The Gail model is validated to estimate breast carcinoma risk. The authors assessed the association of Gail risk scores with screening and cancer risk perception.
The Gail model is validated to estimate breast carcinoma risk. The authors assessed the association of Gail risk scores with screening and cancer risk perception.
Using the 2000 National Health Interview Survey, the authors studied women ages 41–70 without a cancer history. Gail scores ≥ 1.66% defined increased risk. The authors used logistic regression to assess associations between breast carcinoma risk and previous and recent (≤ 1 year) mammography and clinical breast examination (CBE).
Of 6410 women, 15.7% had increased risk. High-risk women more frequently reported previous mammograms (94% vs. 85%; P < 0.0001), previous CBE (93% vs. 88%; P < 0.0001), recent mammograms (70% vs. 54%; P < 0.0001), recent CBE (71% vs. 61%; P < 0.0001), and high cancer risk perception (20% vs. 9%; P < 0.0001). However, 30% of high-risk women had not received a recent mammogram. After adjustment for sociodemographic factors, access to care factors, and cancer risk perception, high-risk women remained more likely to have received recent mammography (adjusted odds ratio [OR], 1.45, 95% confidence interval [95% CI], 1.19–1.77), recent CBE (OR, 1.32; 95% CI, 1.08–1.61]), and previous mammography than average-risk women. The authors observed an interaction between risk and age, with women ages 41–49 years more frequently reporting previous mammography (OR, 4.79; 95% CI, 1.55–4.81) than average-risk, same-age women. For women age ≥ 50 years, the odds of previous mammography were similar regardless of risk.
In a nationally representative sample, 15.7% of women had increased breast carcinoma risk using the Gail model. High-risk women perceived higher cancer risk and more often received screening. However, nearly one in three high-risk women did not receive recent screening and most of these women did not perceive increased risk. Cancer 2004. © 2004 American Cancer Society.
Breast carcinoma is a major cause of morbidity and mortality in the U.S. It is the most commonly diagnosed cancer among women, accounting for approximately 40,000 deaths annually.1 One study showed that screening mammography decreased breast carcinoma mortality among women ages 50–69 years.2 Clinical breast examinations (CBE) help to improve early detection. For example, 5% of breast carcinomas are detected by CBE and not mammography.3 Many organizations recommend initiation of breast carcinoma screening at age 40 years,4–7 although the optimal screening interval remains controversial. Recommendations vary from annually after age 40 years,4, 5 to biennially after age 40 years,6 to biennially between ages 40–50 years and yearly thereafter.7
Past studies have established several modifiers of breast carcinoma screening behavior. These include sociodemographic factors such as age, race, education, and income,8–14 as well as markers of access to care such as a usual source of care, insurance, and frequency of healthcare visits.8–11 Factors shown to increase breast carcinoma risk have also been evaluated with respect to their impact on screening. Most studies have assessed individual risk factors for breast carcinoma12, 15–18 rather than overall risk. In addition, findings from studies of overall risk have been inconsistent.19–21
The Gail model is a validated tool to estimate a woman's overall risk of developing breast carcinoma in the next 5 years. Risk estimates are calculated based on age, age at menarche, age at first live birth, family history of breast carcinoma in first-degree relatives, and history of breast biopsy.22 Gail scores of ≥ 1.66% typically indicate increased risk.23 This threshold is believed to reflect changes in breast carcinoma incidence after age 60 years and to approximate the risk of the average 60-year-old woman in the U.S.24 This threshold has been used in previous studies, such as the Breast Cancer Prevention Trial,23 and the Study of Tamoxifen and Raloxifene,25 to qualify women for enrollment. A similar threshold is employed in clinical practice to determine eligibility for chemopreventive agents.25
In this context, we evaluated the relation between breast carcinoma risk as assessed by the Gail model and breast carcinoma screening practices in a national sample of women. We hypothesized that women at increased risk for developing breast carcinoma would have higher rates of screening than average-risk women. We also evaluated the role of cancer risk perception and provider recommendations concerning mammography use and identified factors associated with failure to undergo screening among high-risk women.
