Académique Documents
Professionnel Documents
Culture Documents
Summary
Correspondence Background Despite the fact that ultraviolet (UV) light exposure is the most impor-
Jochen Schmitt. tant risk factor for cutaneous squamous cell carcinoma (SCC) there is an ongoing
E-mail: Jochen.Schmitt@uniklinikum-dresden.de debate concerning the relationship between cumulative work-related UV expos-
ure and SCC occurrence.
Accepted for publication
6 October 2010 Objectives To analyse comprehensively the relationship between work-related UV
exposure and SCC risk.
Funding sources Methods We conducted a systematic electronic literature search in PubMed (up to
No external funding.
5 May 2010) supplemented by a hand search, which identified 18 relevant stud-
Conflicts of interest ies that were included in the review. Data abstraction and study quality assess-
None declared. ment was done independently by two reviewers. Maximally adjusted odds ratios
(ORs) and corresponding 95% confidence intervals (CIs) of all included studies
DOI 10.1111/j.1365-2133.2010.10118.x were pooled in a random-effects meta-analysis. Sensitivity analysis included
meta-regression on study-specific covariates to explore the robustness of the
results and to identify sources of heterogeneity between studies. Eighteen studies
(six cohort studies, 12 case–control studies) met the eligibility criteria and were
included in the systematic review.
Results Sixteen studies (89%) found an increased risk of SCC in individuals with
occupational UV light exposure compared with individuals without occupational
UV light exposure, reaching statistical significance in 12 studies. Two studies
found no association between occupational UV light exposure and SCC
occurrence. The pooled OR (95% CI) was 1Æ77 (1Æ40–2Æ22) and did not differ
significantly between cohort studies [OR (95% CI): 1Æ68 (1Æ08–2Æ63)] and
case–control studies [OR (95% CI): 1Æ77 (1Æ37–2Æ30)]. Meta-regression analyses
suggested an increasing strength of the association between occupational UV light
exposure and SCC risk with decreasing latitude.
Conclusions In summary, there is consistent epidemiological evidence for a positive
association between occupational UV light exposure and SCC risk.
Cutaneous squamous cell carcinoma (SCC) is among the most UV exposure is considered to be primarily relevant in the aeti-
frequent malignancies worldwide and constitutes a significant ology of other types of skin cancer such as melanoma or basal
public health problem.1 The age-adjusted incidence of SCC has cell carcinoma (BCC).5
grown by 50–200% over the past 20–40 years.2–4 Exposure to An individual’s risk of developing SCC depends on pheno-
ultraviolet (UV) radiation is the most common and most typic, genotypic and environmental factors: compared with
important known cause of cutaneous SCC.5 Cutaneous SCC women, men are about twice as frequently affected; 80% of
occurs mainly on sun-exposed body sites such as the face.6,7 cases occur in people aged 60 years and older. Individual UV
Its frequency can be reduced by sun protective measures.1 It susceptibility is an important phenotypic risk factor for the de-
has been shown that cumulative lifetime UV exposure is most velopment of SCC: Caucasians with fair skin, and light eye
important in the pathogenesis of SCC, whereas intermittent and hair colour are at increased risk.1 Smoking has also been
shown to be an important risk factor for the development of occupation. If not specified in the paper, we considered the
SCC.8,9 Geographic variation in SCC occurrence is mainly following occupations as outdoor occupations with high levels
explained by ambient UV light exposure: SCC incidence is as- of work-related UV light exposure: asphalt workers, construc-
sociated with proximity to the equator.6,10–13 tion workers, farmers, winegrowers, roadmen, fishermen ⁄
Although the biological underpinnings of the causal role of seafarers ⁄deck crew, foresters ⁄lumbermen, or gardeners.20,21
UV light exposure and the development of cutaneous SCC are SCC and BCC are frequently summarized as nonmelanoma skin
well understood and epidemiological data from various cancer (NMSC). Because the most relevant pattern of UV light
sources provided indirect but substantial evidence that solar exposure differs between SCC and BCC (cumulative exposure
UV light exposure is by far the most relevant risk factor for vs. intermittent exposure), we excluded studies that reported
SCC on the population level, there is ongoing debate about only on the occurrence of NMSC without differentiating
the relationship between occupational exposure to solar UV between the exact histopathological entities.5 However, aiming
light and the risk of cutaneous SCC.14–17 Relevant proportions for a very sensitive search strategy we extended the literature
of the working population experience high levels of chronic search beyond SCC and also included search terms for BCC
or intermittent work-related exposure to solar UV light. In and NMSC.
Germany, for example, approximately two million individuals A systematic electronic literature search in PubMed was
are working in typical outdoor occupations, e.g. as construc- done from inception up to 5 May 2010. We searched for
tion workers, farmers, winegrowers, roadmen, seafarers, for- combinations of the following search terms:
esters or lumbermen, or fishermen.15 German law aims to ‘risk’ OR ‘incidence’ OR ‘epidemiologic studies’ [MeSH
protect individuals with relevant work-related exposure result- terms]
ing in adverse medical outcomes.18 Although it is reasonable AND
to assume that outdoor workers are at higher risk of develop- ‘Carcinoma, squamous cell’ OR ‘Carcinoma, basal cell’ OR
ing cutaneous SCC compared with indoor workers, UV light- ‘skin cancer’ [MeSH terms]
induced work-related cutaneous SCC is not officially accepted AND
as an occupational disease in Germany. The major reason is ‘sunlight’ OR ‘ultraviolet rays’ OR ‘ultraviolet light’
the lack of a comprehensive summary of the epidemiological AND
literature regarding the relationship between work-related UV ‘occupational’ OR ‘occupation’ OR ‘outdoor work’ OR ‘work’
exposure and cutaneous SCC. OR ‘workplace’
In order to fill this research gap we conducted a systematic We limited the systematic electronic search to original arti-
review and meta-analysis to analyse the relationship between cles reporting on humans and articles with abstracts, but did
work-related exposure to solar UV radiation and the develop- not apply any language constraints. The electronic search was
ment of cutaneous SCC. We hypothesized that (i) occupational supplemented by hand searching the reference lists of all iden-
UV exposure is a significant independent risk factor for SCC tified articles and other review papers for additional relevant
occurrence; (ii) the strength of the association between occu- publications.
