Kahn HS, Tatham LM, Patel AV, Thun MJ, Heath, Jr CW. Increased Cancer Mortality Following a History of Nonmelanoma Skin Cancer. JAMA. 1998;280(10):910-912. doi:10.1001/jama.280.10.910
From the Department of Family and Preventive Medicine, Emory University School of Medicine (Dr Kahn), and the Department of Epidemiology and Surveillance Research, American Cancer Society, Atlanta, Ga. Dr Tatham is now with the Agency for Toxic Substances and Disease Registry, Atlanta.
Context.— Cancer registries have reported an increased incidence of melanoma and
certain noncutaneous cancers following nonmelanoma skin cancer (NMSC). Whether
these findings were attributable to intensified surveillance, shared risk
factors, or increased cancer susceptibility remains unclear.
Objective.— To determine whether a history of NMSC predicts cancer mortality.
Design.— Prospective cohort with 12-year mortality follow-up adjusted for multiple
Setting.— Cancer Prevention Study II, United States and Puerto Rico.
Participants.— Nearly 1.1 million adult volunteers who completed a baseline questionnaire
Main Outcome Measure.— Deaths due to all cancers and common cancers.
Results.— After adjusting for age, race, education, smoking, obesity, alcohol
use, and other conventional risk factors, a baseline history of NMSC was associated
with increased total cancer mortality (men's relative risk [RR], 1.30; 95%
confidence interval [CI], 1.23-1.36; women's RR, 1.26; 95% CI, 1.17-1.35).
Exclusion of deaths due to melanoma reduced these RRs only slightly. Mortality
was increased for the following cancers: melanoma (RR, 3.36 in men, 3.52 in
women); pharynx (RR, 2.77 in men, 2.81 in women); lung (RR, 1.37 in men, 1.46
in women); non-Hodgkin lymphoma (RR, 1.32 in men, 1.50 in women); in men only,
salivary glands (RR, 2.96), prostate (RR, 1.28), testis (RR, 12.7), urinary
bladder (RR, 1.41), and leukemia (RR, 1.37); and in women only, breast (RR,
1.34). All-cause mortality was slightly increased (adjusted men's RR, 1.03
[95% CI, 1.00-1.06]; women's RR, 1.04 [95% CI, 1.00-1.09]).
Conclusions.— Persons with a history of NMSC are at increased risk of cancer mortality.
Although the biological mechanisms are unknown, a history of NMSC should increase
the clinician's alertness for certain noncutaneous cancers as well as melanoma.
EUROPEAN CANCER registries have demonstrated an increased incidence
of melanoma and certain noncutaneous cancers following an initial diagnosis
of squamous cell skin cancer,1,2
basal cell skin cancer,3 or nonmelanoma and
non–basal cell skin cancer.4 Among the
noncutaneous cancers repeatedly found to be at increased incidence after nonmelanoma
skin cancer (NMSC) were those of the salivary glands, lung, and prostate as
well as non-Hodgkin lymphoma and leukemia. These apparent associations between
NMSC and subsequent cancer incidence may have been due to intensified medical
surveillance in patients with a history of cancer, confounding by lifestyle
risk factors or external carcinogenic exposures, a shared internal pathway
of cancer induction, adverse effects of agents used in the treatment of NMSC,
or chance effects.5 We examined the relationship
between a history of NMSC and subsequent cancer mortality in a large cohort
of Americans for whom baseline data were available on potential confounding
Study participants were drawn from the Cancer Prevention Study II (CPS-II),
a prospective mortality study of 508353 men and 676306 women who completed
a 4-page questionnaire in 1982. Baseline data from CPS-II included personal
identifiers, demographic characteristics, various behavioral and dietary exposures,
and personal and family histories of cancer and other diseases. Participants
were identified and enrolled by American Cancer Society volunteers throughout
the United States and Puerto Rico.6 Enrollment
was restricted to persons aged 30 years and older.
The vital status of study participants was determined through December
31, 1994, using 2 consecutive approaches. In the initial phase, volunteers
made personal inquiries in 1984, 1986, and 1988 to determine whether the enrollees
were alive or dead and to record the date of all deaths. Subsequent to 1988,
we used automated linkage with the National Death Index to extend follow-up
through 1994 and to identify deaths among the few participants lost to follow-up
between 1982 and 1988. A validation study that compared volunteer follow-up
with National Death Index linkage found that the volunteers identified virtually
all participant deaths and linkage with the National Death Index identified
93% of participant deaths.7 At completion of
mortality follow-up in December 1994, 21.1% of the men and 12.8% of the women
had died, 78.7% of the men and 86.9% of the women were presumed to be alive,
and 0.2% of the men and 0.3% of the women had follow-up truncated on September
1, 1988, because of insufficient data for National Death Index linkage. Death
certificates were obtained for 98.3% of all participants known to have died.
A history of NMSC was established at baseline by a positive response
to the question, "Have you ever had cancer?" and further specification that
the cancer was skin cancer but not reported to be melanoma. We excluded from
our analytic cohort participants with a baseline history of melanoma or any
noncutaneous cancer (25242 men and 57107 women), those whose death after enrollment
was attributed to NMSC (98 men and 34 women), and those who claimed a baseline
history of NMSC but did not provide the year of treatment (2719 men and 2553
women). Participants remaining in the analytic cohort are described in Table 1. The 19102 men (4.0%) and 15960
women (2.6%) with a history of NMSC were slightly older, more likely to be
white, more educated, and more likely to have stopped smoking than were participants
without a history of NMSC.
