USING THE GUIDE
The next step in advising the woman is to consider the balance of benefits and harms that she may experience if she proceeds with mammographic screening.
What outcomes must investigators measure to estimate the benefits of a screening program? If treatment is effective, some of those who test positive will experience a reduction in mortality or an increase in quality of life. One can estimate the benefit as an absolute risk reduction or a relative risk reduction (RRR) in adverse outcomes. The number of people needed to invite to screening (NNI) to prevent an adverse outcome provides another way of presenting benefit (see Chapter 9, Does Treatment Lower Risk? Understanding the Results).
When the benefit is a reduction in mortality, it is better if decisions are based on an understanding of reduction in both disease-specific and total mortality (ie, mortality from any and all possible causes). Because the target condition is typically only one of many causes of death, and particularly when the treatment of the screen-detected disease can have life-threatening complications (eg, aortic aneurysm repair), very large studies may be required to reveal reductions in all-cause and disease-specific mortality. So for the most part, policymakers have had to be satisfied with demonstrated reductions in disease-specific mortality only. At a minimum, investigators should collect data on all deaths in screening trials so that any increase in deaths in the screened group can be detected and explored.
In addition to prevention of adverse outcomes, people may also regard knowledge of the presence of an abnormality as a benefit, as in antenatal screening for Down syndrome.
Another potential benefit of screening is the reassurance afforded by a negative test result if a person is experiencing anxiety because a family member or friend has developed the target condition. However, if the screening program itself has generated the anxiety—for instance, through extensive publicity—then claiming reduction in that anxiety as a benefit of screening is very questionable.18,19
USING THE GUIDE
The 2002 USPSTF systematic review includes a meta-analysis of 7 mammography (plain radio—graphs) trials among women aged 39 to 75 years.1 The authors report an RRR of 16% (95% CI, 9%-23%) after a mean of 14 years of follow-up. This is equivalent to an NNI of 1224 to prevent 1 death due to breast cancer.
There is always uncertainty about the benefits of screening. Evaluation of screening generally requires very large-scale trials to reveal a statistically significant effect because we are dealing with asymptomatic individuals who are at low risk of developing a relatively rare disease.
As noted earlier, most of the screening trials use disease-specific mortality as their primary outcome, so we are often left with uncertainty about the impact on all-cause mortality.20
USING THE GUIDE
There is also uncertainty about the generalizability of this evidence base to the clinical scenario in this chapter. Most of the mammography trials were conducted in the 1960s to 1980s. Since then, mammography techniques have improved,21 and the treatments for clinically detected and screen-detected breast cancer have become much more effective, which may reduce the benefit of screening.22 The incidence of DCIS has increased because of screening,23 and the overall mortality rate from breast cancer has decreased.1,2,24 There are also differences between the trials and international screening programs with respect to target age, screening intervals, mammography views, and follow-up time.
In general, the benefits of screening are not fixed in time and decrease as the effectiveness, safety, and availability of treatment for more advanced disease increases. If there is a safe, effective, affordable treatment for advanced disease that is readily available with few adverse effects, then the mortality benefits of screening reduce to near zero. This occurred in the case of testicular cancer, with few countries now continuing to screen. Benefits of screening also decrease as prevalence of the risk factor or disease decreases. In several cases, screening was originally useful but is no longer worthwhile because of reduced underlying prevalence, such as with tuberculosis screening because of the reduction in tuberculosis infections and in abdominal aortic aneurism screening when smoking rates have been substantially reduced.
False-positive results: The USPSTF review1,2 observed that test accuracy data are conventionally reported for a test at a single point, whereas for a screening program, cumulative test-positive data over time are more relevant. The cumulative risk of a false-positive result after 10 mammography examinations was reported to range from 21% to 49%. More recently, a study using Breast Cancer Surveillance Consortium data from the United States found that for women who begin screening at 40 years of age, the cumulative probability of receiving at least 1 false-positive recall after 10 years was 61.3% with annual screening compared with 41.6% with biennial screening.25 The cumulative risk of a false-positive biopsy recommendation was 7.0% with annual screening and 4.8% with biennial screening.