pational UV exposure and SCC risk increases with the dose of
individual occupational UV exposure; and (iii) the geographi-
Data abstraction
cal region is one important source of heterogeneity between
studies, i.e. that the strength of the association between occu- Screening for eligibility of articles, data abstraction from the
pational UV exposure and SCC occurrence increases with articles included, and study quality assessment were done
decreasing latitude. independently by two reviewers. Disagreements were resolved
by discussion between all authors. The data abstraction form
included information on relevant study characteristics such as
Materials and methods
geographical region, study design, source population, number
of participants, data sources, as well as participant characteris-
Study design
tics (age, sex, ethnicity, smoking status, family history of cuta-
We systematically reviewed original articles reporting on cohort neous SCC). The latitude of the geographical region in which
studies or case–control studies to investigate the relationship the individual studies were performed was identified from
between work-related exposure to solar UV light and the occur- Google Maps (http://www.gorissen.info/Pierre/maps/google-
rence of cutaneous SCC. The presented systematic review was MapLocationv3.php).
performed in accordance with the principles of the Meta-analy- With regard to our research question it is critically impor-
sis of Observational Studies in Epidemiology (MOOSE) group.19 tant to differentiate between work-related and recreational
exposure to solar UV light. Imprecise exposure assessment
may introduce nondifferential misclassification bias.22 There-
Eligibility criteria and identification of articles
fore, we abstracted detailed information on exposure assess-
We included all cohort studies or case–control studies report- ment and definition and categorization of occupational UV
ing on the occurrence of cutaneous SCC with regard to previ- exposure. Study results were abstracted both descriptively and
ous work-related exposure to solar UV light and ⁄or outdoor quantitatively with the main outcome of interest being the
association between occupational UV exposure and SCC occur- light exposure and SCC risk. A minority of the included stud-
rence expressed as maximally adjusted odds ratios (ORs) and ies reported rate ratios (RRs) instead of ORs. Because RRs and
corresponding 95% confidence intervals (CIs). We also ORs are almost identical if the prevalence of the outcome of
abstracted information on whether adjustment was performed interest is as low as the prevalence of SCC,22 we decided not
for important potential confounding factors, i.e. age, sex, UV to differentiate between OR and RR in the qualitative and
sensitivity, nonoccupational UV exposure, family history of quantitative analysis.
skin cancer, and smoking status.1,4,5,23 If not reported in the To explore the robustness of our findings a set of sensitivity
paper, ORs were calculated from the numbers presented analyses was carried out. As part of the sensitivity analysis
whenever possible. pooled ORs were calculated separately for cohort studies and
For both the results tables as well as for the quantitative case–control studies. Heterogeneity based on study-specific
analysis, tumour frequencies among outdoor and indoor covariates such as the absolute value of latitude, utilization of
workers were compared with each other whenever possible. cancer registry data for exposure assessment vs. assessment of
In studies in which occupational exposure to UV light was individual work-related UV light exposure, and study quality
quantified differently, workers with the greatest UV exposure criteria was examined by meta-regression. Both crude meta-
were compared with those with the least. If the study regression models and models adjusted for study type were
included an actual stratification by occupational groups of out- calculated. Publication bias was explored by regressing the
door workers, the largest occupational group with a high like- study result (OR) on sample size adjusting for study type.26
lihood to actually perform outdoor work, i.e. asphalt workers, All analyses were carried out using STATA, version 10 (Stata
construction workers, farmers, winegrowers, roadmen, fisher- Corporation, College Station, TX, U.S.A.).
men ⁄seafarers ⁄deck crew, foresters ⁄lumbermen, or gardeners
was taken as the measure.20,21 In case of reporting of different
Results
anatomical localizations of SCC occurrence, localizations with
a high likelihood of UV exposure such as lips, hands or face We reviewed the abstracts of all 194 citations yielded by the
were considered. electronic search, excluded 162 articles after title and abstract
review, and another 21 articles after full text review because
Study quality assessment
Methodological quality was assessed by means of the New-
castle–Ottawa scale (NOS).24 The NOS is a validated and stan- Standardized electronic
dardized instrument for judging the quality of cohort studies literature search (PubMed,
up to 5 May 2010)
and case–control studies by means of eight criteria regarding
194 articles identified by
selection of study participants, comparability, outcome assess- review of abstracts
ment and exposure assessment.24 For the criterion ‘compara-
bility’ full credit was given if age, sex, UV sensitivity and
individual nonoccupational UV exposure was controlled for by 162 articles excluded
the study design or in data analysis. Studies that considered
age, but failed to consider one or more of the other relevant
confounders were given half credit.24 In addition to the study 32 full text articles screened
quality criteria captured by the NOS we considered the fol- for eligibility
lowing items as important for our research question: assess-
ment of individual UV light exposure, exploration of effect
21 articles excluded
modification, and likelihood of recall bias (in case–control
studies). The criterion ‘exploration of effect modification’ was
included in light of an investigation by Marks et al.12 on the 11 studies included
relationship between outdoor work and SCC occurrence that
found this relationship to be significantly modified by age.
