Abstract
Purpose
There is scarce information on whether digital screening mammography performance differs between black and white women.
Methods
We examined 256,470 digital screening mammograms performed from 2005–2010 among 31,654 black and 133,152 white Carolina Mammography Registry participants aged ≥40 years. We compared recall rate, sensitivity, specificity, and positive predictive value (PPV1) between black and white women, adjusting for potential confounders using random effects logistic regression.
Results
Breast cancer was diagnosed in 276 black and 1,095 white women. Recall rates were similar for blacks and whites (8.6% vs. 8.5%), as were sensitivity (83.7% vs. 82.4%), specificity (91.8% vs. 91.9%) and PPV1 (4.8% vs. 5.3%) (all p-values>0.05). Stratified and adjusted models showed similar results. Despite comparable mammography performance, tumors diagnosed in black women were more commonly poorly differentiated and hormone receptor-negative.
Conclusion
Equivalent performance of digital screening mammography by race suggests that efforts to understand tumor disparities should focus on etiologic factors that influence tumor biology.
Keywords: screening mammography, disparities, sensitivity, specificity, positive predictive value, breast cancer
Introduction
Black women experience higher breast cancer mortality rates than white women and typically present with more aggressive tumors and worse prognosis, even after taking stage at diagnosis into account.[1] Mammography screening is commonly associated with the diagnosis of smaller, earlier stage tumors, suggesting that racial differences in screening-related tumor detection may influence racial differences in the tumor characteristics. Only one prior study assessed differences in screening mammography performance among blacks and whites; however, that study was limited to film-screen mammography, which has largely been replaced with digital mammography in the United States.[2] Compared with film-screen mammography, digital mammography detects more ductal carcinoma in situ (DCIS)[3] and has improved performance among women with dense breasts, ages <50 years, or who are premenopausal.[4] Age-specific breast cancer rates and mammographic density vary between black and white women, [5–7] but it is unknown whether there are any racial differences in digital mammography screening performance. Hence, we sought to determine if digital screening mammography performs equally well in black and white women.
Methods
Data Sources
We utilized data from the Carolina Mammography Registry (CMR), a prospective population-based breast imaging registry in 39 counties in North Carolina. This study received Institutional Review Board approval for passive consenting process to enroll participants, link and pool data, and perform analysis. All procedures were Health Insurance Portability and Accountability Act compliant. At the time of the mammogram, women provided information related to demographics and breast health history. For each mammography examination the radiologist recorded the reason for the visit, imaging examination performed, Breast Imaging Reporting and Data System (BI-RADS) breast density [8], BI-RADS mammogram assessment [8], and follow-up recommendations. Patient and radiologist data were linked to breast cancer cases from the North Carolina Central Cancer Registry (NCCCR) and to statewide hospital pathology data. Tumor behavior (in situ or invasive), grade, stage at diagnosis, size, nodal status, estrogen receptor (ER) and progesterone receptor (PR) expression were abstracted from NCCCR and pathology reports.
Study Population and Definitions
In this analysis, we examined all digital screening mammograms performed from 2005 to 2010 among black and white women ages ≥40 years with no personal history of breast cancer or history of breast augmentation. Screening mammograms were defined as those that were: (i) bilateral; (ii) performed in asymptomatic women; (iii) defined as a “routine screen” by the radiologist; and (iv) >9 months after any prior breast imaging. Positive screening mammograms had an initial BI-RADS assessment code of 0 (needs additional imaging evaluation), 4 (suspicious abnormality), 5 (highly suggestive of malignancy), or 3 (probably benign finding) when the 3 was associated with a recommendation for immediate follow-up. Negative screening mammograms had an initial BI-RADS assessment of 1 (negative), 2 (benign finding), or 3 with a recommendation for follow-up of >6 months.[9] Positive disease status was defined by diagnosis of DCIS or invasive breast cancer within 12 months of the screening mammogram. Each mammogram was categorized as true positive, false negative, true negative, and false positive according to the BI-RADS assessment and the cancer outcome.
Statistical Analysis
We computed mammography sensitivity, specificity, positive predictive value (PPV1), and recall rate using standard definitions [10] and compared the statistics for black and white women using an F test. We used a random effects logistic regression model to adjust for differences between radiologists interpreting the images and to account for correlated observations within women who had multiple screening examinations during the study period.[11] We adjusted for age at mammogram, rural/urban residence, education, menopausal status, breast density, prior breast biopsy, family history of breast cancer, and time since last screening examination. We present comparisons overall and stratified by age group and BI-RADS breast density (dichotomized into almost entirely fat or scattered fibroglandular densities versus heterogeneously dense or extremely dense). We also compared tumor characteristics by race using the chi-square test, for all cancers and also stratified by true positive or false negative status.
