- Email: [email protected]
Will Supplemental Screening Ultrasound Increase Breast Cancer Overdiagnosis?
Will Supplemental Screening Ultrasound Increase Breast Cancer Overdiagnosis? Virginia M. Molleran, MD Overdiagnosis refers to the detection of cancers that would never come to light in a patient’s lifetime and are only identified by means of screening. Exactly how much overdiagnosis currently exists with screening mammography is uncertain. Because we do not know for certain which tumors would ultimately lead to death if left untreated and which would not, we cannot directly measure overdiagnosis and how best to estimate it is a matter of controversy. A conservative estimate of overdiagnosis with mammography would be on the order of 10%, but estimates have ranged as high as 54%. We know from multiple studies that ultrasound (US) screening mostly detects small, invasive, node-negative cancers; and in the ACRIN 6666 study, there was a greater tendency for US-only–detected tumors to be low grade than those detected with mammography. However, the population of patients undergoing screening US can be expected to differ from the average screening mammography population in that they will have higher breast density, they will be younger, and they may also have higher breast cancer risk than the population undergoing screening mammography. These factors may be associated with more aggressive tumors. There is no way to know whether we will be increasing overdiagnosis without performing a large randomized controlled study with very long-term follow-up. Even if some cancers are overdiagnosed with US, there will be a greater proportion of lethal breast cancers that are successfully treated because of screening US. The more important task is to learn how to correctly diagnose and appropriately treat nonlethal cancers. Key Words: Overdiagnosis; breast cancer; ultrasound. ªAUR, 2015
I
n 2012, the New England Journal of Medicine published an article by Bleyer and Welch (1) claiming that 31% of all breast cancers are ‘‘overdiagnosed’’ and that screening mammography has only a ‘‘small effect’’ on breast cancer mortality, opening the door to yet another heated debate over the benefit of screening mammography. Ironically, this new debate comes at a time when the medical community is faced with the prospect of an expansion of breast cancer screening in the form of whole-breast screening ultrasound (US) for women with dense breasts. The question that naturally follows is would there be an ultimate benefit from US? Or would it only result in needless cancer treatment? Stemming from the efforts of a breast cancer patient in Connecticut whose tumor was occult on mammography and diagnosed only after US, in 2009, Connecticut became the first state to pass legislation requiring that breast density be included in the mammography report and brought the issue of breast density and mammographic sensitivity into the public limelight. At this time, over half of the states have
Acad Radiol 2015; 22:967–972 From the Division of Breast Imaging, Department of Radiology, University of Cincinnati Medical Center, 234 Goodman St, Cincinnati, OH 45267-0761. Received September 17, 2014; accepted October 25, 2014. Address correspondence to: V.M.M. e-mail: [email protected] ªAUR, 2015 http://dx.doi.org/10.1016/j.acra.2014.10.012
followed suite and either passed or introduced legislation requiring breast density reporting. Similar legislation has also been introduced at the federal level. The laws are intended to make women aware of the limitations of mammography in dense breasts and to open up dialogue between women and their physicians regarding other supplemental screening options which include whole-breast US, magnetic resonance imaging (MRI), and digital breast tomosynthesis. The exact nature of the supplemental screening is not specified in the legislation, but the expectation is that the most women will seek US screening as it is less expensive and in general better tolerated than MRI. This is despite the continued recommendation by the American College of Radiology that only those at high risk for breast cancer (lifetime risk >20%) undergo supplementary screening and that supplementary screening should be performed with MRI rather than US, with US reserved for those high-risk women who cannot undergo MRI (2).
WHAT IS THE TRUE EXTENT OF OVERDIAGNOSIS WITH SCREENING MAMMOGRAPHY ALONE? Overdiagnosis refers to the detection of cancers that would never come to light in a patient’s lifetime and are only identified by means of screening. Therefore, a patient undergoes treatment with its associated harms but never derives the 967
MOLLERAN
benefit of additional life years gained. Overdiagnosis is not unique to breast cancer and is also implicated in other cancers, including lung, thyroid, and renal cancers. These ‘‘overdiagnosed’’ cancers are presumably small, slow-growing, nonpalpable tumors which lend themselves to detection with yearly screening, probably more so by repeated opportunities for detection than by imaging conspicuity. Proponents of overdiagnosis cite two factors in support of their argument (3): First, although there has been a significant increase in breast cancer incidence over the past 20 years, there has been no significant increase in mortality, and the observed decrease in mortality can be explained by better treatment alone. Second, there has been no significant change in the rate of diagnosis of late-stage cancers. In fact, according to Bleyer and Welch (1), the much greater absolute reduction in breast cancer mortality (20 per 100,000) compared to the reduction in the number of late-stage cancers (8 per 100,000) means that the effect of screening mammography is small. Although most experts agree that overdiagnosis exists, its extent is a matter of great debate. Because there is no way to tell exactly which cancers would never go on to be life threatening, the rate of cancer overdiagnosis has to be estimated indirectly from large-scale breast cancer screening and epidemiologic studies. The existence of overdiagnosis can be demonstrated by comparing the incidence in a screened to a nonscreened population. Assuming variables such as age and breast cancer risk are the same between the two populations, the cumulative incidence of breasts cancer should be the same between the two populations. If there is a difference in the cancer incidence in the two groups, it must be that the excess numbers (which usually occurs in the screened group) come from nonclinically important cancers in the screened group that were detected only because of screening and never would have surfaced clinically. Presumably, this excess number of cancers also exists in the control group but has not been detected, or as Zahl (4) suggests, may have spontaneously regressed. How accurate are estimates of overdiagnosis? Screening is expected to find most cancers at a preclinical stage, which causes lead time bias. In other words, in the short term, an increased incidence of cancer is expected in the screened population because cancers are found at an earlier, preclinical stage. If you then compare the two populations during a period after screening has stopped, you should see a reduction in the number of cancers in the screened group compared to the control group because many cancers that would have surfaced in the screened group during this period have already been diagnosed at an earlier age. This is referred to as a compensatory drop. With very long-term follow-up after screening is stopped, the cancer incidence in the two populations should equalize because all cancers developing in the two populations would now have been detected either by screening or clinical means. For this model to work, the follow-up period after cessation of screening has to be very long. Additionally, the control group should never be screened during or after the study period because this could 968
Academic Radiology, Vol 22, No 8, August 2015
