Customize your JAMA Network experience by selecting one or more topics from the list below.
Armstrong K, Micco E, Carney A, Stopfer J, Putt M. Racial Differences in the Use of BRCA1/2 Testing Among Women With a Family History of Breast or Ovarian Cancer. JAMA. 2005;293(14):1729–1736. doi:10.1001/jama.293.14.1729
Author Affiliations: Department of Medicine,
University of Pennsylvania School of Medicine (Dr Armstrong, Mss Micco, Carney,
and Stopfer); Abramson Cancer Center, University of Pennsylvania School of
Medicine (Dr Armstrong); Center for Clinical Epidemiology and Biostatistics,
University of Pennsylvania School of Medicine (Drs Armstrong and Putt); Leonard
Davis Institute of Health Economics, University of Pennsylvania (Dr Armstrong);
Center for Health Equity Research and Promotion, Philadelphia Veterans Affairs
Medical Center (Dr Armstrong), Philadelphia, Pa.
Context Given the current context of racial disparities in health and health
care and the historical context of eugenics, racial disparities in the use
of genetic susceptibility testing have been widely anticipated. However, to
our knowledge there are no published studies examining the magnitude and determinants
of racial differences in the use of genetic susceptibility testing.
Objectives To investigate the relationship between race and the use of BRCA1/2 counseling among women with a family history of breast or ovarian
cancer and to determine the contribution of socioeconomic characteristics,
cancer risk perception and worry, attitudes about genetic testing, and interactions
with primary care physicians to racial differences in utilization.
Design, Setting, and Participants Case-control study (December 1999-August 2003) of 408 women with a family
history of breast or ovarian cancer, of whom 217 underwent genetic counseling
for BRCA1/2 testing (cases) and 191 women did not
(controls). Participants received primary care within a large health system
in greater Philadelphia, Pa.
Main Outcome Measures Probability of carrying a BRCA1/2 mutation,
socioeconomic characteristics, perception of breast and ovarian cancer risk,
worry about breast and ovarian cancer, attitudes about BRCA1/2 testing, and primary care physician discussion of BRCA1/2 testing were measured prior to undergoing BRCA1/2 counseling for cases and at the time of enrollment for controls.
Results African American women with a family history of breast or ovarian cancer
were significantly less likely to undergo genetic counseling for BRCA1/2 testing than were white women with a family history of breast
or ovarian cancer (odds ratio, 0.22; 95% confidence interval, 0.12-0.40).
This association persisted after adjustment for probability of BRCA1/2 mutation, socioeconomic characteristics, breast and ovarian
cancer risk perception and worry, attitudes about the risks and benefits of BRCA1/2 testing, and primary care physician discussion
of BRCA1/2 testing (adjusted odds ratio for African
American vs white, 0.28; 95% confidence interval, 0.09-0.89).
Conclusions Racial disparities in the use of BRCA1/2 counseling
are large and do not appear to be explained by differences in risk factors
for carrying a BRCA1/2 mutation, socioeconomic factors,
risk perception, attitudes, or primary care physician recommendations. The
benefit of predictive genetic testing will not be fully realized unless these
disparities can be addressed.
Mutations in BRCA1 or BRCA2 confer a significantly increased risk of breast and ovarian cancer.
Women with a BRCA1/2 mutation have a lifetime breast
cancer risk of 50% to 85% and a lifetime ovarian cancer risk of 14% to 40%.1 Testing for mutations in BRCA1/2 can provide individual information about breast and ovarian cancer
risk that may affect decisions about prophylactic surgery, chemoprevention,
and screening.2 In addition, the identification
of a BRCA1/2 mutation within a family can allow other
family members to better define their own risk of cancer. BRCA1/2 mutations are rare in the general population; however, women
with a family history of breast and ovarian cancer are at increased risk of
carrying a mutation.3
Although clinical testing for BRCA1/2 mutations
became available in 1996, relatively little is known about the determinants
of testing use in the general population, including potential racial differences
in utilization. Although many prior studies have examined interest in testing
among the general population, the correlation between level of interest and
actual testing use has been poor and the determinants of interest inconsistent.4-8 Most
prior studies of actual use of testing have focused on (1) decisions about
testing once women have participated in genetic counseling; (2) descriptive
characteristics of women undergoing testing; or (3) the uptake of testing
among high-risk families involved in research protocols.9-14 These
studies have generally included mostly white women. However, it is unclear
whether this low prevalence of nonwhite participants represents the populations
being studied (eg, hereditary breast cancer families participating in research
registries) or indicates a racial disparity in the use of testing.
