—lower bound of a 5% mortality reduction. Readers who believe no effect is the more correct estimate can easily substitute 0 as their lower bound. Our aim is to translate these relative risk reductions into absolute risk reductions for a 10-year course of screening mammography. One might expect, however, that a 10-year course of screening might provide some benefit beyond 10 years if a cancer detected early in year 9 does not show up as an averted death until year 12. We make jamainternalmedicine.com
the assumption that a 10-year course of mammography results in mortality reduction extending 15 years. This assumption favors screening because it assumes that the benefit is not delayed at the front end; instead, the reduction in death appears with the first mammogram. To translate a relative risk reduction into absolute risk reduction requires knowing the risk of breast cancer death in women not exposed to screening. The 15-year risk of breast cancer death in the United States is obtained from the Surveillance Epidemiology and End Results (SEER) data: in 2007-2009, for women aged 50 years it is 6.45 per 1000.1 However, this rate includes both women who are screened and those who are not screened for breast cancer. The risk of breast cancer death among women screened and not screened can be calculated, given that the overall risk is the average of the two, weighted by the size of the 2 groups: Overall Risk of Breast Cancer Death = (% Screened × Risk of Death Among Those Screened) + (% Not Screened × Risk of Death Among Those Not Screened).
Table 1 illustrates these calculations. It begins with the absolute risk of dying in the next 15 years for women aged 40, 50, and 60 years obtained from SEER (2007-2009)1 and either the upper or lower bound of the relative mortality reduction. We used the 2008 National Health Interview Survey (NHIS) to determine the proportion of women in each age group who regularly undergo screening.16 We then calculated the 15-year risk of death both with and without screening. The absolute risk reduction is simply the difference in these rates. For women aged 50 years who undergo annual screening mammography for 10 years, the upper bound on mortality reduction is 3.2 per 1000 women and the lower bound is 0.3 per 1000.
How Many False Alarms Are Caused by Screening Mammography? We sought data on false-positive mammograms (those that require repeated or “recall” mammography) and subsequent biopsies. We conducted a systematic Medline search (May 9, 2013) designed with a medical librarian that identified 761 articles containing JAMA Internal Medicine March 2014 Volume 174, Number 3
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a Ndsu Library Periodicals User on 06/02/2015
449
Quantifying the Benefits and Harms of Mammography
Special Communication Clinical Review & Education
Table 3. Lower-Bound Estimates for the Number of Overdiagnosed Breast Cancers Because of a 10-Year Course of Annual Screening Mammograms Using Data From the Long-term Follow-up of the Malmö Trial25 of Screening Mammographya Lower Bound (Observed in Malmö) Data and Estimates
Notation and Calculation
Age 50 y
Reported starting age, median (range), y
49.5 (45-54)
Age 60 y 62 (55-69)
No. of breast cancers 15 y after trial completion Screened
a
540
780
Control
b
507
698
No. of overdiagnosed cancers
c=a−b
Person-years of screening during trial
d
33
82
109 216
127 742
Overdiagnosis incidence rate
e = c/d
0.000302
0.000642
Calculated
f = [1 − exp(−10 × e)] × 1000
3.02
6.40
Rounded
f (rounded down for lower bound)
3
6
10-y Absolute risk of overdiagnosis per 1000
a
An interactive spreadsheet version is available in Supplement 1.
Table 4. Upper Bound Estimates for the Number of Overdiagnosed Breast Cancers Because of a 10-Year Course of Annual Screening Mammogramsa Upper Bound (One-Third of All Breast Cancers Are Overdiagnosed) Data and Estimates
Notation and Calculation
SEER 10-y risk of developing DCIS or invasive breast cancer per 1000 (2007-2009)
a
Age 40 y
Age 50 y
Age 60 y
19.0
30.2
43.8
10-y Risk without screening per 1000
b = a × 2/3
12.7
20.2
29.2
Proportion screened (NHIS 2008), %
c
61
73
75
10-y Risk with screening per 1000
d = [a − b × (1-c)]/c
23.0
33.9
48.6
Calculated
e=d−b
10.4
13.7
19.4
Rounded
e (Rounded up for upper bound)
11
14
20
10-y Absolute risk of overdiagnosis per 1000
Abbreviations: DCIS, ductal carcinoma in situ; NHIS, National Health Interview Survey; SEER, Surveillance Epidemiology and End Results. a
An interactive spreadsheet version is available in Supplement 1.
to calculate the incidence rates for each group. We accumulated each rate (assuming a constant rate) over 10 years to produce our low estimate of the overall risk of overdiagnosis for women starting at age 50 years and age 60 years, respectively. Because there are no comparable data for women starting at age 40 years, their risk of overdiagnosis cannot be estimated.
Upper-Bound Estimate Our upper bound estimate uses recent epidemiologic observations suggesting that approximately a third of all breast cancers diagnosed (or one-half of all screen-detected cancers) represent overdiagnosis.27,28 We start with the observed 10-year risk of being diagnosed as having breast cancer (invasive and ductal carcinoma in situ) in American women aged 40, 50, and 60 years obtained from DevCan, which was developed by the National Cancer Institute to compute the risk of developing or dying from cancer at a specified age using rates from the standard SEER areas.29 This risk reflects the combined effect of 2 groups: those screened and those not screened. Table 4 illustrates our calculations for the upper-bound estimate. If one-third of all breast cancers represent overdiagnosis (and overdiagnosis is solely a consequence of screening), then the 10year risk of being diagnosed as having breast cancer in the absence of screening is simply two-thirds of the currently observed 10-year jamainternalmedicine.com
risk. To estimate the analogous risk of diagnosis for a completely screened population requires the same weighted-average approach used in Table 1 to parcel out the overall risk of death into 2 distinct groups: those screened and those not screened. In this case we know the overall risk of diagnosis (the combination of screened and not screened) and have an estimate of the risk for those not screened. Given the data on current screening penetration (the same data used in the benefit calculations), we calculate the analogous risk of diagnosis in a screened population. The difference is the 10year risk of overdiagnosis, which, for a woman starting annual screening at age 50 years, is 14 per 1000. The Figure summarizes our findings on the benefit-harm tradeoff for annual screening mammography. Reducing the frequency of screening mammography to every 2 years has been demonstrated to reduce the harm of false positives by about half5,6 and would be expected to reduce the harm of overdiagnosis.
