Ethics statement

The primary NHANES I and NHEFS were approved by the National Center for Health Statistics (NCHS) Institutional Review Board.[13] The present study was approved by the Ethics Committee of Zhejiang University School of Public Health.

Participants

The First National Health and Nutrition Examination Survey (NHANES I) was conducted between 1971 and 1975 by the National Center for Health Statistics based on a probability sample of the civilian noninstitutionalized population of the United States. [14], [15], [16] The NHANES I Epidemiologic Follow-up Study (NHEFS) is a longitudinal follow-up of participants in NHANES I who were aged 25 to 74 years at the time of survey. [17] Detailed information on NHANES I and NHEFS are provided elsewhere. [14], [15], [16], [17]

Of the 12, 053 white subjects followed in the NHEFS, we excluded 2760 NHANES I Augmentation Survey participants who were not administered a dermatologic examination[16]; 83 subjects who were pregnant at the time of the baseline examination, and 487 subjects whose demographic, socioeconomic, or other covariates were missing. Among the remaining participants, 251 were lost during the follow-up period, leaving 8472 participants (men, 40.5%; women, 59.5%) who contributed 139,802 person-years of follow-up to the final analyses.

Outcome definitions

Based on interviews, health care facility medical records, and death certificates, NHEFS follow-up data were collected in 4 waves: 1982 to 1984, 1986, 1987, and 1992. [17] Death was ascertained by either a death certificate or a proxy interview, or both, and the underlying cause of death was coded according to the International Classification of Diseases, Ninth Revision (ICD-9). Deaths were analyzed for all causes, cancer (ICD-9 codes 140–208, e.g., lung cancer, colon cancer, breast cancer), and cardiovascular disease (ICD-9 codes 390–448, e.g., ischemic heart disease, cerebrovascular diseases, hypertension).

Actinic skin damage assessment

A complete clinical dermatologic examination of the skin was undertaken in NHANES I to evaluate variations in texture and color, certain manifestations of aging, and all pathologic changes. The dermatologic examinations were performed by a total of 101 dermatologists during the baseline survey. All dermatologists were trained for the dermatologic examination protocol before the commencement of the survey examinations at each location. After examining the skin covering the entire body, the dermatologist recorded an estimate of the presence and severity of overall actinic skin damage and its components, including fine telangiectasia, solar elastosis (senile elastosis), and actinic keratoses. All of these measures were graded as being either absent, minimal, moderate, or severe. [18]

Covariate definitions

Several potential confounding factors derived from the baseline interview and physical examinations were included in the regression models based on findings from previously published studies. [18], [19], [20] These included age, gender, educational attainment (<12^th, 12^th, >12^th grade); family income (<$3000, $3000–$7000, $7000–$10000, $10000–$15000, ≥$15000); combined measure of occupational and recreational physical activity (low, medium, high); body mass index (underweight, normal, overweight, obese); vitamin supplement use (regular or irregular use, no); place of residence (rural, urban); region of residence (Northeast, Midwest, South, West); season of the year when the examination was taken (spring, summer, autumn, winter); eye color as a surrogate of skin type [21] (dark brown, light brown, gray-green-hazel, dark blue, light blue, other); frequency of alcohol consumption (less than once a month or never, once a month to several times a week, almost daily or daily). Menopause was defined by the answer “yes” for the question, “have the cycles entirely stopped?”. Occupation was coded according to the 1970 census of population Alphabetical Index of Industries and Occupations [22]: professional, manager, sales worker, clerical and kindred workers, craftsmen and kindred workers, operatives, laborer, farmer, service works, and unemployment variables for retired, keeping house, and others (disabled, student, looking for work, missing). Smoking was coded as current, past, or never. Past smokers were defined as those who reported that they had smoked ≥100 cigarettes during their lifetimes but did not currently smoke cigarettes. Retrospective smoking data collected during the follow-up were used for persons with missing baseline information. Retrospective information on smoking agreed reasonably well with data collected in NHANES I. [23] Baseline diseases were defined as self-reported history of having cardiovascular diseases, cancer, diabetes, hypertension, ulcer, kidney diseases or hepatitis.

