Antiperspirant Use and the

Risk of Breast Cancer

Dana K. Mirick, Scott Davis,
David B. Thomas

The rumor that antiperspirant use

causes breast cancer continues to circu-
late the Internet. Although unfounded,

Downloaded from https://academic.oup.com/jnci/article/94/20/1578/1802711 by UNIVERSIDAD DE SONORA user on 08 April 2024

there have been no published epidemio-
logic studies to support or refute this
claim. This population-based case–
control study investigated a possible
relationship between use of products
applied for underarm perspiration
and the risk for breast cancer in
women aged 20–74 years. Case pa-
tients (n = 813) were diagnosed be-
tween November 1992 and March
1995; control subjects (n = 793) were
identified by random digit dialing and
were frequency-matched by 5-year
age groups. Product use information
was obtained during an in-person
interview. Odds ratios (ORs) and
95% confidence intervals were esti-
mated by the use of conditional logis-
tic regression. P values were deter-
mined with the Wald ␹ 2 test. All
statistical tests were two-sided. The
risk for breast cancer did not increase
with any of the following activities:
1) antiperspirant (OR = 0.9; P = .23)
or deodorant (OR = 1.2; P = .19) use;
2) product use among subjects who
shaved with a blade razor; or 3) ap-
plication of products within 1 hour of
shaving (for antiperspirant, OR = 0.9
and P = .40; for deodorant, OR = 1.2
and P = .16). These findings do not
support the hypothesis that antiper-
spirant use increases the risk for

Affiliations of authors: D. K. Mirick, Program

in Epidemiology, Division of Public Health Sci-
ences, Fred Hutchinson Cancer Research Center,
Seattle, WA; S. Davis, D. B. Thomas, Program in
Epidemiology, Division of Public Health Sci-
ences, Fred Hutchinson Cancer Research Center,
and Department of Epidemiology, School of Pub-
lic Health and Community Medicine, University
of Washington, Seattle.
Correspondence to: Dana K. Mirick, M.S., Fred
Hutchinson Cancer Research Center, 1100 Fair-
view Ave. North, MP-425, P.O. Box 19024,
Seattle, WA 98109-1024 (e-mail: dmirick@
fhcrc.org).
See “Note” following “References.”
© Oxford University Press

1578 BRIEF COMMUNICATIONS Journal of the National Cancer Institute, Vol. 94, No. 20, October 16, 2002
 breast cancer. [J Natl Cancer Inst procedures for contacting potential par- 30% of control subjects), or to report
2002;94:1578–80] ticipants, obtaining informed consent, application of antiperspirant within 1
and collecting all data. All participants hour of shaving (36% of case patients
In the last decade, the public has been provided written informed consent be- versus 40% of control subjects). Table 1
faced with a seemingly endless number fore participation. displays the results from the regression
of reports that claim another agent in the Several measures of antiperspirant analyses of product use and the risk for
modern environment is associated with use were constructed to evaluate a pos- breast cancer. There was no evidence of
the risk of developing cancer. A news sible relationship to breast cancer, in- an association between the risk of breast
item appearing in the September 20, cluding ever regular antiperspirant use, cancer and any of the three measures of
2000 issue of the Journal (1) highlighted exclusive use of antiperspirant (versus antiperspirant use. Compared with sub-
the increasing prevalence of such re- deodorant or talc products), and applica- jects who did not use antiperspirant,
tion typically within 1 hour of shaving. there was no evidence that subjects who

Downloaded from https://academic.oup.com/jnci/article/94/20/1578/1802711 by UNIVERSIDAD DE SONORA user on 08 April 2024

ports and their widespread circulation
on the Internet. One rumor in particular, Because many subjects reported the use reported the use of a blade razor for un-
that antiperspirant use causes breast can- of deodorants, the three measures of derarm hair removal were at an in-
cer, received such intense interest that a product use listed above were also creased risk for breast cancer from anti-
number of cancer research and informa- evaluated for deodorants. Additional perspirant use, or that subjects who
tion organizations were forced to post analyses were conducted by stratifying reported applying antiperspirant within
statements denying the link between on the use of a blade (i.e., nonelectric) 1 hour of shaving with a blade razor
breast cancer and the use of antiperspi- razor to evaluate whether the relation- were at an increased risk for breast can-
rants (1). Although there are no pub- ship between antiperspirant use and the cer (data not shown).
lished reports in the scientific literature risk for breast cancer differed according Deodorant use was more prevalent
to suggest a biologic mechanism by to this method of underarm hair re- than antiperspirant use: among subjects
which the use of antiperspirants could moval. This analysis was prompted by a who used at least one method of under-
cause breast cancer and no epidemio- concern that small nicks in the skin from arm hair removal, 71% of case patients
logic study of this question has been the use of a blade might facilitate the and 65% of control subjects reported
reported, public concern has persisted. absorption of harmful substances in the having used deodorant regularly. Case
We conducted a population-based products. Odds ratios and 95% confi- patients were more likely to report the
case–control study of breast cancer in dence intervals were used to estimate use of deodorant exclusively compared
western Washington State, described relative risks with conditional logistic with control subjects (43% of case pa-
more fully elsewhere (2,3). Eligible case regression (4) (SAS procedure PHREG, tients versus 38% of control subjects)
patients were women aged 20–74 years SAS/STAT release 6.11; SAS Institute, and were more likely to report applying
who were first diagnosed with breast Inc., Cary, NC). All models were con- deodorant within 1 hour of shaving
cancer from November 1992 through ditional on 5-year age strata, with ad- (49% of case patients versus 43% of
March 1995. Control subjects were justment for a number of factors associ- control subjects). Similar to the results
women without breast cancer, identified ated with the risk for breast cancer for antiperspirant use, there was no evi-
by random-digit dialing from the same previously identified in this study (3). dence of an association between the risk
population as the case patients, who Statistical significance of the odds ratios for breast cancer and any of the three
were frequency-matched to the case pa- was evaluated with the Wald ␹2 test. All measures of deodorant use (Table 1).
tients by 5-year age groups. An in- statistical tests were two-sided. There was also no evidence that subjects
person interview was used to gather Approximately 78% (n ⳱ 813) of who reported using a blade razor were at
information on a large number of past the eligible case patients and 75% (n ⳱ an increased risk for breast cancer from
exposures of interest. During the devel- 793) of the eligible control subjects deodorant use, or that subjects who re-
opment of the questionnaire, we became agreed to participate and were inter- ported applying deodorant within 1 hour
aware of a concern that the use of prod- viewed for this study (2,3). A total of of shaving with a blade razor were at an
ucts for underarm perspiration might be 810 case patients and 793 control sub- increased risk (data not shown).
related to the risk for breast cancer. Spe- jects provided complete information on To our knowledge, this is the only
cifically, there was concern that such underarm hair removal. Nearly all case epidemiologic evidence pertaining to a
products might contain harmful sub- patients and control subjects had at possible association of the risk for breast
stances that could be absorbed via small some point in their lifetime regularly cancer with use of underarm antiperspi-
nicks or abrasions caused by hair re- used at least one method of underarm rants or deodorants, and our results pro-
moval. Consequently, we included a hair removal (94% of case patients and vide no indication that such a relation-
question to ascertain whether the re- 93% of control subjects), with the most ship exists. The strength of these results
spondent regularly shaved under her common method reported as shaving may be limited somewhat by the lack of
arms. For those who responded affirma- with a blade razor. Of the subjects who more detailed information on specific
tively, we asked whether she applied reported the use of at least one method patterns of product use and by the self-
anything for underarm perspiration and, of underarm hair removal, case patients reported nature of the data. However,
if so, which products she used, and were less likely than control subjects to the comprehensive assessment of both
whether any of the products were ap- have used antiperspirant regularly (50% antiperspirant and deodorant use helps
plied within 1 hour of shaving. The Fred of case patients versus 56% of control to address the possibility that subjects
Hutchinson Cancer Research Center subjects), to have used antiperspirant may have reported the use of an antiper-
Institutional Review Board approved all exclusively (24% of case patients versus spirant when, in fact, the product ap-

