Occupational nonsolar risk factors of squamous cell carcinoma of the skin: A population-based case-controlled study
Panagiotis Mitropoulos, Robert Norman DO²
Dermatology Online Journal 11 (2): 5

Introduction

Importance of investigating skin squamous cell carcinoma

There is a widely held perception that skin cancer other than melanoma is a relatively trivial condition. Even though the mortality rate from nonmelanoma skin cancer compared to other cancers is relatively low, it still represents an important source of morbidity, cosmetic damage, and health-care expenditure [1, 2]. The incidence of nonmelanoma skin cancer appears to be increasing with time, and the disease may represent a major health and economic problem in the United States and other parts of the world [2, 3, 4].

The American Cancer Society estimates more than one million new nonmelanoma skin cancer cases annually in the United States, one-fifth of which are squamous cell carcinoma (SCC) [5]. Skin SCC is the second most frequent neoplastic disease in Caucasians [6]. Southeast Arizona has the highest reported skin SCC incidence in the United States, and among the highest in the world [7].

Known environmental risk factors for SCC of the skin are ultraviolet radiation from sunlight exposure; ionizing radiation; arsenic; and the products arising from the combustion and distillation of coal, and petroleum [8]. These confirmed environmental hazards, however, do not account for all of the reported skin SCC cases [9].

Epidemiological studies suggest that individuals with SCC of the skin are more likely to develop other malignancies compared to individuals with no history of nonmelanoma tumors [10, 11, 12]. Patients with previous nonmelanoma tumors face more than three times the risk of a subsequent malignant melanoma, a more severe type of skin cancer [11]. Additionally, cancer registries have reported for patients who have a history of nonmelanoma skin cancer, there is an increased incidence and mortality from leukemia, non-Hodgkin lymphoma, and cancers of the lung, bladder, breast, testis, salivary glands, small intestine, and pharynx [10, 11]. According to several recent epidemiological studies, a history of nonmelanoma skin cancer may be responsible for raising the long-term risk of death from another type of cancer by 20 to 30 percent [10, 11].

Furthermore, the behavior of skin SCC somewhat differs from that of the other major nonmelanoma skin cancer, basal cell skin carcinoma, in that the tumor is capable not only of local invasion but are also associated with a substantial risk of metastasis [13].

Because it is a preventable disease, it is essential to identify all the factors that may lead to development of skin SCC so that protection measures against exposure to those factors may be taken. A more comprehensive understanding of squamous cell cancer biology, disease, and progression could lead to better prevention methods.

Importance of nonsolar risk factors of skin squamous cell carcinoma

The role of sunlight in the etiology of skin SCC has been established, however, there is a growing literature that identifies other factors related to increased risk for SCC of the skin [14]. Nonsolar risks include occupational and environmental exposure to a variety of chemical agents (which may act as complete carcinogens, initiators, or promoters), and physical conditions. Table 1 lists some of the known chemical and environmental skin SCC risk agents.

Although it is possible to separate sunlight from nonsunlight risks for workplace-related skin SCC, most studies have not investigated this interrelationship. In addition, the majority of data that relate certain chemicals to skin SCC are based on animal models. Epidemiological studies that include humans are few, and have been mainly focused on malignant melanoma rather than nonmelanoma skin cancers.

Skin squamous cell carcinoma and occupation

Although sunlight is the major environmental risk factor for skin carcinomas, the relationship of skin cancer with sun exposure is not straightforward, nor is risk consistently higher among persons with outdoor occupations [6]. Occupational factors such as employment in chemical-related industries certainly contribute some fraction to the total reported skin carcinoma cases [6]. Although most studies were designed to investigate malignant melanoma, some have focused on nonmelanoma skin cancers, indicating an association between occupation, work environment, and the risk of carcinoma [15, 16, 17].

Several findings indicate a higher risk of skin cancers among workers in basic chemicals production and the printing industry and among professional, technical, and white-collar workers [6]. Furthermore, associations have been observed for men employed in the brewery and malt-processing industry, and in shoe fabrication from leather and skins [6].

