In the United States, breast cancer is the most commonly diagnosed malignancy in women. In 2012, the American Cancer Society predicts that 226,870 American women will be diagnosed with breast cancer, and 39,510 will die from the disease.1 It is clear that this disease affects African American women differently, and an interplay of both societal and individual factors2 contributes to a 36% higher mortality rate in this population compared with their non-African American counterparts.3
In November 2009, the United States Preventive Service Task Force (USPSTF) published a set of updated guidelines for breast cancer screening that differed markedly from their last update in 2002. These updates no longer support routine screening mammography for women ages 40 to 49 years but, instead, leave such decisions in the hands of the patient and physician. The updated USPSTF guidelines also suggest that women ages 50 to 74 years undergo screening biennially rather than annually, as in previous versions.4 These updated guidelines are divergent from the American Cancer Society (ACS) guidelines; consequently, it is uncertain how this change will affect patient survival, especially among certain subpopulations.
In the current study, we assess the potential impact of biennial screening starting at age 50 years versus annual screening starting at age 40 years in a public hospital. Our concern is that physicians and staff in public hospitals and busy health care clinics may not have the time or resources to prioritize an in-depth discussion of the benefits and risks of screening mammography. However, these health care facilities are seeing and treating a disproportionately high number of lower incomes families and minority women at “high risk” of cancer at a younger age. The objective of this study was to compare the effect of the USPSTF and ACS guidelines on patient disease stage and survival in a predominantly African American population treated in an inner urban public hospital setting.
MATERIALS AND METHODS
Study Population and Study Variables
A cross-sectional, retrospective review of patient charts and tumor registry data was performed to identify patients who were diagnosed with and/or treated for stage I through III breast cancer at the AVON Comprehensive Breast Center at Grady Memorial Hospital during the 2008 calendar year. Because of limitations in the tumor prediction model, in situ disease and distant metastatic disease were omitted (see Tumor Staging Model, below). In total, 151 patients were diagnosed with breast cancer at Grady Memorial Hospital in 2008. The following patients were eliminated for various reasons: recurrent disease (n = 7), metastatic disease (n = 24), in situ disease (n = 31), duplicate patient entries because of bilateral disease (n = 4), and men (n = 1). This left a total sample size of 84 women with stage I through III breast cancer. Of these 84 qualified patients, 3 women were lacking data on clinical disease stage, and an additional 3 women were lacking data on tumor size or grade necessary to calculate predicted stages. Information was abstracted from patient care review summaries, radiology and pathology reports, radiation oncology and medical oncology records, and surgical notes. Data collected included age, race, insurance type, family history, previous screening dates, method of detection, clinical and pathologic stages, biopsy type, and information regarding tumor biology. Measurements on ultrasound imaging were chosen to represent the most accurate tumor size, and diagnostic mammography was substituted if ultrasound imaging was unavailable. “Observed” clinical stages were assigned using the seventh edition of the American Joint Committee on Cancer (AJCC) Cancer Staging Manual5 (AJCC-7) and were corroborated with patient care review summaries and radiologic reports.
Tumor Staging Model
By using previously published data on tumor doubling times of primary breast carcinomas, we estimated prospective and retrospective tumor sizes based on patients' ages and tumor grades. Peers and colleagues used serial mammograms to calculate time-dependent changes in tumor size by assuming an exponential growth pattern. This gave rise to growth rates expressed as tumor volume doubling times. When evaluated by age, women aged <50 years had an average doubling time of 80 days (range, 44-147 days), women ages 50 to 70 years had a doubling time of 157 days (range, 121-204 days), and women aged >70 years had a doubling time of 188 days (range, 120-295 days).6 Assuming that tumor biology also plays a role in the rate of tumor growth, we further stratified each group by pathologic grade (see Table 1). We assumed that grade 1 tumors had longer tumor doubling time ranges to signify slower growth rates and that grade 3 tumors had shorter doubling time ranges to signify faster growth rates.7 The mean of each range was used in prospective and retrospective calculations. This allowed us to develop regression equations to predict tumor size based on patient age, tumor size, and tumor grade.
