Background:

We evaluated differences in diagnostic mammography performance based on women's race/ethnicity.

Methods:

This cohort study included 267,868 diagnostic mammograms performed to evaluate screening mammogram findings at 98 facilities in the Breast Cancer Surveillance Consortium between 2005 and 2017. Mammogram assessments were recorded prospectively and breast cancers occurring within one year were ascertained. Performance statistics were calculated with 95% confidence intervals (CI) for each racial/ethnic group. Multivariable regression was used to control for personal characteristics and imaging facility.

Results:

Among non-Hispanic White (70%), non-Hispanic Black (13%), Asian/Pacific Islander (10%), and Hispanic (7%) women, the invasive cancer detection rate (iCDR, per 1,000 mammograms) and positive predictive value (PPV2) were highest among non-Hispanic White women (iCDR, 35.8; 95% CI, 35.0–36.7; PPV2, 27.8; 95% CI, 27.3–28.3) and lowest among Hispanic women (iCDR, 22.3; 95% CI, 20.2–24.6; PPV2, 19.4; 95% CI, 18.0–20.9). Short interval follow-up recommendations were most common among non-Hispanic Black women [(31.0%; 95% CI, 30.6%–31.5%) vs. other groups, range, 16.6%–23.6%]. False-positive biopsy recommendations were most common among Asian/Pacific Islander women [per 1,000 mammograms: 169.2; 95% CI, 164.8–173.7) vs. other groups, range, 126.5–136.1]. Some differences were explained by adjusting for receipt of diagnostic ultrasound or MRI for iCDR and imaging facility for short-interval follow-up. Other differences changed little after adjustment.

Conclusions:

Diagnostic mammography performance varied across racial/ethnic groups. Addressing characteristics related to imaging facility and access, rather than personal characteristics, may help reduce some of these disparities.

Impact:

Diagnostic mammography performance studies should include racially and ethnically diverse populations to provide an accurate view of the population-level effects.

Racial and ethnic differences in breast cancer incidence and mortality are well documented. In the US, breast cancer incidence is highest among non-Hispanic White and non-Hispanic Black women and breast cancer–related mortality is highest for non-Hispanic Black women, whereas incidence and mortality rates are lowest for Hispanic and Asian/Pacific Islander women (1). Differences are also apparent in the types of breast cancers diagnosed. Non-Hispanic Black women are more likely to be diagnosed with tumors that have poorer prognostic characteristics (e.g., large size, high grade, hormone receptor negative) compared with other groups (1). Hispanic women are more commonly diagnosed with larger breast tumors and regional or distant-stage disease when compared with non-Hispanic White women (2–4). Efforts to determine the root causes of these differences have identified racial/ethnic differences in risk factors, socioeconomic status, health insurance coverage, provider actions/inaction, access to timely treatment, and access to high-quality care as contributors (5, 6). However, differences are not eliminated completely after controlling for these factors (4, 7–9), suggesting that additional causes may be involved.

Mammography plays a major role in breast cancer detection as a screening and diagnostic tool, but its potential contributions to racial differences in breast cancer characteristics and mortality have not been fully established. Although several studies have addressed racial/ethnic differences in the timeliness of breast cancer screening and receipt of care after an abnormal screening mammogram (3, 10–15), little is known about differences in other parts of the diagnostic work-up continuum, particularly outcomes following a diagnostic mammography examination. Studies addressing this have analyzed film screen mammography (a technology that is no longer used) and suggest that diagnostic mammography performance varies across racial/ethnic groups, which could affect the characteristics of cancers diagnosed. Diagnostic film mammography sensitivity is higher for Black women compared with White women and specificity is lower for Black women compared with White women (16). Other data show that rates of abnormal interpretation and cancer detection differ for digital versus film screen mammography (17, 18). However, these data were not reported for racial/ethnic groups, leaving the question of whether the racial and ethnic differences reported for diagnostic performance with film screen mammography persist for digital mammography.

The goal of this study was to evaluate racial/ethnic differences in the performance of diagnostic digital mammography conducted following a recent screening mammogram. Using prospectively collected data from six population-based breast imaging registries, we evaluated diagnostic mammography performance statistics among non-Hispanic White, non-Hispanic Black, Asian/Pacific Islander, and Hispanic women. Examining differences in diagnostic digital mammography performance and tumor characteristic outcomes by race and ethnicity may help to understand why disparities in cancer detection and quality of care may persist for some demographic groups.

Study population

This study was conducted using data from the Breast Cancer Surveillance Consortium (BCSC; ref. 19), a collaborative network of breast imaging registries that collects data on breast imaging examinations, breast procedures, and cancer diagnoses occurring among women seen at participating facilities. This analysis included data from digital diagnostic mammograms [including full-field digital mammograms (FFDM) and digital breast tomosynthesis (DBT)] performed between 2005 and 2017 for women ages ≥18 years at imaging facilities in six registries: Carolina Mammography Registry, Kaiser Permanente Washington Registry, Metropolitan Chicago Breast Cancer Registry, New Hampshire Mammography Network, San Francisco Mammography Registry, and Vermont Breast Cancer Surveillance system. Each BCSC registry and the Statistical Coordinating Center received institutional review board approval for all study procedures. All procedures were Health Insurance Portability and Accountability Act compliant. All registries and the Statistical Coordinating Center received a Federal Certificate of Confidentiality and other protections for the identities of individuals, physicians, and facilities that contributed to this research. Overall, diagnostic exams from 98 imaging facilities were included in our study.

Mammograms

This study included diagnostic digital mammograms (FFDM or DBT) that occurred up to 90 days after a screening mammogram and had an indication of “additional evaluation of a recent mammogram” (Fig. 1). In addition, we included diagnostic mammograms with a missing indication if they occurred up to 90 days after a screening mammogram that had a Breast Imaging and Reporting Data Systems (BI-RADS) assessment of 0 (needs additional imaging evaluation), 3 (probably benign), 4 (suspicious for malignancy), or 5 (highly suggestive of malignancy).

Figure 1.

Diagnostic mammogram inclusion and exclusion criteria. This study included digital diagnostic mammograms performed at an imaging facility affiliated with the Breast Cancer Surveillance Consortium (BCSC) between 2005 and 2017. All diagnostic mammograms were follow-ups to a digital screening mammogram that had occurred within the prior 90 days. NOS, not otherwise specified.

Figure 1.

Diagnostic mammogram inclusion and exclusion criteria. This study included digital diagnostic mammograms performed at an imaging facility affiliated with the Breast Cancer Surveillance Consortium (BCSC) between 2005 and 2017. All diagnostic mammograms were follow-ups to a digital screening mammogram that had occurred within the prior 90 days. NOS, not otherwise specified.

Close modal

The final diagnostic BI-RADS assessment was used to determine the diagnostic mammogram outcome. Diagnostic mammograms assigned a BI-RADS assessment of 4 or 5 were considered positive and mammograms with an assessment of 1 (negative), 2 (benign finding), or 3 were considered negative. If an assessment was BI-RADS 0 or missing, then the diagnostic mammogram was followed for 90 days to determine the final assessment using Materials and Methods described previously (17, 18). For the small number of mammograms (N = 1,487, <1%) where the final BI-RADS assessment during the 90-day period was 0, we imputed a positive/negative result based on age, facility, reader and cancer outcome (20).

Covariates

Data on demographics and health history were self-reported at each imaging visit or collected from electronic health records. These data included self-reported race and ethnicity, age, first-degree family history of breast cancer, history of breast procedures, and time since last mammogram. Whether women received a diagnostic ultrasound and/or MRI within 90 days after the additional evaluation diagnostic mammogram was determined on the basis of registry imaging records. BI-RADS breast density was classified by the interpreting radiologist as almost entirely fatty (i), scattered fibroglandular densities (ii), heterogeneously dense (iii), or extremely dense (iv). Responses to health history questions were used to calculate the predicted risk of developing breast cancer during the next 5 years using the validated BCSC Risk Calculator (21–23). All information was reported at the time of the diagnostic mammogram, except time since a woman's last mammogram and breast density, which were reported at the screening mammogram immediately before the diagnostic mammogram.