The role of non-sunlight related risk factors for squamous cell carcinoma (SCC) and basal cell carcinoma (BCC) was investigated in a population-based, case-control study among males in Alberta [9]. Elevated risks of skin SCC were seen in individuals exposed to insecticides (odd ratio (OR) = 2.8; 95 % confidence interval (CI) = 1.4-5.6), herbicides (OR = 3.9; 95 % CI = 2.2-6.9), and fungicides and seed treatments (OR = 2.4; 95 % CI = 1.4-4.0). The authors considered exposures related to farming activities. Additionally, similar results were reported for exposure to petroleum products and grease [9].

Another study from Sweden and Norway investigated several types of cancers, including nonmelanoma skin cancer [17]. An increased risk for nonmelanoma skin cancer was observed (relative risk (RR) = 2.37; 95 % CI = 1.08-4.50) for creosote-exposed workers, particularly after a latency period of 20 years or more. The study as many other studies, did not distinguish between squamous cell and basal cell carcinoma [17].

In Denmark, an investigation of the potential carcinogenic effect of exposure to pharmaceuticals and other chemicals revealed a 1.5 fold (95 % CI = 1.1-2.1) elevated risk of nonmelanoma skin cancer for pharmacy technicians, and long-term pharmacy assistants (RR = 2.8; 95 % CI = 1.6-4.6).16

Methods

Research objectives

• Investigate potential associations between specific occupations and risk of skin SCC in a case-control study.
• Examine history of specific environmental exposures and determine their potential risk for SCC of the skin.

To meet these research objectives an analysis of a completed population case-control study was done. Data were received from the Southeastern Arizona Health Study that focused on skin SCC cases and randomly selected, population-based controls.

Hypotheses

1. Construction workers, farmers, individuals working with specific chemicals, automobiles or other machines have an elevated risk for developing skin SCC because of the potential exposure to carcinogenic agents involved in their occupation.

2. Repeated or prolonged exposure to certain nonsolar agents (chemical or physical) increases the risk for developing SCC of the skin.

Specific agents investigated

The decision to evaluate the nonsolar agents in this current analysis was based on 1) reports of a relationship to skin SCC risk from literature, and 2) availability of data through the SEAHS-2 case-control study. Table 2 lists these agents that were identified during the case-control study interviews.

Methods

The Southeastern Arizona Health Study-2

Data from the Southeastern Arizona Health Study-2 (SEAHS-2) were used to evaluate the relationship of the nonsolar occupational risks to SCC of the skin. The study was conducted between December 1992 and December 1996 in Southeastern Arizona to primarily assess the risk of SSC of the skin in relation to sun exposure.

The Southeast Arizona Skin Cancer Registry was used to identify the cases. All cases were residents of the Tucson Arizona metropolitan area who had their first pathologically confirmed diagnosis of nonmetastatic skin SCC within four months of ascertainment. Study participation required being at least 30 years of age with no history of skin or nonskin carcinoma. Furthermore, only non-Hispanic and Hispanic Caucasian cases were eligible for inclusion.

Operationally, participants were defined as being Caucasian when parents and grandparents were not born in any Central- or South American country, the Caribbean, Spain, or Portugal. Anyone reporting at least one non-Caucasian parent was defined as non-Caucasian. The exclusion of African Americans, Asian Americans, and other races of color reflect the low incidence and the different histological types of skin carcinomas found in these minority groups.

Upon identification and random selection of an eligible case, a letter was sent to the physician requesting permission to contact their patient regarding study participation. The local institutional review board required that physician consent be obtained. Once physician consent was obtained, a letter describing the study was mailed to potential case participants indicating that an initial contact would soon be made. A SEAHS-2 interviewer would telephone potential case participants to describe the study, determine eligibility, invite participation, and schedule a face-to-face interview. Personal telephone calls and letters and in-person contact with cases were designed to increase participation rates. From the initial 883 potential cases that were identified through the Registry, approximately half were eligible or agreed to participate in the study, bringing the final case sample to a total of 404 individuals.

Population-based controls were selected using a random digit dialing method. Controls were frequency matched to the cases by age category (± 5 years) and by gender. No prior history of skin or any other type of carcinoma was a requirement for eligibility as a control. Cases without a residential telephone were matched with controls without residential telephone living within the same block. From the 3,384 telephone numbers called using the random digit dialing method, a final control sample of 391 was obtained.

All subjects completed an extensive interview that sought information on demographics, personal and family medical history, history of sun exposure and exposure to other environmental and chemical agents, occupational history, history of smoking and alcohol consumption. A trained interviewer using a structured questionnaire conducted these interviews at the participants' homes or at a central clinic.