There are several reasons to believe that racial disparities in the
use of predictive genetic testing are likely to exist. Disparities have been
demonstrated in most areas of health care and may be greatest for new technologies.15 Some populations may be particularly likely to distrust
genetic testing because of historical attempts to label certain racial groups
as genetically inferior.16 Genetic testing
is expensive and may be inaccessible to individuals with fewer financial resources.
Racial differences in attitudes have been demonstrated for many components
of medical care and may exist for genetic testing.17
Given the lack of empirical data about the utilization of genetic testing
in the general population and concern about the potential existence of racial
disparities in this area, the goals of this study were to assess the association
between race and use of genetic counseling for BRCA1/2 testing
among women at risk of carrying a BRCA1/2 mutation
and to evaluate the potential contributions of socioeconomic characteristics,
cancer risk perception and worry, attitudes about genetic testing, and interactions
with primary care physicians to this association.
We conducted a health system population–based case-control study
to identify the determinants of the use of genetic counseling for BRCA1/2 testing among women with a family history of breast or ovarian
cancer. The use of BRCA1/2 counseling was selected
as the outcome of interest because we were interested in the factors associated
with referral for genetic testing rather than the decisions about testing
after genetic counseling. Approximately three quarters of women who underwent
counseling completed BRCA1/2 testing after counseling.
The study protocol was approved by the institutional review board at the University
The reference population was defined as women between the ages of 18
and 80 years who had seen a primary care physician within the University of
Pennsylvania Health System in the 3 years prior to the start of the study.
During the study, the University of Pennsylvania Health System included approximately
225 primary care physicians who saw more than 250 000 primary care patients
annually. The primary care physician population was 42% women; 5.4% African
American, 7.7% Asian American, 1.5% Hispanic, and 85.4% white. The reference
population was identified using billing databases. Women from the reference
population were included in the study if they had a first- or second-degree
relative with a breast or ovarian cancer diagnosis. Because the primary aim
of the study was to identify the determinants of BRCA1/2 counseling among women who would be targeted for primary cancer prevention,
women were excluded if they had a personal diagnosis of breast or ovarian
cancer. In addition, women were excluded if they were identified as being
unable to participate because of illness or mental incapacity by their primary
Cases were all women from the reference population who presented for
genetic counseling at the Cancer Risk Evaluation Program at the University
of Pennsylvania Health System between December 1999 and August 2003. At the
time of the study, the Cancer Risk Evaluation Program was the primary provider
practice of genetic counseling within the University of Pennsylvania Health
System. Of the 449 women who participated in the Cancer Risk Evaluation Program
over the study period, 355 completed data collection for a response rate of
Controls were a random sample of women from the reference population
who had not participated in genetic counseling for BRCA1/2 testing. Controls were accrued every 6 months during the study to
approximate the timing of case accrual. At each 6-month interval, a sample
of controls was selected equal to twice the number of cases accrued over the
prior 6 months. Controls were excluded if they had previously participated
in BRCA1/2 genetic counseling. Of the 956 eligible
controls, 603 were successfully enrolled for a response rate of 63.1%.
It was not possible to assess personal or family cancer history from
the billing database. Thus, family and personal history of cancer was measured
by self-report at the time of data collection and used to define the final
study population. Of the 1062 women who completed data collection, 392 women
were not included in the final sample because they did not have a family history
of breast cancer (375 controls and 17 cases). In addition, 254 women were
excluded because they had a personal diagnosis of breast cancer (38 controls
and 224 cases).
Data were collected using a mailed self-administered questionnaire,
with cases completing the questionnaire prior to being seen at the Cancer
Risk Evaluation Program. Potential participants were informed that participation
was completely voluntary and would have no effect on their medical care, that
their responses would be confidential, and that completion of the questionnaire
represented their consent to participate.
Sociodemographic Characteristics. Age, education,
and annual household income were measured using items from the Behavioral
Risk Factor Surveillance System 1998 questionnaire.18 Race/ethnicity
and religious heritage were measured by self-report using items developed
from the US Census guidelines.19
Probability ofBRCA1/2Mutation. Each woman’s probability of carrying a BRCA1/2 mutation was calculated from a prediction model
developed by Frank et al.20,21 This
model includes family history of breast cancer (number of relatives diagnosed
before age 50 years and number of relatives diagnosed at age 50 years or older),
family history of ovarian cancer, and Ashkenazi Jewish heritage.