We used data from the 2000 National Health Interview Survey (NHIS),27 a nationally representative sample of the civilian, noninstitutionalized U.S. population. The NHIS is an annual survey administered by the National Center for Health Statistics with data collected through in-person interviews. Information on health status and sociodemographics is available from the Sample Adult Core of the Basic Module. In 2000, a cancer control module was appended that included questions regarding cancer history, screening, family history, and other health-related behaviors. This module includes information on the five risk factors in the Gail model. A total of 32,374 adults were interviewed, with an overall response rate of 72%.
We identified 7923 women ages 41–70 years. We selected the lower age limit as many women initiate screening during their 40th year. We selected this upper age limit because breast carcinoma screening guidelines are less consistent for women age > 70 years. We excluded 601 women with a history of cancer (other than nonmelanoma skin carcinoma) and 886 women who were missing data regarding ≥ 1 Gail risk factors. An additional 26 women were excluded because they gave likely inaccurate responses to Gail risk factor questions (e.g., age at menarche of 19–50 years, age at first birth < 10 years). Of the 912 women excluded based on Gail risk factor data, 43% were missing information concerning > 2 risk factors.
To determine breast carcinoma risk, we computed 5-year Gail scores for all women. Women with scores ≥ 1.66% were considered to be at increased risk for developing breast carcinoma and those with lower scores were considered to be at average risk. We categorized each risk factor according to the Gail model22 including age (< 50 years, ≥ 50 years), age at menarche (< 12 years, 12–13 years, ≥ 14 years), age at first live birth (< 20 years, 20–24 years, 25–29 years or nulliparous, ≥ 30 years), number of benign breast biopsies (0, 1, and ≥ 2), and number of first-degree relatives with breast carcinoma (0, 1, and ≥ 2).
Several NHIS questions ascertained breast biopsy history. We considered women to have a history of biopsy if they reported a needle biopsy or lump excision after an abnormal mammogram. Women reporting at least one mammogram were also asked if they ever had a benign breast lump removed and if so, how many. We classified 915 women who reported never having a mammogram (and who were not asked about previous biopsies) as having no history of biopsy because we believed that it was clinically unlikely that a woman would have a biopsy without a previous mammogram.
We assessed both history of and recent breast carcinoma screening with both mammography and CBE. We defined recent screening as screening within 1 year, consistent with recommendations from the American Medical Association and the American Cancer Society.4, 5 We conducted sensitivity analyses using a 2-year screening interval consistent with recommendations of the U.S. Preventive Services Task Force.6 For analyses of mammography outcomes, we excluded 225 women whose most recent study was performed to evaluate specific breast problems because they were not considered screening studies.
We examined the relation between breast carcinoma risk and cancer risk perception. Respondents were asked, “Would you say your risk of getting cancer in the future is low, medium, or high?” We categorized women into two groups comparing women who believed their cancer risk was high with those who did not.
We evaluated the association between risk status and provider recommendations for mammography among women not screened in the previous 2 years. These women were asked why they were not screened and whether their provider recommended mammography within the previous year. Women who attributed failure to undergo screening to a lack of recommendation and women reporting no provider recommendation for mammography within the previous year were considered not to have received a recommendation.
Because women with an increased risk of developing breast carcinoma may be more likely to benefit from screening than average-risk women, unscreened high-risk women may represent an important group on which to focus education and screening efforts. Given this, we further examined women at increased risk to identify sociodemographic and access-to-care factors associated with failure to undergo recent mammography. For this analysis, factors found to be significant on bivariable analyses were initially entered into the model. We used backwards elimination to exclude nonsignificant factors from the model. Eliminated covariates were evaluated for confounding (i.e., a ≥ 20% change in the adjusted odds ratio [aOR]). No confounders were identified.