Table 1. Age-Adjusted and Grade-Adjusted Average Tumor Doubling Timesa
|<50||131 (114-147)||96 (79-113)||61 (44-78)|
|50-70||190.5 (177-204)||163 (150-176)||135 (121-149)|
|>70||243.5 (237-250)||158 (179-237)||149 (120-178)|
For each patient, tumor sizes were predicted at annual intervals before and after the date of biopsy; and, for patients aged <50 years, tumor sizes at age 50 years also were predicted. The threshold for detection by mammography screening was set at 10 mm, and the threshold for palpable or clinically detectable tumors was set at 20 mm.8 We realize that many tumors may present outside our threshold parameters, but these limits were applied to derive a conservative estimate.
”Tumor size” was defined as the largest single dimension of the primary tumor mass identified by ultrasonography or diagnostic mammography. The most accurate tumor size was based on radiographic reports and patient conference reports as determined by a group of dedicated breast imagers. Ultrasound imaging size was the most preferred modality unless otherwise stated by the documentation noted above. The date of diagnosis was defined as the date an invasive cancer was identified by stereotactic or ultrasound-guided biopsy. Tumor size was presumed constant between the time of size measurement (ultrasonography/mammography) and the time of “diagnosis” (biopsy).
Clinical stages were assigned based on measurement from ultrasonography as the preferred modality (70%; n = 59 of 84 patients), diagnostic mammography (15%; n = 13 of 84 patients), or other sources, such as clinical examination (14%; n = 12), to provide the tumor classification (T) element of the AJCC-7 TNM staging system, in which T represents gradations of tumor size, N indicates the involvement of regional lymph nodes, and M indicates the presence of distant metastases.5 The morphologic appearance of lymph nodes on ultrasonography, computed tomography, or positron emission tomography imaging reports was used to provide lymph node status, which also was drawn from patient care documents at the time of diagnosis indicating physical examination findings or from pathology reports of fine-needle aspiration/biopsy results of lymph node sampling.
Tumor Size and Breast Cancer Stage Prediction
ACS guidelines recommend annual screening starting at age 40 years,9 whereas USPSTF guidelines recommend biennial screening starting at age 50 years unless individual patients and practitioners decide that earlier screening is preferred.4 For the purpose of this study, USPSTF-predicted stages are the “worst case” scenario and assume that no discussion or individualization of screening has taken place. Therefore, all USPSTF stages in this study are predicted as those occurring with biennial screening starting at age 50 years.
Through chart review, radiologic reports of imaging performed at Grady Memorial Hospital as well as patient care documents noting past imaging were assessed for the patient's breast imaging history. The date of the patient's last screening mammogram was determined. If the date of most recent imaging study was within the prior 2 years, then hypothetical USPSTF guideline adherence would provide no change from their actual date of diagnosis. If the date of most recent imaging was within the past year, then neither hypothetical ACS guidelines nor USPSTF guidelines would provide a change and, thus, there would be no predicted advantage of either screening schedule.
When prior screening had occurred but at an interval greater than 1 or 2 years, an earlier diagnosis would be predicted by hypothetical adherence to ACS guidelines (actual interval, >1 year), USPSTF guidelines (actual interval, >2 years), or both. Similarly, if the patient underwent no prior screening, then 1 or both sets of guidelines could be expected to confer an advantage. In these cases, retrospective tumor sizes were back-calculated to predict the size of the cancer at the point that it may have been discovered.
For some patients who had had tumors detected by screening, their tumors may have been identified later had they strictly adhered to a 1-year or 2-year screening interval. In these patients, prospective tumor sizes were calculated assuming detection at the next mammogram scheduled according to hypothetical ACS or USPSTF guidelines. For patients screened before age 50 years, their predicted tumor size at age 50 years was calculated for their “USPSTF” predicted stage. In our study, we adjusted for tumor characteristics by grouping women with similar ages, grades, and tumor sizes. In women with the same tumor characteristics, those screened every year would have better survival than those screened every 2 years. This should mitigate against lead time bias. Thus, in this study, because we adjusted for tumor characteristics, if we also adjusted for lead time bias, then we would be doing it twice.10
By using these tumor sizes, clinical stages were assigned based on AJCC-7 criteria; and, because lymph node status and metastasis could not be predicted by the tumor model, all patients were assigned N0 and M0 status. We also assumed that all patients with predicted tumor sizes >20 mm would have presented clinically with palpable mass; therefore, patients were assigned a disease stage of IIA (T2N0M0) unless they actually presented clinically otherwise and the stage at which they presented was used as their assigned clinical stage.
