Page 1
Sunscreen Use and the Risk for Melanoma: A Quantitative Review
Leslie K. Dennis, MS, PhD; Laura E. Beane Freeman, PhD; and Marta J. VanBeek, MD
Background: Originally developed to protect against sunburn,
sunscreen has been assumed to prevent skin cancer. However,
conflicting reports include claims that sunscreen increases risk for
melanoma.
Objective: To examine the strength and consistency of associ-
ations between melanoma and sunscreen use in the published
literature.
Data Sources: A comprehensive MEDLINE search of articles
published from 1966 to 2003 that reported information on sun-
screen use and melanoma in humans.
Study Selection: Analytic studies reporting data on sunscreen
use before diagnosis of melanoma.
Data Extraction: Two independent reviewers extracted data.
Inconsistencies were rereviewed until agreement was achieved.
When necessary, a third party resolved discrepancies.
Data Synthesis: Odds ratios were pooled across studies by
using standard meta-analytic techniques. Pooled odds ratios for
ever use among 18 heterogeneous studies did not support an
association between melanoma and sunscreen use. Variation
among odds ratios was explained by studies that did not adjust
for confounding effects of sun sensitivity. The lack of a dose–
response effect with frequency of use (never, sometimes, or al-
ways) or years of use provided further evidence of a null associ-
ation.
Conclusions: No association was seen between melanoma and
sunscreen use. Failure to control for confounding factors may
explain previous reports of positive associations linking melanoma
to sunscreen use. In addition, it may take decades to detect a
protective association between melanoma and use of the newer
formulations of sunscreens.
Ann Intern Med. 2003;139:966-978.
For author affiliations, see end of text.
www.annals.org
I
United States, and 7600 people (4700 men and 2900
women) will have died of melanoma (1). The incidence of
skin melanoma is increasing faster than that of any other
type of cancer in the United States (2). The age-adjusted
incidence of melanoma reported by the Surveillance Epi-
demiology and End Results (SEER) registries increased
from 7.4 per 100 000 persons in 1973 to 20.1 per 100 000
in 1999 among white persons (3).
Melanoma occurs predominantly in white persons,
and its incidence increases with age. The 2 major etiologic
risk factors for melanoma are sun sensitivity (tendency to
burn, inability to tan, light skin color, or tendency to de-
velop freckles) and exposure to ultraviolet radiation
through sun exposure. Additional risk factors include age,
sex, family history of melanoma, many nevi, and possible
exposure to artificial ultraviolet radiation (through tanning
beds). Various other factors have also been examined, in-
cluding oral contraceptive use, diet, smoking, and alcohol
use, but no consistent associations have been established.
Sunscreens are thought to protect skin from many of
the harmful effects of the sun. Consequently, some profes-
sionals suggest that limiting exposure to ultraviolet radia-
tion through use of sunscreen during childhood can reduce
the lifetime risk for nonmelanoma skin cancers by as much
as 78% (4). In contrast, several recent reports (5–8) have
suggested that sunscreen use may cause melanoma. Sun
sensitivity probably positively confounds the association
between sunscreen use and melanoma, yet few studies have
stratified participants by sun sensitivity or appropriately
controlled for this factor.
We sought to examine the strength and the consis-
tency of the observed associations between melanoma and
n 2003, an estimated 54 200 cases (29 900 in men and
24 300 in women) of melanoma will be diagnosed in the
sunscreen use. Where possible, we examined sunscreen use
stratified by sun sensitivity.
METHODS
Literature Search
To identify relevant studies for sunscreen use and mel-
anoma, we searched the MEDLINE database by using mel-
anoma (epidemiology or etiology) and sunscreening agents as
Medical Subject Headings and melanoma and sunscreen as
key words or text words. The search was repeated period-
ically to account for fluctuations in the results of searches.
We also searched Cancerlit but found no additional arti-
cles. We checked the references of identified articles, in-
cluding bibliographies of the review articles, for additional
relevant studies. We also reviewed 70 articles from the first
author’s files that appeared to be related to melanoma and
sunscreen use, sunburns, or sunlight.
Only cohort and case–control studies that measured
sunscreen use in relation to melanoma were included. Of
the eligible articles, 6 were not in the English language.
These were reviewed by 1 person who was multilingual and
trained in epidemiology; however, none reported on sun-
screen use.
In May 2003, the MEDLINE search for articles pub-
lished from 1966 through April 2003 was repeated by us-
ing melanoma (epidemiology or etiology) and sunscreening
agents as Medical Subject Headings and melanoma and sun-
screen as key words or text words (Figure 1). The latter
search found 138 articles, including 14 case–control stud-
ies; 32 review articles on melanoma; 42 editorials or com-
mentaries; 4 articles on melanoma trends; 41 articles on
habits, behaviors, or prevention; 4 prevention trials that
did not examine melanoma as an outcome; and 1 animal
Article
966 © 2003 American College of Physicians
Page 2
study. Of the 14 case–control studies found, 12 had rele-
vant data.
We then expanded our MEDLINE search to include
melanoma and sunburns as Medical Subject Headings, key
words, and text words, in an attempt to find additional
studies not listed on key or text words for sunscreen. Of
183 articles found through this expanded search, we re-
viewed 67 in detail, which yielded 6 additional articles on
sunscreen use. Three additional articles not found on
MEDLINE were obtained from authors’ files; 2 of these
articles were also discussed in other reviews. Among 3 re-
view articles, 17 studies were discussed, of which 13 were
found on MEDLINE, 2 were found in the authors’ files,
and 2 were not found elsewhere. Two additional studies, 1
of which was an abstract and 1 an unpublished study, were
found in the references of reviews on sunscreen use (9, 10)
but not on MEDLINE. The authors were contacted but
provided no additional data. In general, authors were not
contacted because the response rate was low in a previous
meta-analysis.
Of the 23 relevant analytic articles on sunscreen use
and melanoma, 3 were duplicate publications, leaving 20
independent studies. One of the 20 studies reported only
on tanning lotion.
Selection Criteria
Cohort, case–control, and cross-sectional studies that
included adults or children were considered for this meta-
analysis. However, we found no cohort or cross-sectional
studies of sunscreen use and development of melanoma.
One author and an assistant screened the titles and ab-
stracts obtained from the literature review to exclude case
reports; commentaries; and articles on melanoma trends,
habits and behaviors, and other biological aspects of mela-
noma. If relevance was unclear, the study was reviewed in
detail by 2 independent reviewers. Studies that reported
use of what appeared to be tanning oils or tanning enhanc-
ers rather than sunscreen were independently reviewed and
were later excluded from pooled analyses. We included
studies that reported odds ratios and those that reported
the percentage of cases and controls who used sunscreen.
Because the diagnosis of melanoma is based on histologic
examination, all studies were assumed to have included
histologic confirmation, even if this was not explicitly stated.
Data Extraction
For each study and level of sunscreen use, the natural
log of the odds ratio and its variance were required. Where
available, the variances were calculated on the basis of the
reported CIs (11). Otherwise, we calculated the variances
of the natural log of the odds ratio from reported data. For
studies that reported no association between melanoma
and sunscreen and did not report an odds ratio, we as-
sumed an estimated odds ratio of 1.0. We estimated cor-
responding variances on the basis of the number of partic-
ipants, assuming an average exposure rate that was
calculated from all other studies. When available, covari-
ate-adjusted odds ratios were used.
Reviewers were blinded to the authors, journal of pub-
lication, and introduction and discussion of each article.
Two independent reviewers abstracted data from every ar-
ticle, and the 2 sets of results were compared for concor-
dance and rereviewed if necessary. Inconsistencies were re-
reviewed until agreement was achieved. We sought third-
party resolution of disagreements when necessary. We
classified studies as hospital-based if they used hospital,
clinic, or unidentified control groups. We considered stud-
ies that included neighborhood controls, community con-
trols, and general population–based controls to be popula-
tion based. The most adjusted odds ratios were abstracted,
and we assumed that the authors had properly examined
and adjusted for confounders. Because of heterogeneity, we
examined subgroups of studies that reported adjustment
for sun sensitivity, sunburns, and other sun exposure.