Results
Of 256,470 digital screening mammograms, 56,239 (21.9%) were performed among black women and 200,231 (78.1%) were performed among white women (Table 1). The majority of women were ages 40–59 years, lived in urban areas, were post-menopausal, had no prior breast biopsy, and no family history of breast cancer. The proportion of black women with some college of higher education was 43.9% compared with 60.1% for white women. BI-RADS breast density of heterogeneously or extremely dense was 40.6% for black women versus 48.0% for white women.
Table 1.
Characteristics of women undergoing digital screening mammography by race, Carolina Mammography Registry 2005–2010
| Characteristic | Black Women N=56,239 |
White Women N=200,231 |
||
|---|---|---|---|---|
| N | % | N | % | |
| Age Group | ||||
| 40–49 | 16,391 | 29.2 | 52,034 | 26.0 |
| 50–59 | 18,560 | 33.0 | 56,395 | 28.2 |
| 60–69 | 12,505 | 22.2 | 50,868 | 25.4 |
| 70–79 | 6,815 | 12.1 | 31,037 | 15.5 |
| 80+ | 1,968 | 3.5 | 9,897 | 4.9 |
| Rural/Urban Residence | ||||
| Rural | 14,238 | 25.3 | 58,944 | 29.4 |
| Urban | 41,995 | 74.7 | 141,282 | 70.6 |
| Missing | 6 | --- | 5 | --- |
| Educational Level | ||||
| < High school | 3,380 | 18.8 | 5,726 | 7.3 |
| High school graduate | 6,717 | 37.3 | 25,650 | 32.6 |
| Some college/technical school | 4,574 | 25.4 | 24,194 | 30.8 |
| College graduate | 3,318 | 18.4 | 23,120 | 29.4 |
| Missing | 38,250 | --- | 121,541 | --- |
| Menopausal Status | ||||
| Pre or peri menopausal | 19,877 | 35.5 | 62,131 | 31.1 |
| Post menopausal | 36,162 | 64.5 | 137,467 | 68.9 |
| Missing | 200 | --- | 633 | --- |
| BIRADS Breast Density* | ||||
| Almost entirely fat | 4,958 | 9.2 | 13,316 | 7.0 |
| Scattered fibroglandular densities | 27,199 | 50.3 | 85,963 | 45.1 |
| Heterogeneously dense | 20,063 | 37.1 | 79,992 | 41.9 |
| Extremely dense | 1,877 | 3.5 | 11,558 | 6.1 |
| Missing | 2,142 | --- | 9,402 | --- |
| Prior Breast Biopsy | ||||
| Yes | 10,637 | 23.9 | 46,346 | 25.9 |
| No | 33,954 | 76.2 | 132,758 | 74.1 |
| Missing | 11,648 | --- | 21,127 | --- |
| Family History of Breast Cancer | ||||
| Yes | 6,789 | 12.1 | 28,168 | 14.1 |
| No | 49,241 | 87.9 | 171,643 | 85.9 |
| Missing | 209 | --- | 420 | --- |
BI-RADS breast density refers to the Breast Imaging Reporting and Data System
A total of 1,371 breast cancers were diagnosed, including 231 true positives and 45 false negatives among blacks and 902 true positives and 193 false negatives among whites. The number of false positives was 4,607 for blacks and 16,099 for whites. The overall recall rate was 8.5%, sensitivity was 82.6%, specificity was 91.9%, and PPV1 was 5.1%, similar to digital mammography performance previously reported among women in the Breast Cancer Surveillance Consortium.[12] The recall rate, sensitivity, specificity, and PPV1 were similar for blacks and whites in both crude (Table 2) and adjusted models (p-values for adjusted rates were 0.1773, 0.4869, 0.3194, and 0.3992, respectively). Furthermore, stratification by age group or breast density did not reveal any differences in performance by race.
Table 2.