Bivariable analyses were performed using chi-square tests. We used multivariable logistic regression to evaluate the independent effect of increased breast carcinoma risk on breast carcinoma screening practices beyond known risk factors.8–14 Each model was adjusted for age, race, education, income, number of healthcare visits in the previous year, having a usual source of care, insurance, region, urban residence, and cancer risk perception. Covariates with > 5% missing data were dummy-coded in our models.
We explored two interactions with risk status. First, we evaluated the interaction between breast carcinoma risk and age dichotomized at 50 years. We hypothesized that breast carcinoma risk may be an effect modifier for women in their 40s given the increased controversy regarding breast carcinoma screening for women ages 40–49 years. We also evaluated the interaction between risk status and cancer risk perception. We hypothesized that women at increased risk who perceive themselves at high risk for developing cancer would be more likely to undergo screening. We repeated analyses of screening outcomes, restricting our sample to white women because the Gail model is validated only for white women.28–31 We were unable to assess the association between risk and screening within other racial or ethnic groups, given the small numbers of high-risk women in these groups in the current study.
We used SAS-callable SUDAAN (Version 8.1; SAS, Cary, NC) to account for the complex sampling design and to obtain results weighted to reflect national estimates. For all analyses, significance was determined at P < 0.05. The Committee on Clinical Investigations at Beth Israel Deaconess Medical Center (Boston, MA) approved our use of the 2000 NHIS.
Our final sample included 6410 women (Table 1). Overall, 15.7% of women had Gail scores ≥ 1.66%, representing an estimated 5.95 million women nationwide at increased risk of developing breast carcinoma. As expected, high-risk women were older than average-risk women. The majority of women in both groups were white, although there were more white women in the high-risk group. Compared with average-risk women, women at increased risk had slightly higher education levels and were somewhat less likely to report extremes of income. The majority of women in both groups reported access to care as represented by a usual source of care, insurance, and visits to a healthcare provider in the previous year.
|Characteristics||Breast carcinoma risk (%)c|
|Average risk (n = 5473)||Increased risk (n = 937)|
|Less than high school||6||3|
|Some high school/high school diploma||44||40|
|Some college/Associate degree||41||45|
|College degree or higher||9||12|
|Usual source of carea|
|No. of healthcare visits in past yra|
|Cancer risk perceptiona|
Overall, 86% of women reported a history of mammography and 56% reported a recent mammogram. Table 2 presents the associations between breast carcinoma risk and mammography screening. Compared with average-risk women, high-risk women were more likely to ever have a mammogram (85% vs. 94%) and to have received recent mammography (54% vs. 70%). However, 30% of women at increased risk, representing 1.74 million U.S. women in this age group, have not received a recent mammogram.
|Breast carcinoma risk||Ever mammography|
|No. (weighted percent)||Adjusted ORb (95% CI)|
|Average risk (n = 5473)|
|Ages 41–49 yrs||1889 (79)||1.00 —|
|Age ≥ 50 yrs||2547 (90)||2.51 (2.06–3.06)|
|High risk (n = 937)|
|Ages 41–49 yrs||115 (97)||4.89 (1.58–15.14)|
|Age ≥ 50 yrs||719 (93)||2.83 (2.01–4.00)|
After adjustment, high-risk women remained more likely than average-risk women to have received mammography screening. Without considering our hypothesized interaction terms, an increased risk of breast carcinoma remained significantly associated with receipt of previous mammography (aOR = 1.50; 95% confidence interval [95%CI], 1.08–2.09) and recent mammography (aOR 1.47; 95% CI, 1.20–1.79).
After considering potential interactions for previous mammography, we observed a significant interaction between risk status and patient age (Table 2). Women ages 40–49 years at increased risk for breast carcinoma were substantially more likely to have undergone a previous mammogram than women of the same age at average risk (aOR 4.89; 95% CI, 1.58–15.14). For women age ≥ 50 years, the odds ratios for previous mammography were similar regardless of risk. We found no significant interactions between risk and cancer risk perception for previous mammography.