Study Quality
On the basis of our previous study of melanoma and
meta-analyses, we developed a quality-assessment scoring
system for articles reporting on melanoma and sunscreen
use. Two reviewers applied the system to each of the 20
studies. Because selection of the control group is pertinent,
we scored studies as follows: hospital, cancer, or outpatient
dermatology controls, 0 points; hospital visitors, or other
unclear group, 1 point; and population controls, 2 points.
Studies were given credit for standardizing or pretesting
the questionnaire in a sample similar to their study sample
(2 points) or partial credit for using the same questions
from a structured questionnaire (1 point). Interviewer-
administered studies were assigned a score of 2 points if the
Context
Reports that sunscreen use increases risks for melanoma
have led to controversy.
Contribution
This meta-analysis of 18 case–control studies found no
good evidence for an increased risk for melanoma with
sunscreen use. Several studies did not account for patients’
sensitivity to sunlight, which could increase both sunscreen
use and melanoma. A few studies found protective rela-
tionships between sunscreen use and melanoma.
Implications
Previous reports of increased risk for melanoma with sun-
screen use were misleading.
Cautions
Studies that were reviewed did not evaluate newer sun-
screens with a sun protection factor greater than 15, pro-
tection against ultraviolet A radiation, or water resistance.
–The Editors
Article
Melanoma and Sunscreen Use
www.annals.org
16 December 2003 Annals of Internal Medicine Volume 139 • Number 12 967
Page 3
interviewer was blinded to the status of the patient and 1
point if he or she was not blinded; studies in which ques-
tionnaires were self-administered were assigned a score of 0
points. To help account for potential detection bias, we
gave credit to studies in which the control group had a skin
examination (1 point).
To address the potential confounding of sun sensitiv-
ity in the relationship between sunscreen and melanoma,
we assigned several scores according to the method by
which sun sensitivity was measured. A higher score was
given to studies that adjusted for skin color, skin type,
ability to tan, and tendency to burn (4 points), whereas a
lower score was assigned to those that adjusted only for
hair color, eye color, or freckling (2 points). We gave credit
to studies that adjusted for the potential confounders of
sunburn (2 points) and other sun exposure (2 points) and
to those that partially adjusted for these confounders (1
point). In addition, points were assigned to studies that
adjusted or matched for sex and age (2 points) or were
frequency-matched only (1 point). In examining sunscreen
use, credit was given to studies that reported more detail
beyond ever use: for example, studies that reported years of
use (2 points) and frequency of use (1 point). If any of the
above information was unclear or not stated, the study
received 0 points for that characteristic.
The maximum possible quality score was 19 points.
Figure 1. Literature search for articles on risk for melanoma and sunscreen use.
MeSH ? Medical Subject Heading.
Article
Melanoma and Sunscreen Use
968 16 December 2003 Annals of Internal Medicine Volume 139 • Number 12
www.annals.org
Page 4
After we examined the reliability of the 2 reviews, the re-
viewers discussed the differences and created a final overall
quality score for each study (Appendix Table, available at
www.annals.org). Reliability was examined by using the ?
statistic for categorical measures; total scores were com-
pared by using interclass correlations (12). The reliability
scores varied as follows. The ? value was correlated at 1.0
for whether the survey was completed by an interviewer or
self-administered, selection of control groups, and whether
the study adjusted for sunburns. All other categories had a
? value greater than 0.60 (Table 1). The quality scores,
which have not been validated, are presented to provide an
overall sense of each study’s quality but were not used in
any quantitative analyses.
Statistical Analysis
For dichotomous factors (ever use of sunscreen), we
used fixed-effects and random-effects models to obtain
pooled relative risk estimates (11, 13). Fixed-effects models
provide inferences about the included studies. Random-
effects models assume that the study samples were drawn
from a larger set of possible studies and provide inferences
about all studies in the hypothetical set of studies (11, 13).
Statistical tests of homogeneity (11, 14) were performed to
assess the consistency of associations. To quantify the ex-
tent of heterogeneity among the studies, we estimated the
between-study variance (15). This analysis included report-
ing of the H statistic, which is calculated as the heteroge-
neity statistic Q divided by its expected value in the ab-
sence of heterogeneity. An H statistic of 1.0 indicates
homogeneity. The I2statistic was also calculated to de-
scribe the proportion of total variation in estimates of the
odds ratio that is due to the heterogeneity between studies
(15). Unless otherwise stated, the reported odds ratios and
CIs are based on random-effects models because of heter-
ogeneity.
In addition, we stratified data by type of controls, ad-
justment for sun sensitivity, and sun sensitivity when avail-
able. For studies that did not report odds ratios and CIs,
we estimated ever use on the basis of case and control
distribution for frequency of sunscreen use. For 3 studies
that did not report case and control distribution, we esti-
mated the odds ratio and variance for ever use on the basis
of an average of the odds ratios and variances reported for
frequency of sunscreen use.
For multiple ordinal categories (frequency of sun-
screen use and years of sunscreen use), we used a fixed-
effects dose–response method to evaluate possible linear
relationships (14). This method combines different levels
of exposure in a linear regression of the natural log of the
odds ratio while adjusting for correlated measures within
studies. As suggested by Greenland and Longnecker (14),
for years of sunscreen use, we used the median number of
Table 1. Interrater Reliability and Final Distribution of the Quality Score, by Study Design, Analyses, and Reporting
Item Assessed
? Value for Interrater
Reliability (95% CI)
Points Assigned to Items Distribution of Final
Summary Scoring,
n (%)
Control group1.00 ? hospital, cancer, dermatology
1 ? hospital visitors
2 ? population based
9 (45)
1 (5)
10 (50)
Standardized questionnaire (same in cases
as controls)0.80 (0.58–1.02)*0 ? no, not clear, or not stated
1 ? same questions, structured questionnaire
2 ? standardized or piloted
0 ? self
1 ? interviewer
0 ? no, not applicable
1 ? yes
0 ? no
1 ? yes
0 ? no adjustment
2 ? adjustment for hair color, eye color, or freckling only
4 ? skin color, skin type, ability to tan, tendency to burn
0 ? no
2 ? yes
0 ? no
1 ? partial†
2 ? yes
0 ? no
1 ? frequency matched only
2 ? yes
0 ? only ever use
1 ? frequency of use
2 ? years of use, amount, duration
19 possible points
7 (35)
9 (45)
4 (20)
8 (40)
12 (60)
18 (90)
2 (10)
12 (60)
8 (40)
6 (30)
5 (25)
9 (45)
13 (65)
7 (35)
10 (50)
8 (40)
2 (10)
4 (20)
1 (5)
15 (75)
6 (30)
9 (45)
5 (25)
Interviewer administered 1.0
Interviewer blinded to case status1.0
Control had skin examination 0.68 (0.35–1.01)
Adjusted for sun sensitivity0.82 (0.61–1.02)*
Adjusted for sunburn1.0
Adjusted for other sun exposure0.74 (0.54–0.93)*
Adjusted for age, sex matched 0.77 (0.46–1.09)*
Measured more than ever use 0.75 (0.48–1.0)*
Total
* Weighted ? value.
† Partial adjustment was considered for such items as age at migration, sunny vacations, and sunbathing. Full adjustment for sun exposure was recorded for adjustment for
cumulative hours of sun exposure, recreational hours, occupational hours, or damage from ultraviolet radiation.