Whenever obvious, additional possible methodological limita- Hand search: identification of
tions and the direction of the bias introduced were recorded. 7 additional studies meeting
the eligibility criteria
Quantitative analysis
ORs of all included studies were pooled using random-effects
18 studies included in the
meta-analysis (DerSimonian–Laird method).25 Fully adjusted systematic review
ORs of the individual studies were used for meta-analysis,
because we considered adjusted ORs as the best estimate of
the true independent association between work-related UV Fig 1. Flow chart of study identification.
they did not meet the eligibility criteria. In addition to the 11 the studies found an inverse relationship between occupational
studies identified through the electronic literature search that UV light exposure and SCC occurrence (Table 4). Seventeen
met our predefined eligibility criteria, seven additional eligible (94%) and sixteen studies (89%) considered age and age and
studies were identified through a hand search so that a total sex as important confounders, respectively. Other relevant
of 18 studies could be included in the systematic review confounders such as UV sensitivity, nonoccupational UV ex-
(Fig. 1). posure, and smoking status were considered only in a minor-
ity of studies. None of the studies adjusted for family history
of skin cancer (Table 4).
Qualitative analysis of studies included
Details on study quality assessment are summarized in
The characteristics of the six cohort studies11,12,27–30 and 12 Table 5. None of the studies adequately met all of the pre-
case–control-studies8,9,14,21,31–38 included are summarized in defined quality criteria. Some of the study quality items such
Table 1. Eleven studies were conducted in Europe, five in as assessment of exposure, adequacy of follow-up, absence of
North America, and two in Australia. All except one cohort outcome of interest at baseline, selection of controls ⁄nonex-
study11 were recruited from the general population. In the posed cohort, and representativeness of exposed cohort ⁄cases
case–control studies, cases were identified directly from the were met by the majority of studies. Methodological weak-
general population or from population-based cancer registries nesses included the exploration of effect modification, which
in nine studies and from selected dermatological centres in was done only in the study by Marks et al.12 In this study,
three studies. Eleven studies utilized cancer registry data exclu- significant interaction of the relationship between outdoor
sively or supplemented by questionnaire data, while seven work and SCC risk by age was detected (P = 0Æ034) and there
studies relied exclusively on primary data collected by ques- was an increase in SCC risk only in outdoor workers above
tionnaire, interview and ⁄or clinical examination (Table 1). age 55 years.12 The limitations in the assessment of individual
There was significant heterogeneity in the definition of occupational UV light exposure most likely introduced non-
‘occupational UV exposure’: eight studies looked at specific differential misclassification resulting in bias towards the null
occupations typically involving outdoor work such as farming, effect.22 In six out of 12 case–control studies (50%) we
construction work or seafaring. Five studies estimated occupa- judged recall bias not to be unlikely – a possible source of
tional UV light exposure on a continuous scale and compared bias introducing an overestimation of the true relationship
participants with high vs. low cumulative exposure. Five stud- between occupational UV light exposure and SCC risk
ies reported only whether outdoor work was performed ever (Table 4).22
or at a specific point in time without providing further details
(Table 2). In the majority of the studies included (14 of 18
Quantitative results – meta-analysis
studies; 78%) the reference group was defined as ‘mainly
indoor work’, ‘low occupational UV light exposure’, or Meta-analysis indicated that individuals with outdoor UV light
included occupations actually performing outdoor work, exposure are at significantly increased risk of developing cuta-
mixed indoor ⁄outdoor work, or the general population. This neous SCC. The pooled OR (95% CI) of the random-effects
imprecise definition of the unexposed group is an important meta-analysis including all 18 studies was 1Æ77 (1Æ40–2Æ22)
source of nondifferential misclassification introducing bias (P < 0Æ001) (P for heterogeneity < 0Æ001; moment-based esti-
towards the null effect (Table 2).22 Five studies were restricted mate of between studies variance = 0Æ131). The pooled OR
to males. Information on important characteristics and (95% CI) was virtually identical in cohort studies [n = 6; OR
potential confounders such as ethnicity, smoking status and 1Æ68 (1Æ08–2Æ63); P = 0Æ022] (P for heterogeneity < 0Æ001;
family history of skin cancer were frequently not reported moment-based estimate of between studies variance = 0Æ231)
(Table 2). and case–control studies [n = 12; OR 1Æ77 (1Æ37–2Æ30);
Thirteen studies assessed individual occupational UV light P < 0Æ001] (P for heterogeneity = 0Æ002; moment-based esti-
exposure while five studies exclusively relied on census or mate of between studies variance = 0Æ087) (Fig. 2).
registry data and used the job description as a proxy for occu- Meta-regression analysis identified significant sources of
pational UV light exposure (Table 3). A quantification of heterogeneity between studies. The strength of the association
occupational UV light exposure was given in 10 papers and between occupational UV light exposure and SCC risk in-
was not assessed in the remaining eight papers (Table 3). creased significantly with decreasing latitude of the geographi-
Table 3 reveals information on the measurement of occupa- cal region where the study was conducted [P = 0Æ022 in the
tional and nonoccupational UV light exposure in the 18 stud- adjusted model; P < 0Æ001 in the unadjusted model (Fig. 3)].