Results are similar in women who begin screening at 50 years of age. The US study25 cited above also estimated a false-positive rate of 16.3% in the initial screening round and 9.6% in subsequent rounds. This compares to the lower estimated cumulative risk of false-positive screening results in European women aged 50 to 69 years undergoing 10 biennial screening tests of 19.7%.26
The adverse effects associated with false-positive results are one of the main risks of mammography screening. For example, in a qualitative study of women with false-positive results, participants reported negative psychosocial consequences (eg, anxiety, negative effect on sleep and behavior) in the period after their abnormal screening result until they were declared free of cancer suspicion.27 Negative effects persisted into the longer term.26 In a quantitative longitudinal study, 1, 6, 18, and 36 months after their screening, women who had false-positive screening results had significantly (P < .01) higher (worse) mean scores than women who had not screened positive on, respectively, 12, 6, 9, and 4 of 12 dimensions of a validated questionnaire specific to psychological consequences of breast cancer screening.28,29 There are also physical harms associated with a biopsy for a benign breast lump, including pain and scarring.30,31
Overdetection and overtreatment: A major harm associated with mammography screening is overdetection and the resulting overtreatment of disease that is destined to never manifest clinically. For example, a recent meta-analysis of 3 RCTs of mammography screening found that for women invited to screening, there is a 19% probability that a cancer diagnosed during the screening period is overdetected.32 An observational study based on 30 years of US data reported a 100% increase in early-stage breast cancer, with only a very small (8%) reduction in advanced cancer, strongly suggesting that overdetection is a major consequence of mammography screening (Figure 28.3-2).33 The removal and treatment of breast cancers that were destined to never cause symptoms or death lead to unnecessary surgery, radiotherapy, adjuvant hormone therapy, and chemotherapy, all of which carry important adverse consequences.
Effect of 3 Decades of Screening Mammography on Breast Cancer Incidence
A, Self-reported use of screening mammography and the incidence of stage-specific breast cancer among women 40 years or older. B, Incidence of stage-specific breast cancer among women who generally did not have exposure to screening mammography: those younger than 40 years.
Copyright © 2012 Massachusetts Medical Society. All rights reserved. Reproduced from Bleyer,33 with permission from the Massachusetts Medical Society.
Harms of overdetection also include negative psychological consequences, such as increased anxiety, heightened sense of cancer risk, and negative effect on sleep behavior and sexuality. There are also adverse effects of being labeled as a cancer patient, including feelings of stigma, shame, and guilt, and wider ramifications, such as effects on relationships, family members, and insurance status.
False-negative results (missed cancers): As noted previously, among those who test negative, adverse consequences may include false reassurance and delayed presentation of later symptomatic disease. Mammography screening will detect between 61% and 89% of the cancers that occur in a population of regularly screened women.1,2 Thus, the interval cancer rate (which includes both missed cancers and cancers that develop de novo in the screening interval) is 11% to 39%.
Other harms: The USPSTF review2 reports data from a systematic review of radiation-induced breast cancer. In low-dose radiation exposure, risk was inconsistent, whereas high-dose exposure was associated with increased risk of breast cancer.34 A Canadian analysis found that for every 100 000 women screened annually from ages 40 to 55 years, then biennially to age 74 years, radiation exposure due to mammography would cause 86 breast cancers and 11 breast cancer deaths.35 The introduction of digital mammography (which uses a lower radiation dose and is currently the main form of screening in the United States) should reduce the risk.25
People found by screening to have clinically important disease may experience a benefit from early detection and treatment, but they can also experience earlier physical and psychosocial adverse effects of treatment. Health economists have found that we often consider present benefits, harms, and costs to be more important than those that will occur in the future. That is, we discount future outcomes and costs.36 The preference for the present over the future is more pronounced as people age.37 There is also the societal harm of the opportunity costs of screening if the money could be more cost-effectively spent elsewhere in health. We address cost-effectiveness considerations later in this discussion.
Balancing Benefits and Harms
Now that you are aware of the possible benefits and harms, if the woman proceeds with screening mammography, how can these benefits and harms be balanced against one another, and how can they be evaluated for her particular situation?
Unfortunately, the USPSTF guideline1,2 does not include information about breast cancer screening in a user-friendly format, such as a balance sheet of benefits and harms for people aged 40, 50, and 60 years who are regularly screened or not.38 Such data38 can be obtained by applying the rates of benefits and harms reported in RCTs to local populations and then used to develop decision aids for women considering screening mammography.39,40 We have recently updated earlier estimates38 using this approach for Australian women, and a balance sheet of outcomes for 20 years is given in Table 28.3-2 for women who begin biennial screening at 50 years of age. For example, of 1000 50-year-old women screened every 2 years for 20 years, 467 will be recalled at least once, and of those, 412 will have a false-positive and 122 will have a biopsy. Of the screened women, 73 will have a diagnosis of breast cancer as will 44 who are not screened. In the screened women, of the 73 with diagnosed breast cancer, 55 will have their cancer found by screening. Guideline committees could commission balance sheets like this when making local recommendations about screening to help in discussions with patients about screening.
Estimates of Number of Women Among 1000 Screened Every 2 Years Who Will be Recalled, Number Diagnosed With Breast Cancer, and Number of Deaths and Confidence in Estimates
|Favorite Table|Download (.pdf) TABLE 28.3-2
Estimates of Number of Women Among 1000 Screened Every 2 Years Who Will be Recalled, Number Diagnosed With Breast Cancer, and Number of Deaths and Confidence in Estimates
|Event During 20 Years ||No. of Events Among 1000 50-Year-Old Women Screened Every 2 Years for 20 Years ||No. of Events Among 1000 50-Year-Old Women Who Do Not Receive Screening ||Confidence in Estimates |
|Recalled ||467 || ||High |
|False-positive results ||412 || ||High |
|Have a biopsy ||122 || ||High |
|All breast cancers ||73 ||44 ||High |
|Cancers found by screening ||55 || ||High |
|Invasive cancers ||62 ||43 ||High |
|DCIS ||11 ||1 ||High |
|Interval cancers ||18 || ||High |
|Overdetected cancers ||19 || ||Moderate |
|Total deaths ||86 ||89 ||Moderate |
|Breast cancer deaths ||8 ||12 ||Moderate |
The balance sheet provides perspective on the benefits and harms of breast cancer screening. It shows that screening 1000 women biennially with mammography from 50 years of age will prevent approximately 4 deaths from breast cancer during 20 years but will lead to about 412 women experiencing a false-positive result and 19 women having an overdetected cancer during the same period. We can have high confidence in most of the estimates in the table because they are obtained directly from recent screening program statistics from records of hundreds of thousands of women. However, the estimates of the number of overdetected cancers and the estimates of the number of deaths among screened women are based on RCTs of mammography screening. Although these were reasonably good trials, they were conducted 20 to 50 years ago, and their applicability to current screening practice is uncertain.
When interpreting such balance sheets you need to consider that, unlike the presentation above, they often do not include all relevant benefits and harms that might be important to the patient. For instance, in a recent review, the most important harms of screening—overdetection and false—positive findings—were reported in only 7% and 4% of 57 cancer screening trials, respectively.41
We have already considered the advantage of RCT data: a reduced likelihood of bias. However, there is also a difficulty in applying trial data to real-world settings. Randomized trials often provide a better-quality screening program and subsequent intervention than are available in practice. If screening and interventions in the real world are not of the same quality as those in the trials, the benefits will be smaller and the harms will be greater than the ones calculated from trial data.
How Do Benefits and Harms Compare in Different People and With Different Screening Strategies?
The USPSTF update2 recommends screening for women aged 50 to 74 years but does not assign any rating (strong to weak) to that recommendation. Using the GRADE approach, one would lean toward a weak recommendation because of variability in trial quality and results, uncertainty about the balance of benefit to harm, and variability in women's values and preferences in relation to the benefit and harms of screening. This suggests that informed, individual decision making could be appropriate.
Because breast cancer incidence and mortality increase substantially with age, the magnitude of benefits and harms will vary for women according to their age. Indeed, the USPSTF2 recommends that the decision to start before the age of 50 years should be an individual one and take the woman's context into account, including the woman's values regarding specific benefits and harms. The balance of benefit to harm also depends on other factors, such as screening interval, screening test, and strategy, as the following discussion reveals.
It is important to keep in mind that the benefits of screening are experienced at some point in the future, whereas harms may be experienced at any time, including immediately after the first screening.
The probability that the woman will benefit from screening depends on her underlying risk of disease. Assuming that the RRR is constant over a broad range of risk of disease, benefits will be greater for people at higher risk of disease. For example, mortality from breast cancer increases with age, and the mortality benefit achieved by screening increases accordingly.2 However, the life-years lost to breast cancer are related both to the age at which mortality is highest and the length of life still available.
Factors such as a family history may increase risk of disease and therefore increase the benefits from screening. The USPSTF1,2 focuses only on average-risk people without an identified BRCA1 or BRCA2 mutation. Assessment of the benefits vs harms of screening in women from families at very high risk because of known mutations is very different. Such women, who may have a lifetime risk of breast cancer of 26% to 84%,42 may be referred to genetic counseling clinics or a clinical geneticist for testing and advice on prophylactic options, such as prophylactic mastectomy and oophorectomy. However, BRCA1 and BRCA2 mutations are very rare in the population (identified in less than 1% of the population overall and approximately 5% of women diagnosed as having breast cancer). Guidance on assessment of risk is available for women in this category. See, for example, the US National Cancer Institute's Breast Cancer Risk Assessment Tool.43 This is not relevant to the woman who has asked your advice because she has no family history of breast or ovarian cancer.
The probability that this woman will obtain a net benefit from screening depends on when she begins screening and how often she is screened (the screening interval). As the screening interval gets shorter, the detection rate (sensitivity) of a screening program and hence its potential effectiveness at cancer detection will improve.
Benefits, however, rarely increase in direct inverse proportion to reductions in screening interval. For example, one might expect screening twice as often to potentially double the relative mortality reduction obtainable by screening. In practice, however, the effect is usually much less. Cervical cancer screening, for instance, may reduce the incidence of invasive cervical cancer among women aged 55 to 69 years by 83%, 87%, and 87% if screening is conducted at 5-year, 3-year, and 2-year intervals, respectively.44
In contrast, the frequency of harms tends to increase in direct proportion to the number of screening tests a person receives. The consequence is that the marginal return (if any) of increasing screening intervals decreases as the screening interval is shortened. Ultimately, the marginal harms will outweigh the marginal benefit of further reductions in the screening interval. For example, modeling performed for the USPSTF assessment of the benefits and harms of mammography screening revealed that biennial screening maintained most (80%) of the mortality benefits of annual screening, but with fewer false-positive results and less overdetection.45
Extending the screening interval and/or changing the starting age of screening have been used in some screening programs to reduce the potential harms of screening, including overdetection and overtreatment. For example, the UK National Screening Committee increased the starting age for cervical screening to 25 years to reduce the overdetection of high-grade lesions in young women that would otherwise spontaneously regress. A large, population-based study in the United Kingdom found that screening women aged 20 to 24 years made virtually no difference in cancer rates among women up to aged 30 years when comparing women who were screened with those who were not.46 However, there is evidence of long-term psychosocial harm (eg, worry, anxiety, guilt, and concerns about infertility and relationships), economic costs (eg, treatment and sick leave), and potential physical harm with the treatment of the cervix linked to perinatal mortality and adverse pregnancy outcomes.47-50
Even with the same underlying risk, starting age, and screening interval, the woman seeking your advice will obtain a different net benefit, depending on the sensitivity and specificity of the test used to screen her for breast cancer. This is a particular issue as new tests are released and marketed for screening purposes. If the apparent sensitivity of a new test is greater than that of the test used in existing trials and if it is detecting significant disease earlier, the benefit of screening will increase (see Chapter 18, Diagnostic Tests). It may be, however, that the new, more sensitive test is detecting more cases of clinically irrelevant disease—for example, by detecting more low-grade prostate cancers or more low-grade cervical epithelial abnormalities51—which will increase the potential for harm.51 This increased sensitivity will be misleading if it does not increase the sensitivity to disease that is destined to become clinically manifest. If specificity is improved and testing produces fewer false-positive results, the net benefit will increase and the test may now be useful in groups in which the old test was not as useful.52
A relatively new technology that may improve breast cancer detection (by improving both sensitivity and specificity compared with mammography) is digital breast tomosynthesis or 3-dimensional mammography.53 As yet, however, we do not know how much the increase in sensitivity is related to cancers that will never become clinically manifest (overdetection) vs those that will.53,54 In screening for breast cancer in selected high-risk women, magnetic resonance imaging (MRI) has been found to have increased sensitivity (but lower specificity) compared with mammography. However, there is also emerging evidence that MRI may contribute to overdetection and may not improve patient outcomes.55-57
USING THE GUIDE
For an average-risk woman like the woman who has sought your advice, you would likely not recommend other screening modalities at this time.55,56
The benefits and harms of screening may also change if the threshold used to identify an abnormal result is increased or reduced. For example, in abdominal aortic aneurysm (AAA) screening, authorities have proposed lowering the threshold from 30 to 25 mm in the screening program.58 However, only 15% of the men initially identified as having an aortic diameter of 25 to 29 mm develop an aortic diameter sufficient to require surgery (>54 mm) in 10 years.59 This indicates that lowering the threshold will lead to a substantial increase of overdetection of harmless AAAs, resulting in more than a doubling of the incidence of AAA. This changes the number of individuals identified in groups A through E (Table 28.3-1) and the net balance of benefits and harms.
What Is the Effect of Values and Preferences?
Different people hold different values and preferences (see Chapter 27, Decision Making and the Patient). For example, couples considering fetal screening for Down syndrome may make different choices, depending on the value they place on knowing if their child will have Down syndrome vs the risk of iatrogenic abortion from amniocentesis.60
The woman's values about the range of potential breast screening benefits and harms will inform her own assessment of the best decision. She will also have a preference for the method of decision making and may want to analyze information and choose for herself. She may find decision making challenging and prefer to trust a physician to advise her on what to do, or she may want to engage in shared decision making with support from her physician.61
If she wants to understand the probabilities of harm or benefit in detail and make an independent or shared decision, a high-quality decision aid can provide balanced information about difficult decisions in a format that is easy for her to understand.61 Decision aids have already been widely evaluated for treatment decisions and have been found to increase knowledge and reduce decisional conflict without increasing anxiety (see Chapter 27, Decision Making and the Patient). Increasingly, investigators are developing patient decision aids for screening decisions.39,40,62
What if the woman prefers to be helped by a trusted person to consider an offer of screening? Entwistle and colleagues,63 who developed the Consider an Offer approach, suggest that such a conversation might include the following:
Who made the recommendation or offer?
What is the basis of the recommendation, and what are the main benefits and harms of screening?
Are there any factors that make the screening test more appropriate for some people than others?
Who might gain from screening and how are people protected?
Does this person need more information?
This conversation would provide less detailed epidemiologic information than a decision aid. However, it would help patients guard against screening offers that may not be in their best interests by, for example, contrasting offers made by unscrupulous commercial providers with those made by reputable professional providers or by determining whether the patient is in an age group for which the test is not recommended. The Consider an Offer approach has recently influenced population screening programs in the United Kingdom.64 Consider an Offer invitations for screening do not encourage screening or simply offer information about benefits and harms. Rather, as shown in the list above, the information provided acknowledges other drivers of decision making and recognizes that not accepting the offer of screening can be a reasonable choice.
What Is the Cost-Effectiveness?
Although clinicians will be most interested in the balance of benefits and harms for the individual screening participant, policymakers must consider issues of cost-effectiveness and local resources in their decisions (see Chapter 28.2, Economic Analysis).
Early cost-effectiveness estimates for mammography screening were very favorable—the UK Forrest Report (1987) estimated that screening would cost £3309 per quality-adjusted life-year (QALY) gained (approximately equivalent to £8094 today).65 US cost-effectiveness estimates were $15 000 to $20 000 per life-year saved (LYS) for annual screening for women aged 50 to 69 years in the 1990s.66,67
Incremental cost-effectiveness ratios were higher (ie, less cost-effective) for screening younger women (because of lower incidence and lower effectiveness) and older women (because of competing mortality). These estimates compared favorably with cost-effectiveness estimates for other preventive and therapeutic interventions at the time (eg, the cost-effectiveness estimate for antihypertension medication was $15 000 per LYS67,68; for coronary artery bypass surgery, $28 000 per LYS; and for car seat belts and airbags, $32 000 per LYS).
These early estimates did not, however, include costs of overdetection and overtreatment plus the potential cost from negative psychosocial consequences of false-positive results. They were based on screening delivering large mortality benefits, whereas more recent estimates1,2 have been much more modest. A recent study, including overdetection and overtreatment and a current estimate of mortality reduction, reported a cost-effectiveness ratio for triennial screening (compared with no screening) at £20 800 per QALY gained,65 still a generally acceptable cost-effectiveness ratio (see Chapter 28.2, Economic Analysis).65
CLINICAL SCENARIO RESOLUTION
Returning to our opening clinical scenario, you inform the woman that the USPSTF guidelines recommend biennial screening mammography for women aged 50 to 74 years but that it would be very reasonable to either follow the guidelines and choose to be screened or make an informed decision considering the benefit and harms herself.
The task facing you and her in shared decision making—should she decide to take the second option—is to weigh the benefit of reduced risk of death from breast cancer against the risks of potentially adverse consequences. These adverse consequences include the high probability of a false-positive result arising from the screening, the risk of an overdetected breast cancer and the adverse effects of treatment for the cancer, as well as the cost and anxiety generated by the investigations and treatment.
You could consider helping her clarify her values about the possible outcomes. For example, if the patient is not bothered by the prospect of regular mammograms and is happy to be screened knowing that if an abnormality is detected there is a risk that it could be overdetected and overtreated, she would probably choose to be screened. However, if she places a high value on avoiding unnecessary investigations and treatment, she may prefer to reconsider screening in a few years' time when the benefits will be greater or decrease altogether. Guiding her to online resources to support informed decision making about breast cancer screening is also an option.69,70