Risk Factors for Basal Cell Carcinoma in a Mediterranean Population: Role of Recreational Sun Exposure Early in Life

Objective  To investigate the role of pigmentary traits, different patterns of sun exposure, artificial sources of UV radiation, and lifestyle-related factors on the risk of basal cell carcinoma (BCC) in a Mediterranean population from central-southern Italy.

Design  Hospital-based case-control study.

Setting  A referral dermatological hospital in Rome, Italy.

Patients  A convenience sample of 166 case patients with histologically confirmed BCC and 158 cancer-free control subjects with minor dermatological conditions observed between March 1995 and June 1997.

Results  In the multivariate analysis, the mean number of weeks per year spent at the beach before the age of 20 years was significantly associated with BCC. A dose-response trend was found for subjects who had spent 3 to 4 (odds ratio, 1.8; 95% confidence interval, 0.8-4.4), 5 to 8 (odds ratio, 3.7; 95% confidence interval, 1.5-9.0), or more than 8 (odds ratio, 4.5; 95% confidence interval, 1.9-10.5) weeks per year at the beach (P = .01 for trend). There was a significant association with the presence of actinic keratoses or solar lentigines, whereas no effect was found for skin type, history of sunburns, exposure to nonsolar UV radiation, and lifestyle-related habits such as cigarette smoking, alcohol consumption, and coffee drinking. Subjects reporting a family history of skin cancer had an extremely increased risk of BCC.

Conclusion  The definite association with recreational sun exposure during childhood and adolescence and the strong relation with family history of skin cancer suggest that genetic predisposition and peculiar exposure patterns to UV radiation are key independent risk factors for the development of BCC in a southern European population.

BASAL CELL carcinoma (BCC) is the most common cancer in white populations. Approximately 1 million cases of basal and squamous cell skin cancers were expected to be diagnosed in 1999 in the United States.¹ Although mortality from BCC is low, due to an extremely low metastatizing potential (estimated around 0.1%),² this tumor carries considerable morbidity and costs for health care services.³

The etiology of BCC is still unclear, although UV radiation seems to play a critical role in tumor development and can be considered the primary established risk factor. Cumulative sun exposure has been implicated as the major causative factor in squamous cell carcinoma development,⁴ with risks increasing exponentially above a threshold value.⁵ By contrast, the relation between UV radiation and BCC is complex and still highly controversial, in regard to patterns of sun exposure and exposures in different periods in life that are supposed to be critical for tumor development. Some recent studies^5-9 have suggested that intense sun exposures during childhood and adolescence are relevant risk factors for BCC. Pigmentary characteristics, such as light hair and eye color, tendency to sunburn, and poor tanning ability, have all been associated with a higher risk of BCC. However, the relatively few analytical studies^7,10-13 conducted so far to investigate constitutional and environmental risk factors for BCC have not elucidated the relative importance of these factors, and their results are often inconsistent, due to major differences in study design, assessment of exposure, and data analysis.

In 1995, we undertook a case-control study to investigate the role of several known or potential risk factors on BCC occurrence in a Mediterranean population from central-southern Italy, including host-related factors, different patterns and timing of sun exposure, artificial sources of UV radiation, and lifestyle-related factors.

The study was conducted between March 21, 1995, and June 10, 1997, at the Istituto Dermopatico dell'Immacolata, a large hospital for skin diseases in Rome, Italy, operating as a referral dermatological center for southern and central Italy. Cases were patients attending the dermatological surgery clinic on a random day each week who had histologically confirmed BCC. Controls were selected on the same days among patients attending the dermatological clinic for minor skin conditions, such as intradermal or melanocytic nevi, common warts, androgenetic alopecia, and seborrheic dermatitis. The exclusion criteria for controls were a history of skin cancer, a history of skin diseases for which UV radiation therapy or heliotherapy was indicated, or a history of skin diseases for which sun avoidance was recommended.

All patients gave their informed consent to complete a standardized questionnaire.

Using a standardized questionnaire administered by a trained interviewer, detailed information was collected on demographic factors; personal and family history of any cancer, including skin cancer; skin reaction to sunlight; history of sunburn before and after the age of 20 years; professional and recreational exposure to sunlight; exposure to nonsolar sources of UV radiation; cigarette smoking; and alcohol, coffee, and tea consumption. Lifetime exposure to nonsolar UV radiation sources was explored by asking subjects information on the type of lamps used at home to light rooms in which they spent most of the time. Exposure to fluorescent or halogen lamps at the workplace was assessed by asking subjects to list every indoor job held for 6 months or longer and its duration and asking them to indicate what type of illumination was used. The frequency of use of sun beds and/or sunlamps for either therapeutic or cosmetic purposes was also assessed.