Occupation assessment

During the interview, the study participants were asked specific questions regarding their occupation. Participants were asked to identify their present occupation and most recent occupation. They were then asked whether the reported occupation was the one they had worked the longest in their lifetime and if not, to report their longest occupation. The interviewers then coded the occupation from a list of 585 potential professions.

Finally, the participants were asked whether the reported occupation was primarily an outdoor or indoor occupation and then to report the number of days per year and hours per day spent outdoors at the occupation.

From the list of the 585 potential professions, four specific occupational categories were created: CONSTRUCTION, FARMING, CHEMICALS (professions involving handling of chemicals), and AUTO/MACHINES. Individuals included in these professions were assumed to be at a high risk for developing skin SCC because of the potential exposure to a number of suspected or confirmed carcinogenic agents related to their work. Table 3 lists the specific occupations/professions included in the 'high risk' occupations.

The remaining individuals in the study were assumed to be at lower occupational risk for developing skin SCC, and are part of a group that is described as 'non-high risk' occupations. For the vast majority of the study participants the most recent occupation was also the longest occupation in their lifetime.

Exposure Assessment

During the interview, all participants were shown nine cards of potential chemical and environmental exposures. They were asked whether they had been exposed to any of the listed agents. The definition of exposure was: A DURATION OF AT LEAST 4 HOURS ON AN AVERAGE OF ONCE A WEEK FOR A MONTH OR LONGER, or A DURATION OF AT LEAST 4 HOURS DAILY FOR AT LEAST A WEEK SOLID. Information on occupation during the period of exposure, geographical location, and time period of the exposure was also inquired. The participants had the opportunity to report up to 8 general exposures and subtypes of exposures. The general and specific exposures are listed on Table 4.

Other variables

In addition to occupational history and general and specific exposures, study interviewers asked about skin characteristics, such as ability of skin to tan, skin color prior to sun exposure, amount of freckling, as well as information on family history of cancer, skin diseases, and other chronic diseases. Questions on tanning history involved information on year-round presence of "tan lines", and detailed skin reaction to short-term and prolonged sun exposure. Information on exposure to ultraviolet radiation was sought, as well as use and frequency of use of sunscreens, sunlamps, sunbeds, and tanning booths. Moreover, data on use of any medication, medical treatment involving x-rays, and biopsy history were gathered.

A section of the questionnaire sought information on smoking and alcohol drinking habits. The participants were asked to report on the frequency and amount of cigarettes and drinks consumed on a daily and yearly basis.

Statistical analysis

Intercooled Stata 6.0 software package was used for all statistical analyses. Epi Info 6.0 was used for sample size and power calculations. Descriptive statistics were used to identify outliers and determine normal distribution of the variables. New variables and interaction terms were created and always double-checked for errors in coding.

Multiple logistic regression models were used to assess the risk of skin SCC in relation to occupation and exposures to chemicals and other environmental agents by estimating the odd ratios and 95 percent confidence intervals. Prior analysis of data from the Southeast Arizona Health Study-2 determined several predictors of skin cancer, which could be potential confounding factors. The impact of these factors was evaluated in the modeling. Potential factors included a) history of actinic keratosis, b) current number of freckles on the arm, and c) reaction of the skin to prolonged sun exposure. Moreover, age and gender were considered as potential confounders controlled by inclusion of these variables in the multivariate analysis.

Additional multivariate logistic models were created that included variables such as smoking, outdoor or indoor occupation, and interactions between variables. Since some epidemiological studies have identified smoking as a predictor for skin SCC, its potential confounding effect was evaluated. Furthermore, in order to account for sun and ultraviolet-radiation exposure, the outdoor or indoor occupation variable was considered. The potential interactions between exposures were evaluated because some studies report a synergistic effect between certain chemical and environmental agents. Interaction variables between construction materials, physical agents, and chemical agents were created and included in the modeling for evaluation.

Finally, statistical analyses were repeated to include only male study participants. Since males reported most of the examined occupations and exposures, this would result in higher power compared to the analyses that included both male and female subjects.

Phases of analysis

The statistical analysis was divided into three phases. To target the first research objective (assessing skin SCC risk vs. occupation), four logistic regression models were developed to estimate crude and adjusted odds ratios (OR) with the corresponding 95 percent confidence intervals (CI) for each major high risk occupational group.