Statistical methods

Characteristics of the participants at baseline were presented as the mean or percentage stratified by severity of overall actinic skin damage. The level of statistical significance of differences were examined by using ANOVA (continuous variables) and the χ² test (categorical variables), with Bonferroni adjustment for multiple comparisons.

Cox proportional hazards regression models [24] were used separately to examine the associations of overall actinic skin damage and its components with all cause, cancer, and cardiovascular disease (CVD) mortality. [25] Age was used as the time scale in all time-to-event analyses, and all analyses were stratified by birth cohort using 10-year interval to control for calendar period and cohort effect. [26] Trend tests for the adjusted hazard ratios (HRs) were performed by assigning the integers of 0, 1, 2, 3 to the degrees of severity of actinic skin damage and its components, and treating the resulting score variables as continuous in the models. The Schoenfeld residual [27] test was used to check the proportional hazards assumption. In addition, we also performed the Cox regression model with time-on-study as the time-scale and baseline age as a covariate.

To assess the effect of the complex survey design on the results, Cox proportional hazard regression analyses were performed to incorporate the sample weights, the stratification and clustering in the analysis. The weighted and unweighted models yielded consistent results. Therefore, only unweighted results were presented here. To explore whether the results will be affected by skin cancer, we also performed the analyses excluding those who died from skin cancer during the follow-up period (n = 8) in all three models. Subgroup analyses by age (25–44, 45–65, 65–74 years old) were conducted to assess the robustness of our findings. Cox regressions were also performed excluding subjects with cardiovascular disease, diabetes, or cancer at baseline, and were repeated when including pregnant women. All Cox models were adjusted for the above-mentioned potential confounding factors. In addition, men and women were combined in the regression analysis as no interactions were detected between gender and actinic skin damage.

Cumulative mortality curves were plotted according to overall actinic skin damage severity using the Kaplan-Meier method. [28] The assessment of cumulative mortality rates was limited to subjects who were 40–85 years of age during the follow-up period. Because the calculation of cumulative mortality rates is sensitive to small sample sizes, moderate and severe actinic skin damage were combined into one group. Pairwise comparisons of absent, minimal, and moderate and severe overall actinic skin damage were performed using a two-sided log-rank test [29] for all-cause (n = 8431), cancer (n = 6154) and cardiovascular disease (n = 6981) mortality, respectively. Bonferroni corrections were used to control for type I error, and a two-sided P value of 0.017 was used for each comparison. All analyses were performed using Stata software (version 11.0 for Windows; Stata Corporation, College Station, TX).

Potential mechanisms

There are several possible mechanisms by which actinic skin damage may influence all-cause and cause-specific mortality. The free radical theory states that the appropriate and inappropriate production of oxidants is intricately connected to aging and lifespan. [30], [31] Many environmental stimuli, including ultraviolet (UV) radiation, can generate high levels of reactive oxygen species (ROS). [30] Plasma antioxidants are decreased in people with actinic keratosis, probably due to their long exposure to UV radiation. [32] ROS plays a role in the pathology of sun-induced skin damage, [33], [34] and thus may potentially be involved in mechanisms responsible for an increased risk of mortality in people with moderate to severe actinic skin damage reported in this study. Recent experimental and clinical data support the hypothesis that cigarette smoke exposure increases oxidative stress as a potential mechanism for initiating cardiovascular dysfunction, [35] and histologic analysis of “smoker's skin” reveals elastic fiber thickening and fragmentation, similar to that found in sun damaged skin. [36] Thus smoking and sun-related actinic skin damage may share some common mechanisms resulting in cardiovascular disease. Furthermore, repeated UV radiation can cause DNA damage and apoptosis on human peripheral blood mononuclear cells, which indicate that sunlight exposure may cause internal diseases by affecting blood components. [37], [