Table 3 presents the unadjusted and adjusted associations of risk with CBE. Overall, 89% of women reported ever having CBE whereas 63% had a recent breast examination. Compared with average-risk women, high-risk women were more likely to have ever had a CBE (88% vs. 93%) and to have had a recent CBE (61% vs. 71%). As with mammography, approximately 30% of high-risk women failed to undergo recent screening. After adjustment, high-risk women remained more likely than average-risk women to report a previous CBE (aOR 1.37; 95% CI, 0.99–1.91) and recent CBE (aOR 1.34; 95% CI, 1.10–1.63), although the former did not achieve statistical significance. We found no significant interaction between risk and age or cancer risk perception for previous or recent CBE.
|Characteristics||Clinical breast examination|
|Ever (weighted percent)a||Within 1 yr (weighted percent)a|
|High risk||871 (93)||660 (71)|
|Average risk||4757 (88)||3262 (61)|
|Adjusted ORb (95% CI)||1.37 (0.99–1.91)||1.34 (1.10–1.63)|
First, we repeated our analyses of recent screening using a 2-year interval because recommended breast carcinoma screening intervals vary. We found similar associations between risk status and recent screening with mammography and CBE, although the association with CBE was slightly weaker and no longer achieved significance (aOR 1.23; 95% CI, 0.97–1.56).
Second, we restricted analyses to white women because the Gail model is validated only for white women. We found that risk status was significantly related to all screening outcomes except for previous CBE and that the interaction between risk status and age remained significant for previous mammography. The magnitude of the associations was similar for all outcomes, although the effect of risk on history of mammography among women < 50 years was modestly attenuated (aOR 4.04; 95% CI, 1.26–12.89).
Table 1 presents the association between breast carcinoma risk and cancer risk perception. As expected, women at increased risk more often considered themselves at high risk for developing cancer than average-risk women (20% vs. 9%). Despite this, few women perceived an increased risk of developing cancer. Among the high-risk women, 44% considered themselves to be at low cancer risk and 36% perceived a “medium” risk of cancer in the future. It is important to note that 80% of women at increased breast carcinoma risk, representing an estimated 4.5 million women nationwide, did not perceive their future cancer risk as high.
Among women not screened in the previous 2 years (n = 1827), there was no difference noted between risk groups with regard to the proportion reporting no recent provider recommendations for mammography (57% vs. 52%; P = 0.58).
Among high-risk women, failure to undergo recent screening was associated with lower education, lower income, lack of a usual source of care, lack of insurance, and having no visits to a healthcare provider in the previous year in unadjusted analyses. After adjustment, high-risk women with lower education, lack of insurance, and no visit to a healthcare provider in the past year remained less likely to report recent screening (Table 4).
|Covariates||Recent mammogram (weighted percent)||Adjusted OR (95% CI)|
|Less than high school||55||0.48 (0.21–1.10)|
|Some high school/high school diploma||64||0.59 (0.43–0.83)|
|Some college/Associate degree||74||1.00 —|
|College degree or higher||75||0.99 (0.57–1.70)|
|No. of visitsa|
|> 7||73||1.17 (0.77–1.78)|
To our knowledge, this is the first study to assess the relation between overall breast carcinoma risk as measured by the 5-year Gail score and breast carcinoma screening practices in a nationally representative sample. Overall, 15.7% or an estimated 5.95 million women ages 41–70 are at increased breast carcinoma risk nationwide. Women at increased risk were significantly more likely to have undergone recent breast carcinoma screening with mammography and CBE than average-risk women. Nevertheless, 30% of high-risk women failed to undergo recent mammograms and 29% of high-risk women did not have a recent CBE. Moreover, although increased breast carcinoma risk was associated with higher overall cancer risk perception, 80% of women at increased risk did not perceive a high personal cancer risk.
