Abstract
Background
Risk assessment is important for breast cancer prevention and early detection. We aimed to examine whether common risk factors, mammographic features, and breast cancer risk prediction scores of a woman were associated with breast cancer risk for her sisters.
Methods
We included 53 051 women from the Karolinska Mammography Project for Risk Prediction of Breast Cancer (KARMA) study. Established risk factors were derived using self-reported questionnaires, mammograms, and single nucleotide polymorphism genotyping. Using the Swedish Multi-Generation Register, we identified 32 198 sisters of the KARMA women (including 5352 KARMA participants and 26 846 nonparticipants). Cox models were used to estimate the hazard ratios of breast cancer for both women and their sisters, respectively.
Results
A higher breast cancer polygenic risk score, a history of benign breast disease, and higher breast density in women were associated with an increased risk of breast cancer for both women and their sisters. No statistically significant association was observed between breast microcalcifications and masses in women and breast cancer risk for their sisters. Furthermore, higher breast cancer risk scores in women were associated with an increased risk of breast cancer for their sisters. Specifically, the hazard ratios for breast cancer per 1 standard deviation increase in age-adjusted KARMA, Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA), and Tyrer-Cuzick risk scores were 1.16 (95% confidence interval [CI] = 1.07 to 1.27), 1.23 (95% CI = 1.12 to 1.35), and 1.21 (95% CI = 1.11 to 1.32), respectively.
Conclusion
A woman’s breast cancer risk factors are associated with her sister’s breast cancer risk. However, the clinical utility of these findings requires further investigation.
Breast cancer is the most common cancer in women, with increasing incidence over time in high- and low-income countries (1,2). Although the prognosis of breast cancer is generally good, women diagnosed at later stages have a poor prognosis and receive intensive treatments that could adversely influence their quality of life (1,3). Consequently, regular mammography screening attendance and being able to diagnose cancers at an early stage are important. Nevertheless, in 2017, it was reported that participation in breast cancer mammography screening programs was less than 70% in two-thirds of 27 European countries (4). A similar issue has also been revealed in the United States where 35% of women never attend a breast cancer screening program (5). Given this, there is still a need to motivate women to attend mammography screenings, particularly for those with a moderate or high risk of breast cancer.
Having a first-degree relative with breast cancer is associated with 1.8 times the risk of developing breast cancer (6,7). However, whether a woman’s risk of breast cancer, as measured using breast cancer risk factors or different risk models, can be useful for predicting breast cancer for her sisters is unclear. Because breast cancer risk factors such as benign breast diseases (BBDs), mammographic density, and microcalcifications are inherited or genetically associated with breast cancer (8-10), we hypothesize that these risk factors and risk models incorporating these features can also predict breast cancer for sisters. This hypothesis is important to test, as breast cancer screening based on risk assessment is foreseeable.
We identified healthy women in the prospective Karolinska Mammography Project for Risk Prediction of Breast Cancer (KARMA) screening cohort and, for the first time, linked them to the Swedish Multi-generation Register to identify their sisters. We investigated whether common breast cancer risk factors, mammographic features, and breast cancer risk models in KARMA participants were associated with breast cancer risk for their sisters.
Methods
Data sources
The KARMA study is a population-based screening cohort comprising 70 877 women in Sweden. All women who attended a screening or clinical mammography between October 2010 and March 2013 at 1 of 4 hospitals across Sweden were invited to participate (11). Participants completed a detailed questionnaire including aspects on reproduction, lifestyle, family history of cancer, and other breast cancer–associated risk factors. Approximately 98% of women donated a blood sample at KARMA enrollment. Women in KARMA have also been linked to the Sympathy Medical System to get access to pathology reports. Detailed information on study recruitment, questionnaire content, and follow-up can be found elsewhere (11).
Using the Swedish personal identity number and linkage to the Swedish Multi-Generation Register (12), it was the first time that we were able to identify all sisters of KARMA women independently of KARMA participation. KARMA women and their sisters were linked to the Swedish Cancer, Cause of Death, and Migration registers. Data on birth, cancer diagnosis, death, and migration were available for KARMA women and their sisters, whereas detailed information on breast cancer risk factors was only available for KARMA women.
Study sample
The study included 53 051 KARMA women who were born in Sweden, aged 40-74 years at KARMA enrollment, had completed the questionnaire, and did not have a prior breast cancer or any other diagnosis of invasive cancer (Supplementary Figure 1, available online). In all, 37 998 full-sibling sisters were identified through the Multi-Generation Register. Where multiple sisters were identified for a KARMA participant, all were included. Sisters of KARMA women who had died, migrated, or had a diagnosis of breast cancer or any invasive cancer before the KARMA participant enrolled in the study were excluded, leaving 32 198 sisters eligible for the analyses (Supplementary Figure 1, available online). Of the 32 198 sisters, 5352 were KARMA participants, and 26 846 were not recruited in KARMA.
For analyses of mammographic features and breast cancer risk models, we further excluded 3965 women who had breast surgery (including breast enlargement or reduction) before being included in KARMA. Genetic analyses were performed on a subgroup of 17 835 women with genotype data.
Exposures
From the KARMA baseline questionnaire, we retrieved information on education level, age at menarche, parity, age at first birth, menopausal status, age at menopause, oral contraceptive use, use of systemic hormone replacement therapy, body mass index (BMI), alcohol consumption, tobacco use, and family history of breast cancer. Blood samples from KARMA participants were genotyped using either a custom Illumina iSelect array (iCOGs array) (13) or the OncoArray (14). Detailed information on quality control and imputation has been published (15-17). Weighted breast cancer polygenic risk scores (BC-PRS) were computed for each KARMA participant using 313 breast cancer single nucleotide polymorphisms (SNPs) (18). We classified BC-PRS into 3 groups based on percentages: 0%-10%, 10%-90%, and 90%-100%. Through the Sympathy pathology record system, we obtained information on breast biopsies for KARMA participants between 1979 and 2015, including diagnoses using the Systematized Nomenclature of Medicine (19) and International Classification of Disease–Oncology from core biopsies and cytological fine-needle aspirations. Diagnoses were subdivided into 10 BBD categories based on the latest European breast pathology guidelines (20-22).
