Butler Nattinger A, Hoffmann RG, Howell-Pelz A, Goodwin JS. Effect of Nancy Reagan's Mastectomy on Choice of Surgery for Breast Cancer by US Women. JAMA. 1998;279(10):762-766. doi:10.1001/jama.279.10.762
From the Departments of Medicine (Dr Nattinger) and Biostatistics (Dr Hoffmann and Ms Howell-Pelz) of the Medical College of Wisconsin, Milwaukee, and from the Department of Medicine and Center on Aging, University of Texas Medical Branch, Galveston (Dr Goodwin).
Context.— While the actions of popular figures are believed to influence the behavior
of the general public, including health care decisions, little research has
examined such an effect.
Objective.— To determine whether a temporal association exists between use of breast-conserving
surgery (BCS) for treatment of breast cancer and Nancy Reagan's mastectomy
in October 1987.
Design/Setting.— Population-based observational cohort study.
Patients.— Two sources of data: (1) 82230 women aged 30 years and older who were
included in the Surveillance, Epidemiology, and End Results tumor registry
because of a diagnosis of local or regional breast cancer from 1983 to 1990;
and (2) 80057 female Medicare beneficiaries aged 65 to 79 years who received
inpatient surgery for local or regional breast cancer in 1987 or 1988.
Main Outcome Measure.— Percentage of use of BCS vs mastectomy over time.
Results.— Compared with women undergoing surgery for breast cancer in the third
quarter of 1987 (just prior to Mrs Reagan's mastectomy), women were 25% less
likely to undergo BCS in the fourth quarter of 1987 (odds ratio [OR], 0.75;
95% confidence interval [CI], 0.66-0.85) and in the first quarter of 1988
(OR, 0.76; 95% CI, 0.67-0.86). In subsequent quarters, the rate returned to
the baseline. In multivariate analyses, the decline was significant among
white but not nonwhite women. It was most prominent among women aged 50 to
79 years in the central and southern regions of the country, and most sustained
among women living in areas with lower levels of income and education.
Conclusions.— Celebrity role models can influence decisions about medical care. The
influence appears strongest among persons who demographically resemble the
celebrity, and those of lower income and educational status.
THERE HAS been a long-standing belief in the influence of public figures
on popular behavior. The best-known examples have to do with fashion, but
celebrities are also thought to be capable of influencing health care behaviors.
This perception has led to the use of celebrity endorsements for promoting
health behaviors such as safe sex and avoiding illegal drugs. Remarkably,
though, there is little information beyond the anecdotal documenting an effect
of celebrity role models on medical care or health behaviors.1
During an analysis of the effect of legislative initiatives on the use of
breast-conserving surgery (BCS) vs modified radical mastectomy in women with
early-stage breast cancer,2 we noted a sharp
drop in BCS in late 1987 that was not associated with any publications in
the medical literature or lay press that would call into question the effectiveness
of BCS. It was, however, closely associated with the treatment of breast cancer
in Nancy Reagan, wife of then President Ronald Reagan.
During the routine screening mammography accompanying Mrs Reagan's annual
medical evaluation in early October 1987, a suspicious lesion was detected.3 On October 17 she underwent an open biopsy of the
lesion followed by a modified radical mastectomy during the same operation.
This treatment choice sparked a controversy that was prominently reported
by the lay press. Articles critical of Mrs Reagan's choice of surgery appeared
in major news media under headlines such as "Mastectomy Seen as Extreme Treatment"4 and "Was This Operation Necessary?"5
One authority on breast cancer was quoted in the New York
Times as stating that Mrs Reagan's decision "set us back 10 years."4Time magazine quoted a prominent
breast cancer specialist as stating "it's my opinion that she was probably
overtreated."5 This negative press in turn
stimulated a number of articles defending the right of Mrs Reagan and all
women with breast cancer to make whatever choice of therapy seems best for
In this article, we describe the decline in use of BCS temporally associated
with Mrs Reagan's breast cancer treatment. We believe this to be the first
detailed examination of the influence of a celebrity role model on national
medical practice. We use 2 independent sources of data: the Surveillance,
Epidemiology, and End Results (SEER) tumor registry, which provides population-based
data on diagnosis and treatment of cancer for approximately 10% of population
of the country, and Medicare Part A charge data, which provide information
on hospital treatment for more than 90% of women aged 65 years and older nationally.