Breast cancer ascertainment

Invasive breast carcinoma and ductal carcinoma in situ (DCIS) diagnoses occurring within one year of the diagnostic mammogram were ascertained through linkage with state or regional cancer registries; Surveillance, Epidemiology, and End Results registries; and pathology databases. Anatomic stage at diagnosis was classified using the AJCC 8th edition definitions (24). Tumor size, tumor grade, axillary lymph node status, and hormone receptor status were tabulated among invasive breast cancers only, due to the low frequency with which these characteristics were reported for DCIS. A combined estrogen receptor, progesterone receptor, and HER2 variable was constructed to identify triple-negative breast cancers. Categorizations were chosen based on prior studies that have used IHC markers to approximate breast cancer molecular subtypes (25, 26).

Analytic population

A detailed schematic of how the final study population was identified is in Fig. 1. Briefly, we included diagnostic mammograms performed between January 2005 and June 2017 if: (i) the participant was ≥18 years old (ii); the diagnostic mammogram met our criteria for additional evaluation of a recent screening mammogram (described above); and (iii) there was at least one year of follow-up time after the diagnostic mammogram. The following diagnostic mammograms were excluded if: (i) the individual had a prior breast cancer diagnosis, mastectomy, or breast implants (ii); the individual had a diagnostic mammogram in the preceding 90 days (iii); the screening mammogram before the diagnostic mammogram was not a FFDM or DBT exam; or (iv) the diagnostic mammogram had a final BI-RADS assessment of 6 (biopsy-proven cancer) or missing. Mammograms with race classified as American Indian/Alaska Native, other race, or mixed race were excluded due to the small sample size in each group. Examinations were also excluded if race/ethnicity was missing.

Statistical analysis

Race and Hispanic ethnicity were combined into a single variable and mammogram and breast cancer characteristics were tabulated by racial/ethnic group. Means and standard deviations (SD) were calculated for continuous variables. Differences in categorical variable distributions were assessed using the χ2 test. Diagnostic mammography performance metrics were assessed as recommended by the American College of Radiology and are defined in Table 1 (27). Sensitivity, cancer detection rate, and false-negative rate were also calculated considering invasive breast cancers only, to explore the influence of DCIS diagnoses on the results. Wald 95% confidence intervals (CI) were calculated for each statistic using previously described Materials and Methods (17, 27).

Table 1.

Definitions of diagnostic mammography performance statistics.

StatisticDefinition
Biopsy recommendation proportion Proportion of mammograms that had a positive assessment (BI-RADSa 4 or 5) 
Positive predictive value 2 (PPV2) Proportion of mammograms recommended for biopsy or surgical consultation that were followed by a breast cancer diagnosis within 1 year 
Positive predictive value 3 (PPV3) Proportion of biopsies performed following a positive mammogram that were followed by a diagnosis of breast cancer within 1 year 
Cancer detection rate Number of breast cancers diagnosed following a positive mammogram assessment, per 1,000 mammograms performed 
Sensitivity Number of breast cancers detected by mammography divided by the total number of breast cancers diagnosed 
False-positive rate Number of positive mammograms that were not followed by a breast cancer diagnosis within 1 year, per 1,000 mammograms 
False-negative rate Number of negative mammograms that were followed by a breast cancer diagnosis within 1 year, per 1,000 mammograms 
Short interval follow-up recommendation proportion Proportion of mammograms with a BI-RADSa assessment of 3 
Short interval follow-up cancer yield Proportion of mammograms recommended for short interval follow-up that were followed by a breast cancer diagnosis within 1 year 
StatisticDefinition
Biopsy recommendation proportion Proportion of mammograms that had a positive assessment (BI-RADSa 4 or 5) 
Positive predictive value 2 (PPV2) Proportion of mammograms recommended for biopsy or surgical consultation that were followed by a breast cancer diagnosis within 1 year 
Positive predictive value 3 (PPV3) Proportion of biopsies performed following a positive mammogram that were followed by a diagnosis of breast cancer within 1 year 
Cancer detection rate Number of breast cancers diagnosed following a positive mammogram assessment, per 1,000 mammograms performed 
Sensitivity Number of breast cancers detected by mammography divided by the total number of breast cancers diagnosed 
False-positive rate Number of positive mammograms that were not followed by a breast cancer diagnosis within 1 year, per 1,000 mammograms 
False-negative rate Number of negative mammograms that were followed by a breast cancer diagnosis within 1 year, per 1,000 mammograms 
Short interval follow-up recommendation proportion Proportion of mammograms with a BI-RADSa assessment of 3 
Short interval follow-up cancer yield Proportion of mammograms recommended for short interval follow-up that were followed by a breast cancer diagnosis within 1 year 

aBI-RADS, Breast Imaging Reporting and Data Systems.

To explore the degree to which between-group differences in performance statistics may be influenced by personal or examination-related characteristics, we used logistic regression to model the probability of a given outcome with race/ethnicity as the independent variable and adjusting for additional factors. Regression models were constructed sequentially, first adjusting for mammography registry and age (continuous and age-squared), then additionally adjusting for family history of breast cancer, breast density, prior biopsy, type of diagnostic mammogram (2-D vs. 3-D), type of screening mammogram (2-D vs. 3-D), time since last mammogram, receipt of a diagnostic ultrasound and/or MRI. Imaging facility was entered into the model as a fixed effect. Models included all observations with non-missing data at each step, such that the number of observations in each sequential model varied.

Sensitivity analysis

2.6% of the exams in this analysis were conducted among women who reported a breast problem (e.g., pain, lump, nipple discharge, other problem not specified) at the time of their screening mammogram. We estimated performance statistics excluding these exams to determine whether their inclusion influenced the results. In addition, we constructed logistic regression models using only the subset of observations with non-missing data for all covariates to determine whether variation in the sample used for each model influenced the observed racial/ethnic differences in performance.

Statistical analyses were conducted using SAS v9.4 (SAS Institute). All statistical tests were two-sided and P values less than 0.05 were considered statistically significant.

Data availability

The data analyzed in this study may be accessed through the BCSC (19) upon reasonable request. Restrictions apply to the availability of these data due to patient privacy requirements.

Population characteristics

A total of 267,868 digital diagnostic mammograms performed among 234,818 unique women were included in this study (Fig. 1). The population included exams from individuals who identified as non-Hispanic White (70%), non-Hispanic Black (13%), Asian/Pacific Islander (10%) or Hispanic (7%), with a mean age at mammography of 55 years (SD, 11 years; Table 2). For about half of the exams, ultrasound was used in addition to mammography to determine the BI-RADS assessment, with receipt highest among Hispanic women (55%) and lowest among non-Hispanic Black women (43%; Table 2).

Table 2.

Personal characteristics associated with 267,868 diagnostic mammograms performed as an additional evaluation of a prior screening mammogram among 234,818 women in the Breast Cancer Surveillance Consortium, 2005–2017.