Full-field digital mammograms, from mediolateral oblique and craniocaudal views of both breasts, were collected when entering KARMA. Percent mammographic density was calculated by dividing the total dense area by the total breast area, using the Stratus method (23). We classified the mammographic density (percent) into 4 categories using the Breast Imaging Reporting and Data System (BI-RADS) scores: A (<2%), B (2%-16.9%), C (17%-48.9%), and D (≥49%) (24,25). The computer-generated BI-RADS score is referred to as the cBI-RADS score. We used women with cBI-RADS A and B as the reference group (49% of women). We used the Food and Drug Administration–approved software CAD (M-Vu CAD, Nashua, NH, USA) (26) to measure the number of breast masses and microcalcification clusters on each breast.
We computed the KARMA 2-year, Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA) 5-year, and Tyrer-Cuzick 5-year breast cancer risk scores, modeled using women’s information from the baseline questionnaire and mammograms. In summary, the KARMA risk model (version 1, full model) (25) incorporates mammographic features (ie, density, masses, and microcalcifications), hormone replacement therapy, and family history of breast cancer. The BOADICEA risk model (version 5, now available via the CanRisk web tool: www.canrisk.org) (27-29) incorporates pedigree data on family history of breast cancer while considering age at breast cancer diagnosis or age of last observation of relatives; mammographic density; and hormonal, reproductive, and lifestyle factors. The Tyrer-Cuzick model (version 8) (30) mainly incorporates family history of breast cancer; mammographic density; and hormonal, reproductive, and lifestyle factors. In line with previous literature (31,32), we categorized 5-year breast cancer risk scores as follows: average risk (<1.5%), moderately increased risk (1.5%-2.49%), and high risk (≥2.5%). These categories align with 2-year risk classification of less than 0.6% (average risk), from 0.6% up to (but not including) 1.0% (moderate risk), and 1.0% or higher (high risk).
Breast cancer follow-up
We followed KARMA women and their sisters from the time the KARMA participants enrolled in the study until breast cancer diagnosis, end of follow-up (October 31, 2019), diagnosis of any invasive cancer, emigration, or death, whichever came first. The mean follow-up times for KARMA women and their sisters were 7.3 and 7.4 years, respectively.
Statistical analysis
We used Cox proportional hazard regression models to examine the association between common breast cancer risk factors, mammographic features, and breast cancer risk scores for KARMA women and breast cancer risk for their sisters. Results for KARMA women were shown for comparison. Age was used as the timescale in the Cox models, using the age of KARMA women for themselves and their corresponding sisters’ ages as the timescale for their sisters. For analyses of having a genetic predisposition to breast cancer, we adjusted for genotype array (iCOGs or OncoArray). BBD among KARMA women was modeled as a time-varying exposure in the analyses. When modeling the association between mammographic density and breast cancer risk, age and BMI at entry to the KARMA cohort were included as potential confounders. We analyzed 2614 full-sibling sister pairs from the KARMA cohort to investigate the intraclass correlation between breast cancer risk factors and risk scores, using Spearman correlation coefficient. The cumulative incidence of breast cancer for KARMA women’s sisters, in relation to risk generated through risk models for KARMA women, was plotted as Nelson-Aalen curves.
For estimating hazard ratios (HRs) per standard deviation of continuous risk measures, we transformed mammographic density, KARMA, BOADICEA, and Tyrer-Cuzick risk scores to appropriate normal distributions and normalized these for age (and BMI for mammographic density) using linear regression. Because the BC-PRS was already normally distributed and was independent of age, we did not transform the PRS or adjust it for age. To evaluate the robustness of our results, we conducted 2 sensitivity analyses. First, we conducted analyses separately for KARMA women aged 40-54 and 55-74 years, as well as their corresponding sisters. Second, in cases where a KARMA participant had multiple eligible sisters, we employed a random seed method to select 1 sister for the analysis.
All statistical analyses were performed using Stata (version 17.0), R (version 4.0.5), and SAS software (version 9.4). All P values were 2-sided and considered statistically significant at less than .05. The KARMA prospective cohort study was approved by the ethical review board at Karolinska Institutet (Stockholm, Sweden, dnr 2010/958-31/1), and informed consent was obtained from all individual participants.
Results
The mean age of women at KARMA enrollment was 54.4 (SD = 9.7) years, and more than half of the women were postmenopausal (Table 1). The mean age of sisters at KARMA enrollment was 54.5 (SD = 10.6) years.
Table 1.