A complete skin examination was performed by a dermatologist to assess pigmentary characteristics and the presence of sunlight-induced skin lesions. Hair color was determined by asking the subjects about their natural hair color at the age of 20 years. Fitzpatrick clinical skin typing was used, with type I to IV indicating decreasing sensitivity to sunlight.

Family history of skin cancer was defined as positive if the subject reported any blood relative who had had BCC only or BCC plus other types of skin cancer.

Data analysis was performed using the Statistical Product and Service Solutions for Windows, release 8.0,¹⁴ and STATA for Windows, release 4.0,¹⁵ statistical packages.

Age- and sex-adjusted odds ratios (ORs) and 95% confidence intervals were calculated for known host risk factors (light hair and eye color, fair skin complexion, skin phototype, sunlight-related skin lesions, and family history of skin cancer), occupational and recreational sun exposure, exposure to nonsolar UV radiation, history of sunburn, and lifestyle-related habits (cigarette, alcohol, coffee, and tea consumption), using unconditional logistic regression models.

Logistic regression models were built to allow for multiple adjustments of confounding and, thus, to identify variables independently associated with BCC occurrence. In all models, age was considered a continuous variable.

For categorical variables, such as skin type, point estimates were first computed for each category. Adjacent categories for which the calculated ORs were similar were collapsed into a single category whenever possible, to keep the model as parsimonious as possible. Sun exposure variables and lifestyle-related variables, which were measured on a continuous scale because of the skewness of their distributions, were categorized by dichotomizing or dividing into thirds the distributions of exposed controls, then adjacent categories were collapsed, where appropriate.

For each variable, the reference category was the most favorable level of exposure (eg, youngest age and darkest hair and eye color). The statistical significance of a linear dose-response trend for adjusted data was assessed by the Wald statistic.

A total of 166 case patients with histologically confirmed BCC and 158 control subjects with minor dermatological conditions were enrolled in the study. The demographic characteristics of the study population are shown in Table 1.

Of the BCCs, 71% were localized on the head or neck region, 20% on the trunk, 5% on the limbs, and 4% on an unspecified site. Of the 166 case patients with BCC, 21 (13%) reported a previous BCC. The mean time elapsed between the first BCC and the lesion removed at the outset of the study was 5.3 years (range, 1-12 years).

The controls had intradermal or melanocytic nevi (34%), seborrheic keratosis (10%), viral or fungal infections (10%), androgenetic alopecia (8%), subacute eczema (18%), angioma or dermatofibroma (7%), or other miscellaneous diagnoses (13%).

To exclude a potential bias due to the inclusion in the control group of subjects with melanocytic nevi, for which evidence indicates a relationship with excessive sun exposure, the nevus counts on the arms of controls referred to our clinic for intradermal or melanocytic nevi were compared with those of controls with other skin conditions (this information was available as the same control set was used in a melanoma case-control study). The mean number of melanocytic nevi was 4.3 (SD, 3.1) in patients with nevi and 4.6 (SD, 4.4) in patients with other control diagnoses. Thus, including these patients in the control group is unlikely to have affected the estimates of the ORs associated with sun exposure variables.

The age-adjusted and multivariate ORs for host-related factors are shown in Table 2. The relationship between hair color and eye color was preliminarily explored, to avoid including in a multivariate model potentially interrelated variables. The 2 variables are associated with BCC, with crude ORs of 1.5 for light hair color and 1.8 for light eye color. As expected, a strong association exists also between the 2 variables, and the magnitude of this association is far greater than that between each variable and the outcome (being a BCC case). This suggests the presence of an interaction between, rather than a confounding effect of, the 2 variables.

To take into account the presence of an interaction, a new secondary variable was built, including the 4 possible combinations of the variables hair color and eye color, and grouped into 3 categories: dark hair and eye color, light hair and eye color, and light hair or eye color.

Compared with subjects with dark brown or black hair and eyes, those with lighter hair (light brown and blond or red) and eye (hazel and green or blue) color had an increased age- and sex-adjusted risk of BCC. However, after multiple adjustment, this association was attenuated and lost the statistical significance (Table 2).

A relationship was observed also between freckles on the face and on the arms and between solar lentigines on the face and on the arms. To avoid colinearity problems, only freckles on the face and solar lentigines on the arms were retained in the logistic regression model.

No relation was observed with skin type. Among the variables indicating skin damage by sunlight, that result from a combination of constitutional factors and excessive sun exposure, an increased risk of BCC was found for the presence of solar lentigines on the arms and an even stronger association was observed for the presence of actinic keratoses, after multivariate adjusting for age, sex, pigmentary traits, and family history of skin cancer.