Differences in patient demographics, diagnosis, and tumor characteristics across the 3 guideline schemes were assessed; and differences in stage classification between the observed and ACS guidelines, the observed and USPSTF guidelines, and ACS and USPSTF guidelines were examined using Pearson chi-square contingency statistics. Screening, tumor size differences at biopsy, and pathologic features for the 3 classification guidelines were analyzed by applying the t test for paired data between the guideline groups.
Patient Demographics & Clinical Features
Both the mean age and the median age of patients in our study were 55 years, and 7 of 84 patients (83%) were aged <40 years, 23 patients (27%) were ages 40 to 49 years, 49 patients (58%) were ages 50 to 74 years, and 5 patients (6%) were aged >75 years. Our study population consisted of 4 racial/ethnic groups, and the majority of patients (86%; n = 71 of 84) self-identified as African American, 8% (n = 7) self-identified as Hispanic, 4% (n = 4) self-identified as Asian, and 1% (n = 1) self-identified as white. Insurance types were grouped into 4 categories: almost half (49%; n = 41 of 84) had Medicaid coverage, 23% (n = 19) had Medicare coverage, 19% (n = 16) were self-pay, and 9% (n = 7) had private coverage.
Screening and Biopsy Characteristics
Clinical features and screening characteristics of the study population are presented in Table 2. Within our sample, 58% of patients (n = 49 of 84) had no family history of breast or ovarian cancer (no relatives reported with these cancers), whereas 38% (n = 32 of 84) reported having a relative (any degree) with either breast or ovarian cancer. The majority of patients (51%; n = 42 of 84) presented with a palpable mass, whereas 44% (n = 36) had no symptoms at presentation and had disease detected by screening mammography. Few patients presented with other symptoms, such as nipple discharge, erythema/edema, or breast pain, as described in Table 2. The mammographic presentation of patients with invasive cancer was primarily a mass with or without associated calcifications (91%; n = 76 of 84); although other characteristics, such as isolated calcifications, intracystic lesion, and asymmetry, were documented in a few patients. Within our database, there were 43 patients with documented prior screening mammograms; and, of these, 14 patients had a prior screen within the past year (presumably following ACS guidelines), whereas 11 patients had a prior screen within the past 2 years (coinciding with USPSTF guidelines). There were 19 patients for whom the mammogram in question was reported as a baseline study and 22 patients for whom information on prior screening was unknown.
Table 2. Screening and Biopsy Characteristics, n = 84
|Family history|| |
| Either breast or ovarian cancer||32 (38)|
| No history of breast or ovarian cancer||49 (58)|
| Unknown/missinga||3 (4)|
|Detection method|| |
| Screening mammogram, asymptomatic||36 (43)|
| Palpable mass||42 (50)|
| Nipple discharge||1 (1)|
| Erythema/edema||1 (1)|
| Breast pain||2(2)|
|Biopsy characteristic|| |
| Mass||61 (73)|
| Calcification||4 (5)|
| Mass and calcification||15 (18)|
| Intracystic||1 (1)|
| Asymmetry||1 (1)|
| Unknown/missinga||2 (2)|
|Prior screening|| |
| No prior screening||19 (22)|
| Prior screening||43 (51)|
| Prior screen in 2007 annual||14 (17)|
| Prior screen in 2006, biennial||11 (13)|
| Prior screen in 2005-1995, other||18 (21)|
| Unknown/missingb||22 (26)|
Clinical and Pathologic Features
Of the 84 invasive breast cancers in our database, 43% (n = 36 of 84) were triple-negative (estrogen receptor [ER]-negative, progesterone receptor [PR]-negative, and human epidermal growth factor receptor-2 [HER2]-negative on immunohistologic testing), 44% (n = 37) were ER-positive and/or PR-positive only, and 6% (n = 5) were HER2-positive only. In our population, 12% of cancers (n = 10 of 84) were low grade, 52% (n = 42) were intermediate grade, and 36% (n = 29) were high grade. The predominant cancer type was invasive ductal carcinoma (IDC), which was reported in 83% of patients (n = 70 of 84), although mixed IDC and ductal carcinoma in situ (DCIS) was observed in 8% of patients (n = 7). Invasive lobular cancers and medullary carcinoma, although rare, also were diagnosed. The clinical and pathologic features of the study patients are presented in Table 3.
