|Home | About | Journals | Submit | Contact Us | Français|
Few data have been published on mammography performance in women who are younger than 40 years.
We pooled data from six mammography registries across the United States from the Breast Cancer Surveillance Consortium. We included 117738 women who were aged 18–39 years when they had their first screening or diagnostic mammogram during 1995–2005 and followed them for 1 year to determine accuracy of mammography assessment. We measured the recall rate for screening examinations and the sensitivity, specificity, positive predictive value, and cancer detection rate for all mammograms.
For screening mammograms, no cancers were detected in 637 mammograms for women aged 18–24 years. For women aged 35–39 years who had the largest number of screening mammograms (n = 73335) in this study, the recall rate was 12.7% (95% confidence interval [CI] = 12.4% to 12.9%), sensitivity was 76.1% (95% CI = 69.2% to 82.6%), specificity was 87.5% (95% CI = 87.2% to 87.7%), positive predictive value was 1.3% (95% CI = 1.1% to 1.5%), and cancer detection rate was 1.6 cancers per 1000 mammograms (95% CI = 1.3 to 1.9 cancers per 1000 mammograms). Most (67468 [77.7%]) of the 86871 women screened reported no family history of breast cancer. For diagnostic mammograms, the age-adjusted rates across all age groups were: sensitivity of 85.7% (95% CI = 82.7% to 88.7%), specificity of 88.8% (95% CI = 88.4% to 89.1%), positive predictive value of 14.6% (95% CI = 13.3% to 15.8%), and cancer detection rate of 14.3 cancers per 1000 mammograms (95% CI = 13.0 to 15.7 cancers per 1000 mammograms). Mammography performance, except for specificity, improved in the presence of a breast lump.
Younger women have very low breast cancer rates but after mammography experience high recall rates, high rates of additional imaging, and low cancer detection rates. We found no cancers in women younger than 25 years and poor performance for the large group of women aged 35–39 years. In a theoretical population of 10000 women aged 35–39 years, 1266 women who are screened will receive further workup, with 16 cancers detected and 1250 women receiving a false-positive result.
Little information is available on the performance of mammography in women who are younger than 40 years.
Data were obtained from six registries on mammography among women who were aged 18–39 years at their first mammogram, screening or diagnostic. The women were followed for 1 year to determine the accuracy of their mammograms, the recall rate for screening mammograms, and the sensitivity, specificity, positive predictive value, and cancer detection rate for all mammograms.
Younger women had very low breast cancer rates but experienced high rates of false-positive results, with high recall rates after mammography, high rates of additional imaging, and low cancer detection rates. No cancers were found in women younger than 25 years, and poor performance was found for women aged 35–39 years.
Many young women at low or average risk for breast cancer were exposed to additional imaging with a low probability of cancer detection. Consequently, harms may include additional radiation exposure, anxiety associated with false-positive findings, and costs associated with additional imaging.
There was a sizeable amount of missing data. Complete family pedigree or BRCA1 or BRCA2 status could not be collected, and so women at very high risk could not be identified.
From the Editors
Among US women, breast cancer is the most commonly diagnosed cancer and the second most common cause of cancer-related deaths (1). Data from the Surveillance, Epidemiology, and End Results (SEER) program show that 6.5% of breast cancers are diagnosed in women who are younger than 40 years and only 0.6% are diagnosed in women who are younger than 30 years (2). Among younger women, SEER age-adjusted invasive breast cancer incidence rates per 100000 women in the general population are 0.1 cancers for those aged 15–19 years, 0.7 cancers for those aged 20–24 years, 3.8 cancers for those aged 25–29 years, 12.9 cancers for those aged 30–34 years, and 29.3 cancers for those aged 35–39 years; this rate is 59.1 cancers for women aged 40–44 years (2). Breast cancers diagnosed in young women have poorer prognostic characteristics, higher recurrence rates (3), and higher relative mortality rates than those diagnosed in older women (4,5).