In Phase 2, six multivariate logistic models were developed to assess skin SCC risk for six general exposure categories, i) nonsolar LIGHT (excluded STRONG SUNLIGHT subcategory under INTENSE LIGHT), ii) RADIATION, iii) PAINT/SOLVENTS GENERAL combined with the full list of SOLVENTS, iv) INCECTIDES/PESTICIDES, v) PLASTICS, and vi) CONSTRUCTION AND MACHINERY MATERIALS.

Due to the broadness of the general exposure categories, Phase 3 of analysis focused on exposure to specific agents. This step of analysis included 18 multivariate logistic models. Again, crude and adjusted ORs along with the 95 percent CIs were calculated.

Results

Study population characteristics

Between December 1992 and December 1996, a total of 795 subjects were interviewed, from which 404 were skin SCC cases and 395 were controls. Approximately 60 percent of the study participants were male. Age ranged from 31 to 91 years with subjects being predominantly highly educated (97 % high school and above) Caucasians (97.7 %). Descriptive analysis of the following variables revealed no major differences between cases and controls.

There were a total of 184 different occupations reported. However, only 90 subjects reported a high risk occupation. Males predominantly occupied those professions categorized as high risk. The sex distribution in the non-high risk group was more balanced with 56.15 percent male and 43.85 percent female.

Phase 1 of statistical analysis (occupation)

Because age and gender were design variables, age and gender adjusted odd ratio (OR-1) was considered to be the crude odd ratio. Tables 5a and 5b report the crude and 95% confidence intervals (CI) for the association between occupational categories and skin SCC. In the same table, OR-2 is also reported, where history of actinic keratosis, current number of arm freckles, and reaction of the skin to prolonged sun exposure are included in the model.

On a first look, there seems to be no difference in the risk for skin SCC when comparing the combined high risk occupational groups to the non-high risk occupations. When looking at specific high risk occupational groups, however, there is an indication of an elevated risk of skin SCC for construction workers and machinists. Skin SCC cases are 38 percent more likely to report construction work than controls (OR = 1.38), and 21 percent more likely to report an occupation related to automobile or machine work (OR = 1.21). There was a slightly decreased risk for farmers (OR = 0.87) and chemical-related workers (OR = 0.42). However, none of the results were statistically significant. The 95 percent CIs are large and they include the null value of 1.

Phase 2 of statistical analysis (general exposures)

Phase 2 of the analysis focused on assessing whether reported exposures to six general categories posed a risk for SCC of the skin. The differences in sex distribution for the general exposure categories were still substantial, as in every category males reported more exposure than females by a large margin.

To determine whether the exposures were occurring in occupational groupings, the general exposures were contrasted to occupational groups. Only few individuals in the high risk occupations reported exposure to the general exposures investigated. This would indicate that many of the exposures examined were not occurring in the occupation identified as the most recent or longest.

Table 6 reports the association between skin SCC and the general chemical and physical exposure categories. There were no statistically significant associations between any general exposures and SCC of the skin. A slightly increased risk for skin SCC was noted for exposure to nonsolar light (OR = 1.44), and construction/machinery materials (OR = 1.18). The 95 percent CIs, however, were large and included the null value.

Phase 3 of statistical analysis (specific exposures)

Phase 3 explored the relationship between specific exposures and skin SCC. One of the problems encountered was that few people reported subtype exposures. Most of the exposures were infrequent. For example, only 5 individuals reported prolonged exposure to arsenic, 4 to each ethylene glycol and projectors/enlargers, 7 individuals reported formaldehyde, and turpentine, and 9 reported DTT exposure.

Table 7 shows the strength of the association between specific exposures and SCC of the skin. Although, none were statistically significant, a highly increased skin SCC risk for arsenic and ethylene glycol exposure is noted. Cases were 4.2 (95 % CI = 0.40-43.9) times more likely to report exposure to arsenic than were controls, and 8.5 (95 % CI = 0.77-92.9) times more likely to report exposure to ethylene glycol.

Exposure to fluorescent light (OR = 1.56, 95 % CI = 0.92-2.61) and cement dust (OR = 1.81, 95 % CI = 0.90-3.62) were both associated with a marginally statistically significant increased risk for skin SCC. Carbon tetrachloride exposure was found to be potentially protective against SCC of the skin (OR = 0.36, 95 % CI = 0.14-0.95).