Journal of the National Cancer Institute, Vol. 94, No. 20, October 16, 2002 BRIEF COMMUNICATIONS 1579
 Table 1. Odds ratios (ORs) for breast cancer and self-reported regular use of antiperspirant or plied was actually a deodorant (or vice
deodorant, for subjects who report the use of at least one method of underarm hair removal versa) or the combination of an antiper-
No. of case No. of control
spirant and a deodorant. These findings
Product use patients (%)* subjects (%)* OR† (95% CI)‡ P value§ are based on data collected from a large
population-based study of rigorous de-
Antiperspirant
Exclusively sign, and as such, the absence of any
No 513 (75.1) 472 (69.5) 1.0 (referent) observed associations may help alleviate
Yes 170 (24.9) 207 (30.5) 0.8 (0.6 to 1.0) .12 the concern of many that use of under-
Ever regularly
No 331 (48.5) 296 (43.6) 1.0 (referent)
arm antiperspirants or deodorants could
Yes 352 (51.5) 383 (56.4) 0.9 (0.7 to 1.1) .23 alter their risk for breast cancer.
Regularly within 1 hour of shaving
No 428 (62.8) 399 (58.8) 1.0 (referent)
REFERENCES

Downloaded from https://academic.oup.com/jnci/article/94/20/1578/1802711 by UNIVERSIDAD DE SONORA user on 08 April 2024

Yes 253 (37.2) 279 (41.2) 0.9 (0.7 to 1.1) .40
Deodorant
Exclusively (1) Jones J. Can rumors cause cancer? J Natl Can-
No 400 (58.6) 421 (62.0) 1.0 (referent) cer Inst 2000;92:1469–71.
Yes 283 (41.4) 258 (38.0) 1.1 (0.9 to 1.4) .41 (2) Davis S, Mirick DK, Stevens RG. Night shift
Ever regularly
work, light at night, and risk of breast cancer.
No 206 (30.1) 238 (35.0) 1.0 (referent)
Yes 479 (69.9) 441 (65.0) 1.2 (0.9 to 1.5) .19 J Natl Cancer Inst 2001;93:1557–62.
Regularly within 1 hour of shaving (3) Davis S, Mirick DK, Stevens RG. Residential
No 355 (52.0) 381 (56.4) 1.0 (referent) magnetic fields and the risk of breast cancer.
Yes 327 (48.0) 294 (43.6) 1.2 (0.9 to 1.5) .16 Am J Epidemiol 2002;155:455–62.
(4) Breslow NE, Day NE. Statistical methods in
*Percentages were calculated from the total number of case patients/control subjects who reported the cancer research. Vol 1. The analysis of case-
use of at least one method of hair removal and had complete data on risk factors and the product-use control studies. International Agency for Re-
measure of interest. search on Cancer (IARC) Scientific Publ No.
†Logistic regression models are conditional on 5-year age strata; ORs were adjusted for parity, age at 32. 1980.
first pregnancy, mother/sister breast cancer, double oophorectomy when younger than age 40 years, oral
contraceptive use, ever upper gastrointestinal x-ray series, and ever smoker (all subjects); mother/sister NOTE
breast cancer when younger than age 45 years and alcohol intake (if premenopausal); and hormone
replacement therapy (if postmenopausal).
‡CI ⳱ confidence interval. Manuscript received March 13, 2002; revised
§P values were determined by using a two-sided Wald ␹2 test. July 15, 2002; accepted August 2, 2002.

1580 BRIEF COMMUNICATIONS Journal of the National Cancer Institute, Vol. 94, No. 20, October 16, 2002