Cancer Risk Perception and Worry. Participants
were asked to rate their lifetime risk of developing cancer in 3 ways: (1)
on a scale from 0% to 100%; (2) on a 7-point Likert scale ranging from very
low to very high; and (3) compared with the average woman using a 7-point
Likert scale ranging from much lower to much higher.22 Risk
perception was assessed separately for breast and ovarian cancer. Because
items for each cancer were highly correlated and the results were substantively
the same irrespective of the measure used, we present the analyses using the
comparative 7-point Likert scale for each cancer. Cancer worry was measured
using a previously developed scale that includes 2 items: “How often
do you worry about developing breast (or ovarian) cancer?” and “How
often does worry about breast (or ovarian) cancer interfere with your life?”
with response options on a 7-point Likert scale ranging from not at all to
all of the time.23 Separate breast and ovarian
cancer worry scores were created by averaging the scores for the 2 items for
Attitudes About Genetic Testing. Attitudes
were assessed by asking participants to endorse statements about the risk
and benefits of BRCA1/2 testing using a 7-point Likert
scale ranging from strongly disagree to strongly agree.9,24 For
these analyses, attitudes were grouped into 4 domains: information from testing
(6 items); discrimination from testing (3 items); reassurance from testing
(2 items); and anxiety from testing (2 items). Summative scores were created
for each of these domains by averaging the scores on individual items. The
multi-item measures of information from testing and discrimination from testing
had reasonable internal consistency (Cronbach α, 0.74 and 0.77, respectively)
and item total correlations between 0.34 and 0.76.
Primary Care Physician Visits and Discussions ofBRCA1/2Testing. Individual items
assessed how often the respondent had seen a gynecologist and how often she
had seen a primary care physician other than a gynecologist in the past year.
Response options were 0 times, 1 to 2 times, 3 to 5 times, and more than 5
times. In addition, for each specialty of physician, respondents were asked
if they had discussed BRCA1/2 testing.
Fisher exact tests, χ2 tests, and t tests were used to assess the statistical significance of differences
in the characteristics of cases and controls as indicated. Because the risk
perception, cancer worry and attitudinal measures were ordinal variables and
not normally distributed, we categorized scores into 3 groups based on a priori
cut points: low (scores<4); medium (score of 4); and high (scores>4). These
measures were compared between cases and controls using χ2 tests.
Multiple logistic regression was used to adjust the association between race
and use of BRCA1/2 counseling for potential mediating
and confounding variables, which we specified a priori. Because less than
10% of our study population was from racial/ethnic groups other than African
American and white, we focused the multivariate analyses on the comparison
of African American and white women and excluded women of other racial/ethnic
groups from the analyses. Groups of variables were added sequentially to the
model to observe the effect on the race coefficient. Age, probability of a BRCA1/2 mutation, risk perception, cancer worry, and attitudinal
variables were entered into the model as linear terms. Because some of the
psychological factors (risk perception and worry) were moderately correlated
(Pearson correlation coefficients, 0.19-0.58) and potentially introduced collinearity
into the model, we also constructed models using a composite score for risk
perception/cancer worry. The race coefficient in the model using the composite
score was not substantively different from the model including each individual
measure of risk perception and worry; thus, we present the models with the
individual measures. The final model demonstrated reasonable performance with
a nonsignificant Hosmer-Lemeshow goodness-of-fit statistic (P = .34; 10 groups) and a C statistic
of 0.92. The analyses were conducted using Stata 7.0 software package (Stata
Corp. College Station, Tex). All hypothesis tests were 2-tailed, using a P = .05 level of significance.
The characteristics of the 217 cases and 191 controls are reported in TABLE 1. Among women with a family
history of breast or ovarian cancer, women who underwent BRCA1/2 counseling (cases) were significantly less likely to be African
American than women who did not undergo BRCA1/2 counseling
(7.4% of cases vs 28.8% of controls). In addition, women undergoing BRCA1/2 counseling had a greater probability of carrying
a BRCA1/2 mutation, were younger, more likely to
be Jewish, and more likely to be married. Education and annual household income
levels were associated with use of BRCA1/2 counseling
but type of health insurance coverage was not. Women who underwent BRCA1/2 counseling were more likely to have seen a gynecologist more
than 2 times in the past year, but less likely to have seen a primary care
physician other than a gynecologist. Discussion of BRCA1/2 testing by their gynecologist or primary care physician was associated
with use of BRCA1/2 counseling.