ies included. Despite this statistically significant association, some individual
Sixteen studies (89%) found an increased risk of SCC in studies found a low risk of SCC among outdoor workers
individuals with occupational UV light exposure compared despite low latitude27,36 or indicated high risk of SCC among
with individuals without ⁄with lower occupational UV light outdoor workers despite high latitude (Fig. 3).11,33
exposure. The positive association reached statistical signifi- Studies utilizing registries found significantly weaker associ-
cance in 12 studies. Two studies found no association between ations between occupational UV light exposure and SCC than
occupational UV light exposure and SCC occurrence. None of studies using only primary data and selected cases without the
southern Europe), Turin patients; frequency matched Ca: n = 139 cancer registry
dermatological by age Co: n = 349
clinics
Marehbian et al.38 U.S.A., New Hampshire 1994–95 43 Case–control General population; Controls randomly n = 824 Cancer registry,
cancer registry selected from Ca: n = 290 interview,
driver’s licence Co: n = 534 questionnaire
registry and
Medicare ⁄ Medicaid
services; frequently
matched for
age and sex
Kenborg et al.14 Denmark 1970–2003 55 Case–control General population; Controls randomly n = 9600 Registry data
cancer registry selected from Danish Ca: 4800g (cancer
Central Person registry,
Registry, matched pension fund)
1 : 1 by age and sex
2011 The Authors
Ca, cases; Co, Controls; SCC, squamous cell carcinoma; py, person years; NA, not applicable. aPeriod of time in which data were collected. bTotal number of participants; in case–control study n (cases
with SCC) + n (controls for cases with SCC). ce.g. Interview, questionnaire, repeated physical examination, cancer registry data. dSCC of the lip. eExcluding SCC of the lip. fNonmetastatic SCC. gUnclear,
higher numbers in Table 4.
BJD 2011 British Association of Dermatologists 2011 164, pp291–307
2011 The Authors
Table 2 Characteristics of study population and definition of occupational ultraviolet (UV) exposure
Participants’ characteristicsa
Participants’ characteristicsa
gardeners
Zanetti et al.37 Highest quartile of cumulative occupational No occupational UV exposure 20–75 0 Caucasians NR NR
UV exposure based on sun exposure index 100
Marehbian et al.38 Several individual occupations assessed; All other occupations 25–74 41 Caucasians: NR NR
Standard Occupational Classification (SOC) > 97
used for coding
Kenborg et al.14 Ever ‘outdoor occupation’, i.e. ‡ 40% of All other occupations (including those Mean 63Æ9 0 NR NR NR
employees spend > 50% of work involving outdoor work, but do not meet
time outdoors criteria for ‘outdoor occupation’)
NR, not reported; OR, odds ratio aIn cohort and cross-sectional studies for total study population, in case–control studies separately for cases (Ca) and (Co) controls (if reported).
2011 The Authors
BJD 2011 British Association of Dermatologists 2011 164, pp291–307
2011 The Authors
Individual Individual
exposure Period of Quantification exposure Period of time Quantification
References Data source assessed time considered of exposure Data source assessed considered of exposure
Marks et al.12 Questionnaire Yes NR NA NA NA NA NA
Green et al.27 Questionnaire Yes Present and past occupations NA Questionnaire Yes Ever Hours spent outdoors;
number of painful
sunburns
Pukkala and Saarni28 Pension fund No Age range unclear; observation Years employed NA NA NA NA
period 1960–80 as seafarer
Adami et al.29 Census ⁄ industrial No Current occupation 1960 NA NA NA NA NA
hygienist (outcome SCC 1971–89)
Hakansson et al.11 Industrial hygienist Yes Present occupation Using Likert scale NA NA NA NA
Radespiel-Tröger et al.30 Cancer registry No Longest occupation NA NA NA NA NA
Aubry and Questionnaire Yes 7–37 years prior to assessment Sun exposure index Questionnaire Yes Lifetime Sun exposure index
MacGibbon8 (weighted hours ⁄ (weighted hours ⁄
lifetime) lifetime)
Hogan et al.31 Questionnaire Yes NR NR Questionnaire Yes NR NR
Gafa et al.32 Questionnaire Yes Whole life Years working in Questionnaire Yes NR Hours ⁄ day
agriculture
NA, not assessed; NR, not recorded; SCC, squamous cell carcinoma.
300 Occupational UV exposure and cutaneous SCC, J. Schmitt et al.
Table 4 Qualitative and quantitative results of studies included in the systematic review
Dose–response UV Nonoccupational
References Qualitative result relationship Quantitative result (OR, 95% CI) Age Sex sensitivity UV exposure Smoking
Marks et al.12 Nonsignificant increase in SCC risk among No 1Æ7, 0Æ9–3Æ3; outdoor vs. indoor •
outdoor workers; significant interaction of workers
the relationship between outdoor work
and SCC risk by age (P = 0Æ034): increase
in SCC risk in outdoor workers above age
55 years
Green et al.27 Nonsignificant increase in SCC risk among No SCC: 1Æ4, 0Æ8–2Æ3; outdoor vs. • • •a
outdoor workers; significantly increased indoor workers. AK: 2Æ2, 1Æ5–3Æ1;
AK risk in outdoor workers; significant > 11 vs. 0; AKs in outdoor vs.
self-selection among outdoor workers indoor workers
with underrepresentation of skin types I & II
Pukkala and Saarni28 More cases of SCC in deck officers and deck Yes 2Æ4, 1Æ3–3Æ9; SIR deck officers • •
crew than expected compared with general population
Adami et al.29 No association between outdoor occupation Unclear [Yes (females) Men 1Æ0, 0Æ9–1Æ0. Women 1Æ3, • •
and SCC occurrence in men, but No (males)] 1Æ1–1Æ6
significantly increased SCC risk in female
BJD 2011 British Association of Dermatologists 2011 164, pp291–307
outdoor workers
Hakansson et al.11 SCC of the lip significantly increased in No 2Æ2, 1Æ0–4Æ5 • • •
construction workers with high UV
exposure
Radespiel-Tröger et al.30 SCC risk significantly increased in male and NA Men 2Æ5; 1Æ4–4Æ7. Women 3Æ6, • •
female outdoor workers compared with 1Æ6–8Æ1; RR outdoor vs. indoor
indoor workers worker
Aubry and MacGibbon8 Positive association between lifetime NA 9Æ1, 2Æ0–84Æ5; occupational UV • • • •
occupational UV exposure and SCC exposure modelled as binary
occurrence variable, details NR
Hogan et al.31 Significant association between agricultural NA Men 1Æ5, 1Æ2–1Æ8. Women 1Æ8; • • •
occupation and SCC risk in both men and 1Æ2–2Æ7 Agricultural vs. other
women occupations
Gafa et al.32 Significantly increased risk of SCC in NA 2Æ4, 1Æ0–5Æ6
individuals working in agriculture for
10 years or longer
2011 The Authors
Gallagher et al.33 Significant association between SCC Yes 4Æ0, 1Æ2–13Æ1; high vs. low mean • • •
occurrence and mean annual occupational annual occupational UV exposure
UV exposure in past 10 years; nonsignificant in past 10 years
association between SCC occurrence and
mean annual lifetime occupational
BJD 2011 British Association of Dermatologists 2011 164, pp291–307
2011 The Authors
Table 4 Continued
Dose–response UV Nonoccupational
References Qualitative result relationship Quantitative result (OR, 95% CI) Age Sex sensitivity UV exposure Smoking
34
Rosso et al. Outdoor work significantly associated with Yes 1Æ6, 1Æ0–2Æ5; high vs. low occupational UV • • • •
SCC occurrence exposure
Perea-Milla López et al.9 Significant positive association between No 11Æ9, 1Æ3–108Æ9; highest quartile vs. no • • • •b
cumulative occupational UV exposure and occupational UV exposure
SCC occurrence
Masini et al.35 Outdoor work significantly associated with NA 4Æ3, 1Æ6–11Æ9; outdoor work vs. indoor • •
SCC occurrence work
Mitropoulos and Norman36 No significant association between NA 1Æ4, 0Æ6–3Æ1; construction workers vs. • • •
occupation and SCC occurrence non-
AK, actinic keratosis; CI, confidence interval; NA, not assessed; NR, not recorded; OR, odds ratio; SCC, squamous cell carcinoma; SIR, standardized incidence ratio. aAdjustment only in SCC model;
b
adjusted for leaving cigarette on lip. Bold indicates numbers used for meta-analysis.
302 Occupational UV exposure and cutaneous SCC, J. Schmitt et al.
Table 5 Methodological quality, likelihood of bias, and limitations of studies included in the systematic reviews
Newcastle–Ottawa scale
Outcome
of Adequacy
interest Comparability of follow-
Selection Ascertainment was not of cohorts up of Assessment
Representativeness of of individual present on basis of Assessment Adequate cohorts ⁄ of individual Effect Other Direction
of exposed nonexposed occupational at start of design or of follow-up follow-up nonoccupational modification relevant of bias
References cohort cohort exposure study analysis outcome time rate adequate UV exposure explored limitations introduced
Cohort studies
Marks et al.12 • • • • •s • • s s • None
Green et al.27 • • • • •s • s • • s None
Pukkala and Saarni28 • s s • •s • • • s s None
Adami et al.29 • • s • •s • • • s s Cumulative Towards OR = 1
occupational
BJD 2011 British Association of Dermatologists 2011 164, pp291–307
exposure unclear
Hakansson et al.11 s • • • •s • • • s s Only current Towards OR = 1
occupation at
time of examination
assessed;
change of
occupation
introduces
nondifferential
misclassification
bias
Radespiel-Tröger et al.30 • • • • •s • s • s s 74% missing None
data on
occupational
exposure; missing
at random assumed
2011 The Authors
BJD 2011 British Association of Dermatologists 2011 164, pp291–307
2011 The Authors
Table 5 Continued
Newcastle–Ottawa scale
Comparability
of cases Ascertain- Same method
and controls ment of of exposure Assessment
on basis of individual assessment in of individual Recall Effect
Case definition Representativeness Selection of Definition design and occupational cases and Nonresponse nonoccupational bias modification other relevant Direction of
adequate of cases controls of controls analysis exposure controls rate UV exposure unlikelya explored limitations bias introduced
Case–control studies
Aubry and MacGibbon8 • s s • •• s • s • s s None
Hogan et al.31 • s • s •s s • s • s s Reference group Towards
contains individuals OR = 1
with outdoor occupation
Gafa et al.32 • • s s ss s • s • • s Inappropriate control group, Overestimation of effect
selective reporting of results
Gallagher et al.33 • • • • •s • • s • • s None
Rosso et al.34 • • • • •• s • s • s s None
Perea-Milla López et al.9 • • • s •s s • s • s s None
Masini et al.35 • s s • •s s • s • s s None
Mitropoulos • s • • •s s • s s • s Only longest Towards OR = 1
• Criterion met. s Criterion not met. aRecall bias unlikely if occupation recorded.
304 Occupational UV exposure and cutaneous SCC, J. Schmitt et al.
Combined
Odds ratio
Fig 2. Results of random-effects meta-analysis for squamous cell carcinoma of the skin of individuals with outdoor occupation ⁄ occupational
ultraviolet (UV) exposure vs. indoor occupation ⁄ no occupational UV exposure.
20 30 40 50 60 Discussion
Latitude†
Our paper is the first meta-analysis that comprehensively sum-
marizes the published epidemiological evidence regarding the
Fig 3. Relationship between latitude and strength of association
relationship between work-related UV light exposure and the
between occupational ultraviolet (UV) exposure and squamous cell
risk of cutaneous SCC. The presented results indicate that out-
carcinoma of the skin. The circle corresponding to each study has an
door work constitutes a relevant, independent, and robust risk
area inversely proportional to the variance of the log-odds ratio. The
superimposed line is obtained by weighted regression using a factor for the development of cutaneous SCC. Our findings are
restricted maximum likelihood (REML) estimate of residual highly relevant as currently SCC is among the most frequent
heterogeneity variance.41 *Log-odds ratio (OR) of the relationship malignancies worldwide with rapidly increasing incidence rates
between occupational UV light exposure and squamous cell carcinoma over past decades.1–4 Our results are biologically plausible, as
of the skin. Absolute value of latitude. the cumulative exposure to UV radiation is the most common
and the most important known cause of cutaneous SCC.5
use of a cancer registry (P = 0Æ049; adjusted model). Studies There are several reasons why we believe that the observed
that directly assessed individual occupational UV light expos- increase in the risk for cutaneous SCC among individuals
ure of study participants found stronger associations between with work-related UV light exposure compared with indoor
workers is an underestimation of the true association. Firstly, possible explanation why the Scandinavian studies by Kenborg
only three studies8,34,38 considered the most important con- et al.14 and Adami et al.29 were the only studies that found no
founders of the relationship between occupational UV light association between occupational UV light exposure and SCC
exposure and SCC occurrence, i.e. age, sex, individual UV sen- occurrence. However, both studies that found no association
sitivity, and individual nonoccupational UV exposure by the between occupational UV light exposure and SCC occurrence
study design or in data analysis. Two of these studies found also have other significant limitations: both studies did not
strong, statistically significant associations between occupa- actually measure or assess occupational UV light exposure, but
tional UV light exposure and SCC occurrence with ORs above used job titles or job descriptions as a proxy for individual UV
10,8,38 and the other study34 observed only a weak, but also light exposure, harbouring the risk of introducing nondiffer-
statistically significant association. Overall, studies with incom- ential misclassification bias. Additionally, the observation
plete consideration of confounders found less strong associa- period was inadequate in the study by Kenborg et al.:14 only
tions between occupational UV light exposure and SCC participants aged 25–58 with occupational exposure until age
occurrence suggesting the possible presence of residual con- 48 were considered for the analysis concerning the relation-
founding. Green et al.27 observed significant self-selection ship between occupational UV light exposure and SCC occur-
among outdoor workers, whereby people with fair or med- rence, although the average age at diagnosis of SCC was
ium complexions and a tendency to sunburn were systemati- 64 years in this population. This is problematic as the study
cally underrepresented among those in long-term outdoor by Marks et al.12 reported significant interaction of the rela-
occupations. The assumption of residual confounding leading tionship between outdoor work and SCC risk by age with an
to an underestimation of the true effect is also strengthened increase in SCC risk only in outdoor workers above age
by results from the presented meta-regression analysis indicat- 55 years. Another limitation of this study is missing informa-
ing that studies adjusting for individual UV light sensitivity tion on occupational exposure in elderly participants, also
found significantly stronger relationships between occupational introducing nondifferential misclassification bias towards the
UV light exposure and SCC than studies not adjusting for indi- null effect.22
vidual UV light sensitivity. Another source of bias towards the Our study followed a priori defined procedures, which also
null effect that was present in 14 of the 18 studies included met the criteria in the MOOSE checklist for performing a sys-
in the review was an imprecise definition of the reference tematic review of epidemiological studies.19 Despite all the
group, i.e. individuals not exposed to solar UV light at the efforts undertaken, we cannot guarantee that we identified all
workplace. The reference group was actually not truly unex- relevant articles on the issue. We conducted several sensitivity
posed in the majority of the studies included in the review. analyses that provide evidence for the robustness of our find-
Five studies failed to collect information on individual expos- ings and independence of the observed significant relationship
ure to solar UV light at the workplace and relied exclusively between occupational solar UV light exposure and SCC risk
on census or registry data and ⁄or used the job description as a from the study type and several study quality criteria. Another
proxy for occupational UV light exposure. Adami et al.29 strength of our meta-analysis is the identification of relevant
hypothesized that the potential misclassification of exposures sources of heterogeneity by means of meta-regression analy-
that are inferred in job titles and industry categories is non- ses. One caveat is that the results of meta-regression analyses
differential but often substantial, as it may reduce and obscure should be interpreted cautiously39 in that the associations
true relevant associations. Consistent with this hypothesis we found should be considered hypothesis-generating rather than
were able to show that studies directly assessing individual regarded as a proof of causality.40
occupational UV light exposure found stronger associations As discussed above, some of the individual studies included
between occupational UV light exposure and SCC occurrence in this review have methodological limitations, most of
than studies estimating occupational UV light exposure solely which, however, cannot explain the observed association
from the job description or job title. In addition to these limi- between occupational solar UV light exposure and SCC occur-
tations of individual studies that led to an underestimation of rence, but most likely introduced bias towards the null effect.