Perspective and Limitations Our analysis is limited by the focus on 3 outcomes. There may be other benefits (eg, less invasive therapies for tumors destined to progress, a general sense of reassurance provided by a normal mammogram) and other harms (eg, more radiation exposure, general JAMA Internal Medicine March 2014 Volume 174, Number 3
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a Ndsu Library Periodicals User on 06/02/2015
451
Quantifying the Benefits and Harms of Mammography
Special Communication Clinical Review & Education
Table 3. Lower-Bound Estimates for the Number of Overdiagnosed Breast Cancers Because of a 10-Year Course of Annual Screening Mammograms Using Data From the Long-term Follow-up of the Malmö Trial25 of Screening Mammographya Lower Bound (Observed in Malmö) Data and Estimates
Notation and Calculation
Age 50 y
Reported starting age, median (range), y
49.5 (45-54)
Age 60 y 62 (55-69)
No. of breast cancers 15 y after trial completion Screened
a
540
780
Control
b
507
698
No. of overdiagnosed cancers
c=a−b
Person-years of screening during trial
d
33
82
109 216
127 742
Overdiagnosis incidence rate
e = c/d
0.000302
0.000642
Calculated
f = [1 − exp(−10 × e)] × 1000
3.02
6.40
Rounded
f (rounded down for lower bound)
3
6
10-y Absolute risk of overdiagnosis per 1000
a
An interactive spreadsheet version is available in Supplement 1.
Table 4. Upper Bound Estimates for the Number of Overdiagnosed Breast Cancers Because of a 10-Year Course of Annual Screening Mammogramsa Upper Bound (One-Third of All Breast Cancers Are Overdiagnosed) Data and Estimates
Notation and Calculation
SEER 10-y risk of developing DCIS or invasive breast cancer per 1000 (2007-2009)
a
Age 40 y
Age 50 y
Age 60 y
19.0
30.2
43.8
10-y Risk without screening per 1000
b = a × 2/3
12.7
20.2
29.2
Proportion screened (NHIS 2008), %
c
61
73
75
10-y Risk with screening per 1000
d = [a − b × (1-c)]/c
23.0
33.9
48.6
Calculated
e=d−b
10.4
13.7
19.4
Rounded
e (Rounded up for upper bound)
11
14
20
10-y Absolute risk of overdiagnosis per 1000
Abbreviations: DCIS, ductal carcinoma in situ; NHIS, National Health Interview Survey; SEER, Surveillance Epidemiology and End Results. a
An interactive spreadsheet version is available in Supplement 1.
to calculate the incidence rates for each group. We accumulated each rate (assuming a constant rate) over 10 years to produce our low estimate of the overall risk of overdiagnosis for women starting at age 50 years and age 60 years, respectively. Because there are no comparable data for women starting at age 40 years, their risk of overdiagnosis cannot be estimated.
Upper-Bound Estimate Our upper bound estimate uses recent epidemiologic observations suggesting that approximately a third of all breast cancers diagnosed (or one-half of all screen-detected cancers) represent overdiagnosis.27,28 We start with the observed 10-year risk of being diagnosed as having breast cancer (invasive and ductal carcinoma in situ) in American women aged 40, 50, and 60 years obtained from DevCan, which was developed by the National Cancer Institute to compute the risk of developing or dying from cancer at a specified age using rates from the standard SEER areas.29 This risk reflects the combined effect of 2 groups: those screened and those not screened. Table 4 illustrates our calculations for the upper-bound estimate. If one-third of all breast cancers represent overdiagnosis (and overdiagnosis is solely a consequence of screening), then the 10year risk of being diagnosed as having breast cancer in the absence of screening is simply two-thirds of the currently observed 10-year jamainternalmedicine.com
risk. To estimate the analogous risk of diagnosis for a completely screened population requires the same weighted-average approach used in Table 1 to parcel out the overall risk of death into 2 distinct groups: those screened and those not screened. In this case we know the overall risk of diagnosis (the combination of screened and not screened) and have an estimate of the risk for those not screened. Given the data on current screening penetration (the same data used in the benefit calculations), we calculate the analogous risk of diagnosis in a screened population. The difference is the 10year risk of overdiagnosis, which, for a woman starting annual screening at age 50 years, is 14 per 1000. The Figure summarizes our findings on the benefit-harm tradeoff for annual screening mammography. Reducing the frequency of screening mammography to every 2 years has been demonstrated to reduce the harm of false positives by about half5,6 and would be expected to reduce the harm of overdiagnosis.
Perspective and Limitations Our analysis is limited by the focus on 3 outcomes. There may be other benefits (eg, less invasive therapies for tumors destined to progress, a general sense of reassurance provided by a normal mammogram) and other harms (eg, more radiation exposure, general JAMA Internal Medicine March 2014 Volume 174, Number 3
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a Ndsu Library Periodicals User on 06/02/2015
451
Clinical Review & Education Special Communication
Quantifying the Benefits and Harms of Mammography
Figure. Benefit-Harm Trade-off for a 10-Year Course of Annual Screening Mammography for Women Starting at Age 40, 50, and 60 Years Among 1000 40-year-old women undergoing annual mammography for 10 years: Benefits 0.1-1.6 Woman will avoid dying from breast cancer
Harms 510-690 Women will have at least 1 “false alarm” (60-80 of whom will undergo a biopsy) ?-11 Women will be overdiagnosed and treated needlessly with surgery, radiation, and/or chemotherapy
Among 1000 50-year-old women undergoing annual mammography for 10 years: Benefits
Harms
0.3-3.2 Women will avoid dying from breast cancer
490-670 Women will have at least 1 “false alarm” (70-100 of whom will undergo a biopsy) 3-14 Women will be overdiagnosed and treated needlessly with surgery, radiation, and/or chemotherapy
Among 1000 60-year-old women undergoing annual mammography for 10 years: Benefits 0.5-4.9 Women will avoid dying from breast cancer
Harms 390-540 Women will have at least 1 “false alarm” (50-70 of whom will undergo a biopsy) 6-20 Women will be overdiagnosed and treated needlessly with surgery, radiation, and/or chemotherapy
anxiety about breast cancer) associated with screening mammography. Unfortunately, there is even more uncertainty involved in quantifying these outcomes than the 3 considered herein. Our analysis also does not consider the effect of screening on overall mortality because no study has been able to demonstrate one. Some may be surprised that we did not make use of published data from meta-analyses, decision models, or the recent independent United Kingdom (UK) panel review. The meta-analyses focus on the benefit of screening and are based on randomized trials that largely predate substantial advances in breast cancer treatment. The Early Breast Cancer Trialists’ Collaborative Group estimated that the combination of adjuvant chemotherapy and hormonal therapy has cut the death rate in half for women with estrogen receptor– positive tumors.13 Because the better we are at treating clinically evident disease, the less benefit there is to screening (eg, there is no point to screen for pneumonia, since we can treat it), any estimate of benefit from meta-analyses would be a upper-bound estimate, which we obtained from the most favorable trial (Swedish Two-County7,8). We avoided using estimates from decision models because they are based on explicit and implicit assumptions that are often not accessible to peer review—it is very hard to get inside of a model’s “black box.” While we could have used the extremes from the models, it seemed more prudent to use the extremes actually observed. The independent UK panel also acknowledged the lack of precision of their estimates: “Given the uncertainties around the estimates, the figures quoted give a spurious impression of accuracy.”30(p2207) 452
Reducing the frequency from annual to every 2 years has been demonstrated to substantially reduce the harm of false alarms and would be expected to reduce the harm of overdiagnosis. The Figure is available as “Table 5” in Supplement 1.