Baseline characteristics of the KARMA participants (n = 53 051)
| Characteristics | No. of KARMA women (%) |
|---|---|
| Age at KARMA baseline, y | |
| 40-49 | 19 612 (37.0) |
| 50-59 | 15 643 (29.5) |
| 60-74 | 17 796 (33.5) |
| Age at baseline, mean (SD), y | 54.4 (9.70) |
| Postmenopausal | |
| No | 22 723 (42.8) |
| Yes | 30 328 (57.2) |
| Age at menopause, ya | |
| 45 or younger | 2380 (16.3) |
| 46-49.9 | 3113 (21.3) |
| 50-54.9 | 6783 (46.4) |
| 55 or older | 2330 (16.0) |
| Education, y | |
| ≤9 | 5138 (10.1) |
| 10-12 | 18 187 (35.6) |
| >12 | 27 693 (54.3) |
| Age at menarche, y | |
| Younger than 13 | 17 974 (34.6) |
| 13 or older | 33 928 (65.4) |
| Nulliparous | |
| No | 46 327 (87.5) |
| Yes | 6641 (12.5) |
| No. of children | |
| 0 | 6641 (12.5) |
| 1-2 | 33 075 (62.4) |
| >2 | 13 252 (25.0) |
| Age at first birth, yb | |
| 30 or younger | 34 613 (74.7) |
| Older than 30 | 11 696 (25.3) |
| Contraception use | |
| Never | 6980 (13.3) |
| Ever | 45 493 (86.7) |
| Hormone replacement therapy usea | |
| Never | 19 034 (69.2) |
| Ever | 6787 (24.7) |
| Current | 1703 (6.2) |
| BMI, kg/m2 | |
| <25 | 29 658 (56.2) |
| 25 to <30 | 16 481 (31.2) |
| ≥30 | 6675 (12.6) |
| BMI, mean (SD) | 25.2 (4.22) |
| Smoking | |
| Never | 24 853 (47.0) |
| Ever | 21 723 (41.1) |
| Current | 6328 (12.0) |
| Alcohol consumption, grams/d | |
| 0 | 9385 (17.8) |
| 0.1-10 | 32 620 (61.9) |
| >10 | 10 672 (20.3) |
| Family history of breast cancer | |
| No | 44 357 (85.7) |
| Yes | 7374 (14.3) |
| Benign breast diseasec | |
| No | 48 894 (92.3) |
| Yes | 4081 (7.7) |
| Mammographic percent densityd | |
| cBI-RADS A and B | 20 636 (48.7) |
| cBI-RADS C | 16 829 (39.7) |
| cBI-RADS D | 4886 (11.5) |
| Breast microcalcification clustersd | |
| 0 | 33 105 (83.0) |
| ≥1 | 6792 (17.0) |
| Breast massesd | |
| 0 | 14 654 (36.7) |
| ≥1 | 25 243 (63.3) |
Among postmenopausal women. Column totals may not equal to total number of women because of missing values. BBD = benign breast disease; cBI-RADS = computer-generated Breast Imaging Reporting and Data System; mammographic density (percent) was classified into four categories using the cBI-RADS scores: A (<2%), B (2%-16.9%), C (17%-48.9%), and D (≥49%); BMI = body mass index; KARMA = Karolinska Mammography Project for Risk Prediction of Breast Cancer.
Among women with biological children.
The BBD information was collected from the Sympathy Medical System. A total of 76 women with breast cancer diagnosed within 6 months of a BBD diagnosis were excluded.
Among women without any breast surgeries (including breast enlargements and breast reductions).
Table 2 shows the association between common breast cancer risk factors in KARMA women and breast cancer risk for KARMA women and their sisters. Later age at first birth, later age at menopause, use of hormone replacement therapy, being a smoker, and high alcohol consumption in women were associated with an increased risk of breast cancer for KARMA women but not for their sisters. Having a higher BC-PRS in women was statistically significantly associated with an increased risk of breast cancer for KARMA women and their sisters. BBD in women was associated with an increased risk of breast cancer for KARMA women and their sisters, with adjusted hazard ratios of 1.89 (95% confidence interval [CI] = 1.63 to 2.20) and 1.27 (95% CI = 1.01 to 1.60), respectively. The strongest association was seen for epithelial proliferation with atypia—hazard ratios of 3.85 (95% CI = 3.16 to 4.70) for KARMA women and 1.74 (95% CI = 1.27 to 2.38) for their sisters (Supplementary Table 1, available online).
Table 2.
Hazard ratios (95% CI) of breast cancer for KARMA women and their sisters by KARMA women’s common breast cancer risk factors, along with the intraclass correlation of each risk factor among sister pairs in the subset of KARMA
| KARMA women | Full-sibling sisters | KARMA sister pairsa | |||||
|---|---|---|---|---|---|---|---|
| (n = 53 051) |
(n = 32 198) |
(n pairs = 2614) |
|||||
| Common risk factors | No. of women | No. of cases | HR (95% CI)b | No. of women | No. of cases | HR (95% CI)c | Correlation coefficient |
| Age at first birth, yd | 0.222 | ||||||
| 30 or younger | 34 613 | 939 | 1.00 (Referent) | 21 346 | 510 | 1.00 (Referent) | |
| Older than 30 | 11 696 | 303 | 1.15 (1.00 to 1.31) | 6974 | 147 | 1.04 (0.87 to 1.26) | |
| Age at menopause, ye | 0.131 | ||||||
| 45 or younger | 2380 | 64 | 1.00 (Referent) | 1444 | 34 | 1.00 (Referent) | |
| 46-49.9 | 3113 | 99 | 1.19 (0.87 to 1.64) | 1902 | 59 | 1.32 (0.87 to 2.02) | |
| 50-54.9 | 6783 | 220 | 1.17 (0.88 to 1.56) | 4180 | 134 | 1.32 (0.91 to 1.93) | |
| 55 or older | 2330 | 100 | 1.46 (1.06 to 2.02) | 1422 | 43 | 1.21 (0.77 to 1.90) | |
| Hormone replacement therapy usee | 0.249 | ||||||
| Never | 19 034 | 561 | 1.00 (Referent) | 11 654 | 312 | 1.00 (Referent) | |
| Ever | 6787 | 230 | 0.99 (0.85 to 1.16) | 3992 | 122 | 1.03 (0.83 to 1.28) | |
| Current | 1703 | 90 | 1.78 (1.42 to 2.22) | 1069 | 35 | 1.18 (0.83 to 1.67) | |
| Smoking | 0.278 | ||||||
| Never | 24 853 | 586 | 1.00 (Referent) | 15 068 | 328 | 1.00 (Referent) | |
| Ever | 21 723 | 631 | 1.13 (1.01 to 1.27) | 13 073 | 321 | 1.02 (0.87 to 1.19) | |
| Current | 6328 | 191 | 1.27 (1.08 to 1.49) | 3972 | 93 | 1.00 (0.79 to 1.26) | |
| Alcohol consumption, grams/d | 0.212 | ||||||
| 0 | 9385 | 231 | 1.00 (Referent) | 5866 | 140 | 1.00 (Referent) | |
| 0.1-10 | 32 620 | 849 | 1.07 (0.92 to 1.24) | 19 702 | 436 | 0.95 (0.78 to 1.15) | |
| >10 | 10 672 | 324 | 1.18 (1.00 to 1.40) | 6387 | 162 | 1.01 (0.80 to 1.26) | |
| Breast cancer polygenic risk scoresf | 0.527 | ||||||
| 0%-9.9% | 1784 | 40 | 0.39 (0.28 to 0.53) | 1078 | 19 | 0.71 (0.44 to 1.14) | |
| 10%-89.9% | 14 268 | 813 | 1.00 (Referent) | 8667 | 210 | 1.00 (Referent) | |
| 90%-100% | 1783 | 205 | 2.01 (1.73 to 2.35) | 1057 | 37 | 1.50 (1.06 to 2.13) | |
| BBDg | 0.084 | ||||||
| No | 48 894 | 1154 | 1.00 (Referent) | 29 610 | 666 | 1.00 (Referent) | |
| Yes | 4081 | 184 | 1.89 (1.63 to 2.20) | 2538 | 76 | 1.27 (1.01 to 1.60) | |
The intraclass correlation coefficient was calculated using the Spearman correlation coefficient. All correlation coefficient P values are less than .01, with the exception of age at menopause (P = .026). Column totals may not be equal to the total number of women due to missing values. BBD = benign breast disease; CI = confidence interval; HR = hazard ratio; KARMA = Karolinska Mammography Project for Risk Prediction of Breast Cancer.