Cancer risk perception, cancer worry, and attitudes about BRCA1/2 testing are reported in (TABLE 2). Each measure of risk perception and worry was significantly
higher among cases than controls. Attitudes about genetic testing were also
significantly associated with the use of BRCA1/2 counseling,
with women who underwent counseling being more likely to endorse the benefits
of testing including the potential to gain information and reassurance. Interestingly,
women who underwent BRCA1/2 counseling were more
likely to endorse concerns about anxiety from BRCA1/2 testing.
The estimate of the inverse association between African American race
and BRCA1/2 counseling changed very little after
adjustment for probability of a BRCA1/2 mutation,
socioeconomic characteristics, cancer risk perception and worry, attitudes
about genetic testing, and primary care physician visits and discussions of BRCA1/2 testing (TABLE 3). Adjustment for racial differences in BRCA1/2 mutation probability, sociodemographic factors, and risk perception
led to slight increases in the point estimate of the odds ratio (OR) for the
association between race and BRCA1/2 counseling (OR,
0.22-0.40), with the race coefficient no longer reaching strict statistical
significance in the model including probability of a BRCA1/2 mutation, sociodemographic factors, and psychological factors (OR,
0.40; P = .08). However, further adjustment
for attitudes about genetic testing and primary care physician interactions
yielded a decrease in the point estimate for the race coefficient (OR, 0.40-0.28)
and increase in the statistical significance. In the final multivariate analysis,
use of BRCA1/2 counseling remained strongly inversely
associated with African American race (P = .01).
This association also held when the sample was restricted to women who were
not of Ashkenazi descent (adjusted OR, 0.26; 95% confidence interval [CI],
0.07-0.94). In addition, in the final model, use of BRCA1/2 counseling was positively associated with the probability of BRCA1/2 mutation, younger age, higher breast cancer risk
perception, higher ovarian cancer worry, and attitudes about discrimination
and reassurance from testing. In addition, use of BRCA1/2 counseling was more common among women who had discussed BRCA1/2 testing with their gy necologist or other primary care physician.
This study demonstrates that African American women with a family history
of breast or ovarian cancer are much less likely to undergo genetic counseling
for BRCA1/2 testing than are white women with a family
history of breast or ovarian cancer. Furthermore, this racial disparity is
not explained by differences in the probability of carrying a BRCA1/2 mutation, socioeconomic status, cancer risk perception and
worry, attitudes about BRCA1/2 testing, or primary
care physician discussions of BRCA1/2 testing.
Racial disparities in the clinical use of genetic susceptibility testing
have been widely anticipated but, to our knowledge, this is the first prospective
study to document their occurrence. Prior studies have demonstrated racial
differences in attitudes about genetic testing, as well as racial differences
in cancer risk perception and response to education and counseling about BRCA1/2 testing in a research setting.24-27 We
previously noted a racial difference in the early uptake of BRCA1/2 testing; however, that study was limited by its retrospective
design, small sample size, and inability to examine potential causes of racial
differences in use of BRCA1/2 counseling.28 In addition to providing some of the first empirical
evidence of a racial disparity in utilization of genetic testing in clinical
practice, the results of this study are important because of the magnitude
of the disparity, with white women having almost 5 times the odds of undergoing BRCA1/2 counseling as African American women.
Because the benefit of BRCA1/2 testing for
all women with a family history of breast or ovarian cancer is controversial,
the implications of racial disparities in the use of BRCA1/2 testing are complex. On one hand, BRCA1/2 testing
provides information about breast and ovarian cancer risk that can influence
decisions about cancer risk reduction and save lives. For example, prophylactic
oophorectomy has been demonstrated to reduce the risk of ovarian cancer by
more than 90%, extend life expectancy by more than 3 years, and be widely
acceptable to women with BRCA1/2 mutations who have
completed childbearing.2,29,30 On
the other hand, the risk information provided by BRCA1/2 testing is imprecise, testing is expensive, and BRCA1/2 mutations are relatively rare, even among women with a family
history of breast or ovarian cancer. Furthermore, many of the available breast
cancer risk-reduction interventions involve significant trade-offs and uptake
of interventions such as prophylactic mastectomy or tamoxifen chemoprevention
among mutation carriers is far from universal.29,30 Given
these issues, not all women with a family history of breast or ovarian cancer
will choose to undergo counseling for BRCA1/2 testing.