Others may be surprised that our estimates show a relatively small amount of benefit. However, this is typical of populationbased screening. The potential to avert breast cancer death is bounded by the probability of breast cancer death, which is always below 1% for the 3 starting ages, even when extending the time horizon for benefit to 15 years. And even under the best of conditions, screening can only remove a fraction of that risk. Similarly, the potential for overdiagnosis is bounded by the probability of breast cancer diagnosis, which is always below 5% for the 3 starting ages. Given that the upper bound for diagnosis is higher than that for death, overdiagnosis is more common than avoiding breast cancer death. The ranges presented herein, however, cannot clarify how much more common—a little more common or 1 (or 2) orders of magnitude more common. False alarms, on the other hand, are very common—much more common than either the mortality benefit or the harm of overdiagnosis. The consequences of false-positive results may be much smaller, but the tendency to minimize their effects as transient has been called into question by recent research documenting that anxiety may persist for at least 3 years and produce psychological morbidity roughly in-between the experience of healthy women and women with breast cancer. 31 Again, we emphasize that false-positive results are less common in other countries and the data presented herein are generalizable only to the United States, where false-positive results remain a major problem. Reducing the harms of both overdiagnosis and false alarms was the primary motivation behind efforts by the US Pre-
JAMA Internal Medicine March 2014 Volume 174, Number 3
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a Ndsu Library Periodicals User on 06/02/2015
jamainternalmedicine.com
Quantifying the Benefits and Harms of Mammography
Special Communication Clinical Review & Education
ventive Services Task Force (and others) to lengthen the screening interval from annual to biennial.32
Conclusions We hope that these data are sufficient for some women to make the decision about whether or not to be screened. Some may choose to pursue screening, valuing any potential for benefit as warranting the accompanying harms. Others may choose not to pursue screening, valuing the plausible range for the magnitude of the harms as being too great to justify pursuing the relatively small benefit. ARTICLE INFORMATION Accepted for Publication: October 28, 2013. Published Online: December 30, 2013. doi:10.1001/jamainternmed.2013.13635. Author Contributions: Drs Welch and Passow contributed equally to this manuscript and had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Welch and Passow. Acquisition of data: Welch and Passow. Analysis and interpretation of data: Welch and Passow. Drafting of the manuscript: Welch and Passow. Critical revision of the manuscript for important intellectual content: Welch and Passow. Statistical analysis: Passow. Conflict of Interest Disclosures: None reported. REFERENCES 1. Surveillance, Epidemiology, and End Results Program. SEER*Stat software (seer.cancer.gov/seerstat). Version 8.0.4. April 15, 2013. Bethesda, Maryland: National Cancer Institute. Accessed April 20, 2013. 2. NCI-funded Breast Cancer Surveillance Consortium. Data & Statistics: “Months since previous mammogram.” 2012. http: //breastscreening.cancer.gov/data/mammography _data.html. Accessed October 1, 2013. 3. Welch HG, Hayes KJ, Frost C. Repeat testing among Medicare beneficiaries. Arch Intern Med. 2012;172(22):1745-1751. 4. Elmore JG, Nakano CY, Koepsell TD, Desnick LM, D’Orsi CJ, Ransohoff DF. International variation in screening mammography interpretations in community-based programs. J Natl Cancer Inst. 2003;95(18):1384-1393. 5. Hubbard RA, Kerlikowske K, Flowers CI, Yankaskas BC, Zhu W, Miglioretti DL. Cumulative probability of false-positive recall or biopsy recommendation after 10 years of screening mammography: a cohort study. Ann Intern Med. 2011;155(8):481-492. 6. Braithwaite D, Zhu W, Hubbard RA, et al; Breast Cancer Surveillance Consortium. Screening outcomes in older US women undergoing multiple mammograms in community practice: does interval, age, or comorbidity score affect tumor characteristics or false positive rates? J Natl Cancer Inst. 2013;105(5):334-341. 7. Humphyrey LL, Helfand M, Chan BKS, Woolf SH. Breast Cancer Screening: A Summary of the Evidence. Rockville, MD: Agency for Healthcare
jamainternalmedicine.com
We recognize, however, that for many women these ranges may not be sufficiently precise to make an informed choice. And if these data prove to be insufficient for most women, the next logical question is the following: What level of precision is needed for each outcome—breast cancer deaths avoided, false alarms, and overdiagnosis—to allow most women to make a decision? Our suspicion is that the top priority for most women would be to have a more precise estimate of the benefit in the current treatment era. It has been 50 years since a randomized trial of screening mammography has been done in the United States. Given the exposure of tens of millions American women to this intervention, perhaps we are due for a second look.
Research and Quality. 2002:181-210. AHRQ Publication No. 03-507B. 8. Jørgensen KJ, Keen JD, Gøtzsche PC. Is mammographic screening justifiable considering its substantial overdiagnosis rate and minor effect on mortality? Radiology. 2011;260(3):621-627. 9. Tabár L, Vitak B, Chen TH, et al. Swedish two-county trial: impact of mammographic screening on breast cancer mortality during 3 decades. Radiology. 2011;260(3):658-663. 10. Black WC, Welch HG. Screening for disease. AJR Am J Roentgenol. 1997;168(1):3-11. 11. Miller AB, To T, Baines CJ, Wall C. The Canadian National Breast Screening Study-1: breast cancer mortality after 11 to 16 years of follow-up: a randomized screening trial of mammography in women age 40 to 49 years. Ann Intern Med. 2002;137(5, pt 1):305-312. 12. Miller AB, To T, Baines CJ, Wall C. Canadian National Breast Screening Study-2: 13-year results of a randomized trial in women aged 50-59 years. J Natl Cancer Inst. 2000;92(18):1490-1499. 13. Early Breast Cancer Trialists’ Collaborative Group (EBCTCG). Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the randomised trials. Lancet. 2005;365(9472):1687-1717. 14. Kalager M, Zelen M, Langmark F, Adami H-O. Effect of screening mammography on breast-cancer mortality in Norway. N Engl J Med. 2010;363(13):1203-1210. 15. Autier P, Boniol M, Gavin A, Vatten LJ. Breast cancer mortality in neighbouring European countries with different levels of screening but similar access to treatment: trend analysis of WHO mortality database. BMJ. 2011;343:d4411. 16. Centers for Disease Control and Prevention. National Health Interview Survey. http://www.cdc.gov/nchs/nhis/nhis_2008_data _release.htm. Accessed November 18, 2013. 17. Elmore JG, Barton MB, Moceri VM, Polk S, Arena PJ, Fletcher SW. Ten-year risk of false positive screening mammograms and clinical breast examinations. N Engl J Med. 1998;338(16):10891096. 18. Christiansen CL, Wang F, Barton MB, et al. Predicting the cumulative risk of false-positive mammograms. J Natl Cancer Inst. 2000;92(20):1657-1666. 19. Baker SG, Erwin D, Kramer BS. Estimating the cumulative risk of false positive cancer screenings. BMC Med Res Methodol. 2003;3(11):11.