Table 3. Clinical and Pathologic Features, n = 84a
|Histologic type|| |
| IDC||70 (83)|
| IDC with DCIS||7 (8)|
| ILC||3 (4)|
| ILC with LCIS||1 (1)|
| Medullary carcinoma||1 (1)|
| Unknown/missing||2 (2)|
|Histologic grade|| |
| Low, grade 1||10 (12)|
| Intermediate, grade 2||42 (50)|
| High, grade 3||29 (35)|
| Unknown/missing||3 (4)|
|Hormone receptor status|| |
| ER-positive or PR-positive and Her-2/Neu-negative||37(44)|
| ER/PR-negative and Her-2/Neu-positive||5 (6)|
| Triple negative||36 (43)|
| ER/PR-positive and Her-2/Neu-positive||4 (5)|
| Unknown/missing||2 (2)|
Distribution of Stages
The ACS-predicted stages indicated the most women with stage I cancers (46%; n = 39 of 84) compared with their observed stage (stage I diagnosed, 37%; n = 31) or their USPSTF-predicted stage (stage I diagnosed, 14%; n = 12). If our sample of women had followed the USPSTF recommendations, then our model predicted a larger number of stage II cancers (73%; n = 61 of 84) than the number of women with an observed stage II diagnosis (46%; n = 39) or the number of women predicted to be diagnosed with stage II disease according to the ACS guidelines (43%; n = 36). The distribution of women with stage III cancers was 13% (n = 11 of 84) based on their observed stage at diagnosis, 4% (n = 3) with the predicted ACS-recommended guidelines, and 6% (n = 5) with the predicted USPSTF-recommended guidelines. All results are reported in Table 4. The shifts in disease stages from observed to those predicted by the different guidelines were statistically significant (P < .0001).
Table 4. Distribution of Observed Stages and Predictive Stages According to American Cancer Society and US Preventive Services Task Force Guidelines, n = 84a
| I||31 (37)||39 (46)||12 (14)|| |
| II||39 (46)||36 (43)||61 (73)|| |
| III||11 (13)||3 (4)||5 (6)|| |
| Missing||3 (4)b||6 (7)c||6 (7)c|| |
|Chi-square analysis|| || || || |
| Observed vs ACS|| || || ||132.02 [< .0001]|
| Observed vs USPSTF|| || || ||103.17 [< .0001]|
| ACS vs USPSTF|| || || ||149.61 [< .0001]|
Breast cancer is a diverse disease in which there are significant disparities between racial and ethnic groups. It is believed that this difference is caused by many factors, including socioeconomic status, demographic factors, cultural beliefs, access to health care, coexisting medical conditions, and tumor biology.11 The AVON Comprehensive Breast Center at Grady Memorial Hospital serves a primarily African American, economically disadvantaged, and underserved community; and past initiatives have focused on improving access to screening mammography and clinical trials to decrease this disparity. The result has been an increased number of screening mammograms in the Grady Memorial Hospital population from 11,942 in 2005 to 16,140 in 2008, yielding a 35% increase.12 It is believed that the increased screening service in this population has led to earlier diagnosis. The USPSTF acknowledges that, in the body of evidence reviewed, no adjustments to guidelines are made on the basis of studies that specifically address racial/ethnic disparities in breast cancer. However, the USPSTF does suggest that the decision to start screening before age 50 years should be individualized to address the patient's beliefs and concerns. The time-intensive nature of this approach can be challenging in a public health care setting given the volume of patients treated. Therefore, in the current study, we analyzed the effects of biennial screening starting at age 50 years versus annual screening starting at age 40 years in a patient population distinctive for its racial and socioeconomic under representation in previous studies.
Our results demonstrated a statistically significant difference between predicted ACS and USPSTF stages. A stage migration was observed, with a greater number of women hypothetically being diagnosed with later stage breast cancer using the USPSTF guidelines. The most notable difference was between stages I and II. The predicted ACS stages resulted in a larger number of stage I cancers (46%) compared with the actual (37%) and USPSTF (14%) stages, whereas the predicted USPSTF stages resulted in the largest number of stage II cancers (73%) compared with the ACS (43%) and actual (39%) stages. It is also important to note that the actual stage distribution had the largest percentage of stage III cancers (13%).