A positive family history of skin cancer was the strongest predictor of BCC development, with an age- and sex-adjusted OR of 21.8. The strength of the association was only slightly reduced by multiple adjustment for constitutional risk factors (Table 2) and was not modified by including in the model sun exposure variables (data not shown).

Sixty-four percent of the cases and 55% of the controls had a positive family history of other types of cancer. The age-adjusted OR of BCC for subjects reporting a family history of other cancers was 1.02 (95% confidence interval, 0.6-1.6).

Table 3 shows the age- and sex-adjusted and multivariate-adjusted ORs of BCC for sun exposure variables. Persons who reported outdoor work of longer than 8 years had an age- and sex-adjusted OR of developing BCC of 1.9, which was slightly modified by multiple adjustment, although it lost statistical significance.

Compared with subjects who before the age of 20 had spent less than 3 weeks per year of the summer holidays at the beach, a significant trend toward an increasing multivariate-adjusted risk of BCC was found for those who had spent 3 to 4, or 5 to 8, or more than 8 weeks per year (P = .01 for trend).

No positive association was found with the number of weeks per year spent at the beach after the age of 20 years either in the age- and sex-adjusted analysis or in the multivariate analysis. The possibility of a colinearity problem in the logistic regression model was ruled out by examining the correlation coefficient between the 2 covariates (weeks at the beach before and after the age of 20 years). An r = 0.27 was calculated, indicating a weak correlation, thus unlikely to cause a colinearity phenomenon in the multiple logistic analysis.

A reported history of sunburn before or after the age of 20 years seemed to have no effect on the risk of BCC development. The frequent use of sunscreens showed a tendency to have a protective effect, although not statistically significant, with an OR of 0.6 after adjusting for age, sex, pigmentary traits, outdoor work, recreational sun exposure before the age of 20 years, and family history of skin cancer.

In a subgroup of 152 subjects, including 64 cases and 88 controls, for whom information was available, the role of daily and weekend leisure time spent outdoors was explored, but no association was found, after taking into account age, sex, pigmentary traits, outdoor work, recreational sun exposure before the age of 20 years, and family history of skin cancer. The distribution by age, sex, and educational level of this subset of the study population was all similar to that of the whole study population.

The association between BCC risk and exposure to fluorescent or halogen lamp light at home and at the workplace was estimated by calculating age- and sex-adjusted ORs and confidence intervals (Table 4). A multiple logistic model was run that included pigmentary traits, occupational and recreational exposure to sunlight, and family history of skin cancer. The exposure to fluorescent lamps at the workplace seemed not associated with an increased risk of BCC. No evidence of a positive association with the use of sun beds or sunlamps was found or with exposure to halogen lamps at home and at the workplace.

No increased risk of BCC was observed for subjects who smoked more than 6500 packs of cigarettes in a lifetime, compared with those who had never smoked, after multiple adjustment (Table 5).

A tendency toward a protective effect of wine drinking, observed by comparing drinkers vs nondrinkers, was confirmed by analyzing the quantitative data on wine consumption. Compared with nondrinkers, subjects who had consumed more than 19 000 glasses of wine in a lifetime had an OR of BCC of 0.6, after multiple adjustments for constitutional factors and sun exposure variables. A similar tendency emerged for tea drinkers, compared with nondrinkers, whereas no clear effect was observed for coffee consumption.

The results of this study strongly suggest that recreational exposure to sunlight as a child or teenager may be important in accounting for the occurrence of BCC in adulthood. We observed a statistically significant dose-response trend in the risk of BCC with an increasing number of weeks per year spent sunbathing during childhood and adolescence. A nearly 5-fold increase of risk was calculated for an average summer holiday exposure of more than 8 weeks before the age of 20 years, after adjusting for the potential confounding effect of age, sex, pigmentary traits, working outdoors, and family history of BCC.

The presence of potential biases in the assessment of exposures should be kept in mind when considering these results. First, hospital-based controls might be less healthy in terms of skin conditions and, thus, at higher risk of skin cancer than the general population. This could lead to an underestimate of the role of some risk factors in this study population. Second, recall bias may arise in any study in which subjects are asked to recall events that occurred many years before, with a potential for exposure misclassification. This could well be true when assessing past sun exposure or exposures linked to lifestyle-related habits and diet. However, if an exposure misclassification occurs in a nondifferential way for cases and controls, the net effect is likely an underestimate of the strength of the association. At any rate, the strength of the association found in this study between recreational exposure to sunlight and BCC, and the existence of a dose-response effect, greatly reduces the possibility that a recall bias could have affected our estimates.

The association between recreational "intermittent" sun exposure and BCC has been only recently reported in several studies,^5-7,9 and, on this basis, it has been suggested that BCC might share some risk factors with cutaneous melanoma.