20. Smith-Bindman R, Chu PW, Miglioretti DL, et al. Comparison of screening mammography in the United States and the United Kingdom. JAMA. 2003;290(16):2129-2137. 21. Blanchard K, Colbert JA, Kopans DB, et al. Long-term risk of false-positive screening results and subsequent biopsy as a function of mammography use. Radiology. 2006;240(2): 335-342. 22. Elmore JG, Wells CK, Lee CH, Howard DH, Feinstein AR. Variability in radiologists’ interpretations of mammograms. N Engl J Med. 1994;331(22):1493-1499. 23. Smith-Bindman R, Chu P, Miglioretti DL, et al. Physician predictors of mammographic accuracy. J Natl Cancer Inst. 2005;97(5):358-367. 24. Elmore JG, Jackson SL, Abraham L, et al. Variability in interpretive performance at screening mammography and radiologists’ characteristics associated with accuracy. Radiology. 2009;253(3):641-651. 25. Zackrisson S, Andersson I, Janzon L, Manjer J, Garne JP. Rate of over-diagnosis of breast cancer 15 years after end of Malmö mammographic screening trial: follow-up study. BMJ. 2006;332(7543): 689-692. 26. Welch HG, Black WC. Overdiagnosis in cancer. J Natl Cancer Inst. 2010;102(9):605-613. 27. Bleyer A, Welch HG. Effect of three decades of screening mammography on breast-cancer incidence. N Engl J Med. 2012;367(21):1998-2005. 28. Jørgensen KJ, Gøtzsche PC. Overdiagnosis in publicly organised mammography screening programmes: systematic review of incidence trends. BMJ. 2009;339:b2587. 29. National Cancer Institute. DevCan–Probability of Developing or Dying of Cancer. http://surveillance.cancer.gov/devcan. Accessed March 11, 2012. 30. Marmot MG, Altman DG, Cameron DA, Dewar JA, Thompson SG, Wilcox M. The benefits and harms of breast cancer screening: an independent review. Br J Cancer. 2013;108(11):2205-2240. 31. Brodersen J, Siersma VD. Long-term psychosocial consequences of false-positive screening mammography. Ann Fam Med. 2013;11(2):106-115. 32. US Preventive Services Task Force. Screening for breast cancer: recommendation statement. http://www.uspreventiveservicestaskforce.org /uspstf09/breastcancer/brcanrs.htm. 2009. Accessed November 18, 2013.
JAMA Internal Medicine March 2014 Volume 174, Number 3
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a Ndsu Library Periodicals User on 06/02/2015
453
Special Communication
Quantifying the Benefits and Harms of Screening Mammography H. Gilbert Welch, MD, MPH; Honor J. Passow, PhD Related article page 417
Like all early detection strategies, screening mammography involves trade-offs. If women are to truly participate in the decision of whether or not to be screened, they need some quantification of its benefits and harms. Providing such information is a challenging task, however, given the uncertainty—and underlying professional disagreement—about the data. In this article, we attempt to bound this uncertainty by providing a range of estimates—optimistic and pessimistic—on the absolute frequency of 3 outcomes important to the mammography decision: breast cancer deaths avoided, false alarms, and overdiagnosis. Among 1000 US women aged 50 years who are screened annually for a decade, 0.3 to 3.2 will avoid a breast cancer death, 490 to 670 will have at least 1 false alarm, and 3 to 14 will be overdiagnosed and treated needlessly. We hope that these ranges help women to make a decision: either to feel comfortable about their decision to pursue screening or to feel equally comfortable about their decision not to pursue screening. For the remainder, we hope it helps start a conversation about where additional precision is most needed. JAMA Intern Med. 2014;174(3):448-453. doi:10.1001/jamainternmed.2013.13635 Published online December 30, 2013.
C
ancer screening involves trade-offs. Screening offers the potential benefit of avoiding advanced cancer and subsequent cancer death. It also produces the harms of false alarms, overdiagnosis, and unnecessary treatment. Because different individuals value these benefits and harms differently, there is no single calculation to answer the question of what to do. Instead, each of us needs information about both the benefits and harms to arrive at our own decision. Simply knowing that there are benefits and harms to screening is not sufficient to make the decision; information about their relative magnitude is essential. If 100 people benefit by avoiding a cancer death at the expense of the harms of 50 false alarms and 10 overdiagnoses with the ensuing unnecessary treatments, then the decision is easy. However, if for the same harms, the benefit is only 1 person avoiding a cancer death, the decision may be considerably more difficult. In this article we quantify the benefit-harm trade-off for screening mammography.
Methodological Approach: General Principles Our goal was to produce simple tables that convey the relative magnitude of the benefits and harms of screening mammography. We focus on quantifying 3 outcomes related to screening mammography for women aged 40, 50, and 60 years: reduction in breast cancer death, false-positive results (including subsequent biopsies), and overdiagnosis. We do not estimate the benefit of avoiding metastatic disease or the harm of unnecessary treatment. Fortunately, these 2 outcomes closely mirror those we do quantify: more than 80% of women diagnosed as having metastatic disease die from 448
Supplemental content at jamainternalmedicine.com
Author Affiliations: The Dartmouth Institute for Health Policy and Clinical Practice, The Geisel School of Medicine at Dartmouth, Hanover, New Hampshire. Corresponding Author: H. Gilbert Welch, MD, MPH, The Dartmouth Institute for Health Policy and Clinical Practice, 35 Centerra Pkwy, Lebanon, NH 03766 (H.Gilbert.Welch @dartmouth.edu).
breast cancer, and more than 90% of women diagnosed as having breast cancer are treated.1 We quantify these outcomes in terms of 1000 women screened annually for 10 years, which is long enough for benefit to accrue and short enough to be contemplated by an individual. At the outset, we acknowledge there is no single “right” number to describe the magnitude of either the benefits or harms of screening mammography. There are many sources of variability, such as statistical uncertainty as well as heterogeneity of the populations studied, the mammography intervention itself (eg, frequency of screening and the radiologist’s diagnostic threshold), and the methods and assumptions investigators use to assess the effects of screening. While it is tempting to provide a “best” estimate, we believe that doing so would convey a false sense of certainty and thus be misleading. Our approach is to convey uncertainty using a range, involving both a lower and upper bound. We sought published data to identify the extreme values for each of the 3 outcomes. We strive to make our method simple and transparent (for our 4 Tables and Figure presented herein, interactive spreadsheet versions [Table Calculators] are available in Supplement 1) so that readers who believe they know the single right number (eg, for mortality reduction or rate of overdiagnosis) can quickly recalculate the benefit-harm trade-off. Because our interest is to convey the order of magnitude, not a precise estimate, numbers have been rounded (down for lower bound and up for upper bound). Finally, our focus is on annual screening mammography, since this remains the most common practice in the United States.2,3 Because the risk of false-positive results is considerably lower in other countries4 and with biennial screening,5,6 these data should not be generalized elsewhere.
