The question of who should be screened for breast cancer is not simply an academic debate. The real problem is a practical one: What do we say to the patient we will see this afternoon? How do we explain her risk for breast cancer, the effect that screening will have on that risk, and the harms and costs of screening? How do we help her make an informed decision? We will focus on this issue by reviewing the risk for breast cancer, weighing the evidence of benefits and of the harms and costs of screening, and offering some preliminary ideas about what to say to our patients.

The Risk for Breast Cancer

Age is the most important risk factor for breast cancer. The risk for being diagnosed with invasive breast cancer in the coming year increases with age up to 75 or 80 years of age, increasing more than 15-fold between 30 and 70 years of age Figure 1 [2]. For a woman aged 30 to 34 years, the risk for being diagnosed with breast cancer in the coming year is about 2.7 per 10 000 women; for a woman aged 70 to 74 years, it is about 43.4 per 10 000 women.

View larger version (11K): [in this window] [in a new window]   Figure 1. Breast cancer incidence (\#9632;) and mortality () for women in different age groups. Rates are per 100 000 population. Data from the Surveillance, Epidemiology, and End Results (SEER) program.

Another common way of expressing the risk for being diagnosed with breast cancer is lifetime risk at a given age. As most people know, a white woman's lifetime risk at birth for being diagnosed with invasive breast cancer has increased from 1 in 10 to about 1 in 8 in the past 10 to 20 years [2, 3]. The lifetime risk at birth for a black woman is about 1 in 11. The lifetime risk for being diagnosed with breast cancer remains about the same until 40 years of age when, because fewer years of life remain in which a woman can be diagnosed with breast cancer, the lifetime risk actually declines. At 60 years of age, a white woman's risk for being diagnosed with breast cancer during the rest of her life is about 1 in 10, and at 70 years of age, her risk is about 1 in 14. For a black woman at 70 years of age, the lifetime risk is about 1 in 20 [2, 3].

The problem with these expressions of risk is that screening does not reduce the risk for being diagnosed with breast cancer. It may, in fact, increase this risk, especially if carcinoma in situ is considered to be "cancer." So what is the risk relevant to breast cancer screening, the risk that we want to discuss with our patients? Clearly, screening is meant to reduce the risk for dying of breast cancer, not the risk for getting the disease. And the risk for dying is not reduced immediately for women in any age group. The benefit of screening for breast cancer lies in finding and treating, this year, cancer at an early stage that would have killed a woman a few (or many) years later. The reduction in mortality occurs several years after screening and detection. Thus, in talking to a woman about the decision to undertake a screening program during the following decade, the most relevant risk to discuss is the risk for ultimately dying (defined here as dying within the next 15 to 20 years) of a breast cancer diagnosed in the next 10 years. This is the risk we hope to reduce by screening over the next 10 years. Other, higher expressions of risk may serve to frighten women into overestimating their own risk, but they do not provide useful information for making decisions about screening.

The risk for dying within the next 15 to 20 years of breast cancer (either carcinoma in situ or invasive cancer) diagnosed in the next 10 years for women who are not screened is 2.4/1000 for women age 30, 7.8/1000 for women age 40, 12.9/1000 for women age 50, 19.5/1000 for women age 60, and 25.3/1000 for women age 70. In comparison, a 45-year-old woman's risk for dying of heart disease in the next 20 years is 24 per 1000 women; her risk for dying of cerebrovascular accident in the next 20 years is 6 per 1000 women; and her risk for dying of injury in the next 20 years is 6 per 1000 women [4, 5].

Evidence of the Benefits and the Harms and Costs of Screening

To make an informed decision about screening, women need not only accurate information about their risks but also information about the benefits and the harms and costs of screening. The best information about these issues has come from randomized controlled trials. During the past 30 years, seven randomized controlled trials have compared breast cancer screening with no screening Table 1 [6]. We review the results of these trials and the results from other studies that clarify the harms and costs of screening. We address harms and costs only in relation to health, although we believe that the financial costs of screening are also relevant [7, 8].