Discussion

The purpose of the analysis was to investigate the effect of occupation and environmental exposures on skin SCC. However, the rate of exposure to agents and high risk occupations was low in the population studied. The SEAHS-2 data on occupation were not very efficient since a very small study sample was produced. No statistically significant difference in the risk for skin SCC was found when the potentially high risk occupational groups were combined and compared to the non-high risk occupations (OR = 0.99, 95 % CI = 0.58-1.67). For specific high risk occupational groups, there was an arguable elevated risk of skin SCC for construction workers (OR = 1.38, 95 % CI = 0.61-3.14), and automobile and machine workers (OR = 1.21, 95 % CI 0.48-3.06). These results support the initial hypothesis for increased risk of SCC of the skin associated with these professions, but cannot be considered reliable. Power calculations suggest that the sample size reported for construction workers and automobile abd machine workers had only 16- and 7.5 percent power, respectively, to identify the association of the magnitude shown.

The interview questionnaire did not adequately explore detailed occupational information from the participants. The study sought only information on the most recent and longest occupations. That was insufficient to gain a more complete picture of occupational exposure. Furthermore, individuals in the non-high risk occupational group reported exposure to agents that were associated with the high risk occupational groups. Such discrepancies could have been avoided had the SEAHS-2 included a questionnaire pertaining to the life-long occupation history of its participants.

The statistical analysis of Phase 2 (skin SCC risk vs. general exposures) and Phase 3 (skin SCC risk vs. specific exposures) was also problematic. Again few individuals in the high risk occupational groups reported exposure to any of the general and specific agents. The highly educated, predominantly white-collar population of the SEAHS-2 study was most likely the reason for the low exposures.

A borderline significant increased skin SCC risk was noted for exposure to nonsolar light (OR = 1.44, 95 % CI = 0.92-2.26), and construction and machinery materials (OR = 1.12, 95 % CI = 0.76-1.84). More specifically, there was a borderline elevated skin SCC risk from prolonged exposure to fluorescent light (OR = 1.56, 95 % CI = 0.92-2.61) and cement dust (OR = 1.81, 95 % CI = 0.90-3.62). The suggested protective effect of carbon tetrachloride exposure (OR = 0.36, 95 % CI = 0.14-0.95) might be an artifact from the multiple variables modeling since no plausible reasoning for such an association can be offered. On the other hand, the large odds ratios for arsenic (OR = 4.21) and ethylene glycol (OR = 8.4), although statistically nonsignificant are in accordance with literature. Arsenic has been strongly associated with an increased risk of skin SCC in humans, and studies have confirmed ethylene as a carcinogen [6, 9, 20, 36, 47].

Male study participants reported more of the occupational and environmental exposures. The analyses were repeated to include only males. Such analysis could have detected potential effect modification, however, no significant differences were indicated.

While repeating the analyses with just the male participants did increase the power of the calculations, none of the comparisons were statistically significant.

The adjusted odds ratios included history of actinic keratosis (AK), current number of freckles on arm, and skin reaction to prolonged sun exposure. While these variables were predictive of skin SCC risk, there was a concern of over-adjustment. Actinic keratosis is the precancerous stage of skin SCC. Because AK is part of the biological pathway to developing skin SCC, it is potentially problematic to adjust for it [29]. However, after repeating the multivariate logistic analysis by excluding the history of AK variable, no significant differences were found. Analyses taking into consideration potential interactions between exposures produced no statistically significant changes. Likewise, inclusion of smoking history as a confounder did not change the results, nor did inclusion of whether the reported occupations were primarily indoor or outdoor occupations.

In summary, the etiology of skin SCC is multifactorial. Environmental, occupational, and life-style factors interact in a complex way in the development of SCC of the skin. Despite the relatively benign nature of nonmelanoma skin cancers, rising incidence will result in increased medical costs, morbidity, and mortality. The key to reducing the number of skin SCC cases is through prevention. Since ultraviolet radiation exposure is responsible for most of the cases, reducing exposure to sunlight can help prevent new cases. Nevertheless, a significant number of additional skin SCC risk factors exists, and to better understand the complex etiology of SCC of the skin and nonmelanoma cancers in general, further population-based research is needed with a bigger sample size studies and a larger range of exposures.

See also literature review

© 2005 Dermatology Online Journal