Screening for Breast Cancer FREE

Breast cancer screening, especially with mammography, has been recommended for many decades,¹ and the majority of women older than 40 years in the United States participate in screening activities.^2- 3 Meanwhile, new screening modalities have been introduced, and some of these have been increasingly incorporated into community practice. However, none of the new technologies has been evaluated for its effect on breast cancer mortality.

Community practice of screening may differ from the care provided within randomized clinical trials and is less often discussed in review articles. Reviews of breast cancer screening usually emphasize efficacy and results of randomized trials, particularly those involving screen-film mammography.^4- 7Efficacy of a screening tool is measured in experimental studies under ideal circumstances.⁸ In contrast, effectiveness is defined as the extent to which a specific intervention “when deployed in the field in routine circumstances, does what it is intended to do for a specific population.”⁸

We systematically reviewed what is known about the community practice of mammography, clinical breast examination, and breast self-examination, when possible, comparing the results from community studies with those of randomized clinical trials. In addition, we reviewed what is known about newer screening modalities, specifically digital mammography, computer-aided detection programs for mammography, ultrasound, and magnetic resonance imaging (MRI).

The evaluation of screening modalities, especially in the community setting, is challenging for methodological, clinical, and ethical reasons. Randomized clinical trials are considered the gold standard for evaluating a new screening test. The long-term breast cancer mortality rate of women randomized to receive a new screening test is compared with that of women randomized to receive standard care. However, such trials are difficult to conduct. They require tens of thousands of women who need to be followed up for more than 15 years. Furthermore, because mammography screening has been shown to be effective in some trials, it would likely be even more difficult to demonstrate any additional efficacy of new tests. Finally, as treatment for breast cancer has improved over time,^9- 11 the impact of screening on breast cancer mortality may be increasingly difficult to establish.

Because of these challenges, new screening tests are often first studied by establishing characteristics of the tests themselves, rather than by studying their effect on patient outcome such as breast cancer mortality. Important test characteristics include sensitivity, specificity, safety, cost, simplicity, and patient and clinician acceptability. We review what test characteristics have been studied and the findings for each new modality. We also indicate the study design and the end points studied for each screening test. Although it is important to use resources wisely when considering a screening test for a large segment of the population, cost-effectiveness analyses are not reviewed.

It is important to determine the characteristics of a screening test in a community setting if the test is to be used in that setting. However, test characteristics of new modalities are usually evaluated among women for whom the rate of breast cancer is higher than average, such as women at increased risk of breast cancer or women in a diagnostic setting with breast symptoms or known breast abnormalities. The reported sensitivity and specificity of a test in these high-risk women may be different from the sensitivity and specificity of the same test used in a general screening population.¹² We therefore indicate if a test has been evaluated as a diagnostic or screening test and if as a screening test, whether it has been evaluated in women thought to be at increased risk or in the general population.

For this review, searches of MEDLINE, The Cochrane Library, the National Guideline Clearinghouse Web site, the US Preventive Services Task Force recommendations and reviews,^5,13 and the International Agency for Research on Cancer Handbook of Cancer Prevention (IARC)⁴ were performed to identify English-language articles about breast cancer screening. Examples of search terms included mass screeningandbreast and then the specific modality (eg, mammography, breast examination, digital mammography, computer assisted diagnosis or computer aided detection, magnetic resonance imaging, ultrasound or ultrasonography). The bibliographies of retrieved articles were also scanned to retrieve additional relevant articles. Details regarding search terms are available from the authors.

Eight reported randomized trials have studied mammography’s effectiveness in the United States,^14- 15 Sweden,^16- 20 Canada,^21- 22 and the United Kingdom.²³ Concerns related to flaws of these randomized clinical trials have been raised.^24- 25 In-depth independent reviews of the criticisms of the trials have concluded that these flaws do not negate mammography’s efficacy in reducing breast cancer mortality, especially in women aged 50 to 69 years.^4- 5,7,26 Trials comparing mammography with or without clinical breast examination to usual care (with little or no screening mammography) demonstrated remarkably consistent results for women older than 50 years. Meta-analyses that included all trials demonstrated statistically significant reductions of 20% to 35% in mortality from breast cancer for women aged 50 to 69 years.⁷ The majority of participants in clinical trials of mammography were white and information on BRCA mutation status was not known.