Article
Melanoma and Sunscreen Use
www.annals.org
16 December 2003 Annals of Internal Medicine Volume 139 • Number 12 969
Page 5
years of each category range in calculating an overall linear
? for each study. The iterations required for estimating the
linear ? for each study were performed in SAS IML (SAS
Software, Inc., Cary, North Carolina). Goodness-of-fit
tests for linear and quadratic models were run to determine
whether the linear models were appropriate. Linear models
were appropriate unless otherwise stated. SAS software
(SAS, Inc., Cary, North Carolina) was used for analyses
and plots.
Role of the Funding Source
The funding source had no role in the collection, anal-
ysis, or interpretation of the data or in the decision to
submit the manuscript for publication.
RESULTS
Table 2 describes the 20 studies from 23 published
articles included in the meta-analysis. The earliest pub-
lished case–control study (16) was excluded because it ex-
amined use of sun lotion or oil rather than sunscreen. One
study (17) included hair, sweat, dirt, dead skin, and sun-
screen use as measures of protection against ultraviolet ra-
diation “always when in the sun” (Table 2). Therefore, we
excluded it from the pooled odds ratio because it was not
comparable to other studies. The remaining 18 studies in-
cluded no cohort studies, 9 population-based case–control
studies (5, 6, 8, 18–25) (including 1 population-based
study of adolescents), 7 non–population-based studies (7,
26–32), and 2 case–control studies (9, 33) in which the
control group was not clearly identified. Two of the 18
studies reported odds ratios for both sunscreen and sun
lotions or oils (Table 2); however, only odds ratios for
sunscreen use were pooled in the analyses.
Table 2 shows the period of data collection, location
of the study sample, age range of participants, type of con-
trol group, other exposures measured in the studies, expo-
sures for which the analyses adjusted, and overall measure
of sunscreen use. Because the studies summarized here
were case–control studies, recall bias of sunscreen use is
open to misclassification that could differ between cases
and controls.
Five studies used similar inclusion and exclusion crite-
ria for cases and controls (8, 16, 20, 21, 25, 27), 1 did not
have similar criteria (7), and the remaining 13 were un-
clear. Six of the studies stated that the controls had a skin
examination (Table 2); therefore, they were unlikely to
have undiagnosed melanomas. Five studies used mailed
questionnaires (Table 2). The remaining 15 studies may
have more complete data because of interviewer probing,
but only 2 studies (20, 22) stated that the interviewers were
blinded to case or control status, thus reducing differential
probing by case status (22).
Using the information provided in the studies, we
scored studies on the basis of several quality assessment
items (Table 1). The 20 studies received scores ranging
from 1 to 18 of 19 possible points. The study that was
excluded from pooled analyses because it reported use of
“sun lotion or oil” (16) received a score of 2. A second
excluded study (17) received a score of 10, moderate to
good quality, but measurement of sun protection included
“hair, sweat, dirt, dead skin and sunscreen use as measures
of ultraviolet radiation protection.” Two studies (23, 32)
adjusted for sun sensitivity and sunburns for some measure
of sunscreen use. However, because we had to calculate the
odds ratio for ever use for these studies in the pooled anal-
ysis, the odds ratio was not adjusted and these studies
scored 6 fewer points than they might have otherwise.
Ever Use of Sunscreen
Five studies reported odds ratios and 95% CIs for
“ever” sunscreen use. For the remaining studies, we esti-
mated odds ratios on the basis of frequency of sunscreen
use (Table 2). The pooled odds ratio for the 18 studies on
ever use of sunscreen was 1.0 (95% CI, 0.8 to 1.2) (P value
for heterogeneity, ? 0.001). Figure 2 shows the odds ra-
tios and 95% CIs for the 18 studies, along with the pooled
estimate. To examine heterogeneity among studies, we
pooled data stratified by study design and by confounding
factors that were adjusted for in the original articles. No
difference in odds ratios was seen between type of control
group. We did not believe that adjustment for hair or eye
color without adjustment for skin color, tendency to burn,
or inability to tan represented adequate adjustment for sun
sensitivity. When the 5 studies that adjusted for hair color
only were pooled with the studies that adjusted for sun
sensitivity, no association was seen (odds ratio, 1.0). How-
ever, the odds ratio for melanoma and sunscreen use was
smaller when these 5 studies were excluded. When studies
were restricted to those that adjusted for sun sensitivity, the
pooled odds ratio decreased to 0.8 (CI, 0.6 to 1.0).
Four of the 18 studies stratified participants by skin
sensitivity. In 3 of the studies, sun sensitivity was defined as
Fitzpatrick skin type I or II. In 1 study, persons with fair or
pale skin color were considered sensitive to sun. When
stratified by sun sensitivity, the available data for sun-sen-
sitive persons were homogenous. Two of the 4 studies re-
ported null associations (odds ratios, 1.14 and 1.17), 1 a
marginally protective association (odds ratio, 0.77), and 1 a
significantly protective association (odds ratio, 0.66). Over-
all, among sun-sensitive persons, who are more likely to
use sunscreens and are at higher risk for melanoma, the
association between sunscreen use and melanoma was ho-
mogeneous and nonsignificantly protective (odds ratio, 0.9
[95% CI, 0.7 to 1.2]; P value for heterogeneity, 0.13). For
sun-resistant persons (those with medium or dark skin or
Fitzpatrick skin type III or IV), 1 study reported a null
association between sunscreen use and melanoma (odds
ratio, 1.17), 2 reported increased associations (odds ratio,
1.3 and 1.7), and 1 reported a significant protective effect
(odds ratio, 0.6), indicating heterogeneous associations (P
value for heterogeneity, 0.002).
Article
Melanoma and Sunscreen Use
970 16 December 2003 Annals of Internal Medicine Volume 139 • Number 12
www.annals.org
Page 6
Table 2. Characteristics of Study Methods Reported in Articles of Melanoma and Sunscreen Use, Ordered by First Year of
Data Collection*
Dates of Data
Collection
Setting Study, Year
(Reference)
Age Range, yControls
Studies included in the analysis
1974–1980 New York State Graham et al., 1985 (26)All Patients at the same hospital
1977–1979
2/1978–12/1983
New York State
Stockholm, Sweden
Herzfeld et al., 1993 (19)
Beitner et al., 1990 (18)
?18 Population based using RDD
Population based from registrar
1/1/1980–11/5/1981 Western Australia Holman et al., 1986 (20, 25)
?80 Population based from electoral rolls
1981–1986 San Francisco, CaliforniaHolly et al., 1995 (21, 24) 25–59 Population based using RDD
10/1/1982–3/31/1985Eastern DenmarkOsterlind et al., 1988 (22) 20–79 Population based from registrar
1/1/1987–6/30/1994 Queensland, AustraliaWhiteman et al., 1997 (8)
?15 Population based; randomly selected
from class lists
1987–1994 Queensland, Australia Youl et al., 2002 (23)15–19 Population based from electoral rolls
1987–1989 Connecticut Berwick (unpublished)Not stated Not stated
1/1/1988–6/30/1990 Southern Sweden Westerdahl et al., 1995 (5)15–75 Population based from registrar
1989–1993 East Andalusia, Spain Rodenas et al., 1996 (27)20–79Randomly selected from hospital
visitors
Not stated United States Fisher et al., 1996 (33) Not statedNot stated
1/1990–1/1994 Madrid, Spain Espinosa Arranz et al., 1999
(28)
Autier, 1995, 1998 (29, 30)
21–87 Population based; patients with
hospital emergencies
Not population based; unclear 1991–1992Belgium, France, Germany
?20
6/1992–2/1995ItalyNaldi et al., 2000 (31) Not stated Not population based; patients at the
same hospital
6/1993–7/1994 Styria, Austria Wolf et al., 1998 (7)15–89 Not population based; patients at
dermatology offices
1/1995–6/1997SwedenWesterdahl et al., 2000 (6) 16–80 Population based from registrar
1995–1998 Brazil Bakos et al., 2002 (32)20–84 Random patients at the same hospital
Studies of suntan lotion or other agents (excluded from analyses)
1/1/1974–5/1/1975 Oslo, Norway Klepp and Magnus, 1979 (16)
?20 Controls with cancer in the hospital
1974–1980 New York State Graham et al., 1985 (26)See information in row 1 above.