the true association between work-related UV light exposure Our findings are of the utmost relevance for occupational
and SCC risk, other sources of bias that may have led to an medicine and public health, as substantial proportions of the
overestimation of the true association were also identified in working populations are regularly exposed to solar UV light at
some of the studies included in the review: recall bias was the workplace. Our findings may serve as a robust and reliable
considered to be unlikely in six of the 12 case–control studies. evidence base to implement targeted primary prevention strat-
Exclusion of potential outdoor workers among controls in the egies for outdoor workers. Ideally, such preventive strategies
study by Zanetti et al.37 may have introduced selection bias. should be designed involving all relevant stakeholders, i.e.
Selective outcomes reporting was suspected in the study by outdoor workers, occupational health experts, dermatologists,
Gafa et al.32 health educators and industry representatives.
Our study extends previous research by providing evidence In Germany, our findings have already induced a discussion
that the strength of the association between occupational UV about the introduction of a new occupational disease ‘SCC of
exposure and development of SCC increases with decreasing the skin among outdoor workers’, which is likely to result in
latitude. This is biologically plausible and provides one significant preventive measures for individuals with high levels
of work-related UV light exposure. Hopefully, similar public 12 Marks R, Jolley D, Dorevitch AP et al. The incidence of non-melan-
health and health policy impact will result in other countries. ocytic skin cancers in an Australian population: results of a five-
year prospective study. Med J Aust 1989; 150:475–8.
13 Leiter U, Garbe C. Epidemiology of melanoma and nonmelanoma
What’s already known about this topic? skin cancer – the role of sunlight. Adv Exp Med Biol 2008; 624:89–103.
14 Kenborg L, Jørgensen AD, Budtz-Jørgensen E et al. Occupational ex-
• Cutaneous squamous cell carcinoma (SCC) is among the posure to the sun and risk of skin and lip cancer among male
most prevalent malignancies worldwide with steadily in- wage earners in Denmark: a population-based case–control study.
creasing incidence rates over past decades. Cancer Causes Control 2010; 21:1347–55.
• Although UV light exposure in general is an established 15 Diepgen TL, Drexler H. [Skin cancer and occupational disease].
risk factor for the development of cutaneous SCC there Hautarzt 2004; 55:22–7.
16 Young C. Solar ultraviolet radiation and skin cancer. Occup Med
is an ongoing debate in the epidemiological and derma-
(Lond) 2009; 59:82–8.
tological literature about the relationship between work- 17 Ramirez CC, Federman DG, Kirsner RS. Skin cancer as an occupa-
related UV light exposure and the risk of cutaneous SCC. tional disease: the effect of ultraviolet and other forms of radiation.
Int J Dermatol 2005; 44:95–100.
18 Blome O, Diepgen TL. Hautkrebs durch UV-Licht – eine neue Ber-
What does this study add? ufskrankheit Nr. 5103 der BKV? Teil 1: die rechtlichen Voraussetz-
ungen für die Aufnahme einer Erkrankung in die BK-Liste und der
• This systematic appraisal of the epidemiological litera- Anerkennung nach § 9 Absatz 2 SGB VII. Dermatol Beruf und Umwelt
ture and meta-analysis clearly indicates that occupational 2007; 55:167–76.
UV light exposure is a substantial and robust risk factor 19 Stroup DF, Berlin JA, Morton SC et al. Meta-analysis of observa-
for the development of cutaneous SCC. tional studies in epidemiology: a proposal for reporting. Meta-anal-
ysis of Observational Studies in Epidemiology (MOOSE) group.
• These findings are of significant public health impact as
JAMA 2000; 283:2008–12.
they highlight the need for preventive measures for 20 Kauppinen T, Toikkanen J, Pedersen D et al. Occupational exposure
individuals with high levels of work-related UV light ex- to carcinogens in the European Union. Occup Environ Med 2000;
posure. 57:10–18.
21 Seidler A, Husmann G, Nübling M et al. UV-exponierte Berufe und
Hauttumoren: berufsbezogene Auswertung von Daten des Krebs-
registers Rheinland-Pfalz. Zbl Arbeitsmed 2006; 56:78–90.
22 Szklo M, Nieto J. Epidemiology Beyond the Basics. Sidbury, MA: Jones
References and Barlett Publishers, 2005.
1 Madan V, Lear JT, Szeimies RM. Non-melanoma skin cancer. Lancet 23 Saladi RN, Persaud AN. The causes of skin cancer: a comprehensive
2010; 375:673–85. review. Drugs Today (Barc) 2005; 41:37–53.