JAMA Internal Medicine March 2014 Volume 174, Number 3
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a Ndsu Library Periodicals User on 06/02/2015
jamainternalmedicine.com
Quantifying the Benefits and Harms of Mammography
Special Communication Clinical Review & Education
Table 1. Upper- and Lower-Bound Estimates for the Number of Breast Cancer Deaths Avoided Because of a 10-Year Course of Annual Screening Mammogramsa Lower Bound (5% Reduction) Data and Estimates
Notation and Calculation
SEER 15-y risk of dying from breast cancer per 1000 (2007-2009)1
Upper Bound (36% Reduction)
Age 40 y
Age 50 y
Age 60 y
Age 40 y
Age 50 y
Age 60 y
a
3.27
6.45
9.87
3.27
6.45
9.87
Relative mortality reduction attributable to screening, %
b
5
5
5
36
36
36
Proportion screened (NHIS 2008), %
c
61
73
75
61
73
75
15-y Risk without screening per 1000
d = a/[c × (1 − b) + (1 − c)]
3.37
6.70
10.25
4.19
8.76
13.52
15-y Risk with screening per 1000
e = (1 − b) × d
3.20
6.36
9.74
2.68
5.61
8.66
Calculated
f=d−e
0.17
0.33
0.51
1.51
3.16
4.87
Rounded
f (Rounded down for lower bound and up for upper bound)
0.1
0.3
0.5
1.6
3.2
4.9
10-y Absolute mortality reduction per 1000
Abbreviations: NHIS, National Health Interview Survey; SEER, Surveillance Epidemiology and End Results. a
An interactive spreadsheet version is available in Supplement 1.
How Many Breast Cancer Deaths Are Avoided by Screening Mammography? Upper-Bound Estimate Of the 9 randomized trials of screening mammography, the Swedish Two-County Trial is typically viewed as the most optimistic.7,8 Our high estimate for the number of breast cancer deaths avoided thus comes from the recent 30-year follow-up of this trial, in which 45 geographic clusters of women were randomized to screening vs control.9 The most favorable relative risk reduction in breast cancer mortality over the period was 31%, with a screening adherence rate of 85%. Therefore the estimated breast cancer mortality reduction for those attending screening was 36% (≈0.31/0.85).10
Lower-Bound Estimate At the other extreme, the Canadian trials show no effect on mortality,11,12 suggesting that the lower bound should be no mortality reduction. Whether any of the randomized trials are still relevant, however, is in question, given substantial improvements in breast cancer treatment.13 Epidemiologic evidence demonstrates that the timing of introduction of screening in geographically similar regions has little bearing on ongoing trends of declining breast cancer mortality.14,15 These epidemiologic data may better reflect the effectiveness of current treatment and suggest that screening itself has little or no effect on breast cancer mortality. While we believe that a lower-bound estimate of no mortality reduction could be justified, we find it implausible that no woman is helped by screening. Thus we opt for a more optimistic—and admittedly arbitrary—lower bound of a 5% mortality reduction. Readers who believe no effect is the more correct estimate can easily substitute 0 as their lower bound. Our aim is to translate these relative risk reductions into absolute risk reductions for a 10-year course of screening mammography. One might expect, however, that a 10-year course of screening might provide some benefit beyond 10 years if a cancer detected early in year 9 does not show up as an averted death until year 12. We make jamainternalmedicine.com
the assumption that a 10-year course of mammography results in mortality reduction extending 15 years. This assumption favors screening because it assumes that the benefit is not delayed at the front end; instead, the reduction in death appears with the first mammogram. To translate a relative risk reduction into absolute risk reduction requires knowing the risk of breast cancer death in women not exposed to screening. The 15-year risk of breast cancer death in the United States is obtained from the Surveillance Epidemiology and End Results (SEER) data: in 2007-2009, for women aged 50 years it is 6.45 per 1000.1 However, this rate includes both women who are screened and those who are not screened for breast cancer. The risk of breast cancer death among women screened and not screened can be calculated, given that the overall risk is the average of the two, weighted by the size of the 2 groups: Overall Risk of Breast Cancer Death = (% Screened × Risk of Death Among Those Screened) + (% Not Screened × Risk of Death Among Those Not Screened).
Table 1 illustrates these calculations. It begins with the absolute risk of dying in the next 15 years for women aged 40, 50, and 60 years obtained from SEER (2007-2009)1 and either the upper or lower bound of the relative mortality reduction. We used the 2008 National Health Interview Survey (NHIS) to determine the proportion of women in each age group who regularly undergo screening.16 We then calculated the 15-year risk of death both with and without screening. The absolute risk reduction is simply the difference in these rates. For women aged 50 years who undergo annual screening mammography for 10 years, the upper bound on mortality reduction is 3.2 per 1000 women and the lower bound is 0.3 per 1000.
How Many False Alarms Are Caused by Screening Mammography? We sought data on false-positive mammograms (those that require repeated or “recall” mammography) and subsequent biopsies. We conducted a systematic Medline search (May 9, 2013) designed with a medical librarian that identified 761 articles containing JAMA Internal Medicine March 2014 Volume 174, Number 3
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a Ndsu Library Periodicals User on 06/02/2015
449
Quantifying the Benefits and Harms of Mammography
Special Communication Clinical Review & Education
Table 3. Lower-Bound Estimates for the Number of Overdiagnosed Breast Cancers Because of a 10-Year Course of Annual Screening Mammograms Using Data From the Long-term Follow-up of the Malmö Trial25 of Screening Mammographya Lower Bound (Observed in Malmö) Data and Estimates
Notation and Calculation
Age 50 y
Reported starting age, median (range), y
49.5 (45-54)
Age 60 y 62 (55-69)
No. of breast cancers 15 y after trial completion Screened
a
540
780
Control
b
507
698
No. of overdiagnosed cancers
c=a−b
Person-years of screening during trial
d
33
82
109 216
127 742
Overdiagnosis incidence rate
e = c/d
0.000302
0.000642
Calculated
f = [1 − exp(−10 × e)] × 1000
3.02
6.40
Rounded
f (rounded down for lower bound)
3
6
10-y Absolute risk of overdiagnosis per 1000
a
An interactive spreadsheet version is available in Supplement 1.
Table 4. Upper Bound Estimates for the Number of Overdiagnosed Breast Cancers Because of a 10-Year Course of Annual Screening Mammogramsa Upper Bound (One-Third of All Breast Cancers Are Overdiagnosed) Data and Estimates
Notation and Calculation
SEER 10-y risk of developing DCIS or invasive breast cancer per 1000 (2007-2009)
a
Age 40 y
Age 50 y
Age 60 y
19.0
30.2
43.8
10-y Risk without screening per 1000
b = a × 2/3
12.7
20.2
29.2
Proportion screened (NHIS 2008), %
c
61
73
75
10-y Risk with screening per 1000
d = [a − b × (1-c)]/c
23.0
33.9
48.6
Calculated
e=d−b
10.4
13.7
19.4
Rounded
e (Rounded up for upper bound)
11
14
20
10-y Absolute risk of overdiagnosis per 1000
Abbreviations: DCIS, ductal carcinoma in situ; NHIS, National Health Interview Survey; SEER, Surveillance Epidemiology and End Results. a
An interactive spreadsheet version is available in Supplement 1.
to calculate the incidence rates for each group. We accumulated each rate (assuming a constant rate) over 10 years to produce our low estimate of the overall risk of overdiagnosis for women starting at age 50 years and age 60 years, respectively. Because there are no comparable data for women starting at age 40 years, their risk of overdiagnosis cannot be estimated.
Upper-Bound Estimate Our upper bound estimate uses recent epidemiologic observations suggesting that approximately a third of all breast cancers diagnosed (or one-half of all screen-detected cancers) represent overdiagnosis.27,28 We start with the observed 10-year risk of being diagnosed as having breast cancer (invasive and ductal carcinoma in situ) in American women aged 40, 50, and 60 years obtained from DevCan, which was developed by the National Cancer Institute to compute the risk of developing or dying from cancer at a specified age using rates from the standard SEER areas.29 This risk reflects the combined effect of 2 groups: those screened and those not screened. Table 4 illustrates our calculations for the upper-bound estimate. If one-third of all breast cancers represent overdiagnosis (and overdiagnosis is solely a consequence of screening), then the 10year risk of being diagnosed as having breast cancer in the absence of screening is simply two-thirds of the currently observed 10-year jamainternalmedicine.com
risk. To estimate the analogous risk of diagnosis for a completely screened population requires the same weighted-average approach used in Table 1 to parcel out the overall risk of death into 2 distinct groups: those screened and those not screened. In this case we know the overall risk of diagnosis (the combination of screened and not screened) and have an estimate of the risk for those not screened. Given the data on current screening penetration (the same data used in the benefit calculations), we calculate the analogous risk of diagnosis in a screened population. The difference is the 10year risk of overdiagnosis, which, for a woman starting annual screening at age 50 years, is 14 per 1000. The Figure summarizes our findings on the benefit-harm tradeoff for annual screening mammography. Reducing the frequency of screening mammography to every 2 years has been demonstrated to reduce the harm of false positives by about half5,6 and would be expected to reduce the harm of overdiagnosis.
Perspective and Limitations Our analysis is limited by the focus on 3 outcomes. There may be other benefits (eg, less invasive therapies for tumors destined to progress, a general sense of reassurance provided by a normal mammogram) and other harms (eg, more radiation exposure, general JAMA Internal Medicine March 2014 Volume 174, Number 3
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a Ndsu Library Periodicals User on 06/02/2015
451
Quantifying the Benefits and Harms of Mammography
Special Communication Clinical Review & Education
Table 3. Lower-Bound Estimates for the Number of Overdiagnosed Breast Cancers Because of a 10-Year Course of Annual Screening Mammograms Using Data From the Long-term Follow-up of the Malmö Trial25 of Screening Mammographya Lower Bound (Observed in Malmö) Data and Estimates
Notation and Calculation
Age 50 y
Reported starting age, median (range), y
49.5 (45-54)
Age 60 y 62 (55-69)
No. of breast cancers 15 y after trial completion Screened
a
540
780
Control
b
507
698
No. of overdiagnosed cancers
c=a−b
Person-years of screening during trial
d
33
82
109 216
127 742
Overdiagnosis incidence rate
e = c/d
0.000302
0.000642
Calculated
f = [1 − exp(−10 × e)] × 1000
3.02
6.40
Rounded
f (rounded down for lower bound)
3
6
10-y Absolute risk of overdiagnosis per 1000
a
An interactive spreadsheet version is available in Supplement 1.
Table 4. Upper Bound Estimates for the Number of Overdiagnosed Breast Cancers Because of a 10-Year Course of Annual Screening Mammogramsa Upper Bound (One-Third of All Breast Cancers Are Overdiagnosed) Data and Estimates
Notation and Calculation
SEER 10-y risk of developing DCIS or invasive breast cancer per 1000 (2007-2009)
a
Age 40 y
Age 50 y
Age 60 y
19.0
30.2
43.8
10-y Risk without screening per 1000
b = a × 2/3
12.7
20.2
29.2
Proportion screened (NHIS 2008), %
c
61
73
75
10-y Risk with screening per 1000
d = [a − b × (1-c)]/c
23.0
33.9
48.6
Calculated
e=d−b
10.4
13.7
19.4
Rounded
e (Rounded up for upper bound)
11
14
20
10-y Absolute risk of overdiagnosis per 1000
Abbreviations: DCIS, ductal carcinoma in situ; NHIS, National Health Interview Survey; SEER, Surveillance Epidemiology and End Results. a
An interactive spreadsheet version is available in Supplement 1.
to calculate the incidence rates for each group. We accumulated each rate (assuming a constant rate) over 10 years to produce our low estimate of the overall risk of overdiagnosis for women starting at age 50 years and age 60 years, respectively. Because there are no comparable data for women starting at age 40 years, their risk of overdiagnosis cannot be estimated.
Upper-Bound Estimate Our upper bound estimate uses recent epidemiologic observations suggesting that approximately a third of all breast cancers diagnosed (or one-half of all screen-detected cancers) represent overdiagnosis.27,28 We start with the observed 10-year risk of being diagnosed as having breast cancer (invasive and ductal carcinoma in situ) in American women aged 40, 50, and 60 years obtained from DevCan, which was developed by the National Cancer Institute to compute the risk of developing or dying from cancer at a specified age using rates from the standard SEER areas.29 This risk reflects the combined effect of 2 groups: those screened and those not screened. Table 4 illustrates our calculations for the upper-bound estimate. If one-third of all breast cancers represent overdiagnosis (and overdiagnosis is solely a consequence of screening), then the 10year risk of being diagnosed as having breast cancer in the absence of screening is simply two-thirds of the currently observed 10-year jamainternalmedicine.com
risk. To estimate the analogous risk of diagnosis for a completely screened population requires the same weighted-average approach used in Table 1 to parcel out the overall risk of death into 2 distinct groups: those screened and those not screened. In this case we know the overall risk of diagnosis (the combination of screened and not screened) and have an estimate of the risk for those not screened. Given the data on current screening penetration (the same data used in the benefit calculations), we calculate the analogous risk of diagnosis in a screened population. The difference is the 10year risk of overdiagnosis, which, for a woman starting annual screening at age 50 years, is 14 per 1000. The Figure summarizes our findings on the benefit-harm tradeoff for annual screening mammography. Reducing the frequency of screening mammography to every 2 years has been demonstrated to reduce the harm of false positives by about half5,6 and would be expected to reduce the harm of overdiagnosis.
Perspective and Limitations Our analysis is limited by the focus on 3 outcomes. There may be other benefits (eg, less invasive therapies for tumors destined to progress, a general sense of reassurance provided by a normal mammogram) and other harms (eg, more radiation exposure, general JAMA Internal Medicine March 2014 Volume 174, Number 3
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a Ndsu Library Periodicals User on 06/02/2015
451
Clinical Review & Education Special Communication
Quantifying the Benefits and Harms of Mammography
Figure. Benefit-Harm Trade-off for a 10-Year Course of Annual Screening Mammography for Women Starting at Age 40, 50, and 60 Years Among 1000 40-year-old women undergoing annual mammography for 10 years: Benefits 0.1-1.6 Woman will avoid dying from breast cancer
Harms 510-690 Women will have at least 1 “false alarm” (60-80 of whom will undergo a biopsy) ?-11 Women will be overdiagnosed and treated needlessly with surgery, radiation, and/or chemotherapy
Among 1000 50-year-old women undergoing annual mammography for 10 years: Benefits
Harms
0.3-3.2 Women will avoid dying from breast cancer
490-670 Women will have at least 1 “false alarm” (70-100 of whom will undergo a biopsy) 3-14 Women will be overdiagnosed and treated needlessly with surgery, radiation, and/or chemotherapy
Among 1000 60-year-old women undergoing annual mammography for 10 years: Benefits 0.5-4.9 Women will avoid dying from breast cancer
Harms 390-540 Women will have at least 1 “false alarm” (50-70 of whom will undergo a biopsy) 6-20 Women will be overdiagnosed and treated needlessly with surgery, radiation, and/or chemotherapy
anxiety about breast cancer) associated with screening mammography. Unfortunately, there is even more uncertainty involved in quantifying these outcomes than the 3 considered herein. Our analysis also does not consider the effect of screening on overall mortality because no study has been able to demonstrate one. Some may be surprised that we did not make use of published data from meta-analyses, decision models, or the recent independent United Kingdom (UK) panel review. The meta-analyses focus on the benefit of screening and are based on randomized trials that largely predate substantial advances in breast cancer treatment. The Early Breast Cancer Trialists’ Collaborative Group estimated that the combination of adjuvant chemotherapy and hormonal therapy has cut the death rate in half for women with estrogen receptor– positive tumors.13 Because the better we are at treating clinically evident disease, the less benefit there is to screening (eg, there is no point to screen for pneumonia, since we can treat it), any estimate of benefit from meta-analyses would be a upper-bound estimate, which we obtained from the most favorable trial (Swedish Two-County7,8). We avoided using estimates from decision models because they are based on explicit and implicit assumptions that are often not accessible to peer review—it is very hard to get inside of a model’s “black box.” While we could have used the extremes from the models, it seemed more prudent to use the extremes actually observed. The independent UK panel also acknowledged the lack of precision of their estimates: “Given the uncertainties around the estimates, the figures quoted give a spurious impression of accuracy.”30(p2207) 452
Reducing the frequency from annual to every 2 years has been demonstrated to substantially reduce the harm of false alarms and would be expected to reduce the harm of overdiagnosis. The Figure is available as “Table 5” in Supplement 1.
Others may be surprised that our estimates show a relatively small amount of benefit. However, this is typical of populationbased screening. The potential to avert breast cancer death is bounded by the probability of breast cancer death, which is always below 1% for the 3 starting ages, even when extending the time horizon for benefit to 15 years. And even under the best of conditions, screening can only remove a fraction of that risk. Similarly, the potential for overdiagnosis is bounded by the probability of breast cancer diagnosis, which is always below 5% for the 3 starting ages. Given that the upper bound for diagnosis is higher than that for death, overdiagnosis is more common than avoiding breast cancer death. The ranges presented herein, however, cannot clarify how much more common—a little more common or 1 (or 2) orders of magnitude more common. False alarms, on the other hand, are very common—much more common than either the mortality benefit or the harm of overdiagnosis. The consequences of false-positive results may be much smaller, but the tendency to minimize their effects as transient has been called into question by recent research documenting that anxiety may persist for at least 3 years and produce psychological morbidity roughly in-between the experience of healthy women and women with breast cancer. 31 Again, we emphasize that false-positive results are less common in other countries and the data presented herein are generalizable only to the United States, where false-positive results remain a major problem. Reducing the harms of both overdiagnosis and false alarms was the primary motivation behind efforts by the US Pre-
JAMA Internal Medicine March 2014 Volume 174, Number 3
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a Ndsu Library Periodicals User on 06/02/2015
jamainternalmedicine.com
Quantifying the Benefits and Harms of Mammography
Special Communication Clinical Review & Education
ventive Services Task Force (and others) to lengthen the screening interval from annual to biennial.32
Conclusions We hope that these data are sufficient for some women to make the decision about whether or not to be screened. Some may choose to pursue screening, valuing any potential for benefit as warranting the accompanying harms. Others may choose not to pursue screening, valuing the plausible range for the magnitude of the harms as being too great to justify pursuing the relatively small benefit. ARTICLE INFORMATION Accepted for Publication: October 28, 2013. Published Online: December 30, 2013. doi:10.1001/jamainternmed.2013.13635. Author Contributions: Drs Welch and Passow contributed equally to this manuscript and had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Welch and Passow. Acquisition of data: Welch and Passow. Analysis and interpretation of data: Welch and Passow. Drafting of the manuscript: Welch and Passow. Critical revision of the manuscript for important intellectual content: Welch and Passow. Statistical analysis: Passow. Conflict of Interest Disclosures: None reported. REFERENCES 1. Surveillance, Epidemiology, and End Results Program. SEER*Stat software (seer.cancer.gov/seerstat). Version 8.0.4. April 15, 2013. Bethesda, Maryland: National Cancer Institute. Accessed April 20, 2013. 2. NCI-funded Breast Cancer Surveillance Consortium. Data & Statistics: “Months since previous mammogram.” 2012. http: //breastscreening.cancer.gov/data/mammography _data.html. Accessed October 1, 2013. 3. Welch HG, Hayes KJ, Frost C. Repeat testing among Medicare beneficiaries. Arch Intern Med. 2012;172(22):1745-1751. 4. Elmore JG, Nakano CY, Koepsell TD, Desnick LM, D’Orsi CJ, Ransohoff DF. International variation in screening mammography interpretations in community-based programs. J Natl Cancer Inst. 2003;95(18):1384-1393. 5. Hubbard RA, Kerlikowske K, Flowers CI, Yankaskas BC, Zhu W, Miglioretti DL. Cumulative probability of false-positive recall or biopsy recommendation after 10 years of screening mammography: a cohort study. Ann Intern Med. 2011;155(8):481-492. 6. Braithwaite D, Zhu W, Hubbard RA, et al; Breast Cancer Surveillance Consortium. Screening outcomes in older US women undergoing multiple mammograms in community practice: does interval, age, or comorbidity score affect tumor characteristics or false positive rates? J Natl Cancer Inst. 2013;105(5):334-341. 7. Humphyrey LL, Helfand M, Chan BKS, Woolf SH. Breast Cancer Screening: A Summary of the Evidence. Rockville, MD: Agency for Healthcare
jamainternalmedicine.com
We recognize, however, that for many women these ranges may not be sufficiently precise to make an informed choice. And if these data prove to be insufficient for most women, the next logical question is the following: What level of precision is needed for each outcome—breast cancer deaths avoided, false alarms, and overdiagnosis—to allow most women to make a decision? Our suspicion is that the top priority for most women would be to have a more precise estimate of the benefit in the current treatment era. It has been 50 years since a randomized trial of screening mammography has been done in the United States. Given the exposure of tens of millions American women to this intervention, perhaps we are due for a second look.