In general, breast cancers detected by mammography screening are smaller and have more favorable histological and biological features than tumors detected between mammography screening rounds or tumors found outside of screening.^27- 32 Because the favorable prognoses of women with breast cancer detected by mammography screening may be attributable to selection bias, length bias, lead-time bias, and overdiagnosis,^33- 34 randomized controlled trials with breast cancer mortality as the outcome have been particularly important in excluding such biases.

The benefit of screening women in their 40s is slower to appear and is somewhat less than that of women older than 50 years. Women in their 40s have a lower incidence of disease, denser breast tissue (which can lower the sensitivity of mammography), and, on average, faster-growing cancers.^27,35- 36 A randomized trial of mammography screening for women in their early 40s is under way in the United Kingdom.³⁷ Clinicians and patients are often surprised at the large number of women who need to be screened to prevent 1 death due to breast cancer. For example, it has been estimated that between 500 and 1800 women who are 40 years of age would need to undergo regular screening mammography to prevent 1 breast cancer death after 14 to 20 years.^7,13

Randomized trials have included few or no women older than 70 years. One case-control study found that screening women between 65 and 74 years of age was associated with mortality reduction.³⁸ Pooled data from a community study showed that the sensitivity and specificity of screening mammography were highest in older women, especially those older than 80 years.³⁶ Screening mammography may not be beneficial for women with significant comorbidity or with a life expectancy of less than 5 years, and may actually necessitate work-up that does not result in any benefit.^39- 41 However, if a woman is in reasonably good health and would be eligible for and interested in treatment, continued screening should be supported. Encouraging individualized decisions may be especially appropriate for women older than 70 years.⁴²

The majority of women with abnormalities noted on screening mammograms (≈ 95%) do not have breast cancer with variability based on multiple factors including the radiologist’s assessment and the woman’s age.⁷ Because the risk of breast cancer increases with age, the likelihood of a woman with an abnormal mammogram result having cancer also increases with age.^43- 44 On the other hand, having a normal mammogram result does not rule out the possibility of having breast cancer, because false-negative mammography examination results do occur. In such cases, either the cancer is not visible on mammography examination or the radiologist fails to notice the lesion prospectively.^45- 47

Results from 7 population-based community screening programs in the United States on 463 372 screening mammograms revealed an overall sensitivity of 75.0% and specificity of 92.3%.³⁶ A sensitivity of 75% means that 25% of women (or 25 of 100 women) who were diagnosed with breast cancer had normal mammogram results between 12 and 24 months before their cancer diagnosis (eg, false-negative examination result). This sensitivity of 75% in the community was similar to that reported in the randomized trials (68% to 88%) but the specificity was lower than in most of the trials (range, 82%-93% for Canadian National Breast Screening Study 1 to 98.5% for Health Insurance Plan of New York).⁴⁸

Breast density and age are important predictors of accuracy.³⁶ Adjusted sensitivity ranged from 63% in women with extremely dense breasts to 87% in women with almost entirely fatty breasts; adjusted sensitivity increased with age from 69% in women aged 40 through 49 years to 83% in women aged 80 through 89 years.³⁶ Adjusted specificity increased from 89% in women with extremely dense breasts to 97% in women with almost entirely fatty breasts.³⁶

Guidelines for quality assurance have been issued by several bodies, such as the Commission of the European Communities^49- 50 and the US Mammography Quality Standards Act.⁵¹ The Breast Imaging Reporting and Data System, used in the United States to standardize reports, includes categories for assessment ranging from 0 to 5⁵² (Table 1). The associated likelihood ratios for a breast cancer diagnosis for first screen are shown in the table.⁵⁴ Use of the Breast Imaging Reporting and Data System has not eliminated the variability among radiologists⁵⁵ that had been noted before.^56- 57

Large differences have been noted between the recall rates (or percentage of screening mammograms considered as positive) of community-based mammography programs in the United States and those in other countries. The recall rate in the United States is twice the recall rate in the United Kingdom (eg, 12.5%-14.4% vs 7.6%), with no difference in cancer detection rate.⁵⁸ Elmore et al⁵⁹ noted comparable differences between North American screening programs and those in other locations, which persisted after adjusting for differences such as age of women screened, use of single vs double reading, and use of 1 vs 2 views of each breast for examinations. Other possible reasons for the regional variability noted include differences in the characteristics of the population screened (eg, presence of risk factors or symptoms) and features of the mammography examination (eg, equipment type and year, technician training).^59- 60 The experience of the physician interpreting the mammograms has also been raised as a possible reason for the variability.^60- 62 Recommendations vary regarding the minimum number of mammograms that the physician should interpret yearly, from 480 in the United States⁵⁰ to 5000 in the United Kingdom.⁵¹ Educational training has been shown to improve sensitivity with no change in specificity.⁶³ Finally, features of the health care system, including malpractice concerns, financial incentives, quality control procedures, and auditing procedures, may be related to some of the variability.⁶⁰

Double-reading of films, where 2 or more radiologists interpret each film, is offered in some US screening programs and in about half of the other countries that use mammography screening.^4,58 However, double reading is performed in various ways (eg, 2 interpretations may be completely independent, 2 radiologists may perform interpretations together, or a third reader may serve as a tie-breaker). A systematic review of 10 cohort studies found that double reading increased the cancer detection rate by 3 to 11 per 10 000 women screened; recall rates increased or decreased depending on the method of double reading used.⁶⁴ The heterogeneity of the definitions of double reading and of screening practices precludes an accurate assessment of benefits and harms.

Both screen-film mammography and full-field digital mammography use x-rays to obtain images. With screen-film mammography the image is captured on film; with full-field digital mammography the image is captured digitally. Digital images can then be printed on film for viewing, or the images can be interpreted directly from a computer monitor. The digital acquisition process improves logistics and work flow by allowing electronic transmission, storage, and retrieval. Radiologists viewing the image on a monitor can alter the contrast and brightness of the image and magnify specific areas without additional x-ray exposure. Digital imaging also allows easier use of computer-aided detection software. One disadvantage, however, is the increased cost associated with full-field digital mammography.⁴

Three community-based studies have compared full-field digital mammography to screen-film mammography (Table 2). Two studies found the sensitivity of full-field digital mammography (64% and 74%) to be less than that of screen-film mammography (79%, 90%), but these studies had a small number of women with breast cancer (42 and 31, respectively) and the display systems and experience of radiologists may have improved since these studies.^65- 66,70 A larger randomized study reported similar cancer detection rates (per all screened), with higher recall rates for full-field digital mammography.⁶⁷ A trial now being conducted, which aimed to enroll 49 400 women, will compare the diagnostic accuracy of full-field digital mammography from 4 different manufacturers with that of traditional screen-film mammography.⁷¹

Computer-aided detection programs, which recognize patterns in breast images associated with cancer (Figure 1), may potentially help radiologists improve their diagnostic accuracy, but presently data are limited (Table 2).^{68- 69,72- 73} Computer programs that can mark calcifications, masses, or other potential lesions on the mammogram may increase the number of cancers detected compared with unassisted interpretations. In a study of 12 860 women,⁶⁸ computer-aided detection increased radiologists’ overall screening recall rate from 6.5% to 7.7% while increasing the number of cancers detected from 41 without computer-aided detection to 49 with the technology. However, in the largest clinical series to date, which included 59 139 mammograms interpreted with computer-aided detection and 56 432 interpreted without, recall rates and cancer detection rates did not differ significantly.⁶⁹ Computer-aided detection may prove helpful in reducing variability among radiologists of differing expertise. Such programs are used by a small but growing number of mammography facilities in the United States.^69,74 Medicare and Medicaid allow for additional billing for computer-aided interpretations of mammograms. As technologies continue to improve, larger multisite studies will be needed for more definitive evidence.