Unadjusted digital mammography performance in black and white women in the Carolina Mammography Registry, 2005–2010
| Race
|
||||
|---|---|---|---|---|
| Black | White | |||
|
| ||||
| % | (95% CI) | % | (95% CI) | |
| All mammograms | ||||
| Recall Rate | 8.6 | (8.4, 8.8) | 8.5 | (8.4, 8.6) |
| Sensitivity | 83.7 | (79.3, 88.1) | 82.4 | (80.1, 84.6) |
| Specificity | 91.8 | (91.5, 92.0) | 91.9 | (91.8, 92.0) |
| PPV1 | 4.8 | (4.2, 5.4) | 5.3 | (5.0, 5.6) |
| Stratified by age at mammogram: | ||||
| 40–49 years | ||||
| Recall Rate | 10.9 | (10.4, 11.3) | 10.9 | (10.6, 11.1) |
| Sensitivity | 81.8 | (70.4, 93.2) | 78.8 | (72.4, 85.1) |
| Specificity | 89.3 | (88.9, 89.8) | 89.3 | (89.1, 89.6) |
| PPV1 | 2.0 | (1.4, 2.7) | 2.3 | (1.8, 2.6) |
| 50–59 years | ||||
| Recall Rate | 8.3 | (7.9, 8.6) | 8.3 | (8.0, 8.5) |
| Sensitivity | 76.9 | (67.6, 86.3) | 81.1 | (76.3, 86.0) |
| Specificity | 92.0 | (91.7, 92.4) | 92.1 | (91.8, 92.3) |
| PPV1 | 3.9 | (3.0, 4.9) | 4.3 | (3.8, 4.9) |
| 60–69 years | ||||
| Recall Rate | 7.5 | (7.0, 7.9) | 7.7 | (7.4, 7.9) |
| Sensitivity | 85.6 | (78.6, 92.6) | 84.1 | (80.3, 88.0) |
| Specificity | 93.2 | (92.7, 93.6) | 92.9 | (92.6, 93.1) |
| PPV1 | 8.9 | (7.1, 10.7) | 7.5 | (6.6, 8.3) |
| ≥70 years | ||||
| Recall Rate | 6.8 | (6.3, 7.3) | 6.8 | (6.6, 7.0) |
| Sensitivity | 91.2 | (83.9, 98.6) | 83.2 | (79.3, 87.2) |
| Specificity | 93.8 | (93.3, 94.3) | 93.8 | (93.6, 94.1) |
| PPV1 | 8.7 | (6.5, 11.0) | 10.2 | (9.1, 11.3) |
| Stratified by breast density: | ||||
| Almost entirely fat or scattered fibroglandular densities | ||||
| Recall Rate | 7.4 | (7.1, 7.7) | 7.1 | (7.0, 7.3) |
| Sensitivity | 86.4 | (80.9, 91.9) | 85.1 | (82.0, 88.2) |
| Specificity | 93.0 | (92.7, 93.2) | 93.3 | (93.1, 93.4) |
| PPV1 | 5.3 | (4.4, 6.2) | 6.1 | (5.6, 6.7) |
| Heterogeneously dense or extremely dense | ||||
| Recall Rate | 10.1 | (9.7, 10.5) | 9.9 | (9.7, 10.1) |
| Sensitivity | 80.8 | (73.8, 87.9) | 79.8 | (76.4, 83.2) |
| Specificity | 90.3 | (89.9, 90.7) | 90.5 | (90.3, 90.7) |
| PPV1 | 4.4 | (3.5, 5.2) | 4.7 | (4.3, 5.2) |
PPV1 = positive predictive value; CI = confidence interval
Although performance was similar between blacks and whites, black women were diagnosed with significantly higher proportions of DCIS with comedo necrosis (p-value=0.04), poorly differentiated invasive tumors (p-value=0.024), and ER-negative (p-value<0.001) and PR-negative (p-value=0.004) tumors (Table 3). There was limited power to evaluate whether racial differences in tumor characteristics by race differed if the tumor was detected by mammography or not (true positive versus false negative); however, racial disparities in tumor grade, ER, and PR expression were similar for true positive and false negative tumors.
Table 3.
Pathologic characteristics of breast cancers in black and white women in the Carolina Mammography Registry, 2005–2010
| Pathologic Characteristic | ALL CANCERS | TRUE POSITIVE CANCERS | FALSE NEGATIVE CANCERS | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Black | White | p-value | Black | White | p-value | Black | White | p-value | |||||||
| N | %* | N | %* | N | %* | N | %* | N | %* | N | %* | ||||
| Total cancers | 276 | 1,095 | 231 | 902 | 45 | 193 | |||||||||
| Type | |||||||||||||||
| Invasive | 201 | 72.8 | 840 | 76.7 | 0.041 | 160 | 69.3 | 674 | 74.7 | 0.093 | 41 | 91.1 | 166 | 86.0 | 0.36 |
| Ductal Carcinoma In-Situ | 75 | 27.2 | 255 | 23.3 | 71 | 30.7 | 228 | 25.3 | 4 | 8.9 | 27 | 14.0 | |||
| DCIS | |||||||||||||||
| Grade | |||||||||||||||
| Well differentiated | 9 | 14.8 | 36 | 17.4 | 0.88 | 9 | 15.5 | 30 | 16.0 | 0.90 | n/a | n/a | n/a | ||
| Moderately differentiated | 25 | 41.0 | 80 | 38.7 | 24 | 41.4 | 71 | 38.0 | |||||||
| Poorly differentiated | 27 | 44.3 | 91 | 44.0 | 25 | 43.1 | 86 | 46.0 | |||||||
| Missing | 14 | --- | 48 | --- | 13 | --- | 41 | --- | |||||||
| Histologic Subtype | |||||||||||||||
| Comedo | 20 | 29.0 | 40 | 17.6 | 0.04 | 19 | 29.2 | 35 | 17.2 | 0.036 | n/a | n/a | n/a | ||
| Non-Comedo | 49 | 71.0 | 187 | 82.4 | 46 | 70.8 | 168 | 82.8 | |||||||
| Missing | 6 | --- | 28 | --- | 6 | --- | 25 | --- | |||||||
| Invasive | |||||||||||||||
| Late Stage | |||||||||||||||
| Yes | 49 | 25.3 | 229 | 28.4 | 0.38 | 37 | 23.7 | 176 | 27.2 | 0.37 | 12 | 31.9 | 53 | 33.1 | 0.86 |
| No | 145 | 74.7 | 577 | 71.6 | 119 | 76.3 | 470 | 72.8 | 26 | 68.4 | 107 | 66.9 | |||
| Missing | 7 | 34 | --- | 4 | --- | 28 | --- | 3 | --- | 6 | --- | ||||
| Grade | |||||||||||||||
| Well differentiated | 42 | 22.7 | 233 | 30.2 | 0.024 | 36 | 24.7 | 198 | 31.9 | 0.059 | 6 | 15.4 | 35 | 23.0 | 0.44 |
| Moderately differentiated | 78 | 42.2 | 338 | 43.8 | 64 | 43.8 | 280 | 45.2 | 14 | 35.9 | 58 | 38.2 | |||
| Poorly differentiated | 65 | 35.1 | 201 | 26.0 | 46 | 31.5 | 142 | 22.9 | 19 | 48.7 | 59 | 38.8 | |||
| Missing | 16 | --- | 68 | --- | 14 | --- | 54 | --- | 2 | --- | 14 | --- | |||
| Size | |||||||||||||||
| <=10 mm | 57 | 28.8 | 244 | 29.9 | 0.47 | 50 | 31.9 | 212 | 32.4 | 0.24 | 7 | 17.1 | 32 | 19.9 | 0.14 |
| 11 – 20 mm | 66 | 33.3 | 270 | 33.1 | 58 | 36.9 | 212 | 32.4 | 8 | 19.5 | 58 | 36.0 | |||
| 21 – 30 mm | 36 | 18.2 | 115 | 14.1 | 24 | 15.3 | 82 | 12.5 | 12 | 29.3 | 33 | 20.5 | |||
| >30 mm | 39 | 19.7 | 186 | 22.8 | 25 | 15.9 | 148 | 22.6 | 14 | 34.2 | 38 | 23.6 | |||
| Missing | 3 | --- | 25 | --- | 3 | --- | 20 | --- | 0 | --- | 5 | --- | |||
| Nodal status | |||||||||||||||
| Positive | 56 | 30.11 | 187 | 24.3 | 0.10 | 43 | 28.9 | 137 | 22.2 | 0.087 | 13 | 35.1 | 50 | 32.5 | 0.76 |
| Negative | 130 | 69.89 | 583 | 75.7 | 106 | 71.1 | 479 | 77.8 | 24 | 64.9 | 104 | 67.5 | |||
| Missing | 15 | --- | 70 | --- | 11 | --- | 58 | --- | 4 | --- | 12 | --- | |||
| Estrogen receptor status | |||||||||||||||
| Positive | 127 | 70.6 | 624 | 83.8 | <0.0001 | 105 | 73.4 | 512 | 85.9 | 0.0003 | 22 | 59.5 | 112 | 75.2 | 0.057 |
| Negative | 53 | 29.4 | 121 | 16.2 | 38 | 26.6 | 84 | 14.1 | 15 | 40.5 | 37 | 24.8 | |||
| Missing | 21 | --- | 95 | --- | 17 | --- | 78 | --- | 4 | --- | 17 | --- | |||
| Progesterone receptor status | |||||||||||||||
| Positive | 111 | 62.0 | 541 | 72.8 | 0.0044 | 92 | 64.8 | 445 | 74.9 | 0.015 | 19 | 51.3 | 96 | 64.4 | 0.14 |
| Negative | 68 | 38.0 | 202 | 27.2 | 50 | 35.2 | 149 | 25.1 | 18 | 48.7 | 53 | 35.6 | |||
| Missing | 22 | --- | 97 | --- | 18 | --- | 80 | --- | 4 | --- | 17 | --- | |||
Percentages are of the non-missing
n/a = not applicable as the numbers are too small to report
Conclusions
Our finding of no difference in the performance of digital screening mammography between black and white women, even after controlling for possible confounders, is in agreement with a previous report of no difference in film-screen mammography performance by race.[2] Although we found the performance was similar, the types of tumors identified by digital screening mammography differed by race. As has been reported in previous studies of film-screen detected breast cancers, black women were more likely to be diagnosed with higher grade tumors among both DCIS and invasive lesions and with ER or PR-negative tumors [1], suggesting that these differences are likely not caused by the rate of screening-related detection. Our study is the first to compare the performance of digital mammography by race and included over 160,000 women with over 1,300 breast cancers. Future work seeking to explain racial differences in tumor characteristics in a screening population should focus on etiologic risk factors that may influence tumor biology.
Acknowledgments
Funding
This work was supported by the National Cancer Institute funded Breast Cancer Surveillance Consortium (HHSN261201100031C) and the National Cancer Institute funded Risk-Based Breast Cancer Screening in Community Settings grant (P01CA154292).
The collection of cancer and vital status data used in this study was supported by the North Carolina state public health department and cancer registry. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health. We thank the participating women, mammography facilities, and radiologists for the data they have provided for this study.
Footnotes
Conflict of Interest
The authors declare that they have no conflict of interest.
Ethical Approval
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
References
- 1.Wheeler SB, Reeder-Hayes KE, Carey LA. Disparities in breast cancer treatment and outcomes: biological, social, and health system determinants and opportunities for research. Oncologist. 2013;18(9):986–93. doi: 10.1634/theoncologist.2013-0243. Epub 2013 Aug 12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Gill KS, Yankaskas BC. Screening mammography performance and cancer detection among black women and white women in community practice. Cancer. 2004 Jan 1;100(1):139–48. doi: 10.1002/cncr.11878. [DOI] [PubMed] [Google Scholar]
- 3.Bluekens AM, Holland R, Karssenmeijer N, Broeders MJ, den Heeten GJ. Comparison of digital screening mammography and screen-film mammography in the early detection of clinically relevant cancers: a multicenter study. Radiology. 2012 Dec;265(3):707–14. doi: 10.1148/radiol.12111461. [DOI] [PubMed] [Google Scholar]
- 4.Pisano ED, Gastonis C, Hendrick E, et al. Diagnostic performance of digital versus film mammography for breast-cancer screening. N Engl J Med. 2005 Oct 27;353(17):1773–83. doi: 10.1056/NEJMoa052911. [DOI] [PubMed] [Google Scholar]
- 5.American Cancer Society. Breast Cancer Facts & Figures 2013–2014. Atlanta: American Cancer Society, Inc; 2013. [Google Scholar]
- 6.Habel LA, Capra AM, Oestreicher N, et al. Mammographic density in a multiethnic cohort. Menopause. 2007 Sep-Oct;14(5):891–9. doi: 10.1097/gme.0b013e318032569c. [DOI] [PubMed] [Google Scholar]
- 7.Conroy SM, Butler LM, Harvey D, et al. Physical activity and change in mammographic density: the study of Women’s Health Across the Nation. Am J Epidemiol. 2010 May 1;171(9):960–8. doi: 10.1093/aje/kwq025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.American College of Radiology. The American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) 4. Reston, VA: American Coll Radiology; 2003. [Google Scholar]
- 9.Yankaskas BC, Taplin SH, Ichikawa L, Geller BM, Rosenberg RD, Carney PA, et al. Association between mammography timing and measures of screening performance in the United States. Radiology. 2005;234:363–73. doi: 10.1148/radiol.2342040048. [DOI] [PubMed] [Google Scholar]
- 10.Breast Cancer Surveillance Consortium. [Accessed April 4, 2015];(BCSC): performance benchmarks for screening mammography. http://breastscreening.cancer.gov/data/benchmarks/screening/
- 11.McLean RA, Sanders WL, Stroup WW. A Unified Approach to Mixed Linear Models. The American Statistician. 1991;45:54–64. [Google Scholar]
- 12.Kerlikowske K, Hubbard RA, Miglioretti DL, et al. Comparative Effectiveness of Digital Versus Film-Screen Mammography in Community Practice in the United States: A Cohort Study. Ann Intern Med. 2011;155:493–502. doi: 10.7326/0003-4819-155-8-201110180-00005. [DOI] [PMC free article] [PubMed] [Google Scholar]