In this study, we defined “young” women as being younger than 40 years because this age group is excluded from standard guidelines for screening mammography in the United States (6–8) and other countries (9). Today, the American Cancer Society (1,6), Cancer Genetics Studies Consortium (10), and others (11) advocate screening among young women who have a high-risk profile or known BRCA1 and/or BRCA2 mutations. In the 1980s, the American Cancer Society recommended a baseline mammogram for women at average risk who were aged 35–40 years to provide a comparison image that would be available when regular screening began at age 40 or older years, but this recommendation was dropped in 1992 after a consensus meeting reviewed the evidence on screening recommendations and agreed that there was little evidence to support a benefit of the baseline screening before age 40 years (12). Using data from six counties in California, Kerlikowske et al. (13) reported that the percentage of women receiving a first screening mammogram between the ages of 30 and 39 years increased from 21.8% in 1985 to 28.9% in 1992. However, still in 2005, 28.9% of US women aged 30–39 years reported ever having a mammogram (14). Anecdotal evidence suggests that nearly two decades later, some physicians and women still believe that a baseline mammogram before age 40 years is recommended.
Previous research on mammography use in young women has been mainly confined to retrospective examinations of mammography and ultrasound data (15) from patients with cancer or to examinations of the pathology of the detected cancers. Details relating to mammography in young women are scarce, with particular gaps in our understanding of who has mammography, how accurately mammography is performed, and how the outcomes in women younger than 40 years compare with those of women aged 40 years or older, for whom regular screening is recommended. We address these issues by examining the distribution, performance characteristics, and pathological outcomes of first mammographic examination for screening or diagnostic purposes among young women receiving mammography in community practice in the United States by using data from the Breast Cancer Surveillance Consortium (BCSC).
The BCSC is a National Cancer Institute-funded collaborative network of population-based mammography registries with linkages to pathology and/or tumor registries (16). Analyses in this study are based on pooled data from six registries: North Carolina, Western Washington State, New Hampshire, New Mexico, San Francisco, and Vermont. Each registry links its mammography data to its regional or SEER cancer registry for cancer outcomes data, including tumor size, stage, and molecular markers. Along with the BCSC Statistical Coordinating Center, sites adhere to strict confidentiality procedures, comply with the Health Insurance Portability and Accountability Act, and have a Federal Certificate of Confidentiality and other protection for the identities of women, physicians, and facilities who are subjects of this research.
We included 117738 women from the BCSC who were aged 18–39 years when they had their first screening or diagnostic mammogram during 1995–2005 and followed them for 1 year to determine accuracy of mammography assessment. A mammogram was determined to be the first if the woman reported no previous mammogram at the time of the mammography examination and no previous mammogram was found in the registry. We excluded women with a history of breast cancer and women with missing information on the mammogram interpretation. Mammograms were classified according to the radiologists’ recorded indication—screening (including 5% where women reported a lump) and/or diagnostic—for evaluation of a breast problem (16).
Each woman completed a questionnaire at the time of her visit for breast imaging that included information on age, race or ethnicity, self-reported presence of breast problems, first-degree family (ie, first-degree relatives) and personal histories of breast cancer, and menopausal status. Radiologists and technologists recorded information on imaging performed, indication for the visit, breast parenchymal density, mammography interpretation, and recommendations for follow-up.
Mammograms were determined to be positive or negative on the basis of the interpretation assessment assigned by the radiologist, who used the lexicon from the Breast Imaging Reporting and Data System (BI-RADS), which scores assessment on a scale of 1–5 with increasing probability of cancer and uses a “0” for indeterminate studies that need further evaluation. The BI-RADS scores for assessment are 1 = normal; 2 = benign finding; 3 = probably benign; 4 = suspicious abnormality; and 5 = suspicious for cancer (17). Previous BCSC articles (18,19) describe the use of this scale. An examination was positive if the patient was recalled for further assessment or recommended for biopsy examination (18,19). Standard definitions for true-positive, false-positive, true-negative, and false-negative results were calculated by use of a 12-month follow-up period (20). Cancers in patients were classified as ductal carcinoma in situ or invasive cancer; lobular carcinomas in situ were not classified as cancer.
We calculated the following performance measures: recall rate for screening (ie, the proportion of screening assessments that led to a recommendation for further workup), sensitivity (ie, the probability of a positive mammogram assessment, given a cancer was detected within 1 year), specificity (ie, the probability of a negative mammogram assessment, given no cancer was diagnosed within 1 year), cancer detection rate (ie, the number of cancers diagnosed within 1 year of a positive mammogram per 1000 mammograms), screening positive predictive value (ie, the probability of a cancer diagnosis within 1 year after a positive screening examination), and diagnostic positive predictive value (ie, the probability of cancer, given a recommendation for biopsy examination that was based on diagnostic mammography) (17).