The study was approved by the Human Research Committee of the Medical
College of Wisconsin.
SEER includes a population-based cohort of cancer patients from 9 geographic
sites: the states of Connecticut, Hawaii, Iowa, New Mexico, and Utah, and
the metropolitan areas of Atlanta, Ga, Detroit, Mich, San Francisco-Oakland,
Calif, and Seattle-Puget Sound, Wash.7 Information
collected by SEER includes demographic information, extent of disease, and
initial course of treatment for each patient.
The federal Area Resource File8 was used
to determine the degree of urbanicity, income, and educational status of the
county of residence of the patient. Assigning proxy census indicators of socioeconomic
status has been shown to be a valid approach to analyzing socioeconomic status
in database studies.9
The 1987-1988 national Medicare Part A claims files maintained by the
Health Care Financing Administration were used for some analyses. This database
contains information on virtually all US residents aged 65 years and older,
except for those treated at Veterans Affairs hospitals.10
A record is kept for each hospital discharge and includes the patient's age,
sex, dates of admission and discharge, up to 5 diagnoses, and up to 3 procedures.
Newspaper and magazine articles about breast cancer treatment were identified
by searching 1987 and 1988 in the National Newspaper Index and the Magazine Database (Information Access
Co, Foster City, Calif). Articles relating to surgery for breast cancer were
obtained and analyzed for content relating to choice of BCS vs mastectomy.
While the broadcast media also covered Mrs Reagan's operation, we believe
that the print media databases provided a representative sample of the overall
content of the lay press coverage of these issues.
For each state in the United States, we obtained the percentage of the
popular vote won by President Reagan in 1984.11
We also determined the American Conservative Union's scores12
for that state's members of the House of Representatives, and averaged the
scores for all members in each state for 1982, 1984, 1986, and 1988.
All 121796 women who developed breast cancer between 1983 and 1990 were
selected. Patients were excluded for the following reasons: if there was no
microscopic confirmation of disease (1947); if the stage of disease was in
situ (11917), distant (6879), or unknown (4800); if the breast cancer was
bilateral (149); if the breast cancer was not the patient's first cancer (16764);
if the patient's age was younger than 30 years (813); or if the patient did
not receive either a mastectomy or BCS (7338). Some patients had more than
1 reason for exclusion. An additional 24 patients were excluded because of
a lack of information regarding the month of diagnosis, and 35 were excluded
because of an invalid county code. A cohort of 83250 women aged 30 years or
older who underwent BCS or mastectomy treatment for local or regional breast
cancer diagnosed in 1983 through 1990 remained. For the more detailed analyses,
only the 22574 women diagnosed in 1987 or 1988 were included.
Age at diagnosis was categorized as 30 through 49, 50 through 64, 65
through 79, and 80 years or older. Race was categorized as white or nonwhite.
Women were considered to have received BCS if they underwent segmental mastectomy,
lumpectomy, quadrantectomy, tylectomy, wedge resection, excisional biopsy,
or partial mastectomy.
Several county-level factors were obtained from the Area Resource File.
The size of the metropolitan statistical area of the county of residence of
the patient was categorized as less than 250000 persons, 250000 to 1 million
persons, or more than 1 million persons. The average per capita income and
the percentage of persons aged 25 years or older who had completed 4 or more
years of college were computed for the county of residence of each patient
and were divided into quartiles.
