Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Oct 15.
Published in final edited form as: Int J Cancer. 2022 Apr 4;151(8):1304–1309. doi: 10.1002/ijc.34001

Association of contralateral breast cancer risk with mammographic density defined at higher-than-conventional intensity thresholds

Gordon P Watt 1,*, Julia A Knight 2,3, Tuong L Nguyen 4, Anne S Reiner 1, Kathleen E Malone 5, Esther M John 6, Charles F Lynch 7, Jennifer D Brooks 3, Meghan Woods 1, Xiaolin Liang 1, Leslie Bernstein 8, Malcolm C Pike 1, John L Hopper 4, Jonine L Bernstein 1
PMCID: PMC9420749  NIHMSID: NIHMS1790796  PMID: 35315524

Abstract

Mammographic dense area (MDA) is an established predictor of future breast cancer risk. Recent studies have found that risk prediction might be improved by redefining MDA in effect at higher-than-conventional intensity thresholds. We assessed whether such higher-intensity MDA measures gave stronger prediction of subsequent contralateral breast cancer (CBC) risk using the Women’s Environment, Cancer, and Radiation Epidemiology (WECARE) Study, a population-based CBC case-control study of ≥1 year survivors of unilateral breast cancer diagnosed between 1990 and 2008. Three measures of MDA for the unaffected contralateral breast were made at the conventional intensity threshold (‘Cumulus’) and at two sequentially higher-intensity thresholds (‘Altocumulus’ and ‘Cirrocumulus’) using the CUMULUS software and mammograms taken up to 3 years prior to the first breast cancer diagnosis. The measures were fitted separately and together in multivariable-adjusted logistic regression models of CBC (252 CBC cases and 271 unilateral breast cancer controls). The strongest association with CBC was MDA defined using the highest intensity threshold, Cirrocumulus (odds ratio per adjusted standard deviation (OPERA) 1.40, 95% CI 1.13–1.73); and the weakest association was MDA defined at the conventional threshold, Cumulus (1.32, 95% CI 1.05–1.66). In a model fitting the three measures together, the association of CBC with Cirrocumulus was unchanged (1.40, 95% CI 0.97–2.05), and the lower brightness measures did not contribute to the CBC model fit. These results suggest that MDA defined at a high-intensity threshold is a better predictor of CBC risk and has the potential to improve CBC risk stratification beyond conventional MDA measures.

Keywords: Breast cancer, cancer survivors, mammographic density, risk stratification

Graphical Abstract

graphic file with name nihms-1790796-f0001.jpg

INTRODUCTION

Risk of contralateral breast cancer (CBC) for breast cancer survivors is at least twice the risk of a first primary breast cancer for unaffected women [1]. Quantitative mammographic breast density defined as the “white or bright regions on a mammogram” [2] measured by the intensity-thresholding software, CUMULUS, is an established risk factor for first primary breast cancer [3, 4]. Recently we reported that mammographic density estimated using the CUMULUS software was also associated with an increased risk of CBC [5]. CUMULUS density is based on a binary split between ‘non-dense’ and ‘dense’ regions, so that the ‘dense’ area includes a range of intensity values that appear ‘white or bright’ on the mammogram. In a series of papers, Nguyen and colleagues modified the standard application of CUMULUS by defining mammographic dense area (MDA) at higher intensity, or brightness, thresholds [4, 6–9]. They introduced the term ‘Altocumulus’ to describe the brighter regions as distinct from the white areas, and ‘Cirrocumulus’ to describe the brightest regions within the brighter areas. By studying mammograms received by Australian and Korean women, they consistently found improved prediction of first primary breast cancer risk from MDA defined at the higher brightness thresholds [4, 6–9]. They also found that, except for cancers detected in the interval between screening exams, when the range of density measures were fitted together, the risk association with the conventional MDA measure, Cumulus, was attenuated to the null.

In this study, we evaluated whether MDA defined using higher-intensity thresholds also had stronger CBC risk associations using mammograms taken prior to treatment for a first primary breast cancer. We also studied the CBC risk associations when fitting the different density measures together in the same model.

MATERIALS AND METHODS

The Women’s Environment, Cancer, and Radiation Epidemiology (WECARE) Study is a multicenter, population-based, individually-matched, case-control study of CBC risk [5, 10]. All participants were diagnosed with a first primary, invasive (local/regional), unilateral breast cancer (UBC) before age 55 years. Cases were women diagnosed with asynchronous CBC ≥1 year after their first breast cancer diagnosis. Controls were women with UBC who were individually matched to cases 1:1 on the following factors: time at risk for CBC; age and year of first diagnosis; race/ethnicity; and recruitment site. All participants completed an epidemiological questionnaire and granted access to medical records.