1977–1979 New York StateHerzfeld et al., 1993 (19) See information in row 2 above.
7/1/1979–6/30/1980Queensland, Australia Green et al., 1986 (17)14–86Population based from electoral rolls
* OR ? odds ratio; RDD ? random-digit dialing; SPF ? sun protection factor; UV ? ultraviolet.
† I ? skin type (Fitzpatrick); II ? tendency to burn; III ? ability to tan; IV ? skin color; V ? freckles; VI ? eye color; VII ? hair color.
‡ 1a ? sunburn in childhood; 1b ? sunburn in adolescence; 1c ? sunburn while an adult; 1d ? sunburn during lifetime; 1e ? sunburn ever; 2a ? cumulative sun exposure;
2b ? residential sun exposure; 3 ? recreational sun exposure; 4 ? occupational sun exposure; 5 ? sunny vacations; 6 ? sunbathing; 7 ? sunlamps/beds.
§ The maximum possible quality score is 19 points.
? Calculated from data and therefore not adjusted for covariates.
¶ Estimated from frequency ORs.
** Estimated from the frequency of sunscreen use in cases and controls mentioned in text.
†† Also measured SPF.
Article
Melanoma and Sunscreen Use
www.annals.org
16 December 2003 Annals of Internal Medicine Volume 139 • Number 12 971
Page 7
Table 2—Continued
ReferenceStandardized
Questionnaire?
Interviewer
Administered?
Skin Examination
among Controls?
Measurement of
Sun Sensitivity†
Measurement of UV and
Other Exposures‡
Quality
Score§
Cases/Controls,
n/n
26 NoYes No III, IV, V, VI, VII2a, 2b, 41 404/521
19
18
Yes
Partial
Yes
No
No
No
II, IV, V, VI, VII
I, VI, VII
3, 4, 6
1e, 4, 5, 6, 7
6
8
324/415 men
523/505
20, 25YesYes Yes II, III, IV, VI, VII1d, 2a, 2b, 3, 4, 6, 718 494/494 (511/511)
21, 24YesYesNo II, III, IV, V, VI, VII1a, 1b, 1c, 1d, 2a, 714 452/930 women
22 PartialYesNoV, VII 1a, 1b, 1c, 2b, 3, 4, 5, 6, 711474/926
8 No NoYes II, III, V, VI, VII1a, 1b, 1d, 2a, 5 1052/156
23PartialYes Yes I?, II, III, V, VI, VII 1d, 2a, 2b, 4, 68 201/205
Unpublished––– 2a6 Not stated
5 Partial NoNoV, VI, VII1a, 1b, 1c 11 400/640
27PartialYesYes IV, V, VI, VII 1a, 1b, 1c, 1d, 2a, 3, 4, 5, 713 105/138
33– No– ConstitutionalSolar3 70/109
28 Partial Yes YesII, III, IV, V, VI, VII2a, 2b, 3, 4, 5, 6, 79116/235
29, 30No Yes NoI, VII 1a, 1b, 1c, 2a, 3, 5, 710418/438 (412/445)
31 PartialYesYes II, IV, V, VI, VII 1d, 3, 4, 5, 715 542/538
7–– No I, IV, V, VI, VII1e, 1d, 2b, 3, 4, 5, 6, 7 10193/319
6 PartialNo No I, V, VI, VII 1c, 2b, 4, 5, 6, 712558/891
32 NoYesYes I, V, VI, VII1d6 103/206
16PartialNo No II, III, V, VI, VII3, 4, 5, 6, 72 78/131
26
19
17 Yes YesYes II, III, IV, V, VI, VII1a, 1b, 1c, 1d, 2a, 3, 4, 510 183/183
Article
Melanoma and Sunscreen Use
972 16 December 2003 Annals of Internal Medicine Volume 139 • Number 12
www.annals.org
Page 8
Dose–Response Analyses
Unlike investigators in previous reviews (34), we did
not believe that it was appropriate to pool ever use of
sunscreen with always-use or use for more than 10 years,
because these categories differ markedly in their description
of sunscreen use. Therefore, we conducted detailed dose–
response analyses. Frequency of sunscreen use is typically
reported in ordered categories of “never or rarely,” “some-
times,” or “almost always.” Although such data are not
optimal for accurately quantifying sunscreen use, they are
ordered categories that provide supplemental information
on sunscreen use beyond “ever use.”
Table 2—Continued
OR for Association between
Melanoma and Ever Use (95% CI)
OR for Association between Melanoma
and Frequency of Use (95% CI)
Adjustments
Reported
Men: 2.2 (1.2–4.1)
Women: 1.0 (0.7–1.6)?
0.8 (0.6–1.1)?
Sun protection agents: 1.6 (1.1–2.3)¶
Crude
Crude (restricted to men)
Age, sex, hair color Never: 1.0
Seldom: 1.4 (0.9–2.0)
Often/very often: 1.8 (1.2–2.7)
Never: 1.0
?1/2 of the time: 1.12 (0.74–1.71)
?1/2 of the time: 1.10 (0.76–1.58)
Never: 2.1 (1.5–3.0)
Sometimes: 1.5 (1.1–2.2)
Almost always: 1.0
Never: 1.0
Occasionally: 1.3 (1.0–1.6)
Always: 1.1 (0.8–1.5)
Never/rarely: 1.0
Sometimes: 1.5 (0.3–8.2)
Often: 1.5 (0.3–7.4)
Always: 2.2 (0.4–11.6)
Average lifetime use at home
Often/always: 1.0
Sometimes: 0.9 (0.5–1.7)
Rarely/never: 0.9 (0.5–1.7)
Never: 1.0
Sometimes: 1.1 (0.8–1.5)
Almost always: 1.3 (0.9–1.6)
Never: 1.0
Sometimes: 1.3 (0.9–1.9)
Almost always: 1.8 (1.1–2.8)
Never: 1.0
Sometimes: 0.6 (0.26–1.42)
Always: 0.2 (0.04–0.79)
1.1 (0.8–1.6)¶
Fair pigment, sensitivity to sunlight, ethnic origin, age at arrival to
Australia (sun exposure) (matched on age, sex, and residence; study
examined confounding due to sunburns)
Childhood sunburns, ability to tan, hair color, large nevi, complexion,
maternal ethnicity, history of skin cancer, and age (women only)
0.8 (0.6–1.0)?
0.8 (0.7–1.0)** Constitutional factors, sex, and age
Sunscreen use on holiday: 1.9 (0.7–5.2)?
Sex, school, grade, tanning ability, freckling, number of nevi (matched on
age, sex, and residence)
1.0 (0.7–1.5)?
Crude (matched on age, sex, and residence)
1.2 (0.9–1.6)¶ Sun sensitivity and sun exposure
1.6 (1.1–2.2) Host factors (nevi, hair and eye color, freckling), sunburns, frequency of
sunbathing, outdoor employment (matched on age, sex, and
residence)
Age, skin color, skin type, number of hours of recreational sun exposure,
hours of occupational sun exposure, number of nevi (no difference by
sex)
Constitutional risk and solar exposure
Age, skin type, nevi (matched on age and sex)
0.4 (0.2–0.7)?
Regular lifelong use: 0.3 (0.1–0.8)
0.5 (0.3–0.7)
1.50 (1.09–2.06) Age, sex, skin type, hair color, holiday weeks in sunny resorts, tropical
sunbathing, and adult sunburns
Age; sex; geographic area; education; skin, eye, and hair color; number
of freckles; large nevi; history of sunburns; tanning pattern; sunny
holidays
Skin color, sunbathing, sunburns, age, and sex
1.1 (0.9–1.4)?