Age of KARMA women was used as the timescale in the Cox regression models.
Age of full-sibling sisters was used as the timescale in the Cox regression models.
Among KARMA women with biological children and their sisters.
Among postmenopausal KARMA women and their sisters.
A total of 17 835 women with available genotype data and their sisters were included. Further adjusted for genotyping method.
The BBD information was collected from the Sympathy Medical System. BBD was coded as a time-varying exposure. A total of 76 women with breast cancer diagnosed within 6 months of a BBD diagnosis were excluded.
Table 3 shows the association between mammographic features in KARMA women and breast cancer risk for KARMA women and their sisters. Dense breast (cBI-RADS D) in women was associated with an increased risk of breast cancer for the KARMA women and their sisters, with adjusted hazard ratios of 2.46 (95% CI = 1.99 to 3.05) and 1.62 (95% CI = 1.19 to 2.22), respectively, when compared with KARMA women or sisters of women with nondense breast (cBI-RADS A and B). Breast microcalcifications and masses in women were associated with a higher risk of breast cancer for KARMA women only but not for their sisters.
Table 3.
Hazard ratios (95% CI) of breast cancer for KARMA women and their sisters by KARMA women’s mammographic features, along with the intraclass correlation of each mammographic feature among sister pairs in the subset of KARMA
| KARMA women | Full-sibling sisters | KARMA sister pairsa | |||||
|---|---|---|---|---|---|---|---|
| (n = 49 086) |
(n = 29 767) |
(n pairs = 2614) |
|||||
| Mammographic features | No. of women | No. of cases | HR (95% CI)b | No. of women | No. of cases | HR (95% CI)c | Correlation coefficient |
| Mammographic percent densityd | 0.392 | ||||||
| cBi-RADS A and B | 20 636 | 524 | 1.00 (Referent) | 12 262 | 280 | 1.00 (Referent) | |
| cBi-RADS C | 16 829 | 505 | 1.70 (1.49 to 1.96) | 10 419 | 260 | 1.42 (1.17 to 1.71) | |
| cBi-RADS D | 4886 | 159 | 2.46 (1.99 to 3.05) | 3002 | 65 | 1.62 (1.19 to 2.22) | |
| Breast microcalcification clusters | 0.092 | ||||||
| 0 | 33 105 | 720 | 1.00 (Referent) | 20 387 | 438 | 1.00 (Referent) | |
| ≥1 | 6792 | 368 | 2.34 (2.06 to 2.66) | 3985 | 112 | 1.17 (0.95 to 1.44) | |
| Breast masses | 0.068 | ||||||
| 0 | 14 654 | 336 | 1.00 (Referent) | 8950 | 183 | 1.00 (Referent) | |
| ≥1 | 25 243 | 752 | 1.29 (1.13 to 1.47) | 15 422 | 367 | 1.13 (0.95 to 1.35) | |
The intraclass correlation coefficient was calculated using the Spearman correlation coefficient. All correlation coefficient P values are less than .01. Column totals may not be equal to the total number of women because of missing values. BMI = body mass index; cBi-RADS = computer-generated Breast Imaging Reporting and Data System density score; mammographic density (percent) was classified into four categories using the cBi-RADS scores: A (<2%), B (2%-16.9%), C (17%-48.9%), and D (≥49%); CI = confidence interval; HR = hazard ratio; KARMA = Karolinska Mammography Project for Risk Prediction of Breast Cancer.
Age of KARMA women was used as the timescale in the Cox regression models.
Age of full-sibling sisters was used as the timescale in the Cox regression models.
Adjusted for age and BMI of KARMA women at baseline..
Table 4 shows the association between breast cancer risk scores for KARMA women and breast cancer risk for KARMA women and their sisters. When we stratified the women into average, moderately increased, and high-risk groups for breast cancer using the KARMA, BOADICEA, and Tyrer-Cuzick risk models, we found positive associations between risk groups and breast cancer risk for KARMA women and their sisters. The hazard ratios for the sisters in the high-risk group compared with the average-risk group were 1.35 (95% CI = 1.04 to 1.74), 2.04 (95% CI = 1.37 to 3.01), and 1.78 (95% CI = 1.39 to 2.26) for the KARMA, BOADICEA, and Tyrer-Cuzick risk models, respectively. Correspondingly, we observed that sisters of KARMA women with high risk had a statistically significantly higher 5-year cumulative incidence rate compared with sisters of KARMA women in the average risk group (Figure 1; the average 5-year cumulative incidence rate for all sisters was 1.58%).