We found few studies to date that have documented the prevalence of increased Gail scores in the general population, with estimates ranging from 6% to 16%, perhaps reflecting variability in age and sample criteria across studies.31–33 In a national sample, we found that 15.7% of women ages 41–70 years had 5-year Gail risk estimates of ≥ 1.66%. This estimate is similar to the 15.5% reported by Freedman et al.33 and is higher than the 8% reported by Bastian et al.,32 which may reflect the older age distribution of our study sample.
Evaluating screening practices among women with an increased risk of breast carcinoma is important, because they may be more likely to benefit from screening than average-risk women. To our knowledge, few studies published to date have evaluated overall breast carcinoma risk as a determinant of screening,19–21 even though it may better approximate a woman's risk than individual factors. Moreover, these studies used different methods of assigning risk and evaluated various outcomes, generating inconsistent conclusions regarding the relation between risk and screening.
Taplin et al.19 assigned women to an overall breast carcinoma risk category based on their individual constellation of several risk factors and assessed the response of women considered at increased risk to an invitation to participate in screening. They observed that overall risk status was a stronger predictor of screening than individual factors and that increasing risk status was associated with higher participation rates. Other studies, however, did not report a significant association between overall risk estimates and screening.20, 21
These results are consistent with previous studies using family history of breast carcinoma to assess increased breast carcinoma risk. Using slightly different screening intervals, Daly et al.12 reported that 29% of women age > 35 years with affected first-degree relatives denied recent mammograms. Bastani et al.21 examined women age > 40 years with positive family histories who did not receive a risk intervention and found that 34% reported no mammography screening in the previous year.
Although we found that women at increased risk were more likely to have undergone screening, a substantial number of women at risk had not been screened recently. Failure of high-risk women to undergo recent screening may be due, in part, to unrecognized risk by providers, misunderstanding and lack of knowledge of personal risk, inadequate access to healthcare, or informed decisions not to be screened despite risk. We found that > 50% of unscreened, high-risk women reported no physician recommendation for mammography and that only 20% of all high-risk women perceived their risk as high, suggesting a possible role for unrecognized risk by both providers and patients. Several studies suggest that provider recommendation is an important determinant of mammography use.15, 34–36 In addition, a lack of provider recommendation was an important reason for not having undergone mammography among women with or without a family history of breast carcinoma.20, 37 We also found that high-risk women without health insurance and who had not visited a healthcare provider in the past year were substantially less likely to report recent mammography use. Inadequate access to care has been shown to be associated with decreased utilization of screening for several cancer diagnoses, including breast carcinoma.10, 21, 38
As expected, we observed a significant interaction between age and risk status, but only for ever having mammography. We found that high-risk women in their 40s were nearly 5 times more likely to ever have had a mammogram than average-risk women of the same age. Among women age ≥ 50 years, those at high risk had similar odds ratios of having a mammogram as those at average risk, suggesting that risk status plays less of a role for older women. It is not clear why we found no evidence of an interaction between age and risk status for recent mammography. However, given the controversy regarding screening for women ages 40–49 years, it is possible that providers are more likely to incorporate increased risk status into their decision to begin to screen women age 40–49 years with mammography.