Research and Quality. 2002:181-210. AHRQ Publication No. 03-507B. 8. Jørgensen KJ, Keen JD, Gøtzsche PC. Is mammographic screening justifiable considering its substantial overdiagnosis rate and minor effect on mortality? Radiology. 2011;260(3):621-627. 9. Tabár L, Vitak B, Chen TH, et al. Swedish two-county trial: impact of mammographic screening on breast cancer mortality during 3 decades. Radiology. 2011;260(3):658-663. 10. Black WC, Welch HG. Screening for disease. AJR Am J Roentgenol. 1997;168(1):3-11. 11. Miller AB, To T, Baines CJ, Wall C. The Canadian National Breast Screening Study-1: breast cancer mortality after 11 to 16 years of follow-up: a randomized screening trial of mammography in women age 40 to 49 years. Ann Intern Med. 2002;137(5, pt 1):305-312. 12. Miller AB, To T, Baines CJ, Wall C. Canadian National Breast Screening Study-2: 13-year results of a randomized trial in women aged 50-59 years. J Natl Cancer Inst. 2000;92(18):1490-1499. 13. Early Breast Cancer Trialists’ Collaborative Group (EBCTCG). Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the randomised trials. Lancet. 2005;365(9472):1687-1717. 14. Kalager M, Zelen M, Langmark F, Adami H-O. Effect of screening mammography on breast-cancer mortality in Norway. N Engl J Med. 2010;363(13):1203-1210. 15. Autier P, Boniol M, Gavin A, Vatten LJ. Breast cancer mortality in neighbouring European countries with different levels of screening but similar access to treatment: trend analysis of WHO mortality database. BMJ. 2011;343:d4411. 16. Centers for Disease Control and Prevention. National Health Interview Survey. http://www.cdc.gov/nchs/nhis/nhis_2008_data _release.htm. Accessed November 18, 2013. 17. Elmore JG, Barton MB, Moceri VM, Polk S, Arena PJ, Fletcher SW. Ten-year risk of false positive screening mammograms and clinical breast examinations. N Engl J Med. 1998;338(16):10891096. 18. Christiansen CL, Wang F, Barton MB, et al. Predicting the cumulative risk of false-positive mammograms. J Natl Cancer Inst. 2000;92(20):1657-1666. 19. Baker SG, Erwin D, Kramer BS. Estimating the cumulative risk of false positive cancer screenings. BMC Med Res Methodol. 2003;3(11):11.
20. Smith-Bindman R, Chu PW, Miglioretti DL, et al. Comparison of screening mammography in the United States and the United Kingdom. JAMA. 2003;290(16):2129-2137. 21. Blanchard K, Colbert JA, Kopans DB, et al. Long-term risk of false-positive screening results and subsequent biopsy as a function of mammography use. Radiology. 2006;240(2): 335-342. 22. Elmore JG, Wells CK, Lee CH, Howard DH, Feinstein AR. Variability in radiologists’ interpretations of mammograms. N Engl J Med. 1994;331(22):1493-1499. 23. Smith-Bindman R, Chu P, Miglioretti DL, et al. Physician predictors of mammographic accuracy. J Natl Cancer Inst. 2005;97(5):358-367. 24. Elmore JG, Jackson SL, Abraham L, et al. Variability in interpretive performance at screening mammography and radiologists’ characteristics associated with accuracy. Radiology. 2009;253(3):641-651. 25. Zackrisson S, Andersson I, Janzon L, Manjer J, Garne JP. Rate of over-diagnosis of breast cancer 15 years after end of Malmö mammographic screening trial: follow-up study. BMJ. 2006;332(7543): 689-692. 26. Welch HG, Black WC. Overdiagnosis in cancer. J Natl Cancer Inst. 2010;102(9):605-613. 27. Bleyer A, Welch HG. Effect of three decades of screening mammography on breast-cancer incidence. N Engl J Med. 2012;367(21):1998-2005. 28. Jørgensen KJ, Gøtzsche PC. Overdiagnosis in publicly organised mammography screening programmes: systematic review of incidence trends. BMJ. 2009;339:b2587. 29. National Cancer Institute. DevCan–Probability of Developing or Dying of Cancer. http://surveillance.cancer.gov/devcan. Accessed March 11, 2012. 30. Marmot MG, Altman DG, Cameron DA, Dewar JA, Thompson SG, Wilcox M. The benefits and harms of breast cancer screening: an independent review. Br J Cancer. 2013;108(11):2205-2240. 31. Brodersen J, Siersma VD. Long-term psychosocial consequences of false-positive screening mammography. Ann Fam Med. 2013;11(2):106-115. 32. US Preventive Services Task Force. Screening for breast cancer: recommendation statement. http://www.uspreventiveservicestaskforce.org /uspstf09/breastcancer/brcanrs.htm. 2009. Accessed November 18, 2013.
JAMA Internal Medicine March 2014 Volume 174, Number 3
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a Ndsu Library Periodicals User on 06/02/2015
453