However, although the optimal proportion of women with a family history of
breast or ovarian cancer who should undergo counseling for BRCA1/2 testing remains uncertain, there are no clear clinical reasons
why utilization should differ between African American and white women with
a family history of breast or ovarian cancer. Breast cancer incidence is lower
among African American than white women in the United States, but breast cancer
is diagnosed younger and case fatality is higher among African American women,
resulting in similar benefits from prevention and early detection.31 Although the published data are sparse, several studies
have suggested that the prevalence of mutations in African American hereditary
breast cancer families may be similar to that in white hereditary breast cancer
our sample of women with a first- or second-degree relative with breast or
ovarian cancer, the predicted probability of a BRCA1/2 mutation
differed very little between African American and white women. Thus, racial
differences in utilization of BRCA1/2 counseling
cannot be directly attributed to differences in the clinical utility of BRCA1/2 testing.
The racial disparity in use of BRCA1/2 counseling
in this population was not explained by differences in the probability of
carrying a BRCA1/2 mutation, socioeconomic status,
cancer risk perception and worry, attitudes about the risks and benefits of BRCA1/2 testing, or primary care physician discussions
of BRCA1/2 testing. In fact, in contrast to many
other clinical scenarios, the strength of the association between race and
use of BRCA1/2 counseling changed very little after
multivariate adjustment. Other factors must contribute to this disparity.
Health care related–distrust has been demonstrated to be higher among
African Americans than whites and may serve as a barrier to use of medical
care.37,38 New technologies, such
as genetic testing, may be particularly sensitive to the effects of distrust.
We are currently exploring the role of health care distrust in this and other
areas where racial disparities have been demonstrated in cancer control. Some
groups may be reluctant to undergo genetic testing because of concerns about
how genetic information may be used. We have recently demonstrated that African
Americans are more likely to be concerned about racial discrimination on the
basis of genetic testing.24 Although African
American women were no more likely to report concern about insurance or employment
discrimination on the basis of BRCA1/2 testing, the
specific concern about racial discrimination was not measured in the current
study and may contribute to the disparity in uptake. Alternatively, racial
differences in use of BRCA1/2 counseling may reflect
differences in the characteristics of the primary care physicians of African
American and white women rather than differences in the characteristics of
the women themselves. If women of different races see different types of primary
care physicians who are associated with different probabilities of referral
for BRCA1/2 counseling, the racial difference in
use of BRCA1/2 counseling could arise from the clustering
of African American patients with certain types of physicians. Such types
might be defined by characteristics that are often used to describe physicians,
such as specialty, practice setting, and demographics, or by behavioral or
attitudinal characteristics that may be more predictive of making referrals
for BRCA1/2 counseling, such as the tendency to be
an early or late adopter of new technology.39 This
possibility is supported by recent evidence that African Americans are seen
by a relatively small proportion of primary care physicians across the United
States and that primary care physicians seen by African Americans are less
likely to be board certified and more likely to report difficulty delivering
high quality care than primary care physicians for visits by whites.40 Because the current study was designed to examine
the contribution of patient level characteristics to a potential racial difference
in use of BRCA1/2 counseling, we were unable to assess
the contribution of differences in physician characteristics. However, given
the lack of explanatory power of patient characteristics, differences in physician
characteristics should be an important focus of future research.39
This study also informs the broader research agenda in racial disparities
in health care. Unlike many prior studies investigating the causes of racial
disparities, we were able to test the contribution of a broad range of potential
causes including clinical, socioeconomic, psychological, attitudinal, and
primary care factors. The results of these analyses highlight several issues.
First, although residual racial differences in medical care after adjustment
for clinical and socioeconomic factors are often attributed to racial differences
in attitudes or perceptions of risk, in this study, adjusting for racial differences
in attitudes and risk perception only increased the size of the racial disparity
in utilization.15 Second, the large racial
difference in the use of BRCA1/2 counseling was not
explained by the factors that are traditionally cited as the predominant causes
of racial disparities in health care.15 Third,
the size of the disparity in utilization of BRCA1/2 counseling
is larger than those seen with more established technologies such as cardiac
catheterization.15 Together, these issues emphasize
the growing recognition that all racial disparities are not alike and the
causes of racial disparities are likely to vary substantially across clinical
setting, time, and place.