Performance measures were evaluated separately by indication (screening or diagnostic mammography) and stratified by four age groups (18–25, 26–29, 30–34, or 35–39 years) for descriptive and age characteristics in the performance tables. Measures were stratified by age groups of 18–34 and 35–39 years for assessing family history and pathology characteristics because 84.5% of the women who received screening mammography were aged 35–39 years and because we hypothesized that this unique group most likely received mammography for baseline screening, as described above. For the remaining characteristics, performance measures were age-adjusted via direct standardization. Throughout, 95% confidence intervals (CIs) and P values were estimated by use of a nonparametric bootstrap method (sampling with replacement from the full sample) that was based on 10000 replications (21). We have calculated P values solely for questions related to lump and family history because we expected that these variables may have an effect among young women, particularly on recall rates, cancer rate, and positive predictive value. P values that were calculated used a Wald statistic in which the standard error in the denominator of the test statistic was taken from the bootstrap sampling distribution of each statistic and the data were collapsed across the groups (ie, under the null hypothesis of no association). Pathology characteristics for cancers were reported as counts and percentages (based on non-missing counts) separately by indication for examination. All statistical tests were two-sided , and all analyses were performed with SAS software, version 9.1 (SAS Institute, Inc, Cary, NC).
Our study sample included 117737 women who had received their first mammography, 86781 (73.7%) with a first screening mammogram and 30956 (26.3%) with a first diagnostic mammogram. There were 714 cancers, 188 (26.3%) among first screening mammograms and 526 (73.7%) among first diagnostic mammograms. The overall cancer detection rate was 1.6 cancers per 1000 mammograms (95% CI = 1.3 to 1.9 cancers per 1000 mammograms).
Among the 86781 young women who received screening mammography, 83862 (96.6%) screening mammograms were among women aged 30–39 years, with a striking increase in the number of such mammograms starting at age 35 years (Figure 1 presents the frequency distribution of mammograms for all women in the Carolina Mammography Registry by age). Among the age groups studied, 2919 (3.3%) screened women were aged 18–29 years, 10527 (12.1%) were aged 30–34 years, and 73335 (84.5%) were aged 35–39 years. At age 34 years, approximately 4000 women were screened compared with approximately 17000 women at age 35 years.
The racial distribution among screened women was 67.0% white, 6.9% African American or African American, and 16.0% other (Asian, Alaska Native or American Indian, Hawaiian or Pacific Islander, or mixed) (Table 1). Only 3726 (4.3%) of the 86781 women aged 18–39 years were menopausal, and 3167 (3.6%) women reported use of postmenopausal hormone therapy at the time of their visit. Overall, a family history of breast cancer was reported in 9481 (10.9%) of the 86871 women who received a first screening mammogram, but it was not reported in 67468 (77.7%) (data were missing for 9832 women). There was a statistically significantly higher proportion of positive family history among women who were younger than 35 years (26.4%) than among those who were aged 35–39 years (8.1%) (P < .001) (Table 2). Dense breasts were observed in 35794 (41.2%) women, with 8957 (10.3%) having extremely dense breasts and 26837 (30.9%) having heterogeneously dense breasts; breast density was missing for 27746 (32.0%) women.
The overall cancer rate in this population was 2.2 cancers diagnosed per 1000 mammograms. We observed no cancers in women younger than 25 years. The cancer prevalence rate was low (2.1 cancers per 1000 mammograms) in the oldest age group (ie, 35–39 years). The overall cancer rate was higher in African American women (3.7 cancers per 1000 mammograms) than in white women (2.1 cancers per 1000 mammograms) or in women of other races (1.5 cancers per 1000 mammograms).
Overall, only 4302 (5.0%) of the 86781 women self-reported a lump at screening, yet they had 49 (26.1%) of the 188 cancers. The cancer prevalence rate was 11.4 cancers per 1000 mammograms (ie, 49 cancers per 4302 mammograms) among women reporting a lump, compared with 1.6 cancers per 1000 mammograms (127 cancers per 78924 mammograms) among women with no lump (P < .001). It is not clear why women who reported a lump had screening mammography and not diagnostic mammography.