Cohorts of women undergoing inpatient BCS or mastectomy for invasive
local or regional breast cancer during 1987 and 1988 were constructed.13,14 Briefly, we selected from the annual
inpatient hospital claims files for 1987 and 1988 records containing a breast
cancer International Classification of Diseases, Ninth Revision,
Clinical Modification15 diagnostic (174.XX)
or procedural (85.XX) code. From the 114896 such claims in 1987, and 116398
claims in 1988, we eliminated claims (4846 in 1987 and 5439 in 1988) that
were for men, for skilled nursing admissions, or duplicates. We determined
the earliest hospitalization with the most invasive operation for each patient
and excluded women who had not undergone either an excisional breast procedure
or axillary lymph node dissection (25216 in 1987 and 24311 in 1988). Also
excluded were women who underwent bilateral mastectomy, had distant metastases,
had carcinoma in situ (19154 in 1987 and 20034 in 1988), or had no recorded
diagnosis of primary breast cancer (15189 in 1987 and 15151 in 1988). Some
women had more than 1 reason for exclusion. We then limited this cohort to
white women aged 65 to 79 years, providing a cohort of 39184 in 1987 and 40873
We coded women as having undergone BCS if they had undergone local excision,
quadrantectomy, partial mastectomy, or axillary lymph node dissection without
mastectomy.14,16 The state in
which each patient was treated was determined from the hospital's identification
The percentage use of BCS was computed for each quarter over the period
of observation, based on date (month and year) of diagnosis. The date of treatment
is not available in the SEER database. For patients diagnosed in 1987 or 1988,
a logistic regression model was used to estimate the odds ratio for receipt
of BCS.17 atient ariables at the individual
level (age, race, stage of disease) and the county level (size of metropolitan
statistical area, per capita income, percentage of persons aged ≥25 years
with ≥4 years of college education) were used as covariables to adjust
for potential differences in these factors. Time was modeled as a series of
binary variables18 referenced to the third
quarter of 1987.
For each patient, the date of admission to the hospital was considered
the treatment date for analytic purposes. For each region of the country,
the percentage of reduction in use of BCS during the 6 months following Mrs
Reagan's mastectomy was determined by dividing the proportion of patients
receiving BCS in the fourth quarter of 1987 and first quarter of 1988 by the
proportion of patients receiving BCS during the third quarter of 1987. This
number was subtracted from 1, and multiplied by 100% to provide the percentage
of reduction in use of BCS.
Figure 1 shows the percentage
of use of BCS from 1983 through 1990, by quarter of diagnosis, for women aged
30 years or older with local or regional breast cancer living in 1 of the
9 SEER areas. As previously described,2 there
is a gradual increase in the use of conservative surgery from 1983 through
1985. This is followed by relative stability in the use of this procedure,
with the exception of a rather sharp decrease in use the last quarter of 1987
and the first quarter of 1988. Based on the multivariate logistic regression,
the odds of receipt of BCS were about 25% lower during these 2 quarters than
during the third quarter of 1987 (Table
1). This 25% decrease, if projected to the entire country, suggests
that approximately 3400 fewer women underwent BCS in the 6 months following
Mrs Reagan's surgery than would have been expected based on previous rates
of utilization. The searches of the lay print media databases yielded numerous
articles discussing the diagnosis of Mrs Reagan's cancer and her treatment
choice. No articles were found seriously questioning the validity of BCS treatment.
Table 2 presents the change
in use of BCS after Mrs Reagan's mastectomy among various demographic groups.
The decrease in use of BCS during the fourth quarter of 1987 and the first
quarter of 1988 was apparent among women aged 50 through 64 and 65 through
79 years, but not among older or younger women. The decrease in use of BCS
was apparent among white women, but not among African American women. The
decrease was present among women treated in both urban and rural areas. The
effect was observed during the fourth quarter of 1987 among women of all income
and educational strata, but recovery was more rapid among more affluent and
The fact that the drop in BCS utilization was greatest in women aged
65 through 79 years allowed us to use Medicare data to examine the effect
nationally. Figure 2 presents week
by week data for 1987 and 1988 on the percentage of white women aged 65 through
79 years with local or regional breast cancer who underwent BCS. Mrs Reagan
had her operation during the 41st week of the year (October 15-21), and a
sharp drop in the percentage of use of BCS is apparent beginning with week
42. The percentage of women aged 65 through 79 years with local or regional
breast cancer receiving BCS is higher in the SEER cohort (≈25%, Figure 1) than in the Medicare cohort (≈13%, Figure 2). This is because the Medicare cohort
does not include women treated in outpatient surgery centers, and because
the SEER database overrepresents urban patients, who have greater use of BCS.19
As assessed by analysis of Medicare data, the reduction in use of BCS
during the last quarter of 1987 and the first quarter of 1988 (compared with
the third quarter of 1987 nationally) was 24.5%. As shown in Table 3, there was considerable regional variation; the reduction
in use of BCS following Mrs Reagan's surgery was greatest in the South Atlantic
and East South Central regions, while it was less in the Middle Atlantic region.