This study provides insight into other factors that are associated with
utilization of clinical BRCA1/2 counseling among
women with a family history of breast or ovarian cancer in the general population.
Several of these factors are not surprising. BRCA1/2 counseling
is more common among women who are at increased risk of carrying a mutation
and who have a higher perceived risk of breast cancer. Younger age and having
had a primary care physician or gynecologist discuss BRCA1/2 testing are also associated with use of BRCA1/2 counseling, as well as the belief that BRCA1/2 testing
will provide reassurance and the absence of concern about discrimination from BRCA1/2 testing. Perhaps somewhat more surprisingly, concern
about anxiety from BRCA1/2 testing was higher among
women seeking BRCA1/2 counseling than controls, suggesting
that this concern is not a significant barrier to the use of genetic testing.
This association is supported by a prior study that demonstrated higher cancer
distress among women choosing to undergo testing.12
The results of this study must be considered within its limitations.
We studied women within a large health system in the greater Philadelphia
area and our results may not generalize to women who do not see a primary
care physician or who live in other areas of the United States or elsewhere.
In addition, we were unable to examine use of BRCA1/2 counseling
among other racial or ethnic groups because these groups are not well represented
in our health system population. It is possible that some women who receive
primary care through our health care system may have undergone BRCA1/2 counseling outside of the health care system during the study
period. To assess this possibility, we surveyed health care system primary
care physicians at the study mid point about the use of BRCA1/2 counseling among their patients. None reported having a patient
who had undergone counseling at a different site. Thus, we believe our case
ascertainment to be reasonably complete.
Our response rate was lower among controls than cases and raises concerns
about possible response bias. To assess this possibility, we compared responders
and nonresponders on age and race/ethnicity, the 2 variables available from
the billing database. Among the cases, responders did not differ significantly
from nonresponders in age or race/ethnicity. Among the controls, responders
did not differ from nonresponders in age but were slightly less likely to
be African American than were nonresponders (33% vs 38%; P = .02), suggesting that our assessment of the magnitude
of the racial disparity in use of BRCA1/2 testing
is conservative. We did not collect information about unaffected relatives
and were unable to calculate BRCA1/2 probabilities
using models that include such information.41 However,
the predictive model we used is widely accepted for use in genetic counseling
and represented the standard of care for the majority of the study period.3 As noted previously, uncertainty remains about who
should undergo BRCA1/2 testing and the limited data
about mutation prevalence among some groups of women who are not white may
increase the uncertainty about testing in this population. However, paradoxically,
the pronounced racial disparity in use of genetic testing contributes to this
lack of data about mutation prevalence, creating the potential for a vicious
cycle of lack of information and lower rates of utilization of a potentially
The complex and highly charged relationship between race and genetics
presents a substantial challenge to the translation of advances in human genetics
into improvements in health. Although empirical evidence of racial disparities
in the utilization of BRCA1/2 counseling provides
important information about this challenge, it only further highlights the
need to move forward in developing health and social policy that ensures the
potential benefits of the Human Genome Project are realized for all segments
of the US population.
Corresponding Author: Katrina Armstrong,
MD, MSCE, 1204 Blockley Hall, 423 Guardian Drive, Philadelphia, PA 19104 (firstname.lastname@example.org).
Author Contributions: Dr Armstrong had full
access to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Armstrong.
Acquisition of data: Armstrong, Micco, Carney,
Analysis and interpretation of data: Armstrong,
Drafting of the manuscript: Armstrong, Micco,
Critical revision of the manuscript for important
intellectual content: Armstrong, Carney, Stopfer, Putt.
Statistical analysis: Armstrong, Micco, Putt.
Obtained funding: Armstrong.
Administrative, technical, or material support:
Armstrong, Micco, Carney, Stopfer.
Study supervision: Armstrong.
Financial Disclosures: None reported.
Funding/Support: Dr Armstrong is supported
by the American Cancer Clinical Research Training Grant and the Robert Wood
Johnson Generalist Physician Faculty Scholar Award.
Role of the Sponsors: The funding organizations
had no role in the design and conduct of the study; the collection, management,
analysis and interpretation of the data; or the preparation, review, or approval
of the manuscript.
Acknowledgment: We acknowledge the work of
Barbara Weber, MD, and J. Sanford Schwartz, who provided valuable guidance
with the study design.