We used Cox proportional hazards modeling8
to examine the association between a baseline history of NMSC and mortality
due to various causes. Cox models were stratified on age at enrollment and
controlled for race (white, including 1.5% undetermined, vs all others), education,
smoking status (Table 1), body
mass index (calculated as weight in kilograms divided by the square of height
in meters [kg/m2]), alcohol use, exercise, vegetable and fat intake,
aspirin use, marital status, prevalent diabetes, and (for women) menopausal
status, parity, and use of oral contraceptives and estrogen replacement therapy.
As appropriate to site-specific cancer outcomes, certain multivariate models
were also adjusted for hysterectomy status or family history of colorectal,
prostate, breast, or ovarian cancer.
After 12 years of follow-up, 26622 men and 21084 women in our analytic
cohort had died of cancer. After adjusting for only age and race, a history
of NMSC was associated with all-site (including melanoma) cancer mortality
for men (relative risk [RR], 1.23; 95% confidence interval [CI], 1.17-1.30)
and for women (RR, 1.23; 95% CI, 1.15-1.32). Additional multivariate adjustment
(Table 2) did not attenuate the
association between a history of NMSC and all-site cancer mortality (men's
RR, 1.30; women's RR, 1.26). Excluding from the analyses 427 men and 221 women
who died of melanoma produced minimal change in the association between NMSC
and death from cancer (fully adjusted men's RR, 1.27; women's RR, 1.24; Table 2).
The increased RR for all-site cancer mortality paralleled the increased
RRs that we identified for death from the following cancers: melanoma, pharynx,
lung, non-Hodgkin lymphoma; in men only, salivary glands, prostate, testis,
bladder, and leukemia; and in women only, breast (Table 2). No sites demonstrated a significant decrease in cancer
mortality associated with a history of NMSC.
The small increase in all-cause death rates among participants with
a history of NMSC (fully adjusted men's RR, 1.03; 95% CI, 1.00-1.06; 94754
deaths and women's RR, 1.04; 95% CI, 1.00-1.09; 70401 deaths) was entirely
attributable to cancer mortality. The fully adjusted RRs related to all circulatory
diseases were 0.91 (95% CI, 0.87-0.95; 45210 deaths) for men and 0.96 (95%
CI, 0.90-1.02; 31235 deaths) for women. The fully adjusted RRs related to
all other causes were 0.98 (95% CI, 0.93-1.04; 22922 deaths) for men and 0.96
(95% CI, 0.89-1.05; 18082 deaths) for women.
The proportional hazards assumption was tested and found not to be violated
for models predicting all-cancer mortality, mortality due to the major cancers,
and all-cause mortality.
This large prospective study found death rates from all noncutaneous
cancers to be 20% to 30% higher among participants who reported a history
of NMSC than among participants who did not. Our finding is similar in direction
and magnitude to the results of 3 previous studies based on incidence data
from European cancer registries1,3,4
and the results of a fourth European incidence study when its analysis was
restricted to persons whose initial squamous cell skin cancer was diagnosed
at younger than 65 years.2 Our findings (Table 2) also resemble those from the European
cancer registries with respect to the specific cancers associated with NMSC.
For each of the cancers for which higher death rates were associated with
NMSC in our study, at least 1 of the European studies reported increased cancer
incidence. Our prospective study, like the European incidence studies, found
no increase in cancers of the esophagus, stomach, pancreas, ovary, or colorectum.
Intensified medical surveillance of persons with a history of NMSC is
unlikely to explain the increased cancer mortality found in our study because
we assessed deaths rather than incident cancers. Our participants were equally
subject to ascertainment of death from cancer irrespective of their baseline
histories. However, ascertainment of subsequent incident cancers may not have
been equal in studies comparing persons with cancer history with those in
the general population.
The observed association between NMSC and fatal cancers is unlikely
to be explained by confounding from educational status, smoking, obesity,
alcohol use, exercise, or any of the other measured risk factors for which
we controlled. Although the European incidence studies were unable to control
their analyses for these variables, we found no evidence that the associations
with NMSC changed after multivariate adjustments. This reduces concern about
residual confounding from these measured risk factors.
It is possible that the association between NMSC and subsequent cancers
may reflect one or more unmeasured shared exposures or susceptibilities. For
example, a high cumulative exposure to UV radiation leading to NMSC9,10 might also be associated with systemic
or with the acquisition of somatic mutations (eg, p53 tumor suppressor gene12). Several studies have considered (but not resolved)
the role played by solar UV exposure in the etiology of non-Hodgkin lymphoma.13- 18
Diet in early life,19,20 ionizing
radiation (including NMSC therapy),10,21
various chemicals (including topical anesthetics or treatments),10
or viruses22- 26
might also contribute to the small increase in shared risk. Alternatively,
an underlying susceptibility factor (inherited or acquired) such as DNA repair
deficiency27 could play a role in the observed
The clinical and public health implications of these epidemiological
findings are not yet clear. However, a history of NMSC should increase the
clinician's alertness for selected neoplasms.28
Since submission of our manuscript we have encountered 2 additional
incidence studies reporting malignancies following basal cell skin cancer.
One study29 found an increased incidence of
all-site, noncutaneous cancers and the other30