There are two reasons for using caution when extrapolating the results of the trials to the United States as it is in 1995. The first is that the "background" detection rate (the rate at which women are sensitive to early symptoms, especially small lumps) may vary from time to time and from culture to culture. Any screening program will be more effective (compared with background detection) in a time and culture that is less sensitive to small lumps, and less effective where women themselves routinely bring small lumps to the attention of their physicians. Background detection in the United States in 1995 is probably high, and programs such as the Health Insurance Plan trial (done in the 1960s, when there was probably less background detection) or programs in Europe (where many people have a different outlook on medical care than we have in the United States) may be less effective here and now.

The second reason for using caution when generalizing the results of European studies to the United States is that the design of the screening program affects the program's benefits and its harms and costs, and screening programs in Europe differ from those in the United States. Sweden, for example, where four of the trials were done, has a more organized approach to screening than the United States. In Sweden, the positive predictive value of biopsy (the percentage of biopsies that find cancer) is as high as 70%; in community practice in the United States, this value is closer to 20% (Brown M. Personal communication) [9]. Thus, fewer biopsies are done for every cancer found in Sweden, and the harms and costs for women being screened are lower. The factors responsible for this difference may include the lesser mobility of people in Sweden and the greater fear of delaying the diagnosis of cancer in the United States [10, 11]. The extent to which the Swedish system could be reproduced here is unclear.

The Randomized Controlled Trials: Benefits of Screening

Characteristics of the seven randomized controlled trials are shown in Table 1. None examined screening in women less than 40 years of age. All contributed information on screening women in their 40s (two of these only studied women aged 45 to 49 years); six gave information on screening women in their 50s; five provided data on screening women in their 60s (three of these studied only women aged 60 to 64 years); and only one gave information on women 70 to 74 years of age. The screening modalities and intervals varied from study to study, and even within the two modalities of mammography and clinical breast examination there were variations in quality (mammography techniques used in the Health Insurance Plan trial in the 1960s differed from those used in the Swedish trials of the 1980s) and intensity (some trials used one-view mammography every 24 months, and others used two-view mammography plus clinical breast examination every year). The studies also differed in their methods of randomization, the compliance of those invited to be screened, and whether screening was obtained outside the study by women in the control group [6, 12-15].

Given all of this variation, it is remarkable that there is agreement about a major conclusion: For women aged 50 to 69 years, mammographic screening done every 1 to 2 years with or without clinical breast examination can decrease mortality from breast cancer by 15% to 30% (Table 2). The decrease is evident within the first few years after screening begins [6, 12-15].

The meaning of this reduction in the risk for death is sometimes confused. The 15% to 30% decrease is the difference in the risk for dying of breast cancer during a predetermined period (here, 10 to 18 years) between the women who were offered screening and the controls, divided by the risk in the control group. This, then, is a "relative" risk reduction, but it does not answer the individual woman's question: "By how much is my risk reduced by screening over a period of time?" What this patient is asking for is the "absolute" risk reduction—the simple difference in risk between screened and unscreened groups of women. One may estimate this for an "average"-risk woman by multiplying the relative risk reduction by the risk for dying of breast cancer as shown in Table 3. For example, for a woman aged 50 years, the risk for dying eventually of a breast cancer diagnosed over the next 10 years is about 12.9 per 1000 women (Table 3). If this risk is reduced by 15% to 30%, the new risk is 9 to 11 deaths per 1000 women, and the absolute risk reduction is about 2 to 4 deaths per 1000 women. In other words, 2 to 4 fewer women will ultimately die of breast cancer for every 1000 screened regularly for 10 years.