Table 4.
Hazard ratios (95% CI) of breast cancer for KARMA women and their sisters by KARMA women’s breast cancer risk scores, along with the intraclass correlation of each risk score among sister pairs in the subset of KARMA
| KARMA women | Full-sibling sisters | KARMA sister pairsa | |||||
|---|---|---|---|---|---|---|---|
| (n = 36 703) |
(n = 22 394) |
(n pairs = 2614) |
|||||
| Breast cancer risk prediction scores | No. of women | No. of cases | HR (95% CI)b | No. of women | No. of cases | HR (95% CI)c | Correlation coefficient |
| KARMA 2-year risk scores | 0.240 | ||||||
| Average risk, <0.6% | 29 370 | 511 | 1.00 (Referent) | 18 062 | 378 | 1.00 (Referent) | |
| Moderate risk, 0.6% to <1.0% | 3656 | 121 | 1.70 (1.39 to 2.07) | 2186 | 50 | 0.98 (0.73 to 1.32) | |
| High risk, ≥1.0% | 3677 | 172 | 2.32 (1.94 to 2.77) | 2146 | 70 | 1.35 (1.04 to 1.74) | |
| BOADICEA 5-year risk scores | 0.403 | ||||||
| Average risk, <1.5% | 29 549 | 548 | 1.00 (Referent) | 18 216 | 365 | 1.00 (Referent) | |
| Moderate risk, 1.5% to <2.5% | 6133 | 193 | 1.48 (1.25 to 1.75) | 3614 | 106 | 1.27 (1.02 to 1.58) | |
| High risk, ≥2.5% | 1021 | 63 | 2.98 (2.29 to 3.89) | 564 | 27 | 2.04 (1.37 to 3.01) | |
| Tyrer-Cuzick 5-year risk scores | 0.478 | ||||||
| Average risk, <1.5% | 22 803 | 382 | 1.00 (Referent) | 14 574 | 277 | 1.00 (Referent) | |
| Moderate risk, 1.5% to <2.5% | 9452 | 217 | 1.22 (1.03 to 1.45) | 5557 | 133 | 1.13 (0.92 to 1.39) | |
| High risk, ≥2.5% | 4448 | 205 | 2.38 (1.99 to 2.83) | 2263 | 88 | 1.78 (1.39 to 2.26) | |
The intraclass correlation coefficient was calculated using the Spearman correlation coefficient. All correlation coefficient P values are less than .01. BOADICEA = Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm; CI = confidence interval; HR = hazard ratio; KARMA = Karolinska Mammography Project for Risk Prediction of Breast Cancer.
Age of KARMA women was used as the timescale in the Cox regression models.
Age of full-sibling sisters was used as the timescale in the Cox regression models.
Figure 1.
Cumulative incidence of breast cancer for sisters of KARMA women, by groups of breast cancer risk scores (average risk vs moderate risk vs high risk) for women. The average 5-year cumulative incidence rate for all sisters was 1.58%. BOADICEA = Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm; KARMA = Karolinska Mammography Project for Risk Prediction of Breast Cancer.
In Tables 2, 3, and 4, intraclass correlation coefficients for risk factors and risk scores among sister pairs in the subset of the KARMA cohort are also presented. Generally, the higher the correlation for a given risk factor and the greater the hazard ratio associated with that risk factor for KAMRA women, the greater the hazard ratio for their sisters.
In Figure 2, for the continuous breast cancer risk measures, the association between these age-adjusted measures in KARMA women and the risk of breast cancer for these women and their sisters is presented. We found that per standard deviation increase in age-adjusted KARMA, BOADICEA, and Tyrer-Cuzick risk scores, age- and BMI-adjusted mammographic density, and BC-PRS of women were statistically significantly associated with an increased risk of breast cancer for sisters (HR = ∼1.2).
Figure 2.
Hazard ratios (95% confidence interval) of breast cancer for KARMA women and their full-sibling sisters per standard deviation increase in age-adjusted risk measures in KARMA women. For KARMA women, we used KARMA women’s age as the timescale in the Cox model, and for full-sibling sisters, we used the age of full-sibling sisters as the timescale. a Adjusted for genotyping method in the Cox model. BOADICEA = Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm; CI = confidence interval; KARMA = Karolinska Mammography Project for Risk Prediction of Breast Cancer.
In the sensitivity analyses, we observed consistent hazard ratios per standard deviation of breast cancer across both younger and older age subgroups of KARMA women and their sisters when stratified by the median age of KARMA participants (Supplementary Table 2, available online). Furthermore, when including only 1 sister per woman, the hazard ratios per standard deviation of breast cancer for the sisters remained consistent (Supplementary Table 3, available online).
Discussion
We found that a woman had an increased risk of breast cancer if her sister had a higher PRS, a history of BBD, higher mammographic density for her age and BMI, or a greater risk for age based on any of the risk models used. Similar associations were not found for mammographic features (microcalcifications, masses) or hormone-related, reproductive, or lifestyle factors.
Carriership of high penetrant mutations such as BRCA1 and BRCA2 has long been known to relate to the risk of breast cancer and other malignancies in relatives (33,34). We show that a high PRS in KARMA women, based on 313 SNPs, also relates to an increased risk of breast cancer for their sisters. When contrasting the population norm (10%-90% BC-PRS) to the highest decile of the PRSs, KARMA women in the high-risk group were found to have a more than twofold higher risk of breast cancer. Their untested sisters had a nearly 1.5-fold increased risk, which was approximately half of the corresponding estimate for the KARMA women. It remains to be shown if the identification of additional SNPs will substantially improve the predictive ability by BC-PRS.