For a subset of 252 CBC case women and 271 UBC control women first diagnosed from 2000 and 2008, we collected film and digital mammograms obtained prior to or at the time of diagnosis of the first breast cancer [5]. Film mammograms were centrally digitized using two Kodak Lumisys Digital Scanners set to 12-bit gray-scale resolution. MDA was the focus of this study instead of mammographic percent density (MPD) because it is less negatively confounded by age and especially body mass index (BMI) [11]. MDA was estimated for the unaffected breast (contralateral to the first breast cancer) by three experienced readers (5 to 12 years’ experience using the software) using the CUMULUS software (see Appendix 2) [2] at three intensity thresholds as described previously [4, 6–9]. The threshold for conventional Cumulus captured the entire ‘white or bright’ area of the mammogram; the intensity threshold for Altocumulus was higher and captured the ‘brighter’ area on the mammogram; and the threshold for Cirrocumulus was the highest and included only the ‘brightest’ areas of the mammogram. Figure 1 shows (A) an example digitized film mammogram; (B) MDA defined as all white or bright areas (Cumulus); (C) MDA defined as the brighter areas (Altocumulus); and (D) MDA defined as the brightest areas (Cirrocumulus). The distributions of the three untransformed measures are provided in Figure 2.

Figure 1.

Figure 1.

Open in a new tab

Example of a digitized film mammogram (A) with mammographic dense area estimated at Cumulus (B), Altocumulus (C), and Cirrocumulus (D) intensity thresholds.

Figure 2.

Figure 2.

Open in a new tab

Distribution of mammographic dense area estimated at Cumulus, Altocumulus, and Cirrocumulus intensity thresholds on film and digitized mammograms in the WECARE Study

We estimated the association between CBC risk and MDA defined at three intensity levels by fitting logistic regression models with adjustment for confounders and for variables used to match controls to cases. All MDA measures were log-transformed, then scaled per unit increase in the standard deviation (SD) of residuals adjusted for age and BMI for the UBC controls, as described previously [6, 12]. Transformed and scaled measures are hereafter differentiated using italics. BMI (kg/m2) was estimated at the time of mammogram using linear interpolation based on BMI measurements reported at other times, as in a previous WECARE Study report [5]. We also report the natural risk gradient of log(OR per adjusted SD) = log(OPERA), which is approximately equal to the difference between cases and controls in the means of the transformed, adjusted, and standardized MDA measures [12].

Case-control status was regressed on the standardized MDA measures in unconditional logistic regression models. Models were adjusted for menopausal status at time of mammogram; disease stage (local/regional) and estrogen receptor status of the first diagnosis; receipt of chemotherapy, hormonal therapy, and radiation therapy for the first breast cancer; first-degree family history of breast cancer; age at menarche; type of mammogram (film/digital); and for variables used to match controls to cases, described above. First, we modeled each MDA measure separately, and second, we fit the three MDA measures in the same model.

MDAs defined at the higher-intensity thresholds are nested within MDAs defined at the lower-intensity thresholds (see Figure 1). To evaluate the contribution of each non-overlapping area of increasing brightness when fitting all three MDA measures in the same model, we redefined them to be mutually exclusive regions of density such that overlapping dense areas were accounted for only once in the model. Cirrocumulus was unchanged, representing the ‘brightest’ area; for Altocumulus, we subtracted out the Cirrocumulus (‘brightest) area leaving only the ‘brighter’ area; for Cumulus, we subtracted out the Altocumulus (‘brighter’ and ‘brightest’), leaving only the ‘white’ area. The mutually exclusive measures were log-transformed and standardized, as described above, prior to model fitting. We also evaluated whether using the three standard (overlapping) MDA measures gave different conclusions from the models using the mutually exclusive measures. The goodness of fit for each model was assessed using Akaike’s Information Criterion (AIC), where a lower AIC indicates a better fitting model. For two models with AICi and AICmin, where AICmin is the better fitting model, let AICΔ = AICi – AICmin. We considered the strength of evidence against the better fitting model to be substantial for 0 ≤ AICΔ ≤ 2; considerably less for 4 ≤ AICΔ ≤ 7; and essentially none for AICΔ > 10 [13]. Finally, in sensitivity analyses, we assessed the associations between CBC risk and the MDA measures when (1) restricting the sample to women with digitized film images; (2) fitting MPD measures rather than MDA measures; and (3) fitting the MDA measures as quartiles rather than continuous predictors.