Never: 1.0
Sometimes: 0.97 (0.69–1.35)
Often: 0.80 (0.54–1.17)††
Never: 1.0
Rarely: 1.30 (0.70–2.39)
Often: 3.47 (1.81–6.64)
Never: 1.0
Sometimes: 1.3 (0.9–1.9)
Always/sometimes: 0.9 (0.6–1.5)
Always first, then sometimes: 1.8 (1.1–2.9)††
Measured use of sunscreen as never, SPF ?8, SPF
8–15, or SPF ?15
1.0 (0.6–1.6)?
1.3 (1.0–1.6)?
Hair color, sunburns, frequency of sunbathing, duration of sunbathing
(matched on age, sex, and residence)
0.5 (0.3–0.9)?
Crude (matched on age, sex, and residence)
Sun lotion or oil: 2.1 (1.1–3.8)?
Sun lotion or oil
Almost never: 1.0
Very rarely: 1.5 (0.7–3.0)?
Sometimes: 2.5 (1.1–6.0)?
Quite often: 1.5 (0.4–5.8)?
Almost always: 3.6 (1.5–8.9)?
Crude
Suntan lotion, men: 1.7 (1.1–2.7)Crude
Restricted to men and adjusted for decade of birthSuntan lotion
Always use: 2.58 (1.42–4.69)
UV radiation–blocking agents (hair, sweat, dirt,
dead skin and sunscreen use) always when in
sun: 0.8 (0.4–1.7)
UV radiation dosage and nevi (matched on age, sex, and residence)
Article
Melanoma and Sunscreen Use
www.annals.org
16 December 2003 Annals of Internal Medicine Volume 139 • Number 12 973
Page 9
Using a linear model that assumed equal distance be-
tween the above categories, we pooled data on ordered
categories from the 12 case–control studies and found that
they were heterogeneous, with no apparent association
(odds ratio, 1.1; P ? 0.09) between melanoma and fre-
quency of sunscreen use. Among these 12 studies, 84% of
variation was due to between-study variation (15). How-
ever, 4 of the 12 studies did not adjust for the confounding
effects of sun sensitivity or adjusted only for hair color.
When these 4 studies were excluded, the odds ratio for the
relationship decreased to 0.93 (CI, 0.81 to 1.07). An ad-
ditional 3 studies did not adjust for the potential con-
founding effects of previous sunburns. When the 5 studies
that adjusted for sun sensitivity and sunburns were pooled,
a significant protective association was observed (odds ra-
tio, 0.76 [CI, 0.65 to 0.90]). However, in each of these
analyses, the pooled studies remained heterogeneous: The
variation due to between-study variation remained at 84%
(15). This finding may indicate that the ordered categories
for frequency of sunscreen use may not be comparable
across studies. More likely, however, it reflects differences
in how people quantify their sunscreen use.
Years of sunscreen use may provide a more quantita-
tive measure of the dose of sunscreen used. However,
“years of use” is an imprecise measure that does not ac-
count for how often sunscreen is used. Four studies re-
ported data on years of sunscreen use. Although the data fit
a linear model, there was no association with increasing
years of sunscreen use. Table 3 shows data from the 4
studies and the pooled estimates. The 4 studies that mea-
sured years of use reported homogeneous odds ratios (P ?
0.2).
Figure 2. Odds ratios and 95% CIs for ever use of sunscreen for the 18 studies, sorted by first year of data collection; overall pooled
estimate, based on a random-effects dose–response model; and pooled estimate including only studies that adjusted for sun
sensitivity (n ? 9).
*Unpublished.
Article
Melanoma and Sunscreen Use
974 16 December 2003 Annals of Internal Medicine Volume 139 • Number 12
www.annals.org
Page 10
DISCUSSION
Our results do not support a positive association be-
tween sunscreen use and melanoma. No association was
seen between melanoma and ever use of sunscreen in this
meta-analysis of 18 heterogeneous studies. The heteroge-
neity across studies limits the validity of the pooled results
and must be considered when interpreting the analyses.
However, no association was seen among the homogenous
odds ratios reported for sun-sensitive persons (those with
fair or pale skin color, or Fitzpatrick skin type I or II). The
heterogeneity seen among sun-resistant persons (those with
medium to dark skin color or Fitzpatrick skin type III or
IV) should be explored further but may be related to pro-
longed exposure to sun after maximal tanning.
The inconsistent findings and heterogeneity among
dose–response analyses for frequency of use are not surpris-
ing. Most studies ask participants how often they use sun-
screens according to such categories as “never,” “some-
times,” “often,” or “always,” but they are not asked about
duration of use or reapplication of sunscreens. Data on
years of use were not heterogeneous and had low or incon-
sistent odds ratios across increasing categories of use in all
4 studies. The lack of a dose–response effect for years of
use supports a null association between sunscreen use and
melanoma. However, additional studies that properly ad-
just for confounding effects of sun sensitivity and sunburns
are needed to confirm any reduction in melanoma risk
with sunscreen use. A better exploration of differences in
risk by skin type is also needed.
Previous reports of an increased risk for melanoma
with sunscreen use are misleading. Sun-sensitive persons
typically are more likely to use sunscreens (35) and are at
higher risk for melanoma (9, 20, 25, 36). Thus, reports of
increased risk for melanoma among sunscreen users may
reflect an increased risk among sun-sensitive persons rather
than an increased risk solely due to sunscreen use. There-
fore, lack of control for the confounding effects of sun
sensitivity would provide crude odds ratios that are larger
than odds ratios adjusted for sun sensitivity. Evidence for
such confounding is supported by a decrease in the linear
odds ratio for frequency of use seen in the analyses when
studies that did not control for sun sensitivity were ex-
cluded. In addition, pooled results from the 4 studies that
reported data stratified by sun sensitivity suggest no effect
of sunscreen use. Thus, the lack of overall association in
most of the analyses does not support an increased risk for
melanoma with use of sunscreen.
Sunscreen use may be higher among persons with a
history of sunburns. Sunscreens were initially developed to
reduce erythema or sunburns. Evidence for confounding
due to sunburn history is supported by a decrease in the
linear odds ratio for melanoma and frequency of sunscreen
use when studies that did not control for sunburns were
excluded from analyses. Some people may use sunscreen to
allow prolonged sun exposure by preventing sunburns,
thus receiving a higher amount of cumulative ultraviolet
radiation. Minimal erythema may only be a marker for
damaging effects of sun exposure. Therefore, prevention of
sunburns without reduction in sun exposure may not re-
duce risk for skin cancer (37). Clinically, sunscreen use
may help prevent melanoma by preventing sunburns; how-
ever, people must be warned that sunscreens do not protect
the skin during prolonged exposure to sun.
Sun exposure is a complex confounding factor of sun-
screen use because it can be intermittent or chronic. Sun-
resistant persons may incur risk for melanoma with cumu-
lative sun exposure after maximal tanning (38). Sun-
resistant persons may use sunscreen to protect against
sunburn, thus allowing them to spend longer amounts of
time in the sun. It is unclear whether such persons would
be protected from an increased risk for melanoma owing to
lack of sunburn and sufficient ultraviolet protection from
the sunscreens.
A double-blind, randomized trial of sunscreen use and
Table 3. Years of Sunscreen Use in 4 Studies, and Pooled Risk for Melanoma
Author, Year
(Reference)
Age
Range
Type of ControlsGeographic
Location
Period of Data
Collection
Cases/
Controls
Odds Ratio (95% CI)
for Duration of
Sunscreen Use
Duration of
Sunscreen
Use
Linear Pooled
Odds Ratios
(95% CI)
y n/ny
Holman et al.,
1986 (20, 25)
?80 Population based Western
Australia
1/1/1980–11/5/1981494/494 1.0, never
1.06 (0.71–1.57), ?10 y
1.15 (0.78–1.68), ?10 y
1.0, never
0.97 (0.59–1.61), 1–9 y
1.01 (0.64–1.58), 10–19 y
0.86 (0.58–1.29), ? 20 y
1.0, never
4.3 (0.8–21.9), 1–20 y
1.7 (0.5–5.6), ? 20 y
1.0, never
0 1.0
Naldi et al., 2000
(31)
UnclearNon–population
based
Italy6/1992–2/1995 542/53810 1.06 (0.96–1.17)
Westerdahl et al.,
2000 (6)
16–80 Population basedSweden 1/1995–6/1997 558/89120 1.12 (0.92–1.37)
Osterlind et al.,
1988 (22)
20–79Population basedEastern
Denmark
10/1/1982–3/31/1985 474/926
P value for linear
trend ?0.2
P value for
heterogeneity
?0.2
1.3 (0.9–1.7), ?10 y
1.2 (0.9–1.5), ?10 y
Article
Melanoma and Sunscreen Use
www.annals.org
16 December 2003 Annals of Internal Medicine Volume 139 • Number 12 975
Page 11
duration of sun exposure (39) illustrates the potential for
prolonged sun exposure among sunscreen users. Partici-
pants 18 to 24 years of age randomly received sunscreen
with a sun protection factor (SPF) of 10 or 30 but no
recommendations about sun protection or sun exposure.
The group that received sunscreen with an SPF of 30 had
higher cumulative exposure to sun and longer daily sun-
bathing than did the group that used sunscreen with an
SPF of 10 (39). This finding suggests that sun exposure is
an important confounder of the association between sun-
screen use and melanoma (9, 40, 41).
Although intermittent exposure to sun appears to be a
major risk factor for melanoma (18, 21, 22, 27), some
research suggests that chronic or cumulative exposure (7,
22, 27, 28) is also an important risk factor. In this review,
the odds ratios pooled across sun-sensitive persons were
homogeneous (P ? 0.13), whereas those among sun-resis-
tant persons were heterogeneous (P ? 0.002). The heter-
ogeneity suggests a wider variation in the odds ratios and
CIs among sun-resistant persons in the reviewed articles
and must be considered a limiting factor in interpreting
these analyses. Some sun-resistant or dark-skinned persons
may use sunscreen to prevent sunburns (potentially reduc-
ing their risk for melanoma), whereas others use sunscreen
to prolong their exposure to sun (potentially increasing
their risk for melanoma). The strong correlation between
sun exposure and sunscreen use makes adjustment and in-
terpretation difficult (40).
Differences in sunscreen use patterns and formulations
may also affect any association between sunscreen use on
melanoma development. Since sunscreen was developed in
1928 (42), the type of ultraviolet protection has changed
dramatically, with the evolution of new active ingredients.
A standardized SPF rating system was proposed in 1978
(43). The SPF is the ratio of the minimal erythema dose
needed to induce minimal sunburn to skin protected by a
standard dose of sunscreen (2 mg/cm) compared with un-
protected skin (44). Studies of ultraviolet radiation in the
fish genus Xiphophorus have suggested that ultraviolet A
radiation is more important than ultraviolet B in develop-
ment of cutaneous malignant melanoma (45, 46). Histor-
ically, most sunscreens were inadequate to protect against
ultraviolet A radiation (41, 47). Partial, short-wave, ultra-
violet A radiation–absorbing compounds were added to
selected sunscreens only starting in 1989. Relative protec-
tion against long-wave ultraviolet A radiation was eventu-
ally achieved with the development of Avobenzone and
Parsol 1789 and inclusion of physical blocking agents, such
as titanium dioxide (48). Since the ultraviolet protection
afforded by sunscreens has changed dramatically in the past
50 years, some of the variability seen in the studies ana-
lyzed may reflect changes in specific ultraviolet protection.
Thus, studies reporting no effect between sunscreen use
and melanoma should be interpreted cautiously if effective
sunscreen agents were not widely available when most
study participants received the majority of their sun expo-
sure (20, 49).
Furthermore, none of the reviewed studies considered
or controlled for possible inconsistencies in sunscreen fac-
tors, such as substantivity (the ability of a sunscreen to
adhere to the skin) or water resistance, and only a few
accounted for SPF. Substantivity and water resistance also
influence the relative protection afforded by various sun-
screen products. Because of inconsistencies in sunscreen
substantivity, the amount of protection provided after ap-
plication can vary greatly despite similar SPF ratings. Sim-
ilarly, water resistance can also determine the relative pro-
tection provided by sunscreens with the same SPF (50).
The efficacy of sunscreen is contingent on the adequacy
and proficiency of its application. However, numerous
studies have documented poor compliance with recom-
mendations of regular sunscreen application (35, 51).
Thus, patient education is of clinical importance. Patients
must understand that sunscreen products must be applied
both evenly and in sufficient amounts to achieve adequate
protection (52). Sunscreens with an SPF greater than 15
are recommended by most experts.
Other biases in retrospective case–control studies may
include increased reporting of sunscreen use among pa-
tients with melanoma. This phenomenon reflects patients’
current concern about sun protection (49). A recent case–
control study of melanoma in Iowa in which participants
were asked about current sunscreen use when outside on a
sunny day supports this idea. At the time of the survey,
which occurred 2 to 3 years after diagnosis, 45.9% of pa-
tients with melanoma reported using sunscreen at least half
the time when in the sun compared with only 25.1% of
controls (53). Current concern about sun protection in
patients after diagnosis of melanoma could bias the report-
ing of sunscreen use, especially if care is not taken to ensure
that only prediagnosis use is ascertained (49). Furthermore,
sunscreen use means different things to different persons
(54). The survey instruments used to measure sunscreen
use must therefore be examined for reliability and validity.
Details are needed on how much sunscreen is applied, fre-
quency of application, and whether it is applied before sun
exposure.
Studies to date do not provide evidence of an in-
creased risk for melanoma with sunscreen use. A few stud-
ies suggest a protective effect of sunscreen use against mel-
anoma. Yet, most of the studies reviewed were conducted
before protection against ultraviolet A radiation was devel-
oped and other sunscreen variables, such as substantivity
and water resistance, were standardized. It will probably
take decades to see the potential protective effect of regular
use of sunscreens with SPF greater than 15, protection
against ultraviolet A radiation, or use according to recom-
mended technique of application on risk for melanoma.
The effects of prolonged exposure to sun due to use of
sunscreens to prevent sunburns are unclear, and further
study is needed. Future studies of sunscreen and melanoma
Article
Melanoma and Sunscreen Use
976 16 December 2003 Annals of Internal Medicine Volume 139 • Number 12
www.annals.org
Page 12
also must obtain detailed measurement of sunscreen use
and confounding factors. Studies that do not attempt to
reduce bias or do not account for confounding factors will
only contribute to the misinformation and confusion
about sun-safe habits.
From the College of Public Health and College of Medicine, University
of Iowa, Iowa City, Iowa.
Grant Support: In part by the National Cancer Institute, grant number
1R03CA88834-01.
Potential Financial Conflicts of Interest: None disclosed.
Corresponding Author: Leslie K. Dennis, MS, PhD, Department of
Epidemiology, College of Public Health, University of Iowa, 200
Hawkins Drive, C21H-GH, Iowa City, IA 52242; e-mail, leslie-
dennis@uiowa.edu.
Current author addresses and author contributions are available at www
.annals.org.
References
1. Jemal A, Murray T, Samuels A, Ghafoor A, Ward E, Thun MJ. Cancer
statistics, 2003. CA Cancer J Clin. 2003;53:5-26. [PMID: 12568441]
2. Horn-Ross PL, Holly EA, Brown SR, Aston DA. Temporal trends in the
incidence of cutaneous malignant melanoma among Caucasians in the San Fran-
cisco-Oakland MSA. Cancer Causes Control. 1991;2:299-305. [PMID:
1932542]
3. Ries L, Eisner M, Kosary C, Hankey B, Miller B, Clegg L, et al. SEER
Cancer Statistics Review, 1973-1999. Bethesda, MD: National Cancer Institute;
2002.
4. Stern RS, Weinstein MC, Baker SG. Risk reduction for nonmelanoma skin
cancer with childhood sunscreen use. Arch Dermatol. 1986;122:537-45. [PMID:
3707169]
5. Westerdahl J, Olsson H, Masback A, Ingvar C, Jonsson N. Is the use of
sunscreens a risk factor for malignant melanoma? Melanoma Res. 1995;5:59-65.
[PMID: 7734957]
6. Westerdahl J, Ingvar C, Masback A, Olsson H. Sunscreen use and malignant
melanoma. Int J Cancer. 2000;87:145-50. [PMID: 10861466]
7. Wolf P, Quehenberger F, Mullegger R, Stranz B, Kerl H. Phenotypic mark-
ers, sunlight-related factors and sunscreen use in patients with cutaneous mela-
noma: an Austrian case-control study. Melanoma Res. 1998;8:370-8. [PMID:
9764814]
8. Whiteman DC, Valery P, McWhirter W, Green AC. Risk factors for child-
hood melanoma in Queensland, Australia. Int J Cancer. 1997;70:26-31. [PMID:
8985086]
9. Weinstock MA. Do sunscreens increase or decrease melanoma risk: an epide-
miologic evaluation. J Investig Dermatol Symp Proc. 1999;4:97-100. [PMID:
10537017]
10. Bastuji-Garin S, Diepgen TL. Cutaneous malignant melanoma, sun expo-
sure, and sunscreen use: epidemiological evidence. Br J Dermatol. 2002;146
Suppl 61:24-30. [PMID: 11966729]
11. Woolf B. On estimating the relation between blood group and disease. Ann
Hum Genet. 1955;19:251-3. [PMID: 14388528]
12. Armstrong BK, White E, Saracci R. Principles of exposure measurement in
epidemiology. In: Monographs in Epidemiology and Biostatistics. vol. 21. New
York: Oxford Univ Pr; 1992.
13. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials.
1986;7:177-88. [PMID: 3802833]
14. Greenland S, Longnecker MP. Methods for trend estimation from summa-
rized dose-response data, with applications to meta-analysis. Am J Epidemiol.
1992;135:1301-9. [PMID: 1626547]
15. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis.
Stat Med. 2002;21:1539-58. [PMID: 12111919]
16. Klepp O, Magnus K. Some environmental and bodily characteristics of
melanoma patients. A case-control study. Int J Cancer. 1979;23:482-6. [PMID:
437925]
17. Green A, Bain C, McLennan R, Siskind V. Risk factors for cutaneous
melanoma in Queensland. Recent Results Cancer Res. 1986;102:76-97. [PMID:
3738188]
18. Beitner H, Norell SE, Ringborg U, Wennersten G, Mattson B. Malignant
melanoma: aetiological importance of individual pigmentation and sun exposure.
Br J Dermatol. 1990;122:43-51. [PMID: 2297503]
19. Herzfeld PM, Fitzgerald EF, Hwang SA, Stark A. A case-control study of
malignant melanoma of the trunk among white males in upstate New York.
Cancer Detect Prev. 1993;17:601-8. [PMID: 8275512]
20. Holman CD, Armstrong BK, Heenan PJ. Relationship of cutaneous malig-
nant melanoma to individual sunlight-exposure habits. J Natl Cancer Inst. 1986;
76:403-14. [PMID: 3456458]
21. Holly EA, Aston DA, Cress RD, Ahn DK, Kristiansen JJ. Cutaneous mel-
anoma in women. I. Exposure to sunlight, ability to tan, and other risk factors
related to ultraviolet light. Am J Epidemiol. 1995;141:923-33. [PMID:
7741122]
22. Osterlind A, Tucker MA, Stone BJ, Jensen OM. The Danish case-control
study of cutaneous malignant melanoma. II. Importance of ultraviolet-light ex-
posure. Int J Cancer. 1988;42:319-24. [PMID: 3417359]
23. Youl P, Aitken J, Hayward N, Hogg D, Liu L, Lassam N, et al. Melanoma
in adolescents: a case-control study of risk factors in Queensland, Australia. Int J
Cancer. 2002;98:92-8. [PMID: 11857391]
24. Cress RD, Holly EA, Ahn DK. Cutaneous melanoma in women. V. Char-
acteristics of those who tan and those who burn when exposed to summer sun.
Epidemiology. 1995;6:538-43. [PMID: 8562632]
25. Holman CD, Armstrong BK, Heenan PJ, Blackwell JB, Cumming FJ,
English DR, et al. The causes of malignant melanoma: results from the West
Australian Lions Melanoma Research Project. Recent Results Cancer Res. 1986;
102:18-37. [PMID: 3738185]
26. Graham S, Marshall J, Haughey B, Stoll H, Zielezny M, Brasure J, et al. An
inquiry into the epidemiology of melanoma. Am J Epidemiol. 1985;122:606-19.
[PMID: 4025303]
27. Rodenas JM, Delgado-Rodriguez M, Herranz MT, Tercedor J, Serrano S.
Sun exposure, pigmentary traits, and risk of cutaneous malignant melanoma: a
case-control study in a Mediterranean population. Cancer Causes Control. 1996;
7:275-83. [PMID: 8740740]
28. Espinosa Arranz J, Sanchez Hernandez JJ, Bravo Fernandez P, Gonzalez-
Baron M, Zamora Aunon P, Espinosa Arranz E, et al. Cutaneous malignant
melanoma and sun exposure in Spain. Melanoma Res. 1999;9:199-205. [PMID:
10380943]
29. Autier P, Dore JF, Schifflers E, Cesarini JP, Bollaerts A, Koelmel KF, et al.
Melanoma and use of sunscreens: an EORTC case-control study in Germany,
Belgium and France. The EORTC Melanoma Cooperative Group. Int J Cancer.
1995;61:749-55. [PMID: 7790106]
30. Autier P, Dore JF. Influence of sun exposures during childhood and during
adulthood on melanoma risk. EPIMEL and EORTC Melanoma Cooperative
Group. European Organisation for Research and Treatment of Cancer. Int J
Cancer. 1998;77:533-7. [PMID: 9679754]
31. Naldi L, Gallus S, Imberti GL, Cainelli T, Negri E, La Vecchia C. Sun-
screens and cutaneous malignant melanoma: an Italian case-control study. Int J
Cancer. 2000;86:879-82. [PMID: 10842205]
32. Bakos L, Wagner M, Bakos RM, Leite CS, Sperhacke CL, Dzekaniak KS,
et al. Sunburn, sunscreens, and phenotypes: some risk factors for cutaneous mel-
anoma in southern Brazil. Int J Dermatol. 2002;41:557-62. [PMID: 12358823]
33. Fisher J, Schwartzbaum J, Siegle R. A matched case-control study of the
effects of chemical sunscreen use on cutaneous malignant melanoma [Abstract].
Am J Epidemiol. 1996;143:S9.
34. Huncharek M, Kupelnick B. Use of topical sunscreens and the risk of ma-
lignant melanoma: a meta-analysis of 9067 patients from 11 case-control studies.
Am J Public Health. 2002;92:1173-7. [PMID: 12084704]
35. Geller AC, Colditz G, Oliveria S, Emmons K, Jorgensen C, Aweh GN, et
al. Use of sunscreen, sunburning rates, and tanning bed use among more than
10 000 US children and adolescents. Pediatrics. 2002;109:1009-14. [PMID:
Article
Melanoma and Sunscreen Use
www.annals.org
16 December 2003 Annals of Internal Medicine Volume 139 • Number 12 977
Page 13
12042536]
36. Dubin N, Moseson M, Pasternack BS. Epidemiology of malignant mela-
noma: pigmentary traits, ultraviolet radiation, and the identification of high-risk
populations. Recent Results Cancer Res. 1986;102:56-75. [PMID: 3738187]
37. Naylor MF. Erythema, skin cancer risk, and sunscreens [Editorial]. Arch
Dermatol. 1997;133:373-5. [PMID: 9080899]
38. Armstrong BK. Epidemiology of malignant melanoma: intermittent or total
accumulated exposure to the sun? J Dermatol Surg Oncol. 1988;14:835-49.
[PMID: 3397443]
39. Autier P, Dore JF, Negrier S, Lienard D, Panizzon R, Lejeune FJ, et al.
Sunscreen use and duration of sun exposure: a double-blind, randomized trial.
J Natl Cancer Inst. 1999;91:1304-9. [PMID: 10433619]
40. Garland CF, Garland FC, Gorham ED. Rising trends in melanoma. An
hypothesis concerning sunscreen effectiveness. Ann Epidemiol. 1993;3:103-10.
[PMID: 8287144]
41. Dobak J, Liu FT. Sunscreens, UVA, and cutaneous malignancy: adding fuel
to the fire. Int J Dermatol. 1992;31:544-8. [PMID: 1428442]
42. Kim HJ. Photoprotection in adolescents. Adolesc Med. 2001;12:181-93.
43. Food and Drug Administration. Sunscreen drug products for over-the
counter use. Fed Reg. 1978;43:38206-69.
44. Pruim B, Green A. Photobiological aspects of sunscreen re-application. Aus-
tralas J Dermatol. 1999;40:14-8. [PMID: 10098283]
45. Moan J. UV-A radiation, melanoma induction, sunscreens, solaria and ozone
reduction. J Photochem Photobiol B. 1994;24:201-3. [PMID: 7965416]
46. Setlow RB, Woodhead AD, Grist E. Animal model for ultraviolet radiation-
induced melanoma: platyfish-swordtail hybrid. Proc Natl Acad Sci U S A. 1989;
86:8922-6. [PMID: 2813430]
47. Autier P, Dore JF, Renard F, Luther H, Cattaruzza MS, Gefeller O, et al.
Melanoma and sunscreen use: need for studies representative of actual behaviours.
Melanoma Res. 1997;7 Suppl 2:S115-20. [PMID: 9578426]
48. Bissonnette R. Update on sunscreens. Skin Therapy Letter. 1997;2:5.
49. Roberts LK, Stanfield JW. Suggestion that sunscreen use is a melanoma risk
factor is based on inconclusive evidence [Letter]. Melanoma Res. 1995;5:377-9.
[PMID: 8541730]
50. Stokes R, Diffey B. How well are sunscreen users protected? Photodermatol
Photoimmunol Photomed. 1997;13:186-8. [PMID: 9542755]
51. Banks BA, Silverman RA, Schwartz RH, Tunnessen WW Jr. Attitudes of
teenagers toward sun exposure and sunscreen use. Pediatrics. 1992;89:40-2.
[PMID: 1728018]
52. Stenberg C, Larko O. Sunscreen application and its importance for the sun
protection factor. Arch Dermatol. 1985;121:1400-2. [PMID: 4051527]
53. Beane Freeman LE. Arsenic exposure, artificial tanning and melanoma in
Iowa [Dissertation]. Iowa City, IA: University of Iowa; 2003.
54. Rigel DS, Naylor M, Robinson J. What is the evidence for a sunscreen and
melanoma controversy? Arch Dermatol. 2000;136:1447-9. [PMID: 11115154]
Article
Melanoma and Sunscreen Use
978 16 December 2003 Annals of Internal Medicine Volume 139 • Number 12
www.annals.org
Page 14
Current Author Addresses: Dr. Dennis: Department of Epidemiology,
College of Public Health, University of Iowa, 200 Hawkins Drive,
C21H-GH, Iowa City, IA 52242.
Dr. Beane Freeman: Cancer Prevention Fellowship Program, Division of
Cancer Prevention, NCI, Executive Plaza North, Suite 3109, 6130 Ex-
ecutive Boulevard, MSC 7361, Bethesda, MD 20892-7361.
Dr. VanBeek: Department of Dermatology, College of Medicine, Uni-
versity of Iowa, 200 Hawkins Drive, BT2045-1, Iowa City, IA 52242.
Author Contributions: Conception and design: L.K. Dennis.
Analysis and interpretation of the data: L.K. Dennis, L.E. Beane Free-
man, M.J. VanBeek.
Drafting of the article: L.K. Dennis, L.E. Beane Freeman, M.J. VanBeek.
Critical revision of the article for important intellectual content: L.K.
Dennis, L.E. Beane Freeman, M.J. VanBeek.
Final approval of the article: L.K. Dennis, L.E. Beane Freeman, M.J.
VanBeek.
Provision of study materials or patients: L.K. Dennis.
Statistical expertise: L.K. Dennis.
Obtaining of funding: L.K. Dennis.
Administrative, technical, or logistic support: L.K. Dennis, M.J. Van-
Beek.
Collection and assembly of data: L.K. Dennis, L.E. Beane Freeman, M.J.
VanBeek.
© American College of Physicians E-979
Page 15
Appendix Table. Quality Assessment Scoring of the Melanoma and Sunscreen Studies*
Author, Year (Reference) Geographic LocationCoding of Items
Type of
Controls†
Standardized
Questionnaire‡
Interviewer
Administered§
Graham et al., 1985 (26)
Herzfeld et al., 1993 (19)
Beitner et al., 1990 (18)
Green et al., 1986 (17)
Holman et al., 1986 (20, 25)
Holly et al., 1995 (21, 24)
Osterlind et al., 1988 (22)
Whiteman et al., 1997 (8)
Youl et al., 2002 (23)
Berwick et al., (Unpublished data)
Westerdahl et al., 1995 (5)
Rodenas et al., 1996 (27)
Fisher et al., 1996 (33)
Espinosa-Arranz et al., 1999 (28)
Autier et al., 1995 and 1998 (29, 30)
Naldi et al., 2000 (31)
Wolf et al., 1998 (7)
Westerdahl et al., 2000 (6)
Bakos et al., 2002 (32)
Klepp and Magnus, 1979 (16)
New York State
New York State
Stockholm, Sweden
Queensland, Australia
Western Australia
San Francisco, CA
Eastern Denmark
Queensland, Australia
Queensland, Australia
Connecticut
Southern Sweden
Eastern Andalusia, Spain
United States
Madrid, Spain
Belgium, France, and Germany
Italy
Styria, Austria
Sweden
Brazil
Oslo, Norway
0
2
2
2
2
2
2
2
2
0
2
1
0
0
0
0
0
2
0
0
0
2
1
2
2
2
1
0
1
0
1
1
0
1
0
1
0
1
0
1
1
1
0
1
2
1
2
0
1
0
0
1
0
1
1
1
0
0
1
0
* See Table 2 for a description of individual studies.
† 0 ? hospital based or unclear; 1 ? hospital visitors; 2 ? population based.
‡ 0 ? no, not stated; 1 ? same questions, structured; 2 ? standardized or piloted.
§ 0 ? self; 1 ? nonblinded interviewer; 2 ? blinded interviewers.
? 0 ? no; 1 ? yes.
¶ 0 ? no adjustment; 2 ? hair color, eye color, or freckling; 4 ? skin color, skin type, ability to tan, tendency to burn.
** 0 ? no; 2 ? yes.
†† 0 ? no; 1 ? partial; 2 ? yes. Partial adjustment was recorded for such items as age at migration, sunny vacations, and sunbathing. Full adjustment was recorded for
adjustment for cumulative hours of sun exposure, recreational hours, occupational hours, or ultraviolet radiation damage.
‡‡ 0 ? no; 1 ? frequency matched; 2 ? yes.
§§ 0 ? only ever use; 1 ? frequency of use; 2 ? years of use, amount, or duration.
?? Maximum possible score is 19 points.
E-980
Annals of Internal Medicine Volume • Number
www.annals.org
Download full-text 