Absolute risk reduction sounds much lower than relative risk reduction. Thus, those who wish to persuade women to have screening usually use relative risk reduction and those who oppose regular screening use absolute risk reduction. In fact, absolute risk reduction is more relevant to the individual patient but can be misinterpreted as being lower than is justified. The patient should consider not only the absolute risk reduction that results from screening but also whether this reduction is worth the harm and cost of screening. In addition, the patient should understand that large absolute reductions in the risk for premature death can best be attained by reducing risk for several preventable conditions, such as breast cancer, lung cancer, and atherosclerotic heart disease.

The relative risk reductions for different age groups in the randomized controlled trials are given in Table 2. The four Swedish studies are considered together, using a recent meta-analysis [13]. Although the trials agree about the effectiveness of screening for women 50 to 69 years of age, there has been major disagreement about the effectiveness of screening women 40 to 49 years of age. No single study to date has found a statistically significant reduction in risk from screening women in this age group Table 2, with the possible exception of the Health Insurance Plan trial, where there is disagreement about the statistical significance of the risk reduction [16, 17]. All three studies that have had 10 or more years of follow-up, however, show a statistically nonsignificant trend toward risk reduction for women in their 40s who have been offered screening: 13% in the combined Swedish trials, 25% in the Health Insurance Plan trial, and 23% in the Edinburgh trial (which only included women 45 to 49 years of age). In each case, no benefit (statistically significant or not) was evident until 8 to 10 years after screening was begun.

The results from the Canadian National Breast Screening Study were published late in 1992 and have been controversial [18, 19]. Although the study's screening program, which used annual two-view mammography plus clinical breast examination, was one of the most sensitive in any of the randomized controlled trials [6], the group of women 40 to 49 years of age who were screened had a statistically nonsignificant increase in breast cancer mortality after 7 years of follow-up (odds ratio, 1.36; 95% CI, 0.84 to 2.21). Given the results of the other trials, it is unlikely that this increase in mortality is real; we suspect that it is due to random variation and that longer follow-up will find the survival curves coming together. We will not address the extensive debate about this report, but will note that the Canadian study is consistent with the other trials in finding no risk reduction after 7 years of follow-up; the other trials only showed risk reduction after 8 to 10 years of follow-up.

Thus, the Health Insurance Plan trial (which used both mammography and clinical breast examination), the meta-analysis of the Swedish trials (which used mammography only), and the Edinburgh trial (which used both mammography and clinical breast examination) all found a small, statistically nonsignificant decrease in mortality from breast cancer in women 40 to 49 years of age who were offered screening; the decrease only appeared after 8 to 10 years of follow-up. The statistical tests indicate that chance could explain the findings, but the fact that all three studies found the same result makes us lean toward calling it real. Still questionable are the magnitude of the benefit and whether the benefit balances the harms and costs.

Balancing the Benefits and the Harms and Costs of Mammographic Screening

As we noted initially, screening is a two-edged sword. In our rush to control the terrible problem of breast cancer, it is easy to forget that, in the best of circumstances, the many must be screened to benefit the few. It is important to consider the experience of those who do not benefit from screening [20]. In defining the possible harms and costs of screening, and in considering how they balance against the benefits noted above, it is useful to consider a specific example. Figures 2 and 3 show some "ballpark" estimates of the consequences of a single mammographic screening for 10 000 average-risk women 50 to 70 years of age Figure 2 and for 10 000 average-risk women 40 to 49 years of age Figure 3, assuming that all of the women are taking part in an ongoing program of annual screening [2]. Ten thousand is the approximate number of women in each of these age groups in an average U.S. city with a population of 100 000 to 150 000 persons. These examples should be considered only approximate because of the considerable variation that exists among mammography facilities. We have assumed the use of mammography facilities that are neither the worst nor the best in the country but ones to which readers might reasonably refer patients. In Figures 2 and 3, for example, we use a positivity rate (the percentage of screening mammograms that require some additional follow-up) of 5%, which is indicative of some good mammography practices in the United States (Unpublished data. New Hanover Breast Cancer Screening Study; McCarthy B. Personal communication). Many community radiology practices, however, have positivity rates of more than twice this figure (Brown M. Personal communication). Although we have used the same positivity rate for both younger and older women, some practices report higher positivity rates for younger women (McCarthy B. Personal communication; Unpublished data. New Hanover Breast Cancer Screening Study).

View larger version (22K): [in this window] [in a new window]   Figure 2. Estimated consequences of a single screening mammogram for 10 000 women 50 to 70 years of age who participate in a program of regular annual screening.

View larger version (25K): [in this window] [in a new window]   Figure 3. Estimated consequences of a single screening mammogram for 10 000 women 40 to 49 years of age who participate in a program of regular annual screening.

Between 2.5 and 4 of every 1000 women in their 50s and 60s will be diagnosed clinically with breast cancer in the coming year [2]. Assuming that screening detects cancer in about 85% of these women [6], and that these women have been screened previously (as most U.S. women in this age group have been), we will find between 21 and 34 cancers from screening 10 000 of these women (Figure 2). Thus, 466 to 479 women [or more, if we assume a higher positivity rate] will have "false-positive" results. Unfortunately, not all of the 21 to 34 women with screening-detected breast cancer will have their lives extended by our detection. To show this, we consider each cancer as one of three prognostic types: 1) cancers that are so treatable that they will not cause death, even if they are not detected until the woman finds a lump; 2) cancers that are so untreatable that they will cause death no matter when they are detected; and 3) cancers that can be more effectively treated if found earlier rather than later.

We know from survival data that about 50% of women aged 50 to 70 years who have breast cancer (not found by screening) will still be alive 15 to 20 years after diagnosis (Unpublished data, 1973-1980. Surveillance, Epidemiology and End Results [SEER] program). These women, then, had "type 1" cancers that were slower growing or were more treatable. We also know from the randomized controlled trials Table 2 that screening reduces the risk for dying of breast cancer in women in their 50s and 60s by 15% to 30%. Thus, if we take roughly 15% to 30% of 50% (the percentage who will die), it is clear that 7.5% to 15% of all women with breast cancer (about 9% to 18% of women with screening-detected breast cancer) will have their lives extended by screening. This is 2 to 6 women for every 10 000 screened Figure 2, or 1 life extended for every 1700 to 5000 women screened.

The trials suggest that screening women aged 50 to 70 years every 2 years is as effective in reducing mortality as screening them annually [6]. Thus, it seems reasonable that screening every 2 years will produce about the same benefit with fewer mammograms (in Figure 2, this is 4 to 12 lives extended over 2 years, or about 1 life extended for every 850 to 2500 women screened). The extent to which biennial screening would also reduce a woman's risk for a false-positive result is unclear, although some reduction is likely. Overall, the balance between the benefits and the harms and costs of biennial screening for women aged 50 to 70 years appears to be better than that of annual screening.

Figure 3 provides similar data for women in their 40s. We expect to find about 12 breast cancers (the expected incidence is 16 new cases multiplied by 75% of cancers detected) [2, 6] from screening 10 000 women in this age group. Again, about 50% of these women have "type 1" cancers that will allow long-term survival no matter when they are detected; thus, about 6 of these 12 women will eventually die of breast cancer. Optimistically, if screening provides the same reduction in the risk for dying for these women as for women in their 50s and 60s (15% to 30%), then 1 to 2 women will have their lives extended for every 10 000 women screened. It should be understood that this is a generous calculation of the benefits of screening women in their 40s. It does not, for example, take into consideration the negative results from the Canadian study or the fact that investigators of the randomized controlled trials have found no risk reduction for women in their 40s until 8 to 10 years after screening begins.

The trials also suggest that the length of time by which screening advances diagnosis is shorter for women in their 40s than for women in their 50s and 60s [21]. Thus, assuming that screening is effective in reducing mortality for younger women, it is likely that we will need to screen women in their 40s more frequently than women 50 to 70 years of age, decreasing the benefits of each screening and possibly increasing the overall harms and costs. As shown in Figure 3, fewer lives are extended by screening 10 000 women in their 40s than by screening the same number of older women (Figure 2). The use of the term "marginal" to describe this risk reduction seems justified [6].

The harms and costs that result from extending the lives of 2 to 6 women in their 50s and 60s (4 to 12 women if screening is biennial) or 1 to 2 women in their 40s are substantial. First, false-positive results occur more frequently in the group aged 40 to 49 years than in the group aged 50 to 70 years, but in both groups, the women with false-positive results outnumber the women with cancer by a large margin. The number of false-positive results given in our figures is for only 1 year; more women will have false-positive results every year women are screened. Few studies of mammography programs provide the rate of total invasive procedures (open biopsy, core biopsy, and fine-needle aspiration) produced by screening. The number of invasive procedures given in the figures for one screening (100 to 200; 1% to 2% of all women screened) was extrapolated from a nationally representative study of mammography programs (Brown M. Personal communication). In the Canadian National Breast Screening Study, 12.1% of women aged 40 to 49 years in the screened group (9.4% of those aged 50 to 59 years) had at least one invasive procedure during 4 years of screening; most of these procedures were done in the initial year of the study (Miller AB. Personal communication). The proportion of women who will have had at least one invasive breast procedure after 10 years of screening remains unclear; a reasonable estimate would be about 20%. If noninvasive follow-up procedures (magnification views, ultrasonography, early repeated mammography) are added, the proportion would be much higher, especially in younger women, who have more breast cysts.

Whether false-positive rates differ according to the age of the women being screened is also unclear. As noted above, some mammography programs find that positivity rates (which are directly related to the false-positive rates in a stable population) vary by age; younger women have higher rates. If more frequent screening is associated with higher cumulative false-positive rates (which seems likely), then this factor may also produce higher false-positive rates in younger women. Evidence also indicates that the search for breast cancer after an abnormality is detected requires more procedures and more physician visits for younger women than for older women [22].

Because fewer cancers are detected in younger women, the ratio of procedures done to cancers detected is higher in younger than in older women. A recent study from an excellent academic screening center found that 61 follow-up procedures (such as magnification views, ultrasonography, fine-needle aspiration, and biopsy) were required to detect a single case of breast cancer among women in their 30s who were having their first screening examinations. Forty-four procedures were required for women in their 40s, 22 for women in their 50s, and 10 for women in their 60s [23]. Although all of these numbers would probably be lower for subsequent screenings, the trend would be the same: The number of false-positive results per cancer detected is larger for younger women and smaller for older women.

The effect of false-positive results on women is considerable. First, there is the small complication rate from biopsy, which includes scarring and diminished sensation. More troubling is the strong evidence that most women suffer marked anxiety in the time between an abnormal mammogram and the follow-up test, and that many women who have had a false-positive result have anxiety lasting for months, regardless of whether they had a biopsy or a noninvasive follow-up test [24-27]. In one study, 47% of women who had had a false-positive mammogram and who were told that they did not have cancer reported substantial mammography-related anxiety 3 months later. Twenty-six percent of these women reported that mammography-related worry affected their moods, and 17% reported that it affected their daily functioning [24].

Another harm that can result from screening should be considered. Ductal carcinoma in situ is a pathologic lesion that is being detected much more frequently now as a result of screening programs. In some series, more than half of the "cancers" in women younger than 50 years of age have been ductal carcinomas in situ [23]. The natural history of this lesion is not clear, but in many cases, the lesions appear to grow slowly, becoming invasive and potentially lethal only years later [28]. Evidence suggests that 50% or more of all untreated women with ductal carcinoma in situ never develop invasive cancer [29, 30]. Yet many women with this lesion are being treated with mastectomy because it is difficult to distinguish lesions that will become invasive from those that will not. If it is true that some women diagnosed with ductal carcinoma in situ will never have invasive disease, then they are receiving treatment they do not need as a result of screening. If detection of this lesion is an important factor in reducing risk for women in their 40s, the slow growth of the lesion may allow many cases to be detected and treated just as well by screening after 50 years of age.

Two other harms that screening can cause should be mentioned. First, the women whose cancers were detected but whose lives were not extended by screening have been made to live longer with the knowledge that they have cancer than would otherwise have been the case. Some of these women may be angry with their physicians for not diagnosing the breast cancer earlier or may feel guilty that they themselves did not find it earlier, even though early detection would not have affected their health outcome. In this way, quality of life may be decreased by screening. In addition, some women whose cancers were missed by screening (false-negative results) may have been falsely reassured that they do not have cancer and thus may put off seeking care when they find a small lump.

Screening has numerous potential harms and costs. A woman's cumulative risk for harm from screening is probably greater the earlier in life she begins a regular screening program. Thus, these harms and costs must be weighed against the potential benefits of screening.

Screening with Clinical Breast Examination

The analyses in Figures 2 and 3 focus on screening done using mammography alone. Adding clinical breast examination would probably increase the number of cancers detected by about 10%, but it is unclear whether doing so would extend more lives. Two observations from the randomized controlled trials lead us to believe that clinical breast examination is a valuable screening test. First, the benefit among women in their 40s in the Health Insurance Plan trial may have been due more to clinical breast examination than to mammography. Fifty-eight percent of cancers detected by screening in this age group were found by breast examination alone; only 25% were detected by mammography alone [12]. Second, one part of the Canadian trial (not included in Table 1 compared clinical breast examination with mammography for women 50 to 59 years of age. After 7 years of follow-up, breast cancer mortality did not differ between the two groups [31]. In the Canadian trial, the breast examination was done by specially trained nurses who took extra time and care [32]. It is certainly premature to conclude that a clinical breast examination is as good as mammography in reducing mortality for women in their 50s, but neither would one want to abandon the breast examination as a screening test until further research has evaluated its independent effect. It is clear that the clinical breast examination is often done poorly by practicing physicians and that training can substantially improve performance [33, 34].

Summary of the Evidence: Benefits and Harms of Screening

The evidence from randomized controlled trials shows that benefit from screening in absolute risk reduction (number of lives extended) is "marginal" but probably present for women in their 40s and larger for women 50 to 70 years of age. Because the risk in women less than 40 years of age is low, it is unlikely that screening these women has any sizeable benefit, even screening to establish a "baseline." Because the risk in women 70 years of age or more is high, it is likely that older women who have at least 8 to 10 years to live will benefit from screening. (Women 70 years of age will have an average additional 14 to 15 years of life; women 80 years of age will have an average additional 8 to 9 years of life. Active women of these ages would have longer average life spans [35].) The extent to which screening will benefit "high-risk" women less than 50 years of age is unclear. If "high-risk" women are those with a first-degree relative who was diagnosed with breast cancer before 50 years of age [36]—less than 10% of women are in this category—then they may have enough of an increased risk to expect some benefit. The assumption that these women will benefit from screening, however, is untested.

The harms and costs of screening are still uncertain but are becoming better defined. The earlier regular screening is started, the greater the cumulative risk for harm. Thus, the benefits gradually increase with age and the harms and costs gradually decline. In Table 3, we summarize this tradeoff between the benefits and the harms and costs and present our own clinical strategy. As with many issues in medicine, the point at which benefits outweigh harms (and financial costs) is unclear. Most, but not all, physicians and patients have decided that the balance for women 50 to 70 years of age tips in the direction of screening. We agree with this assessment. Weighing benefits and harms for younger, average-risk women is difficult. Patient involvement is clearly needed in decisions about breast cancer screening.

What to Say to Our Patient: Some Preliminary Ideas

All of this discussion about the benefits and the harms and costs of screening may be interesting, but how do we use it for our patients' benefit? We need a clinical strategy, based as much as possible on evidence, to help us talk to our patients. More research is needed in this area, but there is some information that could be the basis for a clinical strategy.

First, it is clear that women often misperceive their risk for breast cancer. A recent study found that women in their 40s overestimated their risk for dying of breast cancer more than 20-fold and overestimated the benefit of screening 6-fold [37]. Another study found that only 20% of women of any age understood that risk increases with age. Older women were less worried about breast cancer than younger women [38]. Eliciting perceived risk and providing individualized feedback can change "optimistic" or "pessimistic" risk perceptions to more realistic ones [39].

Second, women vary in their preferences for and their "barriers" to screening [40], but physician recommendation is still an important factor in a woman's decision to have mammography. In one study, women whose physicians had discussed screening with them were 4 to 12 times more likely to have had a mammogram [41]. In another study, 90% of eligible women reported that they would have mammography if a physician recommended it, yet nearly half also said that they would not want their physician to remind them when they are due [42]. Clearly, some women are more interested than others. Patients in lower socioeconomic groups are screened less often than other patients, partly because they seem less interested and partly because physicians recommend screening to them less often [43, 44]. Clarifying and addressing barriers (such as concern about pain from breast compression) and making a clear recommendation will often be effective in encouraging screening.

Encouraging screening should not, however, be the only goal. Rather, the goal should be appropriate decision making that takes into account both the health tradeoffs of screening at particular ages and the patient's perspective and values. We suggest that physicians examine the information in Table 3 and, bearing their patients' values in mind, divide the age continuum into three categories: those for whom the benefits clearly outweigh the harms and costs, those for whom the benefits and the harms and costs constitute a "close call," and those for whom the harms and costs clearly outweigh the benefits. The women who most need to be involved in decision making are those in the first two groups.

For those in the "benefits outweigh harms" and the "close call" groups, we suggest the following clinical strategy: 1) Routinely raise the issue of screening by eliciting the patient's perceptions of her risk for breast cancer and of the benefits and the harms and costs of screening; 2) provide information and discussion to encourage an accurate understanding of both breast cancer risk and screening; 3) encourage the patient to use her own values to weigh the balance between the benefits and the harms and costs of screening; 4) point out biases in reasoning [45] and minimize socioeconomic barriers; and 5) when benefits obviously outweigh the harms and costs, make a clear recommendation in favor of screening.

These discussions need not take a long time. In some cases, they may occur over several visits, allowing the patient to discuss the issues with other relatives or advisers. A nurse or clinical assistant may assist with the discussion, and printed material or decision aids [46] may be used to improve communication and save time. The discussions should include simple quantitative information where possible, but qualitative discussions may be necessary at times, comparing the risks, benefits, and harms with other commonly understood analogies.

In these and other discussions with patients, one must be aware of certain biases and shortcuts that people often use to simplify complex decisions [45]. For example, some women may have a tendency to categorize interventions as being either entirely negative or entirely positive and thus may believe that mammography is either worthless or completely effective. Some women may also have concerns about breast cancer that are out of proportion to their objective risk and thus may take an overly zealous approach to screening. Other women may magically believe that if they are not screened they cannot have the disease. Physicians should also know that people are influenced by the way discussions of risk are "framed" [47], whether positively or negatively, and thus that they must be careful to frame discussions in a balanced way, providing information about both relative and absolute risk reduction and about both benefits and harms and costs.

We are still at an early stage in our understanding of how to talk to patients about such issues as breast cancer screening. Our suggested strategy is thus tentative until further research can clarify an optimal approach. Certainly it is easier just to tell a woman that she should have mammography. In the end, though, it may be that the kind of discussions we propose, assisting the patient to weigh the benefits and the harms and costs of screening, will help us all to better understand the real but limited role that screening can play in controlling the problem of breast cancer. That is an important goal in itself.