A previous BBD statistically significantly relates to the risk of a subsequent breast cancer not only for the KARMA woman (HR = 1.89) but also to a lesser extent for her sister (HR = 1.27). In addition, we found that the association between BBD in women and breast cancer risk for their sisters differed by subtype of BBD, with the strongest risk being associated with epithelial proliferation with atypia (HR = 1.74). This result suggests that having a family history of epithelial proliferation with atypia or a family history of breast cancer has the similar prediction ability on a sister’s risk of being diagnosed with breast cancer (6,7).
In this study, we found that mammographic density of KARMA women was associated with risk of breast cancer for their sisters, by using more than 32 000 sisters (including around 27 000 of non-KARMA participants) identified for KARMA women. The results were consistent with our previous study, using approximately 2000 sister pairs among KARMA participants and showing that the heritability for mammographic density was as high as 58% (10).
Microcalcifications; masses; and hormone-related, reproductive, and lifestyle factors in women were not associated with the risk of breast cancer risk for their sisters. These findings are not surprising, as none of these factors are highly correlated between sisters or strongly associated with breast cancer risk (35,36). However, despite having a relatively large dataset, the lack of association might still be because of insufficient statistical power.
Breast cancer risk models are mainly used to identify women at high risk of breast cancer. High-risk women could be offered supplemental breast cancer screening, mutation screening, and/or preventive measures (37-39). Whether an elevated risk score also has a bearing on a sister’s risk has previously not been studied at large scale. Based on our findings, if a woman’s breast cancer–risk score is classified as high risk, her sister may also have an increased risk of breast cancer. In genetic counseling, individuals identified with a rare genetic mutation are encouraged to inform their relatives (40). This disclosure enhances relatives’ disease awareness and increases their likelihood of undergoing genetic testing themselves. Similarly, when a woman is classified as high risk during screening, sharing this information to her sisters may boost disease awareness among them and may motivate those who do not attend mammography screening to participate and obtain risk assessments themselves. This can also be beneficial in improving these sisters’ adherence to screening and promoting a healthier lifestyle. A woman younger than the entry age to the screening programs [which is 50 years of age in most countries (41)] may be more likely to visit the clinic if she knows her breast cancer risk than if she does not.
Despite the potential benefits, future research is essential to explore the extent of our findings’ clinical utility. First, the association between a woman’s breast cancer risk assessment and her sister’s breast cancer risk is relatively modest. Second, our results should also be considered in the context of potential ethical dilemma. We demonstrate that a woman not asking or consenting to risk assessment might indirectly have her risk of breast cancer estimated when her sister receives the results. This is a similar issue faced by women attending oncogenetic clinics. Indeed, collecting information that has a bearing on individuals not consenting is a critical ethical challenge that has to be addressed in the era of precision medicine.
This study is strengthened by a large sample size, detailed information on breast cancer risk factors, high quality of Swedish nationwide registers—the Swedish Multi-Generation Register and the Swedish Cancer Register in particular—and the prospective cohort design. By linking data from the KARMA cohort to several Swedish registers, we have been able to create a family-cancer data set where KARMA women and their sisters are followed prospectively from the time the KARMA participants enrolled in the study.
Our study also has limitations. Firstly, the study was limited to women enrolled in the KARMA cohort and their sisters (11). The participants were better educated and more likely to have a family history of breast cancer than an average Swedish woman (11). Such a selection most likely does not affect the internal validity of our results but could influence the generalizability to other populations. Prevalence of risk factors and their effects differ between populations and are dependent on age distribution, socioeconomic factors, and ethnicity among other factors. Future studies have to challenge our results.
In summary, our results suggest that an individual’s breast cancer risk assessment could be indicative of breast cancer risk not only for the woman herself but also for her sisters, although the strength of the association with her sisters is relatively modest. Because it is common in the United States to assess the individual risk of breast cancer in women attending breast cancer screening and there is a strong interest to do the same among other countries (42,43), our study indicates that the risks of their sisters are indirectly evaluated at the same time. This knowledge may increase sisters’ disease awareness and attendance rate of mammography screening, but it can also be seen as a violation of sisters’ integrity. Our results thus come with ethical challenges that should not be underestimated but correctly handled.
Supplementary Material
Acknowledgement
The funding sources had no role in study design; data collection, analysis, or interpretation; or the decision to approve publication of the finished manuscript.
Contributor Information
Xinhe Mao, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden.
Wei He, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden; Chronic Disease Research Institute, The Children’s Hospital, and National Clinical Research Center for Child Health, School of Public Health, School of Medicine, Zhejiang University, Hangzhou, Zhejiang, China; Department of Nutrition and Food Hygiene, School of Public Health, Zhejiang University, Hangzhou, Zhejiang, China.
Mikael Eriksson, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden.
Linda S Lindström, Department of Oncology-Pathology, Karolinska Institutet and Hereditary Cancer Unit, Karolinska University Hospital, Stockholm, Sweden.
Natalie Holowko, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden; Department of Medicine Solna, Clinical Epidemiology Division, Karolinska Institutet, Stockholm, Sweden.
Svetlana Bajalica-Lagercrantz, Department of Oncology-Pathology, Karolinska Institutet and Hereditary Cancer Unit, Karolinska University Hospital, Stockholm, Sweden.
Mattias Hammarström, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden.
Felix Grassmann, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden; Institute for Clinical Research and Systems Medicine, Health and Medical University, Potsdam, Germany.
Keith Humphreys, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden.
Douglas Easton, Department of Oncology, Centre for Cancer Genetic Epidemiology, University of Cambridge, Cambridge, UK; Department of Public Health and Primary Care, Centre for Cancer Genetic Epidemiology, University of Cambridge, Cambridge, UK.
Per Hall, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden; Department of Oncology, Södersjukhuset, Stockholm, Sweden.
Kamila Czene, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden.
Data availability
Access to datasets of the KARMA study can be requested from https://karmastudy.org/data-access/.
Author contributions
Xinhe Mao, MD, MSC (Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Software; Validation; Visualization; Writing—original draft; Writing—review & editing), Wei He, PhD (Conceptualization; Data curation; Funding acquisition; Investigation; Methodology; Supervision; Writing—original draft; Writing—review & editing), Mikael Eriksson, PhD (Investigation; Software; Writing—review & editing), Linda S. Lindström, PhD (Investigation; Writing—review & editing), Natalie Holowko, PhD (Investigation; Writing—review & editing), Svetlana Bajalica-Lagercrantz, MD, PhD (Investigation; Writing—review & editing), Mattias Hammarström, MSC (Investigation; Writing—review & editing), Felix Grassmann, PhD (Investigation; Writing—review & editing), Keith Humphreys, PhD (Investigation; Methodology; Writing—review & editing), Douglas Easton, PhD (Investigation; Writing—review & editing), Per Hall, MD, PhD (Investigation; Writing—review & editing), and Kamila Czene, PhD (Conceptualization; Data curation; Funding acquisition; Investigation; Methodology; Resources; Supervision; Writing—original draft; Writing—review & editing).
Funding
This work was supported by the Swedish Research Council (grant number: 2022-00584); Swedish Cancer Society (grant number: 22 2207 and 19 0267); the Stockholm County Council (grant number 20200102); and FORTE (grant number: 2018-00877). XM is supported by the China Scholarship Council (grant number: 201806210002). WH is supported by Zhejiang University through the “Hundred Talents Program.”
Conflicts of interest
We declare no competing interests.
References
- 1. Li N, Deng Y, Zhou L, et al. Global burden of breast cancer and attributable risk factors in 195 countries and territories, from 1990 to 2017: results from the Global Burden of Disease Study 2017. J Hematol Oncol 2019;12(1):140. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Fitzmaurice C, Abate D, Abbasi N, et al. Global, regional, and national cancer incidence, mortality, years of life lost, years lived with disability, and disability-adjusted life-years for 29 cancer groups, 1990 to 2017: a systematic analysis for the Global Burden of Disease Study. JAMA Oncol. 2019;5(12):1749-1768. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Ginsburg O, Bray F, Coleman MP, et al. The global burden of women’s cancers: a grand challenge in global health. Lancet. 2017;389(10071):847-860. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Eurostat. Breast cancer screening differs among member states. https://ec.europa.eu/eurostat/web/products-eurostat-news/-/ddn-20200109-1. Accessed March 17, 2022.
- 5. Esserman L, Eklund M, Veer LV, et al. The WISDOM study: a new approach to screening can and should be tested. Breast Cancer Res Treat. 2021;189(3):593-598. [DOI] [PubMed] [Google Scholar]
- 6. Pharoah PD, Day NE, Duffy S, Easton DF, Ponder BA.. Family history and the risk of breast cancer: a systematic review and meta-analysis. Int J Cancer. 1997;71(5):800-809. [DOI] [PubMed] [Google Scholar]
- 7. Collaborative Group on Hormonal Factors in Breast Cancer. Familial breast cancer: collaborative reanalysis of individual data from 52 epidemiological studies including 58,209 women with breast cancer and 101,986 women without the disease. Lancet. 2001;358(9291):1389-1399. [DOI] [PubMed] [Google Scholar]
- 8. Bertelsen L, Mellemkjaer L, Balslev E, Olsen JH.. Benign breast disease among first-degree relatives of young breast cancer patients. Am J Epidemiol. 2008;168(3):261-267. [DOI] [PubMed] [Google Scholar]
- 9. Boyd NF, Rommens JM, Vogt K, et al. Mammographic breast density as an intermediate phenotype for breast cancer. Lancet Oncol. 2005;6(10):798-808. [DOI] [PubMed] [Google Scholar]
- 10. Holowko N, Eriksson M, Kuja-Halkola R, et al. Heritability of mammographic breast density, density change, microcalcifications, and masses. Cancer Res. 2020;80(7):1590-1600. [DOI] [PubMed] [Google Scholar]
- 11. Gabrielson M, Eriksson M, Hammarstrom M, et al. Cohort profile: the Karolinska mammography project for risk prediction of breast cancer (KARMA). Int J Epidemiol. 2017;46(6):1740-1741g. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Ekbom A. The Swedish Multi-generation Register. Methods Mol Biol. 2011;675:215-220. [DOI] [PubMed] [Google Scholar]
- 13. Michailidou K, Hall P, Gonzalez-Neira A, et al. ; for the GENICA (Gene Environment Interaction and Breast Cancer in Germany) Network. Large-scale genotyping identifies 41 new loci associated with breast cancer risk. Nat Genet. 2013;45(4):353-361, 361e1-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Amos CI, Dennis J, Wang Z, et al. The OncoArray consortium: a network for understanding the genetic architecture of common cancers. Cancer Epidemiol Biomarkers Prev. 2017;26(1):126-135. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Michailidou K, Lindstrom S, Dennis J, et al. ; for the ConFab/AOCS Investigators. Association analysis identifies 65 new breast cancer risk loci. Nature. 2017;551(7678):92-94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. O’Connell J, Gurdasani D, Delaneau O, et al. A general approach for haplotype phasing across the full spectrum of relatedness. PLoS Genet. 2014;10(4):e1004234. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Williams AL, Patterson N, Glessner J, Hakonarson H, Reich D.. Phasing of many thousands of genotyped samples. Am J Hum Genet. 2012;91(2):238-251. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Mavaddat N, Michailidou K, Dennis J, et al. ; for the NBCS Collaborators. Polygenic risk scores for prediction of breast cancer and breast cancer subtypes. Am J Hum Genet. 2019;104(1):21-34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. SNOMED International. Leading Healthcare Terminology, Worldwide. https://www.snomed.org. Accessed September 23, 2021.