2 Glass AG, Hoover RN. The emerging epidemic of melanoma and 24 Wells GA, Shea B, O’Connell D et al. The Newcastle–Ottawa scale
squamous cell skin cancer. JAMA 1989; 262:2097–100. (NOS) for assessing the quality of nonrandomised studies in meta-
3 Gray DT, Suman VJ, Su WP et al. Trends in the population-based analyses. Ottawa Hospital Research Institute, 2010. Available at
incidence of squamous cell carcinoma of the skin first diagnosed http://www.ohri.ca/programs/clinical_epidemiology/oxford.htm
between 1984 and 1992. Arch Dermatol 1997; 133:735–40. (last accessed 24 October 2010).
4 Alam M, Ratner D. Cutaneous squamous-cell carcinoma. N Engl J 25 Egger M, Davey Smith G, Altman DG. Systematic Reviews in Health Care.
Med 2001; 344:975–83. Meta-Analysis in Context, 3rd edn. London: BMJ Publishing Group,
5 Armstrong BK, Kricker A. The epidemiology of UV induced skin 2003.
cancer. J Photochem Photobiol B 2001; 63:8–18. 26 Macaskill P, Walter SD, Irwig L. A comparison of methods to
6 Chuang TY, Popescu NA, Su WP et al. Squamous cell carcinoma. A detect publication bias in meta-analysis. Stat Med 2001; 20:641–54.
population-based incidence study in Rochester, Minn. Arch Dermatol 27 Green A, Battistutta D, Hart V et al. Skin cancer in a subtropical
1990; 126:185–8. Australian population: incidence and lack of association with occu-
7 Chuang TY, Reizner GT, Elpern DJ et al. Squamous cell carcinoma pation. The Nambour Study Group. Am J Epidemiol 1996;
in Kauai, Hawaii. Int J Dermatol 1995; 34:393–7. 144:1034–40.
8 Aubry F, MacGibbon B. Risk factors of squamous cell carcinoma of 28 Pukkala E, Saarni H. Cancer incidence among Finnish seafarers,
the skin. A case–control study in the Montreal region. Cancer 1985; 1967–92. Cancer Causes Control 1996; 7:231–9.
55:907–11. 29 Adami J, Gridley G, Nyren O et al. Sunlight and non-Hodgkin’s
9 Perea-Milla López E, Miñarro-Del Moral RM, Martı́nez-Garcı́a C lymphoma: a population-based cohort study in Sweden. Int J Cancer
et al. Lifestyles, environmental and phenotypic factors associated 1999; 80:641–5.
with lip cancer: a case–control study in southern Spain. Br J Cancer 30 Radespiel-Tröger M, Meyer M, Pfahlberg A et al. Outdoor work and
2003; 88:1702–7. skin cancer incidence: a registry-based study in Bavaria. Int Arch
10 Katalinic A, Kunze U, Schafer T. Epidemiology of cutaneous mela- Occup Environ Health 2009; 82:357–63.
noma and non-melanoma skin cancer in Schleswig-Holstein, Ger- 31 Hogan DJ, Lane PR, Gran L et al. Risk factors for squamous cell
many: incidence, clinical subtypes, tumour stages and localization carcinoma of the skin in Saskatchewan, Canada. J Dermatol Sci 1990;
(epidemiology of skin cancer). Br J Dermatol 2003; 149:1200–6. 1:97–101.
11 Hakansson N, Floderus B, Gustavsson P et al. Occupational sunlight 32 Gafa L, Filippazzo MG, Tumino R et al. Risk factors of non-
exposure and cancer incidence among Swedish construction work- melanoma skin cancer in Ragusa, Sicily: a case–control study. Cancer
ers. Epidemiology 2001; 12:552–7. Causes Control 1991; 2:395–9.
33 Gallagher RP, Hill GB, Bajdik CD et al. Sunlight exposure, pigmen- 37 Zanetti R, Rosso S, Martinez C et al. Comparison of risk patterns in
tation factors, and risk of nonmelanocytic skin cancer II. Squamous carcinoma and melanoma of the skin in men: a multi-centre case-
cell carcinoma. Arch Dermatol 1995; 131:164–9. case–control study. Br J Cancer 2006; 94:743–51.
34 Rosso S, Zanetti R, Martinez C et al. The multicentre South 38 Marehbian J, Colt JS, Baris D et al. Occupation and keratinocyte
European study ‘Helios’. II: Different sun exposure patterns in the cancer risk: a population-based case–control study. Cancer Causes Con-
aetiology of basal cell and squamous cell carcinomas of the skin. Br trol 2007; 18:895–908.
J Cancer 1996; 73:1447–54. 39 Davey SG, Egger M, Phillips AN. Meta-analysis. Beyond the grand
35 Masini C, Fuchs PG, Gabrielli F et al. Evidence for the association of mean? BMJ 1997; 315:1610–14.
human papillomavirus infection and cutaneous squamous cell car- 40 Baker WL, White CM, Cappelleri JC et al. Understanding heteroge-
cinoma in immunocompetent individuals. Arch Dermatol 2003; neity in meta-analysis: the role of meta-regression. Int J Clin Pract
139:890–4. 2009; 63:1426–34.
36 Mitropoulos P, Norman R. Occupational nonsolar risk factors of 41 Thompson SG, Sharp SJ. Explaining heterogeneity in meta-analysis:
squamous cell carcinoma of the skin: a population-based case-con- a comparison of methods. Stat Med 1999; 18:2693–708.
trolled study. Dermatol Online J 2005; 11:5.