Race and ethnicity
Overall (N = 267,868)Non-Hispanic White (N = 188,268)Non-Hispanic Black (N = 34,573)Asian/Pacific Islander (N = 27,391)Hispanic (N = 17,636)
CharacteristicN (%)N (%)N (%)N (%)N (%)
Age, y 
 <40 7,771 (2.9) 5,305 (2.8) 1,070 (3.1) 723 (2.6) 673 (3.8) 
 40–49 88,649 (33.1) 60,820 (32.3) 10,189 (29.5) 9,688 (35.4) 7,952 (45.1) 
 50–59 80,741 (30.1) 56,112 (29.8) 10,460 (30.3) 9,195 (33.6) 4,974 (28.2) 
 60–69 57,777 (21.6) 41,606 (22.1) 7,821 (22.6) 5,581 (20.4) 2,769 (15.7) 
 70–79 25,984 (9.7) 19,044 (10.1) 4,064 (11.8) 1,830 (6.7) 1,046 (5.9) 
 ≥80 6,946 (2.6) 5,381 (2.9) 969 (2.8) 374 (1.4) 222 (1.3) 
 Mean (SD) 55.3 (11.3) 55.6 (11.4) 56.2 (11.5) 53.9 (10.2) 52.0 (10.4) 
First-degree family history of breast cancer 
 No 212,929 (82.8) 145,724 (81.0) 28,656 (83.7) 23,592 (90.1) 14,957 (88.2) 
 Yes 44,383 (17.2) 34,188 (19.0) 5,600 (16.3) 2,590 (9.9) 2,005 (11.8) 
Unknown 10,556 8,356 317 1,209 674 
History of breast biopsy or aspiration 
 No 201,632 (78.4) 137,152 (76.4) 27,709 (80.7) 22,241 (84.2) 14,530 (84.6) 
 Yes 55,679 (21.6) 42,266 (23.6) 6,615 (19.3) 4,163 (15.8) 2,635 (15.4) 
Unknown 10,557 8,850 249 987 471 
Type of additional evaluation 
 Digital (2-D) 257,930 (96.3) 180,026 (95.6) 33,914 (98.1) 26,785 (97.8) 17,205 (97.6) 
 DBTa (3-D) 9,938 (3.7) 8,242 (4.4) 659 (1.9) 606 (2.2) 431 (2.4) 
Type of previous screening mammogram 
 Digital (2-D) 255,336 (95.3) 177,527 (94.3) 34,081 (98.6) 26,650 (97.3) 17,078 (96.8) 
 DBTa (3-D) 12,532 (4.7) 10,741 (5.7) 492 (1.4) 741 (2.7) 558 (3.2) 
Time since last mammogram (mo) 
 No prior mammogram 30,760 (12.4) 17,917 (10.2) 4,600 (15.2) 4,678 (18.4) 3,565 (22.1) 
 <12 6,408 (2.6) 4,531 (2.6) 810 (2.7) 691 (2.7) 376 (2.3) 
 12–23 148,535 (59.8) 110,318 (62.5) 17,327 (57.1) 13,121 (51.6) 7,769 (48.2) 
 24–35 31,270 (12.6) 21,952 (12.4) 3,463 (11.4) 3,729 (14.7) 2,126 (13.2) 
 >35 31,413 (12.6) 21,746 (12.3) 4,152 (13.7) 3,228 (12.7) 2,287 (14.2) 
Unknown 19,482 11,804 4,221 1,944 1,513 
Diagnostic US/MRIb within 90 days 
 None 139,024 (51.9) 97,314 (51.7) 19,439 (56.2) 14,403 (52.6) 7,868 (44.6) 
 US only 126,620 (47.3) 89,255 (47.4) 14,848 (42.9) 12,846 (46.9) 9,671 (54.8) 
 MRI (with or without USc2,224 (0.8) 1,699 (0.9) 286 (0.8) 142 (0.5) 97 (0.6) 
Breast density 
 Almost entirely fat 14,986 (6.0) 11,064 (6.3) 2,179 (6.5) 637 (2.6) 1,106 (6.7) 
 Scattered fibroglandular densities 102,455 (40.9) 73,641 (41.7) 15,784 (47.3) 6,645 (27.3) 6,385 (38.8) 
 Heterogeneously dense 113,398 (45.2) 78,209 (44.3) 14,057 (42.1) 13,285 (54.6) 7,847 (47.7) 
 Extremely dense 19,938 (8.0) 13,705 (7.8) 1,354 (4.1) 3,751 (15.4) 1,128 (6.9) 
Unknown 17,091 11,649 1,199 3,073 1,170 
BCSC 5-year predicted breast cancer risk (%) 
 <1 84,929 (36.3) 50,655 (30.9) 11,324 (36.8) 12,561 (53.6) 10,389 (65.5) 
 1.00–1.66 85,279 (36.4) 59,987 (36.6) 11,894 (38.6) 9,037 (38.6) 4,361 (27.5) 
 1.67–2.49 43,090 (18.4) 34,877 (21.3) 5,747 (18.7) 1,537 (6.6) 929 (5.9) 
 2.50–3.99 18,157 (7.8) 16,044 (9.8) 1,652 (5.4) 281 (1.2) 180 (1.1) 
 ≥4.00 2,657 (1.1) 2,444 (1.5) 176 (0.6) 26 (0.1) 11 (0.1) 
Unknown 33,756 24,261 3,780 3,949 1,766 
 Mean (SD) 1.4 (0.8) 1.5 (0.9) 1.3 (0.7) 1.0 (0.5) 0.9 (0.5) 
Race and ethnicity
Overall (N = 267,868)Non-Hispanic White (N = 188,268)Non-Hispanic Black (N = 34,573)Asian/Pacific Islander (N = 27,391)Hispanic (N = 17,636)
CharacteristicN (%)N (%)N (%)N (%)N (%)
Age, y 
 <40 7,771 (2.9) 5,305 (2.8) 1,070 (3.1) 723 (2.6) 673 (3.8) 
 40–49 88,649 (33.1) 60,820 (32.3) 10,189 (29.5) 9,688 (35.4) 7,952 (45.1) 
 50–59 80,741 (30.1) 56,112 (29.8) 10,460 (30.3) 9,195 (33.6) 4,974 (28.2) 
 60–69 57,777 (21.6) 41,606 (22.1) 7,821 (22.6) 5,581 (20.4) 2,769 (15.7) 
 70–79 25,984 (9.7) 19,044 (10.1) 4,064 (11.8) 1,830 (6.7) 1,046 (5.9) 
 ≥80 6,946 (2.6) 5,381 (2.9) 969 (2.8) 374 (1.4) 222 (1.3) 
 Mean (SD) 55.3 (11.3) 55.6 (11.4) 56.2 (11.5) 53.9 (10.2) 52.0 (10.4) 
First-degree family history of breast cancer 
 No 212,929 (82.8) 145,724 (81.0) 28,656 (83.7) 23,592 (90.1) 14,957 (88.2) 
 Yes 44,383 (17.2) 34,188 (19.0) 5,600 (16.3) 2,590 (9.9) 2,005 (11.8) 
Unknown 10,556 8,356 317 1,209 674 
History of breast biopsy or aspiration 
 No 201,632 (78.4) 137,152 (76.4) 27,709 (80.7) 22,241 (84.2) 14,530 (84.6) 
 Yes 55,679 (21.6) 42,266 (23.6) 6,615 (19.3) 4,163 (15.8) 2,635 (15.4) 
Unknown 10,557 8,850 249 987 471 
Type of additional evaluation 
 Digital (2-D) 257,930 (96.3) 180,026 (95.6) 33,914 (98.1) 26,785 (97.8) 17,205 (97.6) 
 DBTa (3-D) 9,938 (3.7) 8,242 (4.4) 659 (1.9) 606 (2.2) 431 (2.4) 
Type of previous screening mammogram 
 Digital (2-D) 255,336 (95.3) 177,527 (94.3) 34,081 (98.6) 26,650 (97.3) 17,078 (96.8) 
 DBTa (3-D) 12,532 (4.7) 10,741 (5.7) 492 (1.4) 741 (2.7) 558 (3.2) 
Time since last mammogram (mo) 
 No prior mammogram 30,760 (12.4) 17,917 (10.2) 4,600 (15.2) 4,678 (18.4) 3,565 (22.1) 
 <12 6,408 (2.6) 4,531 (2.6) 810 (2.7) 691 (2.7) 376 (2.3) 
 12–23 148,535 (59.8) 110,318 (62.5) 17,327 (57.1) 13,121 (51.6) 7,769 (48.2) 
 24–35 31,270 (12.6) 21,952 (12.4) 3,463 (11.4) 3,729 (14.7) 2,126 (13.2) 
 >35 31,413 (12.6) 21,746 (12.3) 4,152 (13.7) 3,228 (12.7) 2,287 (14.2) 
Unknown 19,482 11,804 4,221 1,944 1,513 
Diagnostic US/MRIb within 90 days 
 None 139,024 (51.9) 97,314 (51.7) 19,439 (56.2) 14,403 (52.6) 7,868 (44.6) 
 US only 126,620 (47.3) 89,255 (47.4) 14,848 (42.9) 12,846 (46.9) 9,671 (54.8) 
 MRI (with or without USc2,224 (0.8) 1,699 (0.9) 286 (0.8) 142 (0.5) 97 (0.6) 
Breast density 
 Almost entirely fat 14,986 (6.0) 11,064 (6.3) 2,179 (6.5) 637 (2.6) 1,106 (6.7) 
 Scattered fibroglandular densities 102,455 (40.9) 73,641 (41.7) 15,784 (47.3) 6,645 (27.3) 6,385 (38.8) 
 Heterogeneously dense 113,398 (45.2) 78,209 (44.3) 14,057 (42.1) 13,285 (54.6) 7,847 (47.7) 
 Extremely dense 19,938 (8.0) 13,705 (7.8) 1,354 (4.1) 3,751 (15.4) 1,128 (6.9) 
Unknown 17,091 11,649 1,199 3,073 1,170 
BCSC 5-year predicted breast cancer risk (%) 
 <1 84,929 (36.3) 50,655 (30.9) 11,324 (36.8) 12,561 (53.6) 10,389 (65.5) 
 1.00–1.66 85,279 (36.4) 59,987 (36.6) 11,894 (38.6) 9,037 (38.6) 4,361 (27.5) 
 1.67–2.49 43,090 (18.4) 34,877 (21.3) 5,747 (18.7) 1,537 (6.6) 929 (5.9) 
 2.50–3.99 18,157 (7.8) 16,044 (9.8) 1,652 (5.4) 281 (1.2) 180 (1.1) 
 ≥4.00 2,657 (1.1) 2,444 (1.5) 176 (0.6) 26 (0.1) 11 (0.1) 
Unknown 33,756 24,261 3,780 3,949 1,766 
 Mean (SD) 1.4 (0.8) 1.5 (0.9) 1.3 (0.7) 1.0 (0.5) 0.9 (0.5) 