Women at high risk for breast carcinoma were substantially more likely to perceive themselves at high risk for developing cancer, consistent with findings from other studies.16 Nevertheless, the majority of high-risk women in the current study underestimated their risk. In fact, 44% of high-risk women considered their risk to be low and 36% considered their risk to be medium. These findings may reflect an optimism bias (the belief that one is at lower risk than others) reported in previous studies of risk perception39, 40 and may suggest that many women might benefit from education concerning their cancer risk. Similar to the current study, others have found that the majority of women with increased breast carcinoma risk underestimate their cancer risk. One study reported that 20% of women at increased risk for breast carcinoma by family history perceived high personal breast carcinoma risk.15 Another study found that 31% of women age < 60 years with increased breast carcinoma risk by 5-year Gail score perceived a higher than average risk of dying of breast carcinoma, although the study had a small number of such high-risk women.39 Moreover, studies also find that a substantial proportion of women with increased breast carcinoma risk believe themselves to be at lower risk for developing breast carcinoma than other women.15, 40
Although we found that perceived cancer risk was related to estimated breast carcinoma risk, risk perception did not explain differences in screening between risk groups. Risk status remained associated with nearly all screening outcomes after adjusting for risk perception. We found no evidence of an interaction between objective risk and risk perception, suggesting that it is unlikely that increased screening among high-risk women is mediated through higher risk perception.
The role of provider recommendations in explaining the differences in screening between risk groups is unclear. Among women who were not screened with mammography within 2 years, we found no differences in provider recommendation rates between high-risk and average-risk women. This finding suggests little influence of risk status on provider recommendations, at least among unscreened women. Nevertheless, we cannot directly evaluate the role of provider recommendations in mammography use between risk groups because provider recommendations were not ascertained for screened women. Some evidence suggests that risk may not significantly impact provider recommendations for screening.13
The findings of the current study should be interpreted in light of several limitations. First, we used self-reported data and, therefore, some degree of misclassification and recall bias may have occurred. However, previous evidence suggests that patient recall is a reliable measure of previous cancer screening.41 Second, although the Gail model adjusts to some extent for race/ethnicity by incorporating different baseline composite incidence rates and competing hazard rates for some racial/ethnic groups, to our knowledge it has only been validated among white women. It remains unclear how well these results can be generalized to women of different racial groups. We were unable to evaluate the association between risk status and screening within racial or ethnic subgroups because of small numbers of high-risk black women, Hispanic women, and women of other race in the current study. Third, we had limited information on cancer risk perception because only one question was asked and it was not specific to breast carcinoma. However, this would tend to overestimate the proportion of women who consider their future cancer risk high by capturing women at increased risk for other cancers. However, we found that few women perceived their risk as high. Fourth, the Gail model may underestimate risk for some women, including those with family histories suggestive of genetic mutations. Because family history was limited to first-degree relatives in the NHIS, we were unable to determine how many women in our sample may fall into such categories. However, few women reported more than one affected first-degree relative (< 1%) and all were classified as high risk. Finally, we categorized women without a history of mammography as having no previous breast biopsy because they were not asked about biopsies. It is possible that some women may have undergone a biopsy without mammography. However, this would underestimate the risk for these women, as well as our association between risk status and screening.
Despite these limitations, the current study offers several advantages. We assessed risk using the validated Gail model and analyzed recent data from a large, nationally representative sample. We also included a comparison group of average-risk women and adjusted for socioeconomic and access factors. Moreover, we evaluated self-reported screening practices instead of participants' responses to invitations to participate in screening programs19, 20 or risk interventions concerning breast carcinoma screening.21 Therefore, the current study findings may more likely represent actual screening behavior in the general population.
These findings have important implications. Increased breast carcinoma risk according to 5-year Gail scores is relatively common among U.S. women ages 41–70 years, affecting an estimated 5.95 million women. High-risk status is associated with past mammography use and recent screening with both mammography and CBE. However, nearly one-third of women at increased risk did not receive recent screening with either modality. Moreover, few women at increased risk perceived their risk of developing cancer as high. Given this, future studies are needed to develop and evaluate strategies promoting risk assessment and education regarding risk to reduce the number of missed screening opportunities among high-risk women.
The authors thank the following individuals for their assistance in providing programs for the Gail model: Beverly Rockhill (Department of Epidemiology, University of North Carolina, Chapel Hill, NC); Donna Spiegelman and Ellen Hertzmark (Department of Epidemiology, Harvard School of Public Health, Boston, Massachuestts); and Mitchell Gail (Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD).