Estimating lifetime risk for breast cancer as a screening tool for identifying those who would benefit from additional services among women utilizing mobile mammography

Abstract

Background:

To estimate lifetime risk of breast cancer among women utilizing mobile mammography and to determine the proportion that might benefit from additional services, such as genetic counseling and educational programs.

Methods:

Retrospective analysis of electronic health records for 2214 women screened for breast cancer on a mobile mammography van was conducted. Participants answered questions about their demographic characteristics, breast health, and family history of cancer. Logistic regression analyses were used to assess the odds of being recommended for additional services by the Tyrer-Cuzick (TC) lifetime risk score.

Results:

The average TC ten-year risk score was 2.76 % ± 2.01 %, and the average TC lifetime risk score was 7.30 % ± 4.80 %. Using lifetime risk scores ≥ 10 %, it was determined that 444 patients (20.23 %) could be referred to additional services. Less than one percent of patients had been tested for the BRCA genes previously. The odds of being recommended for additional services by the TC model were significantly greater among those who were eligible for the New York Cancer Services Program (i.e., a proxy for lack of insurance) when compared to those who were ineligible (OR=1.31, 95 % CI: 1.03–1.66). After adjustment, screening borough and race/ethnicity were not significantly associated with being recommended for services.

Conclusion:

Genetic counseling and education are some of the tools available to promote awareness and early detection of breast cancer; however, screening guidelines do not mandate genetic counseling or referrals for individuals at high-risk.

Policy Summary:

Patients and providers should have discussions about predicted TC lifetime risk scores at follow-up breast cancer screening appointments, as this is a missed opportunity to improve care at both fixed sites and mobile clinics.

1. Introduction

Breast cancer is the most diagnosed cancer among women and is the second leading cause of death in the U.S. [1]. Women with a family history of breast and/or ovarian cancer have 1.5 times the risk of developing breast cancer when compared to women with no family history [2]. While early mammography screening can reduce breast cancer mortality by up to 40 % in average risk women, increased detection efforts may be needed to adequately screen women who are at elevated risk for both breast and ovarian cancer [3].

Another tool for combatting breast cancer in high-risk women is genetic testing for breast and ovarian cancer susceptibility genes, BRCA1 and BRCA2. BRCA mutation carriers have a 45–65 % risk for developing breast cancer by age 70 [4]. Recent research has demonstrated the benefits of genetic counseling and testing for those at an elevated risk for breast cancer, yet genetic services referral rates for high-risk individuals remain low, especially among minority and undeserved communities [5]. Ways to increase referral rates and improve access to care include improving patient and provider education on the topic as well as addressing the known barriers of lack of utility, limited knowledge, and concerns about cost and insurance coverage [6,7].

Software-based models, like the Tyrer-Cuzick (TC) model, calculate and predict lifetime cancer risk and can be used to directly combat these barriers. The TC model predicts lifetime risk of breast cancer by accounting for family history and other risk factors related to breast cancer (e.g., having breastfed, menopause, body weight, etc…). If the model predicts a lifetime risk ≥ 10 %, researchers and certain healthcare insurance companies recommend referral to genetic counseling services [8,9]. Furthermore, multiple cancer-related organizations recommend rigorous screening guidelines for women with TC lifetime risk scores > 20 % [9]. Identifying at-risk patients through these models, educating them about their risk, and explaining how risk scores can be used to secure insurance coverage could lead to greater uptake of genetic counseling services and mammography.

Little is known about the lifetime risk rates of women who utilize mobile mammography clinics, which provide access to screening mammograms by eliminating cost and transportation barriers, especially for women in underserved communities [10]. The goal of the current study is to assess lifetime risk for breast cancer in a population of women utilizing mobile mammography and to determine the proportion that might benefit from additional services like genetic counseling, rigorous screening, and education. Better understanding of these trends would enable us to understand and assess the criteria in recommending additional services for at-risk women, especially how to conduct outreach for genetic counseling services to women with elevated lifetime risk for breast cancer.

2. Methods

2.1. Study sample

Retrospective analysis of electronic health records for 2214 women screened on the Mount Sinai Mobile Mammography Van between October 2018 and September 2019 was conducted. Women were eligible to be screened on the van if they were 40 years of age or older, were a New York state resident, and had not been screened for breast cancer in the past year. A majority of patients reported never having had breast or ovarian cancer before (n = 2167, 98.72 %). Given the limitations of the TC risk estimation tool, women who reported having had breast and/or ovarian cancer in the past were excluded from these regression analyses (n = 28, 1.28 %).

2.2. Procedure

Prior to screening, patients completed a questionnaire via an iPad using the Ikonopedia application, which is a breast radiology reporting and patient tracking system designed to improve reporting efficiency and optimize facility operations [11]. Ikonopedia includes a survey tool, which serves as the first-ever web-based version of the TC Breast Cancer Risk Assessment Tool. Risk scores calculated by this tool can inform women about their lifetime risk of developing breast cancer; however, TC risk assessments are not currently returned to patients as part of clinical practice.

2.3. Variable definitions

Participants answered questions about their demographic characteristics, breast health, cancer screening history, and family history of cancer. Information about their biological sex and age was collected. NYC Borough, which was determined based on the patients’ screening location, included the following: Manhattan, Brooklyn, Bronx, Queens, and Staten Island.

Participants were asked whether they identified as Hispanic or Latina and which race they most identified with (i.e. American Indian/Alaska Native, Asian, Black/African American, Native Hawaiian/Pacific Islander, White, Other, and prefer not to answer). For our analysis, race and ethnicity were collapsed into one variable with the following categories: non-Latina (NL) white, NL black, Latina, NL Asian, and NL other.

Patients were enrolled in the New York State Cancer Services Program (CSP) if they lived in New York State, were 250 % or more under the federal poverty line, did not have a mastectomy or breast implants, and had health insurance issues. Enrolling patients visiting our mobile mammography van in CSP improved access and affordability for breast cancer screening, as enrollees either didn’t have health insurance or they had health insurance with a cost share that would prevent them from using screening services. For our analysis, we categorized patients as eligible or ineligible for CSP.

Abnormal mammograms were defined as mammograms resulting in BIRADS 0, 3, 4, or 5. A BIRADS 0 indicates that additional information is required, BIRADS 3 indicates probably benign findings, BIRADS 4 indicates suspicious findings, and BIRADS 5 indicates highly suspicious findings [12,13]. BIRADS 0, 3, 4, and 5 results required additional screening such as a diagnostic mammography, ultrasound, and/or biopsy.

Women were asked Are you having any unusual problems with your breasts? While women with any breast symptoms were not eligible for screening on the van, a minority of women (n = 61) responded yes to this question.

Participants were then asked whether they have given birth to one or more children, how many months total they breastfed their children (if any), have they gone through menopause, if they have Ashkenazi Jewish Inheritance, and if they had any family members been diagnosed with breast or ovarian cancer. In our analysis, whether the patient had given birth was categorized as yes or no. Months breastfeeding had the following categories: Has No Children, Never, 0 to < 6 Months, 6 to < 12 Months, and 12 + Months. Menopause was categorized as: Yes, In Menopause Now, and No. Responses regarding Ashkenazi Jewish inheritance and family history of breast and/or ovarian cancer were categorized as yes or no. Lastly, the question regarding whether the patient had BRCA testing done in the past was classified as follows: Tested, BRCA+; Tested, Normal; No/Unknown.

Based on responses to these variables above, TC 10-year risk and lifetime risk scores were calculated for each patient. If a woman had a lifetime risk score ≥ 10 %, she was considered to be recommended for additional services.

2.4. Data analysis

Descriptive statistics (i.e., frequency/percent and mean/standard deviation) were used to describe the distribution of our sample among responses. Logistic regression analyses were used to assess the odds of being recommended for additional services by the TC lifetime risk score. Manhattan, NL Asians, and CSP ineligible patients were chosen as the reference groups for our analyses. In particular, NL Asians were chosen as a reference group because they had the lowest TC risk scores on average.

For adjusted models, we controlled for variables related to the TC risk estimate: breast problems present at screening, having given birth, months breastfeeding, menopause, Ashkenazi Jewish inheritance, and family cancer history. Multiple models were run to be able to account for all relevant variables. For the race/ethnicity and borough models, we further adjusted for CSP eligibility. For the borough and CSP models, we further adjusted for race/ethnicity. All analyses were conducted using SAS version 9.4 (Research Triangle Institute, Research Triangle Park, NC).

3. Results

A descriptive breakdown of patients’ responses to the survey questions can be found in Table 1. The average age of the patients in our analytic sample (n = 2195) was 56.01 ± 10.19 years old. The majority of screenings took place in Manhattan (n = 1220, 55.58 %), but patients were also screened in Brooklyn (200, 9.11 %), the Bronx (316, 14.40 %), Queens (357, 16.26 %), and Staten Island (102, 4.65 %). Patients primarily reported being Latina (1009, 45.97 %), but we also screened patients who identified as NL white (172, 7.84 %), NL black (675, 30.75 %), NL Asian (209, 9.52 %), and NL other (130, 5.92 %). Almost a quarter of our patients were eligible for CSP (513, 23.38 %), and about one-fifth had an abnormal mammogram reading (445, 20.27 %).

Table 1.

Descriptive statistics for analytical sample of mobile mammography van patients (n = 2195).

Characteristic	Frequency/Mean	Percent/Std. Dev.
Age	56.01	± 10.19
New York City Borough
Manhattan	1220	55.58 %
Brooklyn	200	9.11 %
Bronx	316	14.40 %
Queens	357	16.26 %
Staten Island	102	4.65 %
Race/Ethnicity
Non-Latina White	172	7.84 %
Non-Latina Black	675	30.75 %
Latina	1009	45.97 %
Non-Latina Asian	209	9.52 %
Non-Latina Other	130	5.92 %
Cancer Services Program
Eligible	513	23.38 %
Ineligible	1681	76.62 %
Abnormal Mammogram
Yes	445	20.27 %
No	1750	79.73 %
Breast Problems Present at Screening
Yes	61	2.79 %
No	2124	97.21 %
Has Given Birth
Yes	1841	84.41 %
No	340	15.59 %
Breastfed
Has No Children	340	15.60 %
Never	585	26.85 %
0 to < 6 Months	542	24.87 %
6 to < 12 Months	361	16.57 %
12 + Months	351	16.11 %
Menopause
Yes	1080	49.52 %
In Menopause Now	379	17.38 %
No	722	33.10 %
Ashkenazi
Yes	38	1.74 %
No	2143	98.26 %
Family Cancer History
Yes	435	19.94 %
No	1746	80.06 %
BRCA Testing
Tested, BRCA+	3	0.14 %
Tested, Normal	12	0.55 %
No/Unknown	2173	99.31 %
Tyrer-Cuzick Ten-Year Risk Score	2.76 %	± 2.01
Tyrer-Cuzick Lifetime Risk Score	7.30 %	± 4.80
Tyrer-Cuzick Recommendation for Additional Services
Yes	444	20.23 %
No	1713	78.04 %
Ineligible	38	1.73 %

Abbreviations: Std. Dev., Standard Deviation

Sixty-one (2.79 %) women reported having breast problems present at the time of screening. Most patients screened had given birth before (1841, 84.41 %). Rates of lifetime breastfeeding varied among the women who had children: 585 (26.85 %) reported never having breastfed, 542 (24.87 %) reported having breastfed for 0 to < 6 months, 361 (16.57 %) reported having breastfed for 6 to < 12 months, and 351 (16.11 %) reported having breastfed for 12 + months. Most patients either had gone through menopause (1080, 49.52 %) or were currently perimenopausal (379, 17.38 %), but 722 women (33.10 %) reported not having begun perimenopause. A small minority of patients reported having Ashkenazi Jewish inheritance (38, 1.74 %). Almost one-fifth of patients (435, 19.94 %) reported a family history of breast and/or ovarian cancer.

The average TC ten-year risk score was 2.76 % ± 2.01 %, and the average TC lifetime risk score was 7.30 % ± 4.80 %. Using lifetime risk scores ≥ 10 %, it was determined that 444 patients (20.23 %) could be referred for additional services. Less than one percent of patients had been tested for the BRCA genes previously: 3 (0.14 %) had tested positive, and 12 (0.55 %) had tested negative. Two thousand one hundred and seventy-three (99.31 %) patients reported never having been tested for BRCA or not knowing if they had been tested.

Table 2 displays the unadjusted and adjusted logistic regression analyses for being recommended for additional services by the TC lifetime risk score. The odds of being recommended for additional services by the TC model among those who lived in Brooklyn (OR=1.18, 95 % CI: 0.82–1.70), the Bronx (OR=0.97, 95 % CI: 0.71–1.32), Queens (OR=1.15, 95 % CI: 0.86–1.53), and Staten Island (OR=1.17, 95 % CI: 0.72–1.90) were not significantly different from those who lived in Manhattan. When adjusting for race/ethnicity, breast problems present at screening, having given birth, months breastfeeding, menopause, Ashkenazi Jewish inheritance, and family cancer history, these relationships still did not achieve statistical significance.

Table 2.

Unadjusted and adjusted analyses of Tyrer-Cuzick recommendations for additional services by borough, race/ethnicity, and CSP eligibility.

	Unadjusted Models		Adjusted Models*
	Odds Ratio	95 % CI	Odds Ratio	95 % CI
Borough
Brooklyn	1.18	0.82–1.70	1.14	0.74–1.77
Bronx	0.97	0.71–1.32	1.21	0.84–1.74
Queens	1.15	0.86–1.53	1.17	0.81–1.69
Staten Island	1.17	0.72–1.90	1.31	0.74–2.31
Manhattan	1.00	–	1.00	–
Race/Ethnicity
Non-Latina White	1.82	1.12–2.96	1.59	0.80–3.16
Non-Latina Black	1.37	0.92–2.04	1.37	0.86–2.20
Latina	1.02	0.69–1.50	0.78	0.50–1.23
Non-Latina Other	1.13	0.64–1.99	0.71	0.34–1.50
Non-Latina Asian	1.00	–	1.00	–
CSP Eligibility
Eligible	1.31	1.03–1.66	1.37	1.03–1.81
Ineligible	1.00	–	1.00	–

Abbreviations: CSP, Cancer Services Program; CI, Confidence Interval

^*All models were adjusted for breast problems present at screening, having given birth, months breastfeeding, menopause, Ashkenazi Jewish inheritance, and family cancer history. For the race/ethnicity and borough models, we also adjusted for CSP eligibility. For the borough and CSP models, we also adjusted for race/ethnicity.

The odds of being recommended for additional services by the TC model were significantly greater among NL white patients when compared to NL Asian patients (OR=1.82, 95 % CI: 1.12–2.96). These odds for NL blacks (OR=1.37, 95 % CI: 0.92–2.04), Latinas (OR=1.02, 95 % CI: 0.69–1.50), and NL others (OR=1.13, 95 % CI: 0.64–1.99) were statistically equivalent to those among NL Asians. However, after controlling for CSP eligibility, breast problems present at screening, having given birth, months breastfeeding, menopause, Ashkenazi Jewish inheritance, and family cancer history, none of these relationships achieved statistical significance, including that for whites (Adjusted OR=1.59, 95 % CI: 0.80–3.16).

The odds of being recommended for additional services by the TC model were significantly greater among those who were eligible for CSP when compared to those who were ineligible (OR=1.31, 95 % CI: 1.03–1.66). When adjusting for race/ethnicity, breast problems present at screening, having given birth, months breastfeeding, menopause, Ashkenazi Jewish inheritance, and family cancer history, this relationship was still statistically significant (adjusted OR=1.37, 95 % CI: 1.03–1.81).

4. Discussion

Nearly one-fifth of our patient sample had an abnormal mammogram reading and < 1 % had ever been tested for BRCA gene mutations; however, about 20 % of women screened could have potentially benefited from additional services. Our findings are in line with those from Turbitt et al., which showed that only 1.2 % of women in a nationally-representative sample reported receiving genetic counseling and that 0.8 % of women had genetic testing for hereditary breast cancer risk [14]. Gwyn et al. examined whether women due for routine mammography would want genetic testing and found that approximately 41 % of respondents intended to pursue testing [15]. However, Morman et al. evaluated the utility of breast cancer risk assessment at the time of screening mammogram and found that only about 10 % of women were compliant with referrals to genetic counseling [6]. These findings demonstrate the need for developing programs to combat the known barriers to genetic counseling and testing, which include knowledge and awareness, cost-related concerns, and medical mistrust [7].