aDBT, digital breast tomosynthesis.

bUS, ultrasound; MRI, magnetic resonance imaging.

c379 exams with MRI and US.

Diagnostic mammography assessments

The proportion of exams with a negative assessment was similar across racial/ethnic groups after adjustment for mammography registry (BI-RADS 1 or 2 assessment, 55%–61%; Table 3). An assessment of BI-RADS 3, typically used to denote an abnormality that should be followed with additional imaging within 6 months, was most common among non-Hispanic Black women (31%) and least common among Asian/Pacific Islander women (17%). 48,017 (18%) mammograms were assigned a BI-RADS assessment that indicated a suspicion of cancer, the vast majority of which were BI-RADS 4 assessments.

Table 3.

Comparison of final diagnostic mammography assessments for 267,868 additional evaluations of a recent screening mammogram by race/ethnicity, Breast Cancer Surveillance Consortium 2005–2017.

Race and ethnicity
Non-Hispanic WhiteNon-Hispanic BlackAsian/Pacific IslanderHispanic
BI-RADS assessment categorya,bNRaw (%)cAdjusted (%)dNRaw (%)cAdjusted (%)dNRaw (%)cAdjusted (%)dNRaw (%)cAdjusted (%)d
Negative (i) 48,563 26.0 25.1 5,093 14.8 22.0 10,034 36.8 29.5 3,930 22.3 23.9 
Benign (ii) 63,897 34.2 34.8 12,607 36.6 33.4 6,612 24.2 29.5 6,524 37.1 37.1 
Probably benign (iii) 41,681 22.3 22.1 10,733 31.1 25.5 4,535 16.6 20.8 4,155 23.6 23.0 
Suspicious (iv) 30,792 16.5 16.9 5,763 16.7 18.3 5,795 21.2 19.3 2,837 16.1 15.2 
Highly suggestive of malignancy (v) 2,112 1.1 1.1 265 0.8 0.8 315 1.2 0.9 138 0.8 0.8 
Race and ethnicity
Non-Hispanic WhiteNon-Hispanic BlackAsian/Pacific IslanderHispanic
BI-RADS assessment categorya,bNRaw (%)cAdjusted (%)dNRaw (%)cAdjusted (%)dNRaw (%)cAdjusted (%)dNRaw (%)cAdjusted (%)d
Negative (i) 48,563 26.0 25.1 5,093 14.8 22.0 10,034 36.8 29.5 3,930 22.3 23.9 
Benign (ii) 63,897 34.2 34.8 12,607 36.6 33.4 6,612 24.2 29.5 6,524 37.1 37.1 
Probably benign (iii) 41,681 22.3 22.1 10,733 31.1 25.5 4,535 16.6 20.8 4,155 23.6 23.0 
Suspicious (iv) 30,792 16.5 16.9 5,763 16.7 18.3 5,795 21.2 19.3 2,837 16.1 15.2 
Highly suggestive of malignancy (v) 2,112 1.1 1.1 265 0.8 0.8 315 1.2 0.9 138 0.8 0.8 

aBI-RADS, Breast Imaging Reporting and Data Systems.

b1,487 mammograms with a final assessment of BI-RADS 0 are not shown.

cColumn percentage, unadjusted.

dColumn percentage, adjusted for mammography registry using direct standardization.

Diagnostic mammography performance

Exam performance statistics varied by the racial/ethnic group (Table 4). A diagnostic mammogram resulting in a recommendation for biopsy was most common among Asian/Pacific Islander women (22.4% vs. other groups range, 16.9%–17.5%). Biopsy receipt was slightly lower among Hispanic women compared with other women, but the median time between a positive mammogram and biopsy was similar across groups (Supplementary Table S1). Cancer yield among positive mammograms, as measured by PPV2 and PPV3, was highest among non-Hispanic White women (PPV2: non-Hispanic White 27.8% vs. other groups range, 19.4%–24.3%; PPV3: non-Hispanic White 30.5% vs. other groups range, 20.9%–25.9%). All three statistics were lowest among Hispanic women.

Table 4.

Comparison of mammography performance for 267,868 diagnostic mammograms performed for additional evaluation of a recent screening mammogram among 234,818 women in the Breast Cancer Surveillance Consortium by race and ethnicity, 2005–2017.

Performance statisticNon-Hispanic WhiteNon-Hispanic BlackAsian/Pacific IslanderHispanic
Biopsy     
Biopsy recommendation proportion (%; 95% CIa17.5 (17.4–17.7) 17.5 (17.1–17.9) 22.4 (21.9–22.8) 16.9 (16.3–17.5) 
 # Positiveb mammograms 33,003 6,034 6,122 2,979 
 Total # mammograms 188,268 34,573 27,391 17,636 
PPV2c (%; 95% CIa27.8 (27.3–28.3) 23.4 (22.3–24.5) 24.3 (23.2–25.4) 19.4 (18.0–20.9) 
 # True-positive mammograms 9,182 1,412 1,488 579 
 # Positiveb mammograms 33,003 6,034 6,122 2,979 
PPV3c (%; 95% CIa30.5 (30.0–31.1) 25.9 (24.8–27.2) 25.5 (24.3–26.6) 20.9 (19.4–22.5) 
 # True-positive mammograms with biopsies performed 8,811 1,374 1,376 567 
 # Positiveb mammograms with biopsies performed 28,873 5,295 5,405 2,710 
Cancer Detection     
Cancer detection rate, per 1,000 mammograms (95% CIa48.8 (47.8–49.8) 40.8 (38.8–43.0) 54.3 (51.7–57.1) 32.8 (30.2–35.6) 
 # True-positive mammograms 9,182 1,412 1,488 579 
 Total # mammograms 188,268 34,573 27,391 17,636 
Invasive cancer detection rate, per 1,000 mammograms (95% CIa35.8 (35.0–36.7) 29.5 (27.8–31.4) 31.6 (29.6, 33.8) 22.3 (20.2–24.6) 
 # True-positive (invasive) mammograms 6,746 1,021 866 394 
 Total # mammograms 188,268 34,573 27,391 17,636 
Sensitivity, % (95% CIa92.3 (91.8–92.9) 89.8 (88.2–91.3) 94.2 (93.0–95.3) 92.8 (90.5–94.7) 
 # True-positive mammograms 9,182 1,412 1,488 579 
 # Mammograms followed by breast cancer 9,943 1,572 1,579 624 
Sensitivity for invasive cancer, % (95% CIa92.4 (91.7–93.0) 89.7 (87.8–91.4) 94.4 (92.8–95.8) 92.9 (90.1–95.2) 
 # True-positive (invasive) mammograms 6,746 1,021 866 394 
 # Mammograms followed by invasive breast cancer 7,304 1,138 917 424 
False-positive rate, per 1,000 mammograms (95% CIa126.5 (125.0–128.0) 133.7 (130.1–137.3) 169.2 (164.8–173.7) 136.1 (131.1–141.2) 
 # False-positive mammograms 23,821 4,622 4,634 2,400 
 Total # mammograms 188,268 34,573 27,391 17,636 
False-negative rate, per 1,000 mammograms (95% CIa4.0 (3.8–4.3) 4.6 (3.9–5.4) 3.3 (2.7–4.1) 2.6 (1.9–3.4) 
 # False-negative mammograms 761 160 91 45 
 Total # mammograms 188,268 34,573 27,391 17,636 
False-negative rate for invasive cancer, per 1,000 mammograms (95% CIa3.0 (2.7–3.2) 3.4 (2.8–4.1) 1.9 (1.4–2.4) 1.7 (1.1–2.4) 
 # False-negative mammograms (invasive) 558 117 51 30 
 Total # mammograms 188,268 34,573 27,391 17,636 
Short Interval Follow-Up     
Short interval follow-up recommendation proportion (%; 95% CIa22.1 (22.0–22.3) 31.0 (30.6–31.5) 16.6 (16.1–17.0) 23.6 (22.9–24.2) 
 # Mammograms recommended for short interval follow-up 41,681 10,733 4,535 4,155 
 Total # mammograms 188,268 34,573 27,391 17,636 
Short interval follow-up cancer yield (%; 95% CIa1.3 (1.2–1.4) 1.1 (0.9–1.3) 1.4 (1.1–1.8) 0.8 (0.6–1.1) 
 # Short-interval follow-up mammograms with breast cancer during follow-up 537 113 63 34 
 # Mammograms recommended for short interval follow-up 41,681 10,733 4,535 4,155 
Performance statisticNon-Hispanic WhiteNon-Hispanic BlackAsian/Pacific IslanderHispanic
Biopsy     
Biopsy recommendation proportion (%; 95% CIa17.5 (17.4–17.7) 17.5 (17.1–17.9) 22.4 (21.9–22.8) 16.9 (16.3–17.5) 
 # Positiveb mammograms 33,003 6,034 6,122 2,979 
 Total # mammograms 188,268 34,573 27,391 17,636 
PPV2c (%; 95% CIa27.8 (27.3–28.3) 23.4 (22.3–24.5) 24.3 (23.2–25.4) 19.4 (18.0–20.9) 
 # True-positive mammograms 9,182 1,412 1,488 579 
 # Positiveb mammograms 33,003 6,034 6,122 2,979 
PPV3c (%; 95% CIa30.5 (30.0–31.1) 25.9 (24.8–27.2) 25.5 (24.3–26.6) 20.9 (19.4–22.5) 
 # True-positive mammograms with biopsies performed 8,811 1,374 1,376 567 
 # Positiveb mammograms with biopsies performed 28,873 5,295 5,405 2,710 
Cancer Detection     
Cancer detection rate, per 1,000 mammograms (95% CIa48.8 (47.8–49.8) 40.8 (38.8–43.0) 54.3 (51.7–57.1) 32.8 (30.2–35.6) 
 # True-positive mammograms 9,182 1,412 1,488 579 
 Total # mammograms 188,268 34,573 27,391 17,636 
Invasive cancer detection rate, per 1,000 mammograms (95% CIa35.8 (35.0–36.7) 29.5 (27.8–31.4) 31.6 (29.6, 33.8) 22.3 (20.2–24.6) 
 # True-positive (invasive) mammograms 6,746 1,021 866 394 
 Total # mammograms 188,268 34,573 27,391 17,636 
Sensitivity, % (95% CIa92.3 (91.8–92.9) 89.8 (88.2–91.3) 94.2 (93.0–95.3) 92.8 (90.5–94.7) 
 # True-positive mammograms 9,182 1,412 1,488 579 
 # Mammograms followed by breast cancer 9,943 1,572 1,579 624 
Sensitivity for invasive cancer, % (95% CIa92.4 (91.7–93.0) 89.7 (87.8–91.4) 94.4 (92.8–95.8) 92.9 (90.1–95.2) 
 # True-positive (invasive) mammograms 6,746 1,021 866 394 
 # Mammograms followed by invasive breast cancer 7,304 1,138 917 424 
False-positive rate, per 1,000 mammograms (95% CIa126.5 (125.0–128.0) 133.7 (130.1–137.3) 169.2 (164.8–173.7) 136.1 (131.1–141.2) 
 # False-positive mammograms 23,821 4,622 4,634 2,400 
 Total # mammograms 188,268 34,573 27,391 17,636 
False-negative rate, per 1,000 mammograms (95% CIa4.0 (3.8–4.3) 4.6 (3.9–5.4) 3.3 (2.7–4.1) 2.6 (1.9–3.4) 
 # False-negative mammograms 761 160 91 45 
 Total # mammograms 188,268 34,573 27,391 17,636 
False-negative rate for invasive cancer, per 1,000 mammograms (95% CIa3.0 (2.7–3.2) 3.4 (2.8–4.1) 1.9 (1.4–2.4) 1.7 (1.1–2.4) 
 # False-negative mammograms (invasive) 558 117 51 30 
 Total # mammograms 188,268 34,573 27,391 17,636 
Short Interval Follow-Up     
Short interval follow-up recommendation proportion (%; 95% CIa22.1 (22.0–22.3) 31.0 (30.6–31.5) 16.6 (16.1–17.0) 23.6 (22.9–24.2) 
 # Mammograms recommended for short interval follow-up 41,681 10,733 4,535 4,155 
 Total # mammograms 188,268 34,573 27,391 17,636 
Short interval follow-up cancer yield (%; 95% CIa1.3 (1.2–1.4) 1.1 (0.9–1.3) 1.4 (1.1–1.8) 0.8 (0.6–1.1) 
 # Short-interval follow-up mammograms with breast cancer during follow-up 537 113 63 34 
 # Mammograms recommended for short interval follow-up 41,681 10,733 4,535 4,155 

aCI, confidence interval.

bMammograms with a final BI-RADS assessment of 4 and 5 were classified as positive. Mammograms with a final assessment of 1, 2, or 3 were classified as negative. Imputed positive/negative results for BI-RADS 0 mammograms were used in calculations.

cPPV, positive predictive value.

The cancer detection rate was highest among Asian/Pacific Islander women (54.3 per 1,000) and lowest among Hispanic women (32.8 per 1,000). When considering invasive cancers only, the cancer detection rate was highest among non-Hispanic White women (35.8 per 1,000 mammograms) and lowest among Hispanic women (22.3 per 1,000). Mammograms from Asian/Pacific Islander women also had the highest sensitivity (overall and invasive cancers only) and highest false-positive rate. The false-negative rate (overall and invasive cancers only) was highest among non-Hispanic Black women.

A short interval follow-up recommendation was most common among non-Hispanic Black women (31.0%, vs. other groups range, 16.6%–23.6%). However, among those with mammograms recommended for short interval follow-up, the short interval follow-up cancer yield was similar for non-Hispanic Black, non-Hispanic White, and Asian/Pacific Islander women and lowest among Hispanic women.

When we examined the impact of patient-level and facility-level characteristics on differences in performance statistics across racial/ethnic groups, we found that the greatest influences were receipt of ultrasound and/or MRI during the diagnostic work-up and imaging facility where women received their diagnostic mammogram (Fig. 2; Supplementary Table S2). Adjusting for ultrasound and/or MRI reduced the Black–White difference in cancer detection rates [model 7 odds ratio (OR), 0.84; 95% CI, 0.79–0.90 vs. model 8 OR, 0.98; 95% CI, 0.91–1.05]. Adjusting for imaging facility attenuated Black–White differences in the short interval follow-up recommendation proportion (model 8 OR, 1.39; 95% CI, 1.35–1.43 vs. model 9 OR, 1.03; 95% CI, 0.99–1.07) and Hispanic-White differences in biopsy recommendation proportion (model 8 OR, 0.90; 95% CI, 0.86–0.94 vs. model 9 OR, 1.01; 95% CI, 0.96–1.06). Adjustment for other factors did little to explain racial/ethnic differences in mammography performance. The data were similar when regression models included only the subset of observations with non-missing data for all covariates (Supplementary Table S3) and when women who reported a breast problem at the time of screening were excluded (Supplementary Table S4).

Figure 2.

Adjusted diagnostic mammography performance statistics by race/ethnicity for 267,868 diagnostic mammograms performed among 234,818 women in the BCSC (2005–2017). We compared cancer detection rate (A), invasive cancer detection rate (B) sensitivity (C), false-positive rate (D), short interval follow-up recommendation proportion (E), biopsy recommendation proportion (F), and PPV2 (G) among racial/ethnic groups while adjusting for patient characteristics to evaluate whether the characteristics might explain some of the differences present in the study population. Full definitions of each outcome are in Table 1. In each model, the performance statistic for non-Hispanic Black (circle), Asian/Pacific Islander (square), and Hispanic (triangle) groups was compared with the performance statistic for White individuals (reference group). Adjustment was performed sequentially, where each model also adjusted for the factors listed in the prior model.

Figure 2.

Adjusted diagnostic mammography performance statistics by race/ethnicity for 267,868 diagnostic mammograms performed among 234,818 women in the BCSC (2005–2017). We compared cancer detection rate (A), invasive cancer detection rate (B) sensitivity (C), false-positive rate (D), short interval follow-up recommendation proportion (E), biopsy recommendation proportion (F), and PPV2 (G) among racial/ethnic groups while adjusting for patient characteristics to evaluate whether the characteristics might explain some of the differences present in the study population. Full definitions of each outcome are in Table 1. In each model, the performance statistic for non-Hispanic Black (circle), Asian/Pacific Islander (square), and Hispanic (triangle) groups was compared with the performance statistic for White individuals (reference group). Adjustment was performed sequentially, where each model also adjusted for the factors listed in the prior model.

Close modal

Breast cancer characteristics

A total of 9,027 invasive breast cancers and 3,633 DCIS were detected by diagnostic mammography (Supplementary Table S5). Asian/Pacific Islander women had the highest proportion of DCIS diagnoses (43.0%, vs. other groups range, 27.6%–33.5%). When considering only invasive breast cancers, non-Hispanic Black women had the highest proportions of advanced stage at diagnosis, high tumor grade, hormone receptor negative, triple-negative, and lymph node positive tumors when compared with other racial/ethnic groups. Non-Hispanic White and Asian/Pacific Islander women had more favorable tumor characteristic profiles, with a smaller mean invasive tumor size and similarly high proportions of lymph node negative and low-grade cancers. However, HER2+ tumors were more common among Asian/Pacific Islander women (14.5%, vs. other groups range, 10.4%–12.1%). Hispanic women were more likely to have tumors >20 mm in size compared with other women (25.9%, vs. other groups range, 20.1%–24.3%).

The role that diagnostic mammography may play in contributing to racial and ethnic differences in breast cancer incidence and breast cancer survival is understudied. We evaluated diagnostic mammography performance among four racial/ethnic groups to clarify how differences in diagnostic work-up after a recent screening may influence those disparities. We found variation in diagnostic mammography performance by racial/ethnic group, and in most cases the variation was not explained by women's individual characteristics. Imaging facility and concurrent use of ultrasound or MRI during the diagnostic process were the only two factors that explained differences in performance. Although access to care has been identified as a barrier to breast cancer care in other settings, it is important to note that the women in this study had undergone breast cancer screening and were engaged with a breast imaging facility. These data suggest interventions that target the imaging facility and use of additional imaging modalities may have success in reducing some diagnostic disparities.

Our data suggest that non-Hispanic Black women may experience more potential harms related to diagnostic mammography than other groups. Exams conducted among non-Hispanic Black women had the lowest sensitivity and highest false-negative rate, indicating that this group may experience a delayed diagnosis more often. Others have shown that delayed diagnosis due to longer intervals between screening examinations results in higher proportions of advanced-stage, high-grade, and lymph node positive cancers among non-Hispanic Black women compared with other racial/ethnic groups (3). Consistent with that finding, the non-Hispanic Black women in our study had the highest proportion of tumors that were diagnosed at an advanced-stage or were high-grade or lymph node positive. Non-Hispanic Black women also had the highest proportion of recommendations for short interval follow-up. This assessment requires women to return for additional imaging in approximately 6 months. However, the short interval follow-up cancer yield for non-Hispanic Black women was not higher than any other group, suggesting that the additional financial costs, time, and stress associated with a short interval follow-up assessment were not balanced by benefits in additional cancers detected.

Our model-based analysis suggests that imaging facility characteristics may play a role in the higher frequency of short interval follow-up that was observed among non-Hispanic Black women. We did not have detailed information on the characteristics of each imaging facility in this study. However, prior studies have shown that Black women are less likely to obtain mammograms at facilities accredited by the National Consortium of Breast Centers or American College of Radiology's Breast Imaging Center of Excellence programs and therefore may receive a lower quality of care (28). Facility characteristics, including whether the facility was at an academic institution, a center of breast excellence, and had dedicated radiologists, were stronger predictors of a recently screened woman presenting with breast cancer symptoms than socioeconomic factors such as income, education level, and insurance coverage (29). Structural factors have also been shown to have a negative association with breast cancer mortality, though this association is likely mediated by more than just the diagnostic process. Additional efforts are needed to identify the specific structural inequities that lead to disparities in diagnostic mammography performance. Enacting system-level interventions to address those inequities has the potential to reduce some of the disparities we observed in this study. However, because Black women experience delays at every step of the detection, diagnosis, treatment, and care pathway (30), larger coordinated efforts may be needed to see a meaningful impact on known mortality disparities.

Many performance statistics were lowest among Hispanic women. This group had the lowest biopsy recommendation proportion, PPVs, and cancer detection rate, indicating that Hispanic women were least likely to receive a biopsy recommendation, and that if they did receive a biopsy recommendation, they were least likely to have breast cancer. Adjustment for imaging facility partially explained Hispanic-White differences in cancer detection rates, but detection rates remained significantly lower for the Hispanic group in the final multivariable model. Sensitivity for Hispanic women was comparable with that of non-Hispanic White women and higher than that of non-Hispanic Black women, suggesting that the low cancer detection rate and PPVs are not due to a failure to detect breast cancer when it is present. It is possible that there are simply fewer breast cancers to be detected among Hispanic women compared with the other women in this study. This hypothesis is supported by the fact that the Hispanic women were younger and had the lowest predicted breast cancer risk of the four groups. However, Hispanic women had the largest mean tumor size and the second highest proportion of advanced-stage tumors. This result suggests a need for earlier breast cancer detection among Hispanic women, although any efforts to increase early detection must be balanced with the potential risks of over-detection and overdiagnosis.

In the BCSC, Asian/Pacific Islander women were the most likely to receive a recommendation for a biopsy following a diagnostic mammogram. Although cancer was often diagnosed (Asian/Pacific Islander women had the highest cancer detection rate and sensitivity), other statistics suggest that the high frequency of biopsies may result in an unnecessary work-up. Asian/Pacific Islander women had the highest false-positive rate and the highest proportion of DCIS among cancers diagnosed. When considering invasive cancers only, Asian/Pacific Islander women no longer had the highest detection rate. The reasons for these patterns among Asian/Pacific Islander women are unclear. High rates of breast cancer screening do not explain the high proportion of DCIS diagnoses—data from a nationally representative survey show that Asian/Pacific Islander women do not participate in screening more often than non-Hispanic White or non-Hispanic Black women (31). Furthermore, all participants in our analysis received a screening mammogram before the diagnostic mammogram and thus had made the individual decision to seek preventive care.

Our analysis had limitations that may affect the interpretation of the results. Because of the time period covered by our study, DBT examinations were a small proportion of the mammograms studied. Thus, generalizability of these results to diagnostic performance at the growing number of imaging facilities that use DBT today may be limited. Data from studies of screening mammography suggest that DBT results in higher positive predictive values and higher cancer detection rates when compared with 2-D digital mammography alone (32–36). However, whether these differences are similar for diagnostic imaging or whether they vary by race or an ethnic group remains to be seen. Diagnostic mammography indication was not reported in detail for some registries. For these records, we inferred the indication of additional evaluation of a recent mammogram based on the sequence and findings of prior exams. This approach may have failed to classify some women as having an additional evaluation if details of the screening exam findings were incomplete. However, only 3.7% of exams without a specific indication did not meet the criteria of being an additional evaluation, thus the number of potentially misclassified exams is low. We required that exams have a record of a prior screening mammogram, which may have excluded women who received their screening mammogram at an imaging facility not affiliated with a BCSC registry. We lacked information about why some participants did not receive recommended diagnostic procedures. Additional research that directly elicits patient and provider feedback on why care was not provided is needed to fully understand the causes of the differences described in this study. We were not able to compare characteristics of breast cancers diagnosed after a false-negative mammogram due to the small numbers of cases, particularly among Asian/Pacific Islander and Hispanic women. We did not adjust for interpreting radiologist in our multivariable models. Future studies should examine the effect of the interpreting radiologist characteristics, which may be distinct from facility effects.

The strengths of this study include the analysis of data from more than 260,000 diagnostic mammograms, which enabled precise estimation of performance statistics, even for less populous racial/ethnic groups such as Asian/Pacific Islander and Hispanic women. Ninety-eight breast-imaging facilities from across the US contributed data, increasing the generalizability of our results. Participant, radiologist, and examination data were collected using a standardized format, ensuring that the interpretation of each data element was comparable across registries. Data were collected prospectively, eliminating the risk of recall bias, and included patient-reported and radiologist-reported items, allowing for a comprehensive evaluation of the diagnostic work-up process. We included information on diagnostic breast ultrasound and MRI imaging that may have contributed to the final assessment, providing additional insight into the differential results for statistics related to short-term follow-up and cancer detection. Examination outcomes were based on information from multiple sources, including pathology databases and cancer registries, to increase the chances that all newly diagnosed breast cancers were ascertained.

In conclusion, diagnostic mammography performance varied among the four racial/ethnic groups in this study. Our findings indicate that factors associated with the imaging facility, rather than individual characteristics, may explain some of these differences, particularly since the study population consisted of women who had already received a screening mammogram. Performance did not fall below the minimal acceptable standards proposed by Carney and colleagues (37) for any group, but the differences suggest that approaches for maximizing early detection of cancer while limiting the burden of false-positive exams and overdiagnosis differ based on women's race and ethnicity. This underscores the need to ensure that the studies that provide the evidence base for mammography recommendations include populations with sufficient racial and ethnic diversity.

S.J. Nyante reports grants from National Cancer Institute during the conduct of the study. L. Abraham reports grants from National Cancer Institute during the conduct of the study. C.I. Lee reports grants from the National Cancer Institute during the conduct of the study and personal fees from GRAIL, Inc., McGraw Hill, Inc., Oxford University Press, Wolters Kluwer, and from American College of Radiology outside the submitted work. K. Kerlikowske reports grants from National Cancer Institute and Patient-Centered Outcomes Research Institute during the conduct of the study; and reports as an unpaid consultant for GRAIL for the STRIVE study. D.L. Miglioretti reports grants from NIH and PCORI during the conduct of the study. B.L. Sprague reports grants from NIH, PCORI, and Lake Champlain Cancer Research Organization during the conduct of the study. L.M. Henderson reports grants from NIH/NCI during the conduct of the study. No disclosures were reported by the other authors.

All statements in this article, including its findings and conclusions, are solely those of the authors and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute, its Board of Governors or Methodology Committee, nor those of the National Cancer Institute, the National Institutes of Health, or the Agency for Health Research and Quality.

S.J. Nyante: Conceptualization, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. L. Abraham: Data curation, software, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. E.J. Aiello Bowles: Funding acquisition, investigation, methodology, writing–review and editing. C.I. Lee: Funding acquisition, investigation, methodology, writing–review and editing. K. Kerlikowske: Investigation, funding acquisition, methodology, writing–review and editing. D.L. Miglioretti: Formal analysis, funding acquisition, investigation, methodology, writing–review and editing. B.L. Sprague: Funding acquisition, investigation, methodology, writing–review and editing. L.M. Henderson: Conceptualization, supervision, funding acquisition, investigation, methodology, writing–review and editing.

We thank the participating individuals, mammography facilities, and radiologists for the data they have provided for this study. We also thank Chris Tachibana of Kaiser Permanente Washington for assistance with article editing. This work was supported by the Breast Cancer Surveillance Consortium with funding from the National Cancer Institute of the National Institutes of Health under award numbers P01CA154292 (to K. Kerlikowske, D.L. Miglioretti, L. Abraham, E.J. Aiello Bowles, C.I. Lee, B.L. Sprague, and L.M. Henderson), R01CA266377 (to C.I. Lee, D.L. Miglioretti, K. Kerlikowske, L.M. Henderson, and B.L. Sprague), U54CA163303 (to B.L. Sprague, D.L. Miglioretti, and K. Kerlikowske), and R50CA211115 (to E.J. Aiello Bowles). More information about the Breast Cancer Surveillance Consortium may be found at: http://www.bcsc-research.org/. Data collection was additionally supported by funding from the Patient-Centered Outcomes Research Institute under award number PCS-1504–30370 (to D.L. Miglioretti, K. Kerlikowske, L.M. Henderson, and B.L. Sprague) and the Agency for Health Research and Quality under award number R01HS018366. Cancer data collection was supported by several state public health departments and cancer registries. For a full description of these sources, please see (https://www.bcsc-research.org/about/work-acknowledgement).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
DeSantis
CE
,
Ma
J
,
Gaudet
MM
,
Newman
LA
,
Miller
KD
,
Goding Sauer
A
, et al
.
Breast cancer statistics, 2019
.
CA Cancer J Clin
2019
;
69
:
438
51
.
2.
American Cancer Society
.
Cancer Facts & Figures for Hispanics/Latinos 2018–2020
.
Atlanta
:
American Cancer Society, Inc.
;
2018
.
Available from:
https://www.cancer.org/research/cancer-facts-statistics/hispanics-latinos-facts-figures.html.
3.
Smith-Bindman
R
,
Miglioretti
DL
,
Lurie
N
,
Abraham
L
,
Barbash
RB
,
Strzelczyk
J
, et al
.
Does utilization of screening mammography explain racial and ethnic differences in breast cancer?
Ann Intern Med
2006
;
144
:
541
53
.
4.
Iqbal
J
,
Ginsburg
O
,
Rochon
PA
,
Sun
P
,
Narod
SA
.
Differences in breast cancer stage at diagnosis and cancer-specific survival by race and ethnicity in the United States
.
JAMA
2015
;
313
:
165
73
.
5.
Blackman
DJ
,
Masi
CM
.
Racial and ethnic disparities in breast cancer mortality: are we doing enough to address the root causes?
J Clin Oncol
2006
;
24
:
2170
8
.
6.
Gerend
MA
,
Pai
M
.
Social determinants of Black–White disparities in breast cancer mortality: a review
.
Cancer Epidemiol Biomarkers Prev
2008
;
17
:
2913
23
.
7.
Newman
LA
,
Griffith
KA
,
Jatoi
I
,
Simon
MS
,
Crowe
JP
,
Colditz
GA
.
Meta-analysis of survival in African American and white American patients with breast cancer: ethnicity compared with socioeconomic status
.
J Clin Oncol
2006
;
24
:
1342
9
.
8.
Albain
KS
,
Unger
JM
,
Crowley
JJ
,
Coltman
CA
Jr.
,
Hershman
DL
.
Racial disparities in cancer survival among randomized clinical trials patients of the Southwest Oncology Group
.
J Natl Cancer Inst
2009
;
101
:
984
92
.
9.
Parise
CA
,
Caggiano
V
.
Disparities in race/ethnicity and socioeconomic status: risk of mortality of breast cancer patients in the California cancer registry, 2000–2010
.
BMC Cancer
2013
;
13
:
449
.
10.
Caplan
LS
,
Helzlsouer
KJ
,
Shapiro
S
,
Wesley
MN
,
Edwards
BK
.
Reasons for delay in breast cancer diagnosis
.
Prev Med
1996
;
25
:
218
24
.
11.
Chang
SW
,
Kerlikowske
K
,
Nápoles-Springer
A
,
Posner
SF
,
Sickles
EA
,
Pérez-Stable
EJ
.
Racial differences in timeliness of follow-up after abnormal screening mammography
.
Cancer
1996
;
78
:
1395
402
.
12.
Jones
BA
,
Dailey
A
,
Calvocoressi
L
,
Reams
K
,
Kasl
SV
,
Lee
C
, et al
.
Inadequate follow-up of abnormal screening mammograms: findings from the race differences in screening mammography process study (United States)
.
Cancer Causes Control
2005
;
16
:
809
21
.
13.
McCarthy
BD
,
Yood
MU
,
Boohaker
EA
,
Ward
RE
,
Rebner
M
,
Johnson
CC
.
Inadequate follow-up of abnormal mammograms
.
Am J Prev Med
1996
;
12
:
282
8
.
14.
Rojas
M
,
Mandelblatt
J
,
Cagney
K
,
Kerner
J
,
Freeman
H
.
Barriers to follow-up of abnormal screening mammograms among low-income minority women. Cancer Control Center of Harlem
.
Ethn Health
1996
;
1
:
221
8
.
15.
Jones
BA
,
Kasl
SV
,
Curnen
MG
,
Owens
PH
,
Dubrow
R
.
Can mammography screening explain the race difference in stage at diagnosis of breast cancer?
Cancer
1995
;
75
:
2103
13
.
16.
Yankaskas
BC
,
Gill
KS
.
Diagnostic mammography performance and race: outcomes in black and white women
.
Cancer
2005
;
104
:
2671
81
.
17.
Sprague
BL
,
Arao
RF
,
Miglioretti
DL
,
Henderson
LM
,
Buist
DS
,
Onega
T
, et al
.
National performance benchmarks for modern diagnostic digital mammography: update from the breast cancer surveillance consortium
.
Radiology
2017
;
283
:
59
69
.
18.
Sickles
EA
,
Miglioretti
DL
,
Ballard-Barbash
R
,
Geller
BM
,
Leung
JW
,
Rosenberg
RD
, et al
.
Performance benchmarks for diagnostic mammography
.
Radiology
2005
;
235
:
775
90
.
19.
Breast Cancer Surveillance Consortium (BCSC)
.
Available from:
https://www.bcsc-research.org/.
20.
White
IR
,
Royston
P
,
Wood
AM
.
Multiple imputation using chained equations: issues and guidance for practice
.
Stat Med
2011
;
30
:
377
99
.
21.
Tice
JA
,
Miglioretti
DL
,
Li
CS
,
Vachon
CM
,
Gard
CC
,
Kerlikowske
K
.
Breast density and benign breast disease: risk assessment to identify women at high risk of breast cancer
.
J Clin Oncol
2015
;
33
:
3137
43
.
22.
Frank
RD
,
Winham
SJ
,
Vierkant
RA
,
Frost
MH
,
Radisky
DC
,
Ghosh
K
, et al
.
Evaluation of 2 breast cancer risk models in a benign breast disease cohort
.
Cancer
2018
;
124
:
3319
28
.
23.
Tice
JA
,
Bissell
MCS
,
Miglioretti
DL
,
Gard
CC
,
Rauscher
GH
,
Dabbous
FM
, et al
.
Validation of the breast cancer surveillance consortium model of breast cancer risk
.
Breast Cancer Res Treat
2019
;
175
:
519
23
.
24.
Amin
MB
,
Edge
S
,
Greene
F
,
Byrd
DR
,
Brookland
RK
,
Washington
MK
, et al
.,
editors
.
AJCC Cancer Staging Manual
. 8th ed.
New York
:
Springer International Publishing
;
2017
.
25.
Carey
LA
,
Perou
CM
,
Livasy
CA
,
Dressler
LG
,
Cowan
D
,
Conway
K
, et al
.
Race, breast cancer subtypes, and survival in the carolina breast cancer study
.
JAMA
2006
;
295
:
2492
502
.
26.
Nielsen
TO
,
Parker
JS
,
Leung
S
,
Voduc
D
,
Ebbert
M
,
Vickery
T
, et al
.
A comparison of PAM50 intrinsic subtyping with immunohistochemistry and clinical prognostic factors in tamoxifen-treated estrogen receptor-positive breast cancer
.
Clin Cancer Res
2010
;
16
:
5222
32
.
27.
Sickles
EA
,
D'Orsi
CJ
.
ACR BI-RADS Follow-up and Outcome Monitoring
.
In: ACR BI-RADS Atlas, Breast Imaging Reporting and Data System
.
Reston, VA, American College of Radiology
;
2013
.
28.
Molina
Y
,
Silva
A
,
Rauscher
GH
.
Racial/ethnic disparities in time to a breast cancer diagnosis: the mediating effects of health care facility factors
.
Med Care
2015
;
53
:
872
8
.
29.
Mortel
M
,
Rauscher
GH
,
Murphy
AM
,
Hoskins
K
,
Warnecke
RB
.
Racial and ethnic disparity in symptomatic breast cancer awareness despite a recent screen: the role of tumor biology and mammography facility characteristics
.
Cancer Epidemiol Biomarkers Prev
2015
;
24
:
1599
606
.
30.
Elmore
JG
,
Nakano
CY
,
Linden
HM
,
Reisch
LM
,
Ayanian
JZ
,
Larson
EB
.
Racial inequities in the timing of breast cancer detection, diagnosis, and initiation of treatment
.
Med Care
2005
;
43
:
141
8
.
31.
Sauer
AG
,
Siegel
RL
,
Jemal
A
,
Fedewa
SA
.
Updated review of prevalence of major risk factors and use of screening tests for cancer in the United States
.
Cancer Epidemiol Biomarkers Prev
2017
;
26
:
1192
208
.
32.
Bernardi
D
,
Macaskill
P
,
Pellegrini
M
,
Valentini
M
,
Fantò
C
,
Ostillio
L
, et al
.
Breast cancer screening with tomosynthesis (3D mammography) with acquired or synthetic 2D mammography compared with 2D mammography alone (STORM-2): a population-based prospective study
.
Lancet Oncol
2016
;
17
:
1105
13
.
33.
Cohen
EO
,
Weaver
OO
,
Tso
HH
,
Gerlach
KE
,
Leung
JWT
.
Breast cancer screening via digital mammography, synthetic mammography, and tomosynthesis
.
Am J Prev Med
2020
;
58
:
470
72
.
34.
Hofvind
S
,
Holen
ÅS
,
Aase
HS
,
Houssami
N
,
Sebuødegård
S
,
Moger
TA
, et al
.
Two-view digital breast tomosynthesis versus digital mammography in a population-based breast cancer screening programme (To-Be): a randomised, controlled trial
.
Lancet Oncol
2019
;
20
:
795
805
.
35.
Houssami
N
,
Bernardi
D
,
Pellegrini
M
,
Valentini
M
,
Fantò
C
,
Ostillio
L
, et al
.
Breast cancer detection using single-reading of breast tomosynthesis (3D-mammography) compared to double-reading of 2D-mammography: evidence from a population-based trial
.
Cancer Epidemiol
2017
;
47
:
94
99
.
36.
Libesman
S
,
Zackrisson
S
,
Hofvind
S
,
Seidler
AL
,
Bernardi
D
,
Lång
K
, et al
.
An individual participant data meta-analysis of breast cancer detection and recall rates for digital breast tomosynthesis versus digital mammography population screening
.
Clin Breast Cancer
2022 Feb 6 [Epub ahead of print]
.
37.
Carney
PA
,
Parikh
J
,
Sickles
EA
,
Feig
SA
,
Monsees
B
,
Bassett
LW
, et al
.
Diagnostic mammography: identifying minimally acceptable interpretive performance criteria
.
Radiology
2013
;
267
:
359
67
.

Supplementary data