RESULTS

Table 1 shows that, from multivariable models of the association between CBC and individual MDA measures, Cirrocumulus had the strongest association with CBC (Model #3, OPERA = 1.40, 95% CI 1.13–1.73) and this model had the lowest AIC, followed by Altocumulus (Model #2, 1.37, 95% CI 1.10–1.72) and Cumulus (Model #1, 1.32, 95% CI 1.05–1.66). Therefore, the CBC risk gradient was 21% greater for Cirrocumulus (log(OPERA) = 0.34) compared with Cumulus (log(OPERA) = 0.28). When fitting the three MDA measures together using the mutually exclusive definitions (model #4), the OPERAs were attenuated to the null for the white and brighter areas and neither was statistically significant, whereas the OPERA remained unchanged for the brightest area.

Table 1.

Multivariable models of the association between CBC risk and mammographic dense area measured at three intensity levels in the WECARE Study using mammograms taken prior to or at the time of diagnosis of the first primary breast cancer

N, CBC cases = 252
N, UBC controls = 271
Dense area measure(s)a included in model OPERA
(95%CI)b 		P-valuec AIC
Model 1: Cumulus only
Cumulus (‘white’, ‘brighter’, and ‘brightest’ areas) 1.32 (1.05–1.66) 0.016 739.4
Model 2: Altocumulus only
Altocumulus
(‘Brighter and brightest areas’)		 1.37 (1.10–1.72) 0.0049 737.3
Model 3: Cirrocumulus only
Cirrocumulus
(‘Brightest area’)		 1.40 (1.13–1.73) 0.0016 735.2
Model 4: All three measures d
White area + 0.98 (0.68–1.42) 0.93 739.2
Brighter area + 1.01 (0.61–1.68) 0.98
Brightest area 1.40 (0.97–2.05) 0.076
Open in a new tab

Abbreviations. CBC, contralateral breast cancer; WECARE, Women’s Environment, Cancer, and Radiation Epidemiology Study; UBC, unilateral breast cancer; OPERA, odds per age- and BMI-adjusted standard deviation for the controls; CI, confidence interval; AIC, Akaike Information Criterion

a

Mammographic dense area (MDA) estimated at three intensity thresholds as described previously [6] and shown in Figure 1. Cumulus represents conventional mammographic dense area and includes all ‘white or bright’ areas within the mammogram; Altocumulus includes the ‘brighter’ and ‘brightest’ areas; Cirrocumulus includes only the ‘brightest’ areas. Measures were log-transformed and scaled per unit of age- and BMI-adjusted standard deviation for the controls [12].

b

OPERAs additionally adjusted for menopausal status at time of mammogram; family history of breast cancer; stage and estrogen receptor status of first breast cancer diagnosis; chemotherapy, radiation therapy, and hormone therapy received for first breast cancer; age at menarche; type of mammogram (film/digital); and matching variables: recruitment cancer registry, follow-up time, race/ethnicity, and diagnosis year [12].

c

P-value for likelihood ratio test comparing multivariable model with the MDA measure compared to model without the measure.

d

In the model with all three MDA estimates, the measures were modified to be mutually exclusive prior to transformation and standardization; see Appendix 1 for details.

As shown in Appendix 1, Table 1, when fitting each of these mutually exclusive measures in separate models, the risk gradients were lower and not statistically significant for the white area (log(OPERAwhite) = 0.22) compared with standard Cumulus (log(OPERACumulus) = 0.28); similar for the brighter area (log(OPERAbrighter) = 0.30) compared with standard Altocumulus (log(OPERAAltocumulus) = 0.31); and, by definition, the brightest area was equivalent to standard Cirrocumulus. Fitting the standard overlapping MDA measures together in one model reduced the precision of the estimates compared with the model using mutually exclusive measures but gave the same model fit. Regardless of the MDA measures used, the best model fit based on AIC was the model including Cirrocumulus alone.