We did not find any differences in recommendations for additional services using TC lifetime risk estimates by NYC borough or race/ethnicity. While those in different boroughs and those of different races/ethnicities may have an equal need for additional services, research has shown that there are a variety of barriers to genetic counseling that impact varied communities inequitably, such as issues securing transportation and cultural barriers. While screening patients with the mobile mammography van helps minimize time or geographic barriers, abnormal readings and genetic counseling referrals would require patients to visit a fixed site, reducing access to healthcare. Moreover, there are a variety of systemic barriers that prevent minority populations from accessing genetic counseling services. A review by Williams et al. revealed lower genetic counseling and/or testing rates for eligible racial/ethnic minorities among family members of high-risk individuals [16]. Forman et al. reported that barriers to genetic services for racial/ethnic and underserved populations included socioeconomic barriers (e.g., time limitations, access to knowledgeable providers, and language/cultural barriers) as well as psychosocial barriers (e.g., medical mistrust and fears of discrimination) [17].

The TC model has been shown to be well calibrated across most racial/ethnic groups with the exception of Latinas, for whom it overestimates breast cancer risk [18]. Given that Latinas are less likely to develop breast cancer than NL women [19] and that our study found similar odds for recommending additional services across racial/ethnic groups, it is likely that this overestimation is present in our study. Researchers suggest one way to improve the TC model for this population is to include the presence of minor allele rs140068132 at 6q25, a protective single-nucleotide polymorphism commonly found in Latinas [18]; however, this information is not readily available without genetic testing and would need to be simulated using population prevalence data. Nevertheless, for our purposes, this overestimation would only result in more women having access to potentially life-saving services and cancer education.

Furthermore, those who were eligible for CSP, a proxy for lack of insurance, were significantly more likely to be recommended for additional services based on TC lifetime risk estimates when compared to those who were ineligible. While this finding may be attributable differences in factors considered by the TC model between these populations (e.g., demographic and lifestyle factors), the association between poverty and insurance status with breast cancer risk includes environmental stressors on breast cancer biology as well as barriers to proper screening and treatment methods [20]. Williams et al. examined environmental factors stemming from low SES and their increased risk for early menarche, which is associated with higher risk for developing breast cancer; these factors included prenatal smoke exposure, obesity, low fruit and vegetable intake, and chronic stress [16].

Hormone receptor (HR) status, specifically estrogen receptor-negative and progesterone receptor-negative status, is also associated with higher breast cancer risk and lower survival rates. Andaya et al. found that high poverty areas tended to have a greater prevalence of HR-negative and HR-unknown tumors compared to more affluent areas [21]. Moreover, Quang et al. found that patients presenting with breast cancer and patients who were uninsured were more likely to present with a breast-specific complaint rather than an abnormal mammogram, and thus were more likely to have advanced disease [22], representing the lower likelihood for uninsured women to utilize screening and genetic counseling modalities. Environmental stressors and barriers to care may be related to the higher recommendation for additional services in the CSP population.

Our findings were consistent with previous literature in this area of research. We sampled a diverse population across the five boroughs of NYC, and our study had a relatively large sample size which increased the strength of the analyses. Our analyses also adjusted for lifestyle and family history factors that are a part of the TC lifetime risk model; in doing so, we were able to elucidate other factors related to lifetime risk for breast cancer. However, we did not discriminate between cisgender and transgender individuals in our analyses, as no one has examined the utility of the TC model in males or transgender individuals to our knowledge.

Another limitation of our study is that self-reported measures are vulnerable to social desirability and recall biases, leading to inaccurate reporting on the questionnaire administered through Ikonopedia. Moreover, due to the reading level proficiency and language barriers present in the population under study, translation services could have led to potential miscommunications. Furthermore, using cross-sectional analyses for our study gives way to temporal limitations in assessing any exposure or outcomes. Lastly, our sample may not depict NYC racial and ethnic population exactly and our study may not have included other important variables that are analyzed in the literature (e.g., BRCA test results, intentions for genetic testing, etc…).

Despite breast cancer being the second leading cause of death for women in the U.S., there is no requirement for women who present as high-risk to be referred for additional services. We argue that estimating lifetime risk using models like TC can improve clinical care by identifying patients who could benefit from additional services, such as genetic counseling, more rigorous or more frequent cancer screenings, and cancer education. Through early detection and awareness, women are more likely to identify cancerous lesions early, thereby reducing the morbidity and mortality associated with developing breast cancer. Patients and providers should have discussions about patients’ average TC lifetime risk at follow-up breast cancer screening appointments, in addition to providing referrals to genetic counseling services when necessary, as this is a missed opportunity for both mobile mammography clinics and fixed sites, including those at our own medical center. Future studies could develop intervention programs that provide high-risk patients with direct linkage to additional services like genetic counseling and education to explore these recommendations.