In further analyses, the drop in BCS use in each state following Mrs Reagan's
surgery was not related to the percentage vote received by President Reagan
in the 1984 election (r=0.09) or to the average rating
of all the members of Congress in each state on a conservative-liberal scale
In this article, we have demonstrated a 25% reduction in the use of
BCS as opposed to mastectomy among women with local or regional breast cancer
diagnosed during the last quarter of 1987 or first quarter of 1988. The observed
decrease in use of BCS was not associated temporally with any medical literature
or lay print media reports questioning the technique, but was associated with
a great deal of publicity in the lay press regarding the choice of the president's
wife to undergo mastectomy for her own breast cancer.
As might be expected, the effect of Mrs Reagan's surgery was greatest
among women who were demographically similar to her, white women aged 50 through
79 years, as opposed to older or younger women or nonwhite women. The effect
was more prominent in the Central and Southern regions of the country, and
in counties with lower levels of education and income. There was no correlation
between the decrease in use of BCS in each state and popular vote for President
Reagan or a rating of the state's members of Congress on a liberal-conservative
It might be argued that the drop in use of BCS was not related to the
publicity regarding Mrs Reagan. However, the tight temporal relationship shown
in the week by week Medicare data (Figure
2) is impressive. In addition, the fact that the decrease was most
prominent in women demographically similar to Mrs Reagan lends credibility
to this explanation, as does the lack of any other widely publicized information
during this period regarding treatment of breast cancer. Although some potential
for misclassification of BCS exists in these databases, we do not believe
any misclassification would be selective in such a way as to account for the
observed temporal decline in use of BCS.
To date, the most completely studied "celebrity illness" and its effect
on health behaviors was the announcement in November 1991 by Earvin "Magic"
Johnson that he was positive for the human immunodeficiency virus. By chance,
there were several ongoing surveys of sexual knowledge, attitudes, and practices
that overlapped that announcement.20- 23
There were changes in attitudes20 and self-reported
health behaviors23 in some populations, but
not in others.21,22
Several studies have examined temporal relationships between the diagnosis
of breast cancer in a celebrity and use of mammography.24- 26
Lane et al24 reported that a small percentage
(1%-2%) of women studied in Long Island, NY, attributed their decision to
have a first mammogram in late 1987 to the publicity surrounding Mrs Reagan's
breast cancer diagnosis, and others25,26
have reported increases in mammography25 and
other screening tests26 following the diagnoses
of breast cancer in the wives of the president and vice president in 1977.
This increase in screenings was accompanied by an increase in breast cancer
diagnoses.27,28 There may have
been a similar situation following Ronald Reagan's diagnosis of colon cancer,
with an increase in use of endoscopic screening tests.29
Was the effect on choice of surgery due more to physician effect or
patient effect? During the 1980s there was concern that BCS was being underutilized,
and consensus statements30 and legislative
initiatives2,31 were produced
in an attempt to remedy this perceived underutilization. On the other hand,
the 2 treatments, BCS followed by radiation and modified radical mastectomy,
are equivalent in terms of survival and patient satisfaction.32
The major factor influencing successful adaptation after breast cancer treatment
is participation by the patient in the treatment choice.32
We assume that the influence of Mrs Reagan's choice was acting mostly via
the patient rather than on the surgeon, because women most demographically
similar to Mrs Reagan were most affected.
In conclusion, medical care can be influenced substantially by the behavior
of celebrity role models. The influence is strongest among persons who demographically
resemble the celebrity, and those of lower income and educational status.
One can sympathize with public figures facing difficult personal medical decisions,
because they have to deal with the reality that their decisions may very well
influence the behavior of thousands of others. However, this study provides
support for the concept of targeted celebrity role models as a strategy to
influence public health behaviors.