From the sensitivity analyses, restriction to digitized film images did not impact results (Supplementary Table 1). The results were likewise similar when using MPD instead of MDA (Supplementary Table 2). When fitting quartiles, rather than continuous values, of the MDA measures, the best model fit was also obtained when fitting only Cirrocumulus (Supplementary Table 3). The OR for the 4th Cirrocumulus quartile compared with the 1st quartile (OR4th) was 2.71 (95% CI 1.57–4.75) in the Cirrocumulus model; 1.81 (95% CI 1.18–3.50) for Altocumulus; and 1.80 (95% CI 1.05–3.10) for Cumulus. In the model fitting all three quartile measures together, the OR4th for the brightest area (Cirrocumulus) was 2.76 (95% CI 0.98–7.89), while the OR4th for the brighter and white areas were attenuated and not significant.

DISCUSSION

More predictors of CBC risk are needed given the high risk of developing second primary cancers in a growing population of breast cancer survivors. Considering the widespread use of digital mammography in the breast cancer prevention and care continuum, mammography-based predictors of CBC risk could be particularly impactful for risk stratification. Two prior studies have used non-centralized, clinical mammographic density measurements based on the Breast Imaging Reporting and Data System (BI-RADS) to test the association between CBC risk and mammographic density. The first study of 847 CBC case women and 2,541 UBC control women from the Breast Cancer Surveillance Consortium found statistically significant ORs of 1.65, 2.10, and 2.32 for scattered, heterogeneously dense, and extremely dense breast tissue, respectively [14]. The second study comprising 229 CBC case women and 451 UBC control women seen at MD Anderson Cancer Center found an OR of 1.80 (95% CI 1.22–2.64) for heterogenous/extremely dense vs almost entirely fat/scattered density [15]. In a previous WECARE Study report, the association between CBC and blinded, quantitative MPD≥50% (vs 25 to <50%) was OR=1.17 (95% CI 0.73, 1.87) for 253 case women and 271 control women with mammograms taken prior to treatment for the first breast cancer (p for trend 0.12) [5].

The present study builds upon the use of a blinded, quantitative approach by measuring MDA at higher-intensity definitions of dense area and directly adjusting the measures for age and BMI. Whereas quantitative mammographic density measured at conventional CUMULUS intensity thresholds quantifies the amount of white or bright regions in a mammogram, we found that the adjusted MDA measure including only the brightest regions produced a stronger association with CBC risk. Notably, training users of CUMULUS to estimate these higher-intensity MDAs is straightforward and the measures have good inter- and intra-observer agreement [9]. Therefore, the use of higher-intensity MDA is a promising and accessible new marker of CBC risk for women diagnosed with a first primary breast cancer.

From the model fitting all three MDA measures, the addition of the white and brighter areas did not improve the fit of the model beyond that with only the brightest area, suggesting that Cirrocumulus captured most, if not all, of the variation in CBC risk associated with mammographic brightness. Other studies of these three MDA measures found a similar pattern for first primary breast cancers [4, 6, 7, 9], particularly for screen-detected breast cancers [8], indicating that Cirrocumulus may be less prone to confounding by the tumor-masking effect of conventional mammographic density [16].

This study’s strengths included adjustment for clinical and epidemiological risk factors, standardization of MDA measures, and blinded assessment of MDA measures in the unaffected breast. While MDA measurement was not fully automated and is therefore subject to minor variability, any between-image differences are not expected to be differential with respect to case-control status. We also note that this study was primarily based on the analysis of digitized film mammograms which have largely been replaced by full-field digital mammograms; however, previous studies using higher-intensity density estimates from CUMULUS have reported similar results for both film [4] and digital [6] mammograms.

In conclusion, for women ≤55 years when diagnosed with a first primary breast cancer, MDA measured using a higher-intensity threshold could improve prediction of CBC risk.

Supplementary Material

supinfo

Statement of Novelty and Impact.

Breast cancer survivors have a high risk of a subsequent contralateral breast cancer (CBC). In this study, we found that a modified, higher-intensity measure of mammographically dense area had a stronger association with contralateral breast cancer risk than conventional mammographic density. These results further improve our understanding of mammographic density as a risk factor for CBC and provide a potential new tool for CBC risk stratification for survivors of breast cancer.

ACKNOWLEDGMENTS.

The authors would like to thank the participants of the WECARE Study for making this work possible. This work was supported by the United States National Cancer Institute award numbers R01 CA168339, R01 CA097397, P30 CA008748, U01 CA083178, R01 CA129639, R01 CA114236, and P30CA086862; by the Victoria Cancer Council award numbers 1139144 and TP900783; and by the Cure Cancer Australia Foundation award number 1159399.

List of Abbreviations:

AIC

Akaike’s Information Criterion

BMI

body mass index

CBC

contralateral breast cancer

CI

confidence interval

LRT

likelihood ratio test

MDA

mammographic dense area

MPD

mammographic percent density

OPERA

odds per adjusted standard deviation

OR

odds ratio

SD

standard deviation

UBC

unilateral breast cancer

WECARE

Women’s Environment, Cancer, and Radiation Epidemiology Study

Footnotes

CONFLICT OF INTEREST. The authors declare no competing interests related to this work.

ETHICS STATEMENT. The WECARE Study received ethical approval at each recruitment site and each participant provided written informed consent to participate.

DATA AVAILABILITY STATEMENT.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

REFERENCES

  • 1.Chen Y, Thompson W, Semenciw R, et al. Epidemiology of contralateral breast cancer. Cancer Epidemiol Biomarkers Prev 1999;8(10):855–61. [PubMed] [Google Scholar]
  • 2.Byng JW, Boyd NF, Fishell E, et al. The quantitative analysis of mammographic densities. Phys Med Biol 1994;39(10):1629–38. [DOI] [PubMed] [Google Scholar]
  • 3.Jeffers AM, Sieh W, Lipson JA, et al. Breast Cancer Risk and Mammographic Density Assessed with Semiautomated and Fully Automated Methods and BI-RADS. Radiology 2017;282(2):348–355. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Nguyen TL, Aung YK, Evans CF, et al. Mammographic density defined by higher than conventional brightness thresholds better predicts breast cancer risk. International Journal of Epidemiology 2016;46(2):652–661. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Knight JA, Blackmore KM, Fan J, et al. The association of mammographic density with risk of contralateral breast cancer and change in density with treatment in the WECARE study. Breast Cancer Res 2018;20(1):23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Nguyen TL, Aung YK, Evans CF, et al. Mammographic density defined by higher than conventional brightness threshold better predicts breast cancer risk for full-field digital mammograms. Breast Cancer Res 2015;17:142. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Nguyen TL, Aung YK, Evans CF, et al. Mammographic density defined by higher than conventional brightness thresholds better predicts breast cancer risk. Int J Epidemiol 2017;46(2):652–661. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Nguyen TL, Aung YK, Li S, et al. Predicting interval and screen-detected breast cancers from mammographic density defined by different brightness thresholds. Breast Cancer Res 2018;20(1):152. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Nguyen TL, Choi YH, Aung YK, et al. Breast Cancer Risk Associations with Digital Mammographic Density by Pixel Brightness Threshold and Mammographic System. Radiology 2018;286(2):433–442. [DOI] [PubMed] [Google Scholar]
  • 10.Bernstein JL, Langholz B, Haile RW, et al. Study design: Evaluating gene-environment interactions in the etiology of breast cancer - the WECARE study. Breast Cancer Res 2004;6(3):R199–R214. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Nguyen TL, Schmidt DF, Makalic E, et al. Explaining Variance in the Cumulus Mammographic Measures That Predict Breast Cancer Risk: A Twins and Sisters Study. Cancer Epidemiology Biomarkers & Prevention 2013;22(12):2395–2403. [DOI] [PubMed] [Google Scholar]
  • 12.Hopper JL. Odds per Adjusted Standard Deviation: Comparing Strengths of Associations for Risk Factors Measured on Different Scales and Across Diseases and Populations. American Journal of Epidemiology 2015;182(10):863–867. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Burnham KP, Anderson DR. Multimodel Inference:Understanding AIC and BIC in Model Selection. Sociological Methods & Research 2004;33(2):261–304. [Google Scholar]
  • 14.Chowdhury M, Euhus D, O’Donnell M, et al. Dose-dependent effect of mammographic breast density on the risk of contralateral breast cancer. Breast Cancer Res Treat 2018;170(1):143–148. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Raghavendra A, Sinha AK, Le-Petross HT, et al. Mammographic breast density is associated with the development of contralateral breast cancer. Cancer 2017;123(11):1935–1940. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Hopper JL, Nguyen TL, Schmidt DF, et al. Going Beyond Conventional Mammographic Density to Discover Novel Mammogram-Based Predictors of Breast Cancer Risk. Journal of Clinical Medicine 2020;9(3):627. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

supinfo
NIHMS1790796-supplement-supinfo.docx (46.5KB, docx)

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

RESOURCES