We believe this disparity between predicted and observed stages may have been caused in part by a limitation of our staging model: Because the tumor model threshold for palpable tumors was 20 mm, the most advanced stage predicted was stage IIA. However, the disparity also may have been related to the nature of the population screened, in which many patients presented with palpable masses.
The effectiveness of mammography screening is largely undisputed in women ages ≥50 years. However, in women ages 40 to 49 years, there is still much debate.13 A large Swedish study recently compared breast cancer mortality between counties that did and did not invite women for mammography service screening beginning at age 40 years. The results from that study indicted a 26% reduction (relative risk, 0.74; 95% confidence interval, 0.66-0.83) in breast cancer mortality for counties with service screening; and, when researchers further limited their analysis to those women who actually received screening mammograms, the reduction in risk was even greater at 29% (relative risk, 0.71; 95% confidence interval, 0.62-0.80).14 The correlation between breast cancer stage and patient survival is observed in the available statistics from the National Cancer Data Base, which reflects an approximately 88% 5-year relative survival rate for patients with stage I disease, a 74% to 81% 5-year relative survival rate for patients with stage II disease, and a 41% to 67% 5-year relative survival rate for patients with stage III disease.15 This suggests that the stage migration between stages I and II observed in our patients may have a substantial negative impact on patient survival. This information is particularly concerning for African American women, who have a 3-fold increased risk of developing triple-negative tumors regardless of age or body mass index.16 They also have a higher incidence rate of developing cancer before age 45 years3 with reports up to 52% higher among women aged <30 years.17 African American women also are at increased risk of dying from this disease at every age.3 These studies stress the concept that African American women have a host of clinical, societal, and pathologic factors that separates them from national norms. Our results, however, suggest that early and more frequent screening may be an important factor in reducing the risk of death from breast cancer in African American-predominant populations living in areas served by urban public hospitals.
Restrictions of our tumor model may represent the largest limitation of our study. Because we were unable to predict in situ disease, the impact of screening on stage 0 cancers was not demonstrated in this study. This is particularly important because a significant benefit of mammographic screening is related to the ability to identify in situ disease. The threshold size for radiographic detection of malignancy (10 mm) represents an additional limitation. In the American College of Radiology Imaging Network trial of digital versus film mammography (DMIST), 335 invasive cancers were detected. Of these, 82 tumors (24%) were ≤10 mm. However, that study included patients who had tumors detected by both digital and film mammography, and detection limits also depend on breast tissue density. For the purposes of modeling, a conservative estimate was used, but we recognize that lowering this threshold may demonstrate an even stronger shift in stages at diagnosis, providing a greater advantage to screening. Similarly, because of our tumor size threshold (20 mm) for palpable cancers and our inability to predict lymph node status and metastatic disease, the number of predicted stage III and stage IV cancers may have been underestimated. Our regression equation was based on patient age, tumor size, and tumor grade, which were identified in previous studies as significant predictors of tumor growth.6 We do acknowledge that other clinical and pathologic characteristics that may influence growth and disease progression, such as family history and hormone receptor status, were omitted.
Further limitations include our small sample size and our inability to measure associated harms. We also recognize that, in our study, although we adjusted for tumor characteristics, this may not adjust sufficiently for lead time bias for all patients. Because of these limitations and the unique nature of our patient population, we are cautious in generalizing our findings to dissimilar populations.
We plan to address some of these limitations in future studies by expanding the data set to include patient groups spanning 2006 to 2010. We also will apply National Cancer Data Base survival estimates by stage over our predictions of stage distribution, which will allow us to analyze the impact of each strategy on mortality figures. We also can use data on health care costs by stage to communicate the implications of stage migration for public health policy. Finally, by completing a panel review of serial mammograms in our patients, we can better understand the natural history of this disease in our patients by tailoring the tumor doubling times to be more accurate and improving our staging model.
Our current results emphasize the importance of clinical judgment and patient context in evaluating the revised breast screening guidelines. We provide evidence that annual mammography screening should begin at earlier ages, because screening may allow the detection of disease at an earlier clinical stages, especially in populations with social, economic, and racial/ethnic features similar to those in our study population. We encourage practitioners to prioritize a discussion regarding the benefits and risks of mammographic screening among high-risk women and to support earlier and more frequent screening intervals, which we believe ultimately will lead to a more favorable outcome in terms of patient mortality from breast cancer.
This study was supported by an AVON Foundation for Women Center grant and by the Georgia Cancer Coalition.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.