lead to underestimation of overdiagnosis. Duffy and Parmar (5) have shown that you need at least 30 years of follow-up to nullify the effect of lead time. In reality, there is no study that meets these criteria. Even among 11 large randomized controlled trials, there are only three in which the control group was not offered screening at the termination of the study period (Malmo I and Canada I and II) (6). The follow-up period for the Malmo I trial was 12 years and that for the Canadian trails was only 6 years. How to compensate for the effect of lead time is probably the most hotly debated issue among experts on overdiagnosis. Because there is no study with a sufficient control population and follow-up period to nullify the effect of lead time, some compensation must be made. This can be done by making a statistical adjustment which essentially shifts the time of diagnosis among the screen-detected cancers to a later age (5). Furthermore, lead time itself is an estimated quantity and almost certainly highly variable among cancers. What serves as a control population is at issue as well. This is particularly true if data from service screening studies are being used because a contemporary nonscreened population may not be available and data from an earlier period before the start of screening have to be used. This means the background incidence of breast cancer in the two populations will not necessarily be the same (since the incidence of breast cancer increased during the 1900’s) and will also have to be estimated (5). Other issues in the determination of overdiagnosis pertain to the source of the data: Is it better to use data from the large randomized controlled studies or from more recent service screening studies which had the benefit of more modern equipment and practices. These latter studies would be more reflective of the current environment of screening mammography. Whether DCIS is included in the estimation will have an effect on the result as well, being higher if ductal carcinoma in situ (DCIS) is included. Whether the prevalence round of screening is included is another factor because most overdiagnosed cancers would be expected to be found in this round. The screening interval will also have an effect, with more overdiagnosis expected to occur with more frequent screening. Finally, there is no standard format to express overestimation. It is frequently expressed as a ratio with the numerator being the number of overdiagnosed cancers, but the denominator may be the total number of cancers in the population or only the number of screen-detected cancers (7). There have been multiple published reports regarding breast cancer overdiagnosis but two somewhat more controversial reports have been published by Jorgenson and Gotzsche in 2009 (8) and Bleyer and Welch in 2012 (1). Jorgensen and Gotzsche reviewed data from multiple countries and observed that the compensatory drop in cancer cases in older patients above the age of screening was not enough to account for the excess incidence seen during screening. They concluded that 52% of cancers in the screened group were overdiagnosed. This report has been criticized heavily for methodologic flaws (9). More recently,
Academic Radiology, Vol 22, No 8, August 2015
US SCREENING AND BREAST CANCER OVERDIAGNOSIS
Bleyer and Welch looked at SEER data (Surveillance, Epidemiology, and End Results) from 1976 to 1978 which was before widespread screening and compared incidence of cancer with that seen from 2006 to 2008, a period during which approximately 60% of women were being screened (10). Assuming a 0.25% per year increase in the background incidence of breast cancer for the control population, they found a 31% greater than expected incidence of CA in the latter population based on this estimate of background increase. Flaws have also been found with this analysis as well (10,11), primarily related their estimation of the background incidence of cancer. A more conservative estimate was provided in a review by Duffy and Puliti (7). They reviewed 13 service screening studies across Europe. When adjustment for lead time bias was absent or incomplete, estimates of overdiagnosis ranged from 0% to 54%. When appropriate adjustments were made for breast cancer risk and lead time, estimates ranged from 1% to 10%. A review performed by an independent panel in the United Kingdom (6) estimated an overdiagnosis rate of 19% of screen-detected cancers.
associated axillary lymph node metastases versus 33% of those detected with mammography alone. A major limitation of this study is that there was no control group who did not undergo US screening so the impact on mortality and clinical outcomes could not be assessed. It should be noted that the population of women in this study is demographically dissimilar to the general screening mammography population. The median patient age at enrollment in this study was 55, with 29% of women in the study aged <50 years. Over 50% had a personal history of breast cancer. The ACRIN 666 study included only women with dense breasts and increased breast cancer risk—thus a population expected to derive maximal benefit from the addition of supplemental screening US, and indeed, there was a slightly higher rate of cancer detection with US alone in the ACRIN 6666 study compared to other studies of screening US. In widespread screening, the performance would be expected to be slightly lower. Accordingly, two reports of screening US from Connecticut after passage of the breast density notification law showed cancer detection rates of 0.18% (14) and 0.32% (15), respectively. In the latter study, the three malignant lesions detected with US alone were all <1 cm in size and node negative. In the former study, tumor size ranged from 4 to 15 mm (mean, 9.7 mm) and sentinel lymph node biopsy was negative in seven, positive in three, and not reported in one. Interestingly, in both studies, only a fraction (30% and 16%, respectively) of the women invited to US screening (those with heterogeneously or extremely dense tissue) actually returned for screening US, and less than half of those who underwent the first round of screening US returned for screening US the following year. Although the vast majority of cancers detected with US are invasive, US can detect DCIS as well. On US, DCIS can present as a mass, ductal abnormality or calcifications alone. Whether additional DCIS cases detected only with US would be clinically important remains to be seen. Although US can certainly detect both high- and low-grade DCIS (16,17), there is very little data regarding the characteristics of DCIS detected with US alone. Of the US-only–detected cancers in the ACRIN study, only 6.2% were noninvasive. The smaller proportion of DCIS seen with US may suggest that there would be fewer overdiagnosed cancers among USdetected lesions because the proportion of DCIS detected with mammography is approximately 20% versus only 6.2% in the ACRIN study, and DCIS would almost certainly make up a not-insignificant portion of overdiagnosed cases. Other recent multi-institutional studies have also shown low rates of DCIS detection of 0.27% (18) and 4.3% (19), respectively. Of the invasive cancers detected with US alone in the ACRIN 6666 study, low-grade tumors were disproportionally represented, making up almost half (45.8%) of all invasive ductal carcinomas but only representing 17.6% of those detected with mammography alone and 25% of those detected with both mammography and US. Tumor grade is not consistently reported in other studies, but in another large study
DOES MORE IMAGING NECESSARILY MEAN MORE OVERDIAGNOSIS? With advancement in screening imaging technology or addition of a new imaging modality for screening, there would be a natural expectation to detect more lesions. Would an increase in the number of detected lesions necessarily lead to an increase in the rate of overdiagnosis? It would be expected that the additional lesions detected after technological improvements would be small, less-conspicuous lesions, but whether these are clinically important lesions is unknown. We know from several studies of screening US (12) that the expected additional cancer detection rate with US would be 3.5 cancers per 1000 women screened. Furthermore, most (94%) of the US-detected cancers are small (#1.0 cm), invasive, node-negative cancers. These results were corroborated by the ACRIN 6666 study (13), which involved 2662 women with heterogeneously or extremely dense breasts and at least one other risk factor for breast cancer. Twenty-one institutions were involved in this study. Three annual screening rounds were performed with both mammography and US and interpreted by separate radiologists blinded to the results of the other study. During the prevalence round of this study, the incremental cancer detection rate with US alone was 5.3 per 1000. Subsequent incidence rounds showed an incremental cancer detection rate of 3.7 per 1000 with US alone. Average tumor size detected with US alone was small (10 mm) and slightly smaller than that detectable with mammography alone (11.5 mm). Of 32 cancers detected with US alone, only two were DCIS (one high grade and one intermediate grade). Of the 30 invasive cancers, five were invasive lobular carcinoma and one was mixed invasive and lobular carcinoma; the rest were invasive ductal carcinomas. Only one cancer (4%) detected with US alone had
969
MOLLERAN
using automated whole-breast US in 4419 women at eight facilities in the United States (19), the proportion of low-grade tumors detected with US alone was slightly lower than the proportion detected with both US and mammography (38.1% vs. 38.5%). The presence or absence of axillary lymph node metastases is the most important prognostic factor in breast cancer, and tumor size is also a major determinant of prognosis. Tumor grade is also an important factor, but because of the somewhat subjective nature of the assessment and the multiplicity of grading systems in use, it may not be as reliable as size and nodal status (20). The results of the screening US studies therefore suggest that most cancers detected with US performed as a supplement to screening mammography would have a good prognosis. However, it should not be assumed that these cancers are overdiagnosed because there is no nonscreened control group for comparison. Presumably, at least some of these US-only tumors would have either presented later as interval cancers or mammographically detected cancers at a larger size and potentially later stage. Interval cancers are cancers presenting after a normal screening mammogram and before the next screening examination is due. They usually present because of clinical symptoms and therefore cannot be considered to be overdiagnosed. Interval cancers may be cancers that are not detected mammographically because of masking by dense tissue, but they may also represent more aggressive cancers that have grown rapidly since the last screening. As such, more interval cancers are expected in dense breasts. One study demonstrated similar interval cancer rates between a population of women with dense breasts who underwent supplemental US screening and a population of women with nondense breasts who underwent mammographic screening alone (21). The authors conclude that screening US is beneficial in detecting clinically important cancers because more interval cancers would be expected in the population with dense breasts. The relatively high proportion of low-grade tumors in the ACRIN 6666 study may merely reflect an expected increase in detection of indolent tumors with the incidence screening round. Only three screening rounds were performed in the ACRIN study which is relatively few. Based on current recommendations, the average women would be expected to undergo at least 12 screening rounds. Most indolent tumors would be expected to be found in early rounds with later rounds being more likely to show more aggressive tumors. No other studies report the results of incidence rounds separately. It is possible that we would see a greater proportion of non–low-grade tumors if we looked at later screening rounds. A major limitation of the ACRIN study is that there was no control group who did not undergo US screening, and so far, there have been no published studies evaluating the impact of US screening on breast cancer mortality, so the impact on mortality and clinical outcomes could not be assessed. As such, there is no way to know how many of these cancers would have been lethal. There is an ongoing, large, randomized, controlled trial of screening US in Japan, but 970
Academic Radiology, Vol 22, No 8, August 2015
the results of this study may not be generalizable to US women because it involves younger women (40–49 years) with a 2-year screening interval (12). Also, the distribution of breast density across the population differs for Asian women and the peak age of breast cancer diagnosis occurs in the late 40s (22). ARE THERE FEATURES OF THE PATIENT POPULATION THAT WOULD BE SCREENED THAT MAY LEAD TO A HIGH INCIDENCE OF MORE AGGRESSIVE CANCERS? The population of women who would undergo screening US may be demographically different from the population undergoing. The legislation regarding breast density requires radiologists to directly notify women when their breast tissue is dense. The term ‘‘dense’’ includes those with heterogeneously dense and extremely dense breasts (23). Thus, it can be expected that most women undergoing screening US will have dense breasts. It follows from this that they will on average be younger than the general screening population because breast density decreases with age. Furthermore, there may be a greater tendency for high-risk women to choose supplemental screening compared with average-risk women. In one study performed in Connecticut after passage of the breast density legislation, 9% of the study population was at high risk, 16% was at intermediate risk, and 66% were low or average risk. Breast cancer risk was unknown in 9% (15). It is well known that increased breast density lowers the sensitivity of mammography, and we know from multiple screening studies that additional cancers can be detected with screening US, mostly in those with dense breasts. This is largely due to a masking effect: normal breast tissue can obscure cancers. However, breast density is an independent risk factor for cancer, and therefore, women with dense breasts are not only more likely to have their breast cancer missed by mammography but also at higher risk of cancer than those with nondense breasts. The contribution of breast density to breast cancer risk is also a matter of debate. Approximately, 40% of US women are judged to have heterogeneously dense breasts (50%–75% of the breast consisting of fibroglandular tissue) and 10% are judged to have extremely dense breasts (75%–100% fibroglandular tissue). As others have noted (24), if high breast density imparts a high risk of breasts cancers, then 50% of all screen-eligible women would be at high risk of breast cancer simply on the basis of high breast density. However, if you compare those with heterogeneously dense breasts to those with average breast density, the risk is much lower, 1.2 times greater, and the risk for those with extremely dense breasts compared to those with average breast density would be 2.1 times greater (2). Are cancers in women with dense breasts likely to be more aggressive? It has been established that tumors seen in BRCA1 carriers are more likely to be grade III and of the medullary subtype than sporadic cancers (25), but there are little data on association of other risk factors and histologic features. Reports on the associations of breast density with prognostic
Academic Radiology, Vol 22, No 8, August 2015
US SCREENING AND BREAST CANCER OVERDIAGNOSIS
factors such as tumor grade, estrogen receptor (ER) status, and HER2 status have been inconclusive (26,27), although one recent study did show an association between breast density and high-grade tumors, ER-negative tumors, and in situ versus invasive tumors (27). Breast density has also been associated with an increased risk of local and locoregional recurrence (28) but not with distant metastasis or survival. This suggests that the increased risk of recurrence may be attributable more to incomplete identification of the entire extent of disease at the time of diagnosis secondary to the masking effect than to tumor aggressiveness. Breast cancers in young women tend to be biologically different from those in older women. Women aged <40 years with breast cancer tend to have a worse prognosis, in part because of later stage at diagnosis but also because of more aggressive features such as high grade, hormone-receptor negative status, high proliferation fraction, and lymphovascular invasion (29).
entiate life-threatening from non–life-threatening cancers and to promote more appropriate treatment. Although there is no current way to definitively distinguish the two, genomic tests such as Mammaprint and Oncotype DX offer promise. Once appropriately characterized, more tailored treatment options can be offered. There may be some cases in which the best treatment option may be simply no treatment and more of a ‘‘wait-and-see’’ approach, as has been done with prostate cancer. Certainly, we need to inform our patients that the breast cancer that we detect with screening may not ultimately be a life-threatening disease, but we also need to make sure they understand that detecting a small cancer with mammography or US could save their lives.
BY ADDING SCREENING US, ARE WE LIKELY TO INCREASE THE RATE OF OVERDIAGNOSIS? In summary, exactly how much overdiagnosis currently exists with screening mammography is uncertain, but certainly it exists to some extent. Because we do not know for certain which tumors would ultimately lead to death if left untreated and which would not, we cannot directly measure overdiagnosis. It has to be estimated from large-scale studies of breast cancer screening, but there are no screening trials with appropriate design to estimate overdiagnosis and it is very unlikely that any such study would be performed in the future. Overdiagnosis can be estimated from available data, but this requires other estimations and assumptions that increase the degree of uncertainty. Tumor size, grade, lymph node status, and other histologic features could be used as surrogates for cancer aggression, but this would be an imprecise measure because, even among small, low-grade, node-negative cancers, there is no good way currently to differentiate potentially lifethreatening cancers and non–life-threatening ones. It should not be assumed that simply finding more cancers would increase the overdiagnosis. The results of the ACRIN study would suggest a trend toward a greater proportion of lowgrade tumors diagnosed with US alone, but with only 32 cancers diagnosed with US alone, the numbers from this study are relatively small. Features of the population that would be screened may suggest an underlying tendency to more aggressive tumors, but this is also unclear. The more important question is ‘‘What should be done about it?’’ Contrary to some opinions, the potential harm of overdiagnosis should not be considered a reason to discontinue screening. Even if 30% of women diagnosed with breast cancer through screening are overdiagnosed, 70% of detected cancers are still deemed worthy of potentially life-saving treatment. As has been proposed by Gur and Sumkin (30), the solution is not to avoid detection but to work toward methods to differ-
REFERENCES 1. Bleyer A, Welch HG. Effect of three decades of screening mammography on breast cancer incidence. N Engl J Med 2012; 367(21):1998–2005. 2. Price ER, Hargreaves J, Lipson JA, et al. The California breast density information group: a collaborative response to the issues of breast density, breast cancer risk, and breast density notification legislation. Radiology 2013; 269(3):887–892. 3. Ceugnart L, Deghaye M, Vennin P, et al. Organized breast screening: answers to recurring controversies. Diagn Interv Imaging 2014; 95(4): 355–359. 4. Zahl PH, Maehlen J, Welch HG. The natural history of invasive breast cancers detected by screening mammography. Arch Intern Med 2008; 168(21):2311–2316. 5. Duffy SW, Parmar D. Overdiagnosis in breast cancer screening: the importance of length of observation period and lead time. Breast Cancer Res 2013; 15(3):R41. 6. Independent UK Panel on Breast Cancer Screening. The benefits and harms of breast cancer screening: an independent review. Lancet 2012; 380(9855):1778–1786. 7. Puliti D, Duffy SW, Miccinesi G, et al., EUROSCREEN Working Group. Overdiagnosis in mammographic screening for breast cancer in Europe: a literature review. J Med Screen 2012; 19(Suppl 1):42–56. 8. Jørgensen KJ, Gøtzsche PC. Overdiagnosis in publicly organised mammography screening programmes: systematic review of incidence trends. BMJ 2009; 339:b2587. 9. Kopans DB, Smith RA, Duffy SW. Mammographic screening and ‘‘overdiagnosis’’. Radiology 2011; 260(3):616–620. 10. Feig SA. Pitfalls in accurate estimation of overdiagnosis: implications for screening policy and compliance. Breast Cancer Res 2013; 15(4):105. 11. Kopans DB. Point: The New England Journal of Medicine article suggesting overdiagnosis from mammography screening is scientifically incorrect and should be withdrawn. J Am Coll Radiol 2013; 10(5):317–319. 12. Merry GM, Mendelson EB. Update on screening breast ultrasonography. Radiol Clin North Am 2014; 52(3):527–537. 13. Berg WA, Zhang Z, Lehrer D, et al., ACRIN 6666 Investigators, . Detection of breast cancer with addition of annual screening ultrasound or a single screening MRI to mammography in women with elevated breast cancer risk. JAMA 2012; 307(13):1394–1404. 14. Parris T, Wakefield D, Frimmer H. Real world performance of screening breast ultrasound following enactment of Connecticut Bill 458. Breast J 2013; 19(1):64–70. 15. Hooley RJ, Greenberg KL, Stackhouse RM, et al. Screening US in patients with mammographically dense breasts: initial experience with Connecticut Public Act 09-41. Radiology 2012; 265(1):59–69. 16. Rauch GM, Kuerer HM, Scoggins ME, et al. Clinicopathologic, mammographic, and sonographic features in 1,187 patients with pure ductal carcinoma in situ of the breast by estrogen receptor status. Breast Cancer Res Treat 2013; 139(3):639–647. 17. Wang LC, Sullivan M, Du H, et al. US appearance of ductal carcinoma in situ. Radiographics 2013; 33(1):213–228. 18. Corsetti V, Houssami N, Ferrari A, et al. Breast screening with ultrasound in women with mammography-negative dense breasts: evidence on
971
MOLLERAN
19.
20.
21.
22.
23.
24.
incremental cancer detection and false positives, and associated cost. Eur J Cancer 2008; 44(4):539–544. Kelly KM, Dean J, Lee SJ, et al. Breast cancer detection: radiologists’ performance using mammography with and without automated wholebreast ultrasound. Eur Radiol 2010; 20(11):2557–2564. Chang JC, Hilsenbeck SG. Prognostic and predictive markers. In: Harris JR, Lippman ME, Morrow M, et al., eds. Diseases of the breast. Philadelphia PA: Lippincitt Williams & Wilkins, 2010; 443–457. Corsetti V, Houssami N, Ghirardi M, et al. Evidence of the effect of adjunct ultrasound screening in women with mammography-negative dense breasts: interval breast cancers at 1 year follow-up. Eur J Cancer 2011; 47(7):1021–1026. Chae EY, Kim HH, Cha JH, et al. Evaluation of screening whole-breast sonography as a supplemental tool in conjunction with mammography in women with dense breasts. J Ultrasound Med 2013; 32(9):1573–1578. Ho JM, Jafferjee N, Covarrubias GM, et al. Dense breasts: a review of reporting legislation and available supplemental screening options. AJR Am J Roentgenol 2014; 203(2):449–456. D’Orsi CJ, Sickles EA. To seek perfection or not? That is the question. Radiology 2012; 265(1):9–11.
972
Academic Radiology, Vol 22, No 8, August 2015
25. Liebens FP, Carly B, Pastijn A, et al. Management of BRCA1/2 associated breast cancer: a systematic qualitative review of the state of knowledge in 2006. Eur J Cancer 2007; 43(2):238–257. 26. Eriksson L, Czene K, Rosenberg L, et al. The influence of mammographic density on breast tumor characteristics. Breast Cancer Res Treat 2012; 134(2):859–866. 27. Yaghjyan L, Colditz GA, Collins LC, et al. Mammographic breast density and subsequent risk of breast cancer in postmenopausal women according to tumor characteristics. J Natl Cancer Inst 2011; 103(15): 1179–1189. 28. Wang AT, Vachon CM, Brandt KR, et al. Breast density and breast cancer risk: a practical review. Mayo Clin Proc 2014; 89(4):548–557. 29. Partridge AH, Goldhirsch A, Gelber S, et al. Breast cancer in younger women. In: Harris JR, Lippman ME, Morrow M, et al., eds. Diseases of the breast. Philadelphia PA: Lippincitt Williams & Wilkins, 2010; 1073–1082. 30. Gur D, Sumkin JH. Screening for early detection of breast cancer: overdiagnosis versus suboptimal patient management. Radiology 2013; 268(2):327–328.
I
n 2012, the New England Journal of Medicine published an article by Bleyer and Welch (1) claiming that 31% of all breast cancers are ‘‘overdiagnosed’’ and that screening mammography has only a ‘‘small effect’’ on breast cancer mortality, opening the door to yet another heated debate over the benefit of screening mammography. Ironically, this new debate comes at a time when the medical community is faced with the prospect of an expansion of breast cancer screening in the form of whole-breast screening ultrasound (US) for women with dense breasts. The question that naturally follows is would there be an ultimate benefit from US? Or would it only result in needless cancer treatment? Stemming from the efforts of a breast cancer patient in Connecticut whose tumor was occult on mammography and diagnosed only after US, in 2009, Connecticut became the first state to pass legislation requiring that breast density be included in the mammography report and brought the issue of breast density and mammographic sensitivity into the public limelight. At this time, over half of the states have
Acad Radiol 2015; 22:967–972 From the Division of Breast Imaging, Department of Radiology, University of Cincinnati Medical Center, 234 Goodman St, Cincinnati, OH 45267-0761. Received September 17, 2014; accepted October 25, 2014. Address correspondence to: V.M.M. e-mail: [email protected] ªAUR, 2015 http://dx.doi.org/10.1016/j.acra.2014.10.012
followed suite and either passed or introduced legislation requiring breast density reporting. Similar legislation has also been introduced at the federal level. The laws are intended to make women aware of the limitations of mammography in dense breasts and to open up dialogue between women and their physicians regarding other supplemental screening options which include whole-breast US, magnetic resonance imaging (MRI), and digital breast tomosynthesis. The exact nature of the supplemental screening is not specified in the legislation, but the expectation is that the most women will seek US screening as it is less expensive and in general better tolerated than MRI. This is despite the continued recommendation by the American College of Radiology that only those at high risk for breast cancer (lifetime risk >20%) undergo supplementary screening and that supplementary screening should be performed with MRI rather than US, with US reserved for those high-risk women who cannot undergo MRI (2).
WHAT IS THE TRUE EXTENT OF OVERDIAGNOSIS WITH SCREENING MAMMOGRAPHY ALONE? Overdiagnosis refers to the detection of cancers that would never come to light in a patient’s lifetime and are only identified by means of screening. Therefore, a patient undergoes treatment with its associated harms but never derives the 967
MOLLERAN
benefit of additional life years gained. Overdiagnosis is not unique to breast cancer and is also implicated in other cancers, including lung, thyroid, and renal cancers. These ‘‘overdiagnosed’’ cancers are presumably small, slow-growing, nonpalpable tumors which lend themselves to detection with yearly screening, probably more so by repeated opportunities for detection than by imaging conspicuity. Proponents of overdiagnosis cite two factors in support of their argument (3): First, although there has been a significant increase in breast cancer incidence over the past 20 years, there has been no significant increase in mortality, and the observed decrease in mortality can be explained by better treatment alone. Second, there has been no significant change in the rate of diagnosis of late-stage cancers. In fact, according to Bleyer and Welch (1), the much greater absolute reduction in breast cancer mortality (20 per 100,000) compared to the reduction in the number of late-stage cancers (8 per 100,000) means that the effect of screening mammography is small. Although most experts agree that overdiagnosis exists, its extent is a matter of great debate. Because there is no way to tell exactly which cancers would never go on to be life threatening, the rate of cancer overdiagnosis has to be estimated indirectly from large-scale breast cancer screening and epidemiologic studies. The existence of overdiagnosis can be demonstrated by comparing the incidence in a screened to a nonscreened population. Assuming variables such as age and breast cancer risk are the same between the two populations, the cumulative incidence of breasts cancer should be the same between the two populations. If there is a difference in the cancer incidence in the two groups, it must be that the excess numbers (which usually occurs in the screened group) come from nonclinically important cancers in the screened group that were detected only because of screening and never would have surfaced clinically. Presumably, this excess number of cancers also exists in the control group but has not been detected, or as Zahl (4) suggests, may have spontaneously regressed. How accurate are estimates of overdiagnosis? Screening is expected to find most cancers at a preclinical stage, which causes lead time bias. In other words, in the short term, an increased incidence of cancer is expected in the screened population because cancers are found at an earlier, preclinical stage. If you then compare the two populations during a period after screening has stopped, you should see a reduction in the number of cancers in the screened group compared to the control group because many cancers that would have surfaced in the screened group during this period have already been diagnosed at an earlier age. This is referred to as a compensatory drop. With very long-term follow-up after screening is stopped, the cancer incidence in the two populations should equalize because all cancers developing in the two populations would now have been detected either by screening or clinical means. For this model to work, the follow-up period after cessation of screening has to be very long. Additionally, the control group should never be screened during or after the study period because this could 968
Academic Radiology, Vol 22, No 8, August 2015
lead to underestimation of overdiagnosis. Duffy and Parmar (5) have shown that you need at least 30 years of follow-up to nullify the effect of lead time. In reality, there is no study that meets these criteria. Even among 11 large randomized controlled trials, there are only three in which the control group was not offered screening at the termination of the study period (Malmo I and Canada I and II) (6). The follow-up period for the Malmo I trial was 12 years and that for the Canadian trails was only 6 years. How to compensate for the effect of lead time is probably the most hotly debated issue among experts on overdiagnosis. Because there is no study with a sufficient control population and follow-up period to nullify the effect of lead time, some compensation must be made. This can be done by making a statistical adjustment which essentially shifts the time of diagnosis among the screen-detected cancers to a later age (5). Furthermore, lead time itself is an estimated quantity and almost certainly highly variable among cancers. What serves as a control population is at issue as well. This is particularly true if data from service screening studies are being used because a contemporary nonscreened population may not be available and data from an earlier period before the start of screening have to be used. This means the background incidence of breast cancer in the two populations will not necessarily be the same (since the incidence of breast cancer increased during the 1900’s) and will also have to be estimated (5). Other issues in the determination of overdiagnosis pertain to the source of the data: Is it better to use data from the large randomized controlled studies or from more recent service screening studies which had the benefit of more modern equipment and practices. These latter studies would be more reflective of the current environment of screening mammography. Whether DCIS is included in the estimation will have an effect on the result as well, being higher if ductal carcinoma in situ (DCIS) is included. Whether the prevalence round of screening is included is another factor because most overdiagnosed cancers would be expected to be found in this round. The screening interval will also have an effect, with more overdiagnosis expected to occur with more frequent screening. Finally, there is no standard format to express overestimation. It is frequently expressed as a ratio with the numerator being the number of overdiagnosed cancers, but the denominator may be the total number of cancers in the population or only the number of screen-detected cancers (7). There have been multiple published reports regarding breast cancer overdiagnosis but two somewhat more controversial reports have been published by Jorgenson and Gotzsche in 2009 (8) and Bleyer and Welch in 2012 (1). Jorgensen and Gotzsche reviewed data from multiple countries and observed that the compensatory drop in cancer cases in older patients above the age of screening was not enough to account for the excess incidence seen during screening. They concluded that 52% of cancers in the screened group were overdiagnosed. This report has been criticized heavily for methodologic flaws (9). More recently,
Academic Radiology, Vol 22, No 8, August 2015
US SCREENING AND BREAST CANCER OVERDIAGNOSIS
Bleyer and Welch looked at SEER data (Surveillance, Epidemiology, and End Results) from 1976 to 1978 which was before widespread screening and compared incidence of cancer with that seen from 2006 to 2008, a period during which approximately 60% of women were being screened (10). Assuming a 0.25% per year increase in the background incidence of breast cancer for the control population, they found a 31% greater than expected incidence of CA in the latter population based on this estimate of background increase. Flaws have also been found with this analysis as well (10,11), primarily related their estimation of the background incidence of cancer. A more conservative estimate was provided in a review by Duffy and Puliti (7). They reviewed 13 service screening studies across Europe. When adjustment for lead time bias was absent or incomplete, estimates of overdiagnosis ranged from 0% to 54%. When appropriate adjustments were made for breast cancer risk and lead time, estimates ranged from 1% to 10%. A review performed by an independent panel in the United Kingdom (6) estimated an overdiagnosis rate of 19% of screen-detected cancers.
associated axillary lymph node metastases versus 33% of those detected with mammography alone. A major limitation of this study is that there was no control group who did not undergo US screening so the impact on mortality and clinical outcomes could not be assessed. It should be noted that the population of women in this study is demographically dissimilar to the general screening mammography population. The median patient age at enrollment in this study was 55, with 29% of women in the study aged <50 years. Over 50% had a personal history of breast cancer. The ACRIN 666 study included only women with dense breasts and increased breast cancer risk—thus a population expected to derive maximal benefit from the addition of supplemental screening US, and indeed, there was a slightly higher rate of cancer detection with US alone in the ACRIN 6666 study compared to other studies of screening US. In widespread screening, the performance would be expected to be slightly lower. Accordingly, two reports of screening US from Connecticut after passage of the breast density notification law showed cancer detection rates of 0.18% (14) and 0.32% (15), respectively. In the latter study, the three malignant lesions detected with US alone were all <1 cm in size and node negative. In the former study, tumor size ranged from 4 to 15 mm (mean, 9.7 mm) and sentinel lymph node biopsy was negative in seven, positive in three, and not reported in one. Interestingly, in both studies, only a fraction (30% and 16%, respectively) of the women invited to US screening (those with heterogeneously or extremely dense tissue) actually returned for screening US, and less than half of those who underwent the first round of screening US returned for screening US the following year. Although the vast majority of cancers detected with US are invasive, US can detect DCIS as well. On US, DCIS can present as a mass, ductal abnormality or calcifications alone. Whether additional DCIS cases detected only with US would be clinically important remains to be seen. Although US can certainly detect both high- and low-grade DCIS (16,17), there is very little data regarding the characteristics of DCIS detected with US alone. Of the US-only–detected cancers in the ACRIN study, only 6.2% were noninvasive. The smaller proportion of DCIS seen with US may suggest that there would be fewer overdiagnosed cancers among USdetected lesions because the proportion of DCIS detected with mammography is approximately 20% versus only 6.2% in the ACRIN study, and DCIS would almost certainly make up a not-insignificant portion of overdiagnosed cases. Other recent multi-institutional studies have also shown low rates of DCIS detection of 0.27% (18) and 4.3% (19), respectively. Of the invasive cancers detected with US alone in the ACRIN 6666 study, low-grade tumors were disproportionally represented, making up almost half (45.8%) of all invasive ductal carcinomas but only representing 17.6% of those detected with mammography alone and 25% of those detected with both mammography and US. Tumor grade is not consistently reported in other studies, but in another large study
DOES MORE IMAGING NECESSARILY MEAN MORE OVERDIAGNOSIS? With advancement in screening imaging technology or addition of a new imaging modality for screening, there would be a natural expectation to detect more lesions. Would an increase in the number of detected lesions necessarily lead to an increase in the rate of overdiagnosis? It would be expected that the additional lesions detected after technological improvements would be small, less-conspicuous lesions, but whether these are clinically important lesions is unknown. We know from several studies of screening US (12) that the expected additional cancer detection rate with US would be 3.5 cancers per 1000 women screened. Furthermore, most (94%) of the US-detected cancers are small (#1.0 cm), invasive, node-negative cancers. These results were corroborated by the ACRIN 6666 study (13), which involved 2662 women with heterogeneously or extremely dense breasts and at least one other risk factor for breast cancer. Twenty-one institutions were involved in this study. Three annual screening rounds were performed with both mammography and US and interpreted by separate radiologists blinded to the results of the other study. During the prevalence round of this study, the incremental cancer detection rate with US alone was 5.3 per 1000. Subsequent incidence rounds showed an incremental cancer detection rate of 3.7 per 1000 with US alone. Average tumor size detected with US alone was small (10 mm) and slightly smaller than that detectable with mammography alone (11.5 mm). Of 32 cancers detected with US alone, only two were DCIS (one high grade and one intermediate grade). Of the 30 invasive cancers, five were invasive lobular carcinoma and one was mixed invasive and lobular carcinoma; the rest were invasive ductal carcinomas. Only one cancer (4%) detected with US alone had
969
MOLLERAN
using automated whole-breast US in 4419 women at eight facilities in the United States (19), the proportion of low-grade tumors detected with US alone was slightly lower than the proportion detected with both US and mammography (38.1% vs. 38.5%). The presence or absence of axillary lymph node metastases is the most important prognostic factor in breast cancer, and tumor size is also a major determinant of prognosis. Tumor grade is also an important factor, but because of the somewhat subjective nature of the assessment and the multiplicity of grading systems in use, it may not be as reliable as size and nodal status (20). The results of the screening US studies therefore suggest that most cancers detected with US performed as a supplement to screening mammography would have a good prognosis. However, it should not be assumed that these cancers are overdiagnosed because there is no nonscreened control group for comparison. Presumably, at least some of these US-only tumors would have either presented later as interval cancers or mammographically detected cancers at a larger size and potentially later stage. Interval cancers are cancers presenting after a normal screening mammogram and before the next screening examination is due. They usually present because of clinical symptoms and therefore cannot be considered to be overdiagnosed. Interval cancers may be cancers that are not detected mammographically because of masking by dense tissue, but they may also represent more aggressive cancers that have grown rapidly since the last screening. As such, more interval cancers are expected in dense breasts. One study demonstrated similar interval cancer rates between a population of women with dense breasts who underwent supplemental US screening and a population of women with nondense breasts who underwent mammographic screening alone (21). The authors conclude that screening US is beneficial in detecting clinically important cancers because more interval cancers would be expected in the population with dense breasts. The relatively high proportion of low-grade tumors in the ACRIN 6666 study may merely reflect an expected increase in detection of indolent tumors with the incidence screening round. Only three screening rounds were performed in the ACRIN study which is relatively few. Based on current recommendations, the average women would be expected to undergo at least 12 screening rounds. Most indolent tumors would be expected to be found in early rounds with later rounds being more likely to show more aggressive tumors. No other studies report the results of incidence rounds separately. It is possible that we would see a greater proportion of non–low-grade tumors if we looked at later screening rounds. A major limitation of the ACRIN study is that there was no control group who did not undergo US screening, and so far, there have been no published studies evaluating the impact of US screening on breast cancer mortality, so the impact on mortality and clinical outcomes could not be assessed. As such, there is no way to know how many of these cancers would have been lethal. There is an ongoing, large, randomized, controlled trial of screening US in Japan, but 970
Academic Radiology, Vol 22, No 8, August 2015
the results of this study may not be generalizable to US women because it involves younger women (40–49 years) with a 2-year screening interval (12). Also, the distribution of breast density across the population differs for Asian women and the peak age of breast cancer diagnosis occurs in the late 40s (22). ARE THERE FEATURES OF THE PATIENT POPULATION THAT WOULD BE SCREENED THAT MAY LEAD TO A HIGH INCIDENCE OF MORE AGGRESSIVE CANCERS? The population of women who would undergo screening US may be demographically different from the population undergoing. The legislation regarding breast density requires radiologists to directly notify women when their breast tissue is dense. The term ‘‘dense’’ includes those with heterogeneously dense and extremely dense breasts (23). Thus, it can be expected that most women undergoing screening US will have dense breasts. It follows from this that they will on average be younger than the general screening population because breast density decreases with age. Furthermore, there may be a greater tendency for high-risk women to choose supplemental screening compared with average-risk women. In one study performed in Connecticut after passage of the breast density legislation, 9% of the study population was at high risk, 16% was at intermediate risk, and 66% were low or average risk. Breast cancer risk was unknown in 9% (15). It is well known that increased breast density lowers the sensitivity of mammography, and we know from multiple screening studies that additional cancers can be detected with screening US, mostly in those with dense breasts. This is largely due to a masking effect: normal breast tissue can obscure cancers. However, breast density is an independent risk factor for cancer, and therefore, women with dense breasts are not only more likely to have their breast cancer missed by mammography but also at higher risk of cancer than those with nondense breasts. The contribution of breast density to breast cancer risk is also a matter of debate. Approximately, 40% of US women are judged to have heterogeneously dense breasts (50%–75% of the breast consisting of fibroglandular tissue) and 10% are judged to have extremely dense breasts (75%–100% fibroglandular tissue). As others have noted (24), if high breast density imparts a high risk of breasts cancers, then 50% of all screen-eligible women would be at high risk of breast cancer simply on the basis of high breast density. However, if you compare those with heterogeneously dense breasts to those with average breast density, the risk is much lower, 1.2 times greater, and the risk for those with extremely dense breasts compared to those with average breast density would be 2.1 times greater (2). Are cancers in women with dense breasts likely to be more aggressive? It has been established that tumors seen in BRCA1 carriers are more likely to be grade III and of the medullary subtype than sporadic cancers (25), but there are little data on association of other risk factors and histologic features. Reports on the associations of breast density with prognostic
Academic Radiology, Vol 22, No 8, August 2015
US SCREENING AND BREAST CANCER OVERDIAGNOSIS
factors such as tumor grade, estrogen receptor (ER) status, and HER2 status have been inconclusive (26,27), although one recent study did show an association between breast density and high-grade tumors, ER-negative tumors, and in situ versus invasive tumors (27). Breast density has also been associated with an increased risk of local and locoregional recurrence (28) but not with distant metastasis or survival. This suggests that the increased risk of recurrence may be attributable more to incomplete identification of the entire extent of disease at the time of diagnosis secondary to the masking effect than to tumor aggressiveness. Breast cancers in young women tend to be biologically different from those in older women. Women aged <40 years with breast cancer tend to have a worse prognosis, in part because of later stage at diagnosis but also because of more aggressive features such as high grade, hormone-receptor negative status, high proliferation fraction, and lymphovascular invasion (29).
entiate life-threatening from non–life-threatening cancers and to promote more appropriate treatment. Although there is no current way to definitively distinguish the two, genomic tests such as Mammaprint and Oncotype DX offer promise. Once appropriately characterized, more tailored treatment options can be offered. There may be some cases in which the best treatment option may be simply no treatment and more of a ‘‘wait-and-see’’ approach, as has been done with prostate cancer. Certainly, we need to inform our patients that the breast cancer that we detect with screening may not ultimately be a life-threatening disease, but we also need to make sure they understand that detecting a small cancer with mammography or US could save their lives.