No positive association was found with the number of weeks per year spent at the beach after the age of 20 years in the age- and sex-adjusted and the multivariate-adjusted models. The possibility of a colinearity problem in the multivariate model was considered. Colinearity occurs when variables that are highly correlated are put together in a multiple logistic regression model. This might yield "unstable" estimates of coefficients, obscuring or reversing the direction of an effect seen by univariate analysis. This is clearly not the case in our data. In fact, the 2 covariates, weeks per year spent at the beach before and after the age of 20 years, are only weakly correlated (r = 0.27), and the multivariate-adjusted ORs and their 95% confidence intervals are virtually the same as those that are age- and sex-adjusted.

We did not find a significant association of BCC with a history of outdoor work, although a tendency toward an increased risk emerged for subjects who reported more than 8 years of occupational sun exposure.

A remarkably strong association emerged for family history of skin cancer, emphasizing the importance of genetic predisposition to BCC. A recall bias in measurements such as family history of BCC, if differential, due to enhanced recall in cases might cause an exposure misclassification and, therefore, an overestimate of the risk associated. However, it is unlikely for an estimated OR of 17.8 to be entirely explained by a recall bias. In addition, a familial tendency to develop BCC is well documented in several hereditary syndromes, including xeroderma pigmentosum and basal cell nevus syndrome, in which a defective DNA repair mechanism of UV radiation–induced damage and inactivating mutations of the PTCH gene (a gene involved in a signaling pathway required for correct embryonic development), respectively, underlie the clinical manifestations.

Pigmentary traits, including hair and eye color, freckling, skin type, and history of sunburns before and after the age of 20 years, did not seem to affect significantly the risk of BCC in this study population, with only a tendency toward an increase of risk for subjects with blue, green, or hazel eyes and face freckling. The results of most published studies^5,7-13,16-18 aimed at investigating the relative importance of phenotypic traits, sun sensitivity, and different indicators of sun exposure in the development of BCC are largely inconsistent. These differences may be accounted for by several factors. Many measures of skin pigmentation and skin sensitivity to sunlight are self-reported, have a low reproducibility,¹⁹ are subject to recall bias, and might, therefore, introduce a misclassification of cases and controls in risk categories. It is not clear, in fact, how, in reporting sun sensitivity, subjects weight their burning tendency against tanning ability, and whether they can effectively figure out what happens to their skin after exactly an hour of unprotected exposure the first time in the year.

We found a strong association between BCC and skin lesions, such as solar lentigines and actinic keratoses, that result from a combined effect of sun exposure and skin pigmentation characteristics. This relation of BCC with "objective" marks of prolonged sun exposure combined with skin pigmentation has been reported in most studies and adds further evidence to the fact that the inability to consistently demonstrate the role of skin sensitivity or skin characteristics in BCC occurrence may be accounted for by flaws in exposure measurement or by insufficient statistical power.

The only 2 large prospective studies exploring risk factors for BCC in a cohort of female nurses¹⁸ and in a cohort of male physicians⁹ in the United States showed that red or light hair color, light eye color, north European ancestry, tendency to sunburn, and the number of severe sunburns were all associated with an elevated risk of BCC.

Regarding the role of artificial UV radiation sources, this study did not provide any evidence of increased BCC risk due to exposure to fluorescent or halogen light at home and/or at the workplace or to sun beds or sunlamps. Similar results have been reported in the only published article dealing with this issue.²⁰ Paradoxically, we found a significant "protective" effect for the exposure to fluorescent lighting at the workplace. A likely explanation for this finding is that occupational exposure to fluorescent light implies an indoor workplace and an associated reduction in sun exposure. Alternatively, the absence of risk could be due to the fact that exposure to UV radiation from fluorescent or halogen lamps is small compared with that from sunlight, and the latter is likely to overwhelm any effect due to artificial sources of UV radiation.

Finally, cigarette smoking and coffee or tea consumption did not seem to influence the risk of BCC. Interestingly, subjects in the highest category of cumulative wine consumption in a lifetime showed a tendency toward a reduced risk of BCC compared with nondrinkers. However, the potential protective role of diet components needs further evaluation in larger ad hoc designed studies.

In conclusion, the definite association found in this study with recreational sun exposure during childhood and adolescence and the strong relation with family history of skin cancer suggest that genetic predisposition and peculiar exposure patterns to UV radiation are key independent risk factors for the development of BCC, and support the importance of sun protection and avoidance, especially early in life, for the prevention of skin cancer.

Accepted for publication January 16, 2001.

This study was funded by the Italian Ministry of Health, contract ICS-70.4/RF95.246.

Corresponding author and reprints: Rosamaria Corona, DSc, MD, Istituto Dermopatico dell'Immacolata, Via dei Monti di Creta 104, 00167 Rome, Italy (e-mail: r.corona@idi.it).