- 20. Ellis I, Al-Sam S, Anderson N, et al. Pathology Reporting of Breast Disease is Surgical Excision Specimens Incorporating the Dataset for Histological Reporting of Breast Cancer. London: The Royal College of Pathologists; 2016. [Google Scholar]
- 21. Swedish Society of Pathology. SNOMED Codes in numerical order. https://svfp.se/kvast/ovrigt-utskarningsmallar-snomed-etc/snomed-numerisk-ordning/. Accessed September 23, 2021.
- 22. Johansson A, Christakou AE, Iftimi A, et al. Characterization of benign breast diseases and association with age, hormonal factors, and family history of breast cancer among women in Sweden. JAMA Netw Open. 2021;4(6):e2114716. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Eriksson M, Li J, Leifland K, Czene K, Hall P.. A comprehensive tool for measuring mammographic density changes over time. Breast Cancer Res Treat. 2018;169(2):371-379. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Spak DA, Plaxco JS, Santiago L, Dryden MJ, Dogan BE.. BI-RADS(®) fifth edition: a summary of changes. Diagn Interv Imaging. 2017;98(3):179-190. [DOI] [PubMed] [Google Scholar]
- 25. Eriksson M, Czene K, Pawitan Y, Leifland K, Darabi H, Hall P.. A clinical model for identifying the short-term risk of breast cancer. Breast Cancer Res. 2017;19(1):29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Profound AI. iCAD. https://www.icadmed.com/profoundai.html. Accessed February 25, 2021.
- 27. Antoniou AC, Pharoah PP, Smith P, Easton DF.. The BOADICEA model of genetic susceptibility to breast and ovarian cancer. Br J Cancer. 2004;91(8):1580-1590. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Lee A, Mavaddat N, Wilcox AN, et al. BOADICEA: a comprehensive breast cancer risk prediction model incorporating genetic and nongenetic risk factors. Genetics in Medicine. 2019;21(8):1708-1718. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Archer S, Babb de Villiers C, Scheibl F, et al. Evaluating clinician acceptability of the prototype CanRisk tool for predicting risk of breast and ovarian cancer: a multi-methods study. PLoS One. 2020;15(3):e0229999. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Brentnall AR, Harkness EF, Astley SM, et al. Mammographic density adds accuracy to both the Tyrer-Cuzick and Gail breast cancer risk models in a prospective UK screening cohort. Breast Cancer Res. 2015;17(1):147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. National Institute for Health and Clinical Excellence: Guidance. Familial Breast Cancer: Classification and Care of People at Risk of Familial Breast Cancer and Management of Breast Cancer and Related Risks in People with History of Breast Cancer. NICE Guideline CG164. London, England: National Collaborating Center for Cancer; 2013. [PubMed] [Google Scholar]
- 32. Kerlikowske K, Zhu W, Tosteson AN, et al. ; for the Breast Cancer Surveillance Consortium. Identifying women with dense breasts at high risk for interval cancer: a cohort study. Ann Intern Med. 2015;162(10):673-681. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. King MC, Marks JH, Mandell JB; for the New York Breast Cancer Study Group. Breast and ovarian cancer risks due to inherited mutations in BRCA1 and BRCA2. Science. 2003;302(5645):643-646. [DOI] [PubMed] [Google Scholar]
- 34. Ford D, Easton DF, Bishop DT, Narod SA, Goldgar DE.. Risks of cancer in BRCA1-mutation carriers. Breast Cancer Linkage Consortium. Lancet. 1994;343(8899):692-695. [DOI] [PubMed] [Google Scholar]
- 35. McPherson K, Steel CM, Dixon JM.. ABC of breast diseases. Breast cancer-epidemiology, risk factors, and genetics. BMJ. 2000;321(7261):624-628. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Rojas K, Stuckey A.. Breast cancer epidemiology and risk factors. Clin Obstet Gynecol. 2016;59(4):651-672. [DOI] [PubMed] [Google Scholar]
- 37. Monticciolo DL, Newell MS, Moy L, Niell B, Monsees B, Sickles EA.. Breast cancer screening in women at higher-than-average risk: recommendations From the ACR. J Am Coll Radiol. 2018;15(3, pt A):408-414. [DOI] [PubMed] [Google Scholar]
- 38. Visvanathan K, Hurley P, Bantug E, et al. Use of pharmacologic interventions for breast cancer risk reduction: American Society of Clinical Oncology clinical practice guideline. J Clin Oncol. 2013;31(23):2942-2962. [DOI] [PubMed] [Google Scholar]
- 39. Freedman AN, Graubard BI, Rao SR, McCaskill-Stevens W, Ballard-Barbash R, Gail MH.. Estimates of the number of US women who could benefit from tamoxifen for breast cancer chemoprevention. J Natl Cancer Inst. 2003;95(7):526-532. [DOI] [PubMed] [Google Scholar]
- 40. Forrest LE, Delatycki MB, Skene L, Aitken M.. Communicating genetic information in families–a review of guidelines and position papers. Eur J Hum Genet. 2007;15(6):612-618. [DOI] [PubMed] [Google Scholar]
- 41. Winch CJ, Sherman KA, Boyages J.. Toward the breast screening balance sheet: cumulative risk of false positives for annual versus biennial mammograms commencing at age 40 or 50. Breast Cancer Res Treat. 2015;149(1):211-221. [DOI] [PubMed] [Google Scholar]
- 42. Evans DG, Warwick J, Astley SM, et al. Assessing individual breast cancer risk within the U.K. National Health Service Breast Screening Program: a new paradigm for cancer prevention. Cancer Prev Res (Phila). 2012;5(7):943-951. [DOI] [PubMed] [Google Scholar]
- 43. Britt KL, Cuzick J, Phillips KA.. Key steps for effective breast cancer prevention. Nat Rev Cancer. 2020;20(8):417-436. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Access to datasets of the KARMA study can be requested from https://karmastudy.org/data-access/.