BY ADDING SCREENING US, ARE WE LIKELY TO INCREASE THE RATE OF OVERDIAGNOSIS? In summary, exactly how much overdiagnosis currently exists with screening mammography is uncertain, but certainly it exists to some extent. Because we do not know for certain which tumors would ultimately lead to death if left untreated and which would not, we cannot directly measure overdiagnosis. It has to be estimated from large-scale studies of breast cancer screening, but there are no screening trials with appropriate design to estimate overdiagnosis and it is very unlikely that any such study would be performed in the future. Overdiagnosis can be estimated from available data, but this requires other estimations and assumptions that increase the degree of uncertainty. Tumor size, grade, lymph node status, and other histologic features could be used as surrogates for cancer aggression, but this would be an imprecise measure because, even among small, low-grade, node-negative cancers, there is no good way currently to differentiate potentially lifethreatening cancers and non–life-threatening ones. It should not be assumed that simply finding more cancers would increase the overdiagnosis. The results of the ACRIN study would suggest a trend toward a greater proportion of lowgrade tumors diagnosed with US alone, but with only 32 cancers diagnosed with US alone, the numbers from this study are relatively small. Features of the population that would be screened may suggest an underlying tendency to more aggressive tumors, but this is also unclear. The more important question is ‘‘What should be done about it?’’ Contrary to some opinions, the potential harm of overdiagnosis should not be considered a reason to discontinue screening. Even if 30% of women diagnosed with breast cancer through screening are overdiagnosed, 70% of detected cancers are still deemed worthy of potentially life-saving treatment. As has been proposed by Gur and Sumkin (30), the solution is not to avoid detection but to work toward methods to differ-
REFERENCES 1. Bleyer A, Welch HG. Effect of three decades of screening mammography on breast cancer incidence. N Engl J Med 2012; 367(21):1998–2005. 2. Price ER, Hargreaves J, Lipson JA, et al. The California breast density information group: a collaborative response to the issues of breast density, breast cancer risk, and breast density notification legislation. Radiology 2013; 269(3):887–892. 3. Ceugnart L, Deghaye M, Vennin P, et al. Organized breast screening: answers to recurring controversies. Diagn Interv Imaging 2014; 95(4): 355–359. 4. Zahl PH, Maehlen J, Welch HG. The natural history of invasive breast cancers detected by screening mammography. Arch Intern Med 2008; 168(21):2311–2316. 5. Duffy SW, Parmar D. Overdiagnosis in breast cancer screening: the importance of length of observation period and lead time. Breast Cancer Res 2013; 15(3):R41. 6. Independent UK Panel on Breast Cancer Screening. The benefits and harms of breast cancer screening: an independent review. Lancet 2012; 380(9855):1778–1786. 7. Puliti D, Duffy SW, Miccinesi G, et al., EUROSCREEN Working Group. Overdiagnosis in mammographic screening for breast cancer in Europe: a literature review. J Med Screen 2012; 19(Suppl 1):42–56. 8. Jørgensen KJ, Gøtzsche PC. Overdiagnosis in publicly organised mammography screening programmes: systematic review of incidence trends. BMJ 2009; 339:b2587. 9. Kopans DB, Smith RA, Duffy SW. Mammographic screening and ‘‘overdiagnosis’’. Radiology 2011; 260(3):616–620. 10. Feig SA. Pitfalls in accurate estimation of overdiagnosis: implications for screening policy and compliance. Breast Cancer Res 2013; 15(4):105. 11. Kopans DB. Point: The New England Journal of Medicine article suggesting overdiagnosis from mammography screening is scientifically incorrect and should be withdrawn. J Am Coll Radiol 2013; 10(5):317–319. 12. Merry GM, Mendelson EB. Update on screening breast ultrasonography. Radiol Clin North Am 2014; 52(3):527–537. 13. Berg WA, Zhang Z, Lehrer D, et al., ACRIN 6666 Investigators, . Detection of breast cancer with addition of annual screening ultrasound or a single screening MRI to mammography in women with elevated breast cancer risk. JAMA 2012; 307(13):1394–1404. 14. Parris T, Wakefield D, Frimmer H. Real world performance of screening breast ultrasound following enactment of Connecticut Bill 458. Breast J 2013; 19(1):64–70. 15. Hooley RJ, Greenberg KL, Stackhouse RM, et al. Screening US in patients with mammographically dense breasts: initial experience with Connecticut Public Act 09-41. Radiology 2012; 265(1):59–69. 16. Rauch GM, Kuerer HM, Scoggins ME, et al. Clinicopathologic, mammographic, and sonographic features in 1,187 patients with pure ductal carcinoma in situ of the breast by estrogen receptor status. Breast Cancer Res Treat 2013; 139(3):639–647. 17. Wang LC, Sullivan M, Du H, et al. US appearance of ductal carcinoma in situ. Radiographics 2013; 33(1):213–228. 18. Corsetti V, Houssami N, Ferrari A, et al. Breast screening with ultrasound in women with mammography-negative dense breasts: evidence on
971
MOLLERAN
19.
20.
21.
22.
23.
24.
incremental cancer detection and false positives, and associated cost. Eur J Cancer 2008; 44(4):539–544. Kelly KM, Dean J, Lee SJ, et al. Breast cancer detection: radiologists’ performance using mammography with and without automated wholebreast ultrasound. Eur Radiol 2010; 20(11):2557–2564. Chang JC, Hilsenbeck SG. Prognostic and predictive markers. In: Harris JR, Lippman ME, Morrow M, et al., eds. Diseases of the breast. Philadelphia PA: Lippincitt Williams & Wilkins, 2010; 443–457. Corsetti V, Houssami N, Ghirardi M, et al. Evidence of the effect of adjunct ultrasound screening in women with mammography-negative dense breasts: interval breast cancers at 1 year follow-up. Eur J Cancer 2011; 47(7):1021–1026. Chae EY, Kim HH, Cha JH, et al. Evaluation of screening whole-breast sonography as a supplemental tool in conjunction with mammography in women with dense breasts. J Ultrasound Med 2013; 32(9):1573–1578. Ho JM, Jafferjee N, Covarrubias GM, et al. Dense breasts: a review of reporting legislation and available supplemental screening options. AJR Am J Roentgenol 2014; 203(2):449–456. D’Orsi CJ, Sickles EA. To seek perfection or not? That is the question. Radiology 2012; 265(1):9–11.
972
Academic Radiology, Vol 22, No 8, August 2015
25. Liebens FP, Carly B, Pastijn A, et al. Management of BRCA1/2 associated breast cancer: a systematic qualitative review of the state of knowledge in 2006. Eur J Cancer 2007; 43(2):238–257. 26. Eriksson L, Czene K, Rosenberg L, et al. The influence of mammographic density on breast tumor characteristics. Breast Cancer Res Treat 2012; 134(2):859–866. 27. Yaghjyan L, Colditz GA, Collins LC, et al. Mammographic breast density and subsequent risk of breast cancer in postmenopausal women according to tumor characteristics. J Natl Cancer Inst 2011; 103(15): 1179–1189. 28. Wang AT, Vachon CM, Brandt KR, et al. Breast density and breast cancer risk: a practical review. Mayo Clin Proc 2014; 89(4):548–557. 29. Partridge AH, Goldhirsch A, Gelber S, et al. Breast cancer in younger women. In: Harris JR, Lippman ME, Morrow M, et al., eds. Diseases of the breast. Philadelphia PA: Lippincitt Williams & Wilkins, 2010; 1073–1082. 30. Gur D, Sumkin JH. Screening for early detection of breast cancer: overdiagnosis versus suboptimal patient management. Radiology 2013; 268(2):327–328.