Among the 30956 diagnostic mammograms identified in this study, women aged 30–39 years received 25164 (81.3%) of them and women aged 18–29 years received 5792 (18.7%). A lump at the time of their diagnostic mammogram was reported by 20392 (65.9%) women (Table 1), and 14789 (47.8%) had heterogeneously or very dense breasts. Breast density was, however, missing for 8742 (28.2%) women. A positive family history of breast cancer was reported by only 2032 (6.6%). The overall cancer prevalence rate among all diagnostic mammograms was 17.0 cancers per 1000 mammograms.
Recall rates decreased with increasing age, from 17.6% (95% CI = 14.8% to 20.6%) for the age group 18–24 years to 12.7% (95% CI = 12.4% to 12.9%) for the age group 35–39 years (P < .001) (Table 3). Age-adjusted recall rates were statistically significantly lower among women with a lump (11.8%, 95% CI = 11.6% to 12.1%) than among those without a lump (31.1%, 95% CI = 29.6% to 32.8%) (P < .001). Women with a family history of breast cancer had a slightly higher age-adjusted recall rate (15.0%, 95% CI = 14.2% to 15.8%) than women without a family history of breast cancer (13.5%, 95% CI = 13.2% to 13.8%). Recall rates did not differ across racial and/or ethnic groups.
Age-adjusted sensitivity was 76.5% (95% CI = 70.1% to 82.5%), and age-specific sensitivity varied by age, with no cancers in the age group of 18–24 years (66.7%, 95% CI = 20.0% to 100%, in women aged 25–29 years; 81.5%, 95% CI = 65.4% to 95.5%, in women aged 30–34 years; and 76.1%, 95% CI = 69.2% to 82.6%, in women aged 35–39 years) (Table 3). Age-adjusted sensitivity was highest among African American women at 81.2% (95% CI = 62.9% to 96.3%) compared with 78.6% (95% CI = 71.0% to 85.8%) among white women and 67.0% (95% CI = 42.8% to 88.9%) among women of other races. Sensitivity was not associated with breast density or with a family history of breast cancer. Age-adjusted sensitivity was statistically significantly higher among the 4302 (5.0%) women with a lump (84.7%, 95% CI = 73.4% to 94.6%) than among those without a lump (74.8%, 95% CI = 66.9% to 82.2%) (P = .022).
Age-adjusted specificity was 87.1% (Table 3). Specificity increased with increasing age, from 82.4% (95% CI = 79.4% to 85.2%) in the age group of 18–24 years to 87.5% (95% CI = 87.2% to 87.7%) in that of 35–39 years. Specificity was higher among women with fatty breasts (91.5%, 95% CI = 90.4% to 92.4%) than among women with more dense categories (86%–89%). Specificity was 19% lower among women with a lump (69.6%, 95% CI = 68.0% to 71.2%) than among women without a lump (88.3%, 95% CI = 88.0% to 88.5%) (P < .001). Specificity was not associated with race or with family history of breast cancer.
Age-adjusted positive predictive value of screening was 1.3% (95% CI = 1.1% to 1.5%), with a slight variation with age (Table 4). Positive predictive value of screening was 3.7% (95% CI = 2.5% to 4.9%) among women with a self-reported lump compared with 1.0% (95% CI = 0.8% to 1.2%) among women without a lump (P < .001). A higher positive predictive value of screening was observed among African American women than among white women, and a statistically significantly higher positive predictive value was observed among women with a positive family history (1.2%, 95% CI = 1.0% to 1.5%) than among women without such history (0.9%, 95% CI = 0.4% to 1.5%) (P = .032).
The overall age-adjusted cancer detection rate was 1.7 cancers per 1000 mammograms (95% CI = 1.4 to 1.9 cancers per 1000 mammograms). The cancer detection rate essentially did not change with age and was not associated with family history, particularly in the age group of 35–39 years (Table 2). African American women had higher cancer detection rates than white women. The cancer detection rate was higher among women with a lump (11.4 cancers per 1000 mammograms, 95% CI = 7.8 to 15.4) than among women without a lump (1.2 cancers per 1000 mammograms, 95% CI = 1.0 to 1.4) (P < .001).