Although two thirds of US women older than 40 years receive regular screening clinical breast examinations,² few data about the efficacy of clinical breast examinations alone are available from randomized clinical trials. Four randomized trials of mammography included the clinical breast examination in the screened group.^15,21- 23 One of these trials, the Canadian National Breast Screening Study 2 of women aged 50 through 59 years at entry, compared the results of an annual standardized 10- to 15-minute clinical breast examination and breast self-examination with the results of an annual standardized clinical examination and breast self-examination plus mammography (in other words, there was no typical control group that received no screening).²¹ The trial found that breast cancer mortality was similar in the 2 groups of women although yearly mammography in addition to physical examination and breast self-examination detected more small and lymph node–negative breast cancers than did screening with physical examination alone.²¹ Although the sensitivity of screening clinical breast examination was highest at 63% in the National Breast Screening Study 2,⁷⁵ an overall estimate based on all randomized trials calculated sensitivity at 54% (95% confidence interval [CI], 48%-60%) and specificity at 94% (95% CI, 90%-97%).⁷⁶

We suspect that few clinicians in the community setting perform examinations as carefully as those in the Canadian trial, so accuracy may be lower in the community setting. Results reported from community practices showed sensitivity ranging from 28% to 36%^77- 79 (Table 3). In one study, two fifths of physicians (34 out of 80) who performed a screening breast examination on manufactured breast models used no discernible systematic search pattern at all.⁸⁰ Sensitivity of examinations improved by spending more time^80- 81 and by using a thorough, systematic technique.^76,82 However, the number of false-positive examinations may increase with training.⁸¹

In randomized controlled trials, noting an abnormality on a screening clinical breast examination in an asymptomatic average-risk woman increased the likelihood of breast cancer (likelihood ratio [LR], 10.6; 95% CI, 5.8-19.2).⁷⁶ However, in community practice, an abnormal screening breast examination result was associated with an LR of 2.1,⁷⁶ substantially lower than that of women in the same practice presenting with a breast abnormality (LR, 24).⁸³ Noting a suspicious abnormality on a screening mammogram was associated with LRs ranging from 7 to 2200 (Table 1).⁸⁴

Breast self-examination is appealing as a patient-centered, noninvasive procedure that allows women to become comfortable with their own bodies. However, the extent of current practice is thought to be low.⁴ About one third of US women regularly perform breast self-examination, and the estimated sensitivity is low (20% to 30%).⁸⁵ Among respondents to the Women Physicians’ Health Study (N = 4501), only 21% reported performing monthly breast self-examination.⁸⁶

Training in breast self-examination, while associated with increased accuracy of detection of lumps in breast tissue, has been associated with increased rates of false-positive findings and thus diminished specificity.^87- 90 In addition, there is evidence casting doubt on the benefits. A large randomized controlled trial in Shanghai, China, of 266 064 women working in textile factories provided half of the women with intensive initial instruction, including practice with breast models, as well as regular reminders and practice examinations under supervision once every 6 months for 5 years. This study found no positive effect of breast self-examination on breast cancer mortality after 10 years of follow-up but almost double the rate of biopsies due to false-positive findings (1.8% of women in the instruction group vs 1.0% of women in the control group).^88- 89 The study results should be interpreted with caution because approximately 40% of women in the trial were in their 30s. Also, it is possible that a 10-year follow-up was not long enough to see an effect on breast cancer mortality.

A meta-analysis of the effect of regular breast self-examination on breast cancer mortality or rates of advanced breast cancer (a marker of death) was performed on 20 observational studies and 3 clinical trials.⁹⁰ Bias and confounding may affect the results of studies of women with breast cancer who reported practicing self-examination before diagnosis. No difference in death rate was noted in studies of women who detected their cancers during self-examination (pooled relative risk [RR], 0.90; 95% CI, 0.72-1.12), and no mortality differences were noted in trials of training (pooled RR, 1.01; 95% CI, 0.92-1.12).

Although screen-film mammography and full-field digital mammography are the only imaging tools explicitly approved or grandfathered in for breast cancer screening by the US Food and Drug Administration (FDA), other modalities are under study.⁷³ Those approved by the FDA for diagnostic purposes (not screening) include MRI, ultrasound, scintimammography, thermography, and electrical impedance imaging. Although mammography uses x-ray and sonography uses sound waves to create images, MRI produces images from the combination of a strong magnetic field, radio waves, and computer processing (Figure 2).

Screening MRI may be helpful for women for whom mammography is not optimal, such as young women at substantially increased risk for breast cancer because of known BRCA1 or BRCA2 mutations. Available data are limited to studies of test characteristics in women at high risk (Table 4),^91- 98 and the impact on breast cancer mortality has not been determined. Both retrospective and prospective cohorts have been described. The small number of cases with breast cancer in these studies (the range among studies was 3 to 45 women) means that estimates of sensitivity were not precise. Nevertheless, every study reported higher sensitivity for breast MRI than for mammography, ultrasound, or both.^91- 98 The largest study reported on 1909 women at increased risk in the Netherlands, with 45 women diagnosed with cancer who had all screening examinations.⁹⁶ The sensitivity of clinical breast examination, mammography, and MRI in this study was 17.9%, 40%, and 71%, respectively. The overall discriminating capacity of MRI was significantly better compared with mammography as assessed by receiver operating characteristic curves (area under the curve 0.827 for MRI vs 0.686 for mammography).

Specificity of MRI tends to be lower than that of mammography; however, data are not consistently presented and specificity is not always easy to calculate. In the study of 1909 women in the Netherlands, the specificity of clinical breast examination, mammography, and MRI was 98.1%, 95.0%, and 89.8%, respectively,⁹⁶ and the authors noted that screening with MRI led to twice as many unneeded additional examinations (420 vs 207) and 3 times as many unneeded biopsies (24 vs 7) as did screening with mammography.⁹⁶ Warner et al⁹⁸ reported a substantial recall rate in the first round of MRI screening (26%), which decreased to 10% in the third round of MRI screening. Additional studies of MRI screening among high-risk women are under way.⁹⁹

Magnetic resonance imaging has not been studied in the general population as a screening tool, and the results from MRI screening of high-risk women may not apply to women at average risk. The high cost of MRI (approximately 10 times the cost of mammography) and its relatively low specificity (compared with mammography) probably prohibit its routine use for screening general populations. Also, MRI is time-consuming, requires intravenous contrast administration, and may be problematic for claustrophobic patients.

Ultrasound, frequently used as a targeted diagnostic examination focusing on a specific area of concern,¹⁰⁰ may help distinguish between cyst and solid masses and also between benign and malignant masses.¹⁰¹ Breast ultrasound data are available from diagnostic populations, with screening studies limited to women with dense breasts on mammography or at increased risk for breast cancer.¹⁰² Although ultrasound may detect 3 to 4 additional breast cancers per 1000 women in these increased-risk populations,^{93,100- 101,103- 108} there are no data on the use of screening ultrasound in the general population. Breast ultrasound has limitations as a potential screening tool because it requires a well-trained skilled operator. Examination techniques are not standardized, interpretation criteria are variable, and breast ultrasound does not consistently detect microcalcifications. Preliminary data suggest a higher rate of false-positive examination results with ultrasound than with mammography alone.^{100,104- 106} For example, the false-positive rate (based on solid lesion for ultrasound) ranged from 2.4% to 12.9% for ultrasound and 0.7% to 6% for mammography.¹⁰²

All groups recommend screening mammography for women aged 50 through 69 years.⁷ Within the United States, all recommend it for women in their 40s, but vary in the screening intervals recommended and encourage “informed decision making” with all women about the choice. Being older than 70 years should not preclude women from continuing to undergo screening; however, decisions regarding continued screening should include life expectancy and health status.^7,41 International policies differ with respect to the target age group to be screened, the intervals between screening, the number of mammographic views taken per breast, and the screening modalities recommended.⁴ Clinical breast examination is recommended by some,¹⁰⁹ but not all, groups. Most national groups no longer recommend breast self-examination, but some encourage women to become familiar with the contour of their own breasts.^42,109 Other imaging modalities, such as MRI and ultrasound, are not recommended for screening the general population. (See http://www.Guidelines.gov.)

The primary goal of breast cancer screening is to reduce subsequent breast cancer mortality through early detection. Theoretically this should translate into reduced morbidity from the disease.¹¹⁰ In addition, many women report feeling reassured by screening,¹¹¹ especially after having a so-called normal screen result.¹¹²

Possible harms include pain and discomfort, especially noted during compression of breast tissue during mammography. Compression of breast tissue reduces motion artifact and improves image quality. Reports of the level of discomfort, however, vary widely.^113- 115 Anxiety about screening is another concern.^116- 117Breast cancer screening yields both false-positive and false-negative results. False-positive results have been associated with anxiety, additional costs, and morbidity.^117- 118 After 10 years of annual screening in the United States, it is estimated that 1 in 2 women will have at least 1 false-positive mammogram result, and 1 in 5 women will have at least 1 false-positive clinical breast examination result.¹¹⁸ False-negative mammography examinations occur in approximately 20% to 40% of women with breast cancer.³⁶

Overdiagnosis and overtreatment of clinically insignificant disease is possible, especially ductal carcinoma in situ noted by mammography.^119- 120 Theoretical concerns about radiation-induced breast cancer from exposure to repeated mammography have been raised, but the potential benefits are thought to outweigh the risks. For example, the benefit-to-harm ratio is estimated to be 48.5 lives saved per 1 life lost due to radiation exposure.¹²¹ A mortality paradox has been noted in women aged 40 through 49 years, whereby increased mortality is noted among women screened for the first 3 to 10 years after the initiation of screening.^122- 124 Tumor dissemination after needle biopsy has also been suggested although the clinical significance is unclear.¹²⁵

Observer variability among radiologists who interpret mammography examinations has been noted both in a test situation and in community practices.^55- 57,126 A decision to perform a breast biopsy may depend heavily on the radiologist’s interpretation; therefore, interpretive variability can directly affect patient management.¹²⁶

The benefit-to-harm ratio of screening increases as women age because screening accuracy improves and prevalence of breast cancer increases. Younger women, however, have more potential years of life to be gained from screening.

Effective communication of information on benefits and harms is challenging. Multiple studies document inaccurate or incomplete comprehension of risk information, cognitive biases that affect how patients process risk information, and poor communication skills on the part of physicians.^127- 130 Use of frequencies with specific reference groups (“23 out of 1000 women your age”) instead of percentages (“0.23%”) may facilitate the comparison of small risks.^131- 133 Presentation of both positive (“23 in 1000 women your age will develop breast cancer”) and negative framing (“977 in 1000 women your age will not develop breast cancer”) can reduce biases in decision making.^130,132 Visual aids, such as bar graphs or pie charts, can increase the comprehension and saliency of information.¹³⁴ Risk-prediction models can be used to calculate the probability of being diagnosed with breast cancer (Box). Effective physician-patient partnership, for which gaps in comprehension are frequently assessed and resolved may have the greatest impact on patients’ understanding of information.¹³⁵

Reviews of breast cancer screening usually concentrate on the results of randomized trials of mammography to reduce breast cancer mortality. We have emphasized data on effectiveness in the community setting among the general population of women, which can often be different from the ideal setting of a randomized trial or the setting of a study among high-risk women. We have also emphasized the challenge of evaluating new screening modalities. Newer screening tests such as MRI and ultrasound have been studied in women at increased risk of breast cancer (eg, carriers of BRCA1 or BRCA2 mutations). None of the newer tests has been evaluated for its effect on breast cancer mortality in the general population and no data support screening the general population with these technologies. Careful evaluation of newer modalities in the populations for which they will be used is critical, especially since these modalities are usually more expensive than current approaches and the risk of increased false-positives is present. An overview of breast cancer screening modalities is shown in Table 5.

Most national groups recommend screening with mammography, with or without clinical breast examination, beginning at 40 years of age. Data on clinical breast examination as performed in the community suggest a lower level of cancer detection than would be anticipated from trials. Breast self-examination is no longer recommended by most expert groups. Limitations and potential harms have been identified for all existing screening tools. Quality control needs to be emphasized for established screening methods.