Age-adjusted diagnostic sensitivity was high (85.7%, 95% CI = 82.7% to 88.7%) (Table 3). Sensitivity was higher among women who reported a lump (87.5%, 95% CI = 84.3% to 90.5%) than among women who reported other symptoms (71.5%, 95% CI = 59.8% to 82.7%) (P = .001). Sensitivity was lower among African American women (81.5%, 95% CI = 71.5% to 90.0%) than among white women (86.6%, 95% CI = 82.9% to 90.0%) or women of other races (89.0%, 95% CI = 80.7% to 95.7%).
Age-adjusted specificity was 88.8% (95% CI = 88.4% to 89.1%) and was not influenced by family history or race. Specificity was lower in the youngest age group, 18–24 years (83.8%, 95% CI = 81.9% to 85.6%), than in the other three age groups (88%–90%) and was lower among women who reported a lump (86.6%, 95% CI = 86.2% to 87.1%) than among those who did not (93.0%, 95% CI = 92.5% to 93.6%) (P < .001). Across the levels of breast density, specificity was highest among women with fatty breasts (93.2%, 95% CI = 91.6% to 94.7%) and decreased with increasing breast density (91.3% scattered fibroglandular densities, 88.1% heterogeneously dense and 89.2% extremely dense).
The age-adjusted positive predictive value for diagnostic mammography was 14.6% (95% CI = 13.3% to 15.8%) (Table 4). The positive predictive value of diagnostic mammography increased with each age group from 2.3% (95% CI = 0.5% to 4.5%) in those aged 18–24 years to 18.6% (95% CI = 16.6% to 20.6%) in those aged 35–39 years and was statistically significantly higher among women with a lump (16.2%, 95% CI = 14.8% to 17.7%) than among women without a lump (8.3%, 95% CI = 6.0% to 10.6%) (P < .001). In addition, this value was lower (10.6%, 95% CI = 8.1% to 13.3%) among women in the other racial group than among African American women (18.1%, 95% CI = 14.0% to 22.4%) and among white women (15.2%, 95% CI = 13.6% to 16.8%).
Overall, the age-adjusted cancer detection rate was 14.3 cancers detected per 1000 mammograms (95% CI = 13.0 to 15.7). It was higher for African American women (23.5 cancers detected per 1000 mammograms, 95% CI = 17.7 to 29.8) than for white women (14.0 cancers detected per 1000 mammograms, 95% CI = 12.4 to 15.6) or women of other races (13.0 cancers detected per 1000 mammograms, 95% CI = 9.8 to 16.4). Women with a positive family history of breast cancer (20.4 cancers detected per 1000 mammograms, 95% CI = 14.2 to 27.0) had a higher cancer detection rate than those without such a family history (13.5 cancers detected per 1000 mammograms, 95% CI = 12.1 to 15.0). Women who reported a lump had a higher cancer detection rate (19.5 cancers detected per 1000 mammograms, 95% CI = 17.7 to 21.5) than women who did not (4.7 cancers detected per 1000 mammograms, 95% CI = 3.4 to 6.2).
In this study population, no cancers were diagnosed after a first screening mammogram among women aged 18–24 years (Table 5). Tumor characteristics of women aged 25–34 years were associated with poorer prognosis than those of women aged 35–39 years, including a higher proportion of invasive cancers (81.8% vs 70.3%, respectively), stage II or higher (45.4% vs 39.3%), poorly differentiated or undifferentiated (54.5% vs 47.1%), greater than 2 cm (39.4% vs 32.9%), estrogen receptor negative and progesterone receptor negative (30.3% vs 20.6%), and positive lymph nodes (45.4% vs 29.7%).
Women aged 18–34 years, compared with those aged 35–39 years, had higher proportions of invasive cancers (94.8% vs 91.1%, respectively), cancers that were stage II or higher (57.8% vs 55.2%), poorly differentiated or undifferentiated cancers (54.9% vs 51.7%), and estrogen receptor–negative and progesterone receptor–negative cancers (23.7% vs 21.3%). However, we observed little difference between the two age groups in the proportion of cancers that were greater than 2 cm in diameter (42.7% vs 42.6%) or proportion of patients with positive lymph nodes (44.1% vs 45.4%).
To our knowledge, this study was the first comprehensive community-based evaluation of the first mammography experience, including both screening and diagnostic mammography, among women younger than 40 years. In our population, a substantial percentage of young women received screening mammography, but few breast cancers were found, regardless of their specific age, race, or individual characteristics. Yet, these women experience high recall rates with high rates of additional imaging. The sensitivity, specificity, and screening positive predictive value of screening mammography were poor, and cancer detection rates were very low in these young women, who are not yet in an age group for which national organizations recommend regular screening mammography. Harms need to be considered, including radiation exposure because such exposure is more harmful in young women (22,23), the anxiety associated with false-positive findings on the initial examination, and costs associated with additional imaging. If we consider a theoretical population of 10000 women aged 35–39 years, then from our results, an average of 1266 women who are screened will receive further workup, with 16 cancers being detected and 1250 false-positive examinations. This cancer detection rate corresponds to an estimated 79 workups for each cancer detected. Using the same methods that were outlined above, we also considered two theoretical populations of women who underwent their first mammography at the ages of 40–44 years or 45–49 years. The number of workups per cancer detected was estimated to be 66 in the age group of 40–44 years and 40 in the age group of 45–49 years. Performance characteristics were good for diagnostic mammography, as expected because most young women undergoing diagnostic mammography reported the presence of a breast lump and because diagnostic mammography performance improved when a lump was reported.
It is unclear why the first mammogram for some women with a breast lump was classified as a screening mammogram, and for others, it was classified as a diagnostic mammogram. Women reporting a breast lump constituted a small proportion of the screening mammography population (5%) but were different from the women reporting a lump in the diagnostic mammography population (ie, the specificity and the cancer detection rate among women with a lump who received a screening mammogram was observed to be lower than that among women with a lump who received diagnostic mammography). Unfortunately, we could not evaluate when the lumps were detected. For example, it is possible that among women with a breast lump, some made a screening appointment and then developed a lump that was then reported at the time of screening, whereas others with a lump may have sought an immediate consultation for a new breast symptom, without having scheduled a mammogram. The subgroup of women arriving for screening with a lump need further study because they appear to be a clinically distinct group.
The majority of women who were younger than 40 years at their first screening mammogram (67468 [77.7%]) had no family history of breast cancer in a first-degree relative, in particular women aged 35–39 years (80.7%). We caution against screening young women with a positive family history of breast cancer, unless there are characteristics that are associated with an inherited predisposition to breast cancer at a very young age; we make this recommendation because performance characteristics were similar in women with a family history of breast cancer and in women without such history. Women with a family history of breast cancer, compared with those without such history, had higher recall rates but equal cancer rates and cancer detection rates. Some women and/or their providers may be requesting a baseline mammogram, most likely a holdover from historical American Cancer Society guidelines (12,24). It is also possible that the women and/or their providers think that they are at high risk and can benefit from screening. Kapp et al. (25) used data from the National Health Interview Survey and found that women aged 35–39 years who had a mammogram reported a physician recommendation for a mammogram regardless of risk factors. Risk factors, however, were more likely to explain reported phyisician recommendations for mammography for women aged 30–34 years.
To understand the level of performance reported in this study, we compared these results in young women (aged 18–39 years) with results from women in the BCSC who were aged 40–44 years and 45–49 years and who had their first screening mammogram in the BCSC. Mammography performance measures in young women (aged 18–39 years), except for specificity, were inferior to those in women aged 40–49 years. Age-adjusted sensitivity of first screening mammography was lower among young women (76.5%) than among women aged 40–44 years (82.4%) or 45–49 years (87.3%). The cancer detection rate among women aged 18–39 years was 1.7 cancers per 1000 mammograms, among those aged 40–44 years was 2.3 cancers per 1000 mammograms, and among women aged 45–49 years was 4.3 cancers per 1000 mammograms. This cancer detection rate in young women may be higher than expected in comparison to incidence rates because data were from first mammograms. The poor performance of screening mammography among younger women is likely attributed to the very low cancer prevalence in this group. Most of the young women have dense parenchymal tissue that may mask tumors, and so breast density may also contribute to the lower performance of mammography in young women.
Consistent with previously published work (3,26), pathology results indicated that tumors from most young women compared with tumors from older women have poorer prognostic characteristics, higher stage, higher proportion of positive lymph nodes, and higher proportion with an estrogen receptor–negative and progesterone receptor–negative status. Although the young women in our population had a higher proportion of cancers with poor prognostic characteristics, they also had poor mammography performance indicators with low sensitivity, specificity, and cancer detection rate. These results pose challenges to screening decisions. African American women have higher cancer rates and poorer outcomes than white women (27–30). Young African American women are screened at a higher rate than white women (31), which could be explained by physicians giving greater importance to the relative risk of a diagnosis before age 40 years than to the absolute risk and by the perception that because a high level of cancers with poor prognosis are found among young African American women, earlier screening is required for this group (32). Our findings support a need for serious discussion about the appropriateness of mammography in women without the presence of symptoms. The results for diagnostic mammography are much better for all young women perhaps because 70% report the presence of a lump at the time of testing. Symptomatic women should receive mammography.
Few community-based studies have been conducted to evaluate the performance of mammography among young women. The Oregon Breast and Cervical Cancer Program (BCCP), which targeted women at the poverty level, has reported that 21% of women who were younger than 40 years were asymptomatic but that 79% of their women who were sent mammography had been examined for a breast problem, which is a much smaller proportion than were examined by screening in our population (33). Discrepancies between that study and this study might be that data for this study were from all women in participating practices, whereas data in the BCCP study were from women who were eligible for that study (ie, those with incomes of up to 250% of the Federal Poverty Level and who were uninsured or underinsured). Thus, mammographic screening of asymptomatic women is not supported and should be reserved for young women with symptoms.
Breast ultrasound and breast magnetic resonance imaging are increasingly used in young women (34), yet to date there are no community-based evaluations of the performance characteristics or demonstrations of a reduction in morbidity and mortality. Both of these modalities would be important in lowering the radiation dose to young women.
This study has several limitations. First, because our data reflect clinical practice, we had a sizeable amount of missing data for breast density and pathology characteristics. Density data are largely missing as a result of practice-specific nonreporting, and pathology characteristics are missing when they are not reported on the pathology reports to the cancer registries and/or are specific to a hospital or pathology laboratory. Second, we could not collect the complete family pedigree or BRCA1 or BRCA2 status of those in our study, which prevents us from identifying women at very high risk. Finally, we do not know the outmigration of women from the region in which they were diagnosed within 12 months of their mammogram. This percentage is expected to be small, and most women with a positive mammogram complete their workup within a month. The major strengths of our study include the representative community-based nature of the BCSC, the large sample size, and prospective cancer follow-up through linkages with local registry and/or pathology databases.
Future research should explore the association of symptoms with accuracy and health-seeking behavior. In addition, research should provide guidance on defining which young women might benefit from screening mammography and on the role and use of advanced technologies to detect breast cancers in younger women.
In conclusion, young women have received screening mammography, but few cancers have been detected, regardless of their specific age, race, or other individual characteristics. With high recall and low cancer detection rates, many young women, who are at low or average risk for breast cancer, are having additional imaging as a result of undergoing screening mammography, with a low probability of cancer detection. Who should be screened for breast cancer at younger ages and how best to screen them remain important research questions. Cancers in young women, across the board, have characteristics associated with more rapid tumor growth and poorer prognosis than those in older women. Tumor characteristics with poorer prognostic characteristics pose a dilemma: It is important to find breast cancers in young women early because of the underlying tumor biology of such cancers; yet current screening performance is challenged by the high breast density, which is generally found in young women. Research results evaluating the performance of the addition of ultrasound and/or breast magnetic resonance imaging in young women at high risk for breast cancer are needed to show whether these modalities perform better than mammography alone in young women.
National Cancer Institute-funded Breast Cancer Surveillance Consortium cooperative agreements (U01CA63740, U01CA86076, U01CA86082, U01CA63736, U01CA70013, U01CA69976, U01CA63731, and U01CA70040). The collection of cancer incidence data used in this study was supported in part by several state public health departments and cancer registries throughout the United States. For a full description of these sources, please see http://breastscreening.cancer.gov/work/acknowledgement.html.
All authors had full responsibility for all activities related to this manuscript.
We thank the Breast Cancer Surveillance Consortium participating mammography facilities and radiologists for the data they have provided for this study. We thank Robert A. Smith, PhD, for edits to the final revisions of the manuscript. We also thank Molly Jarman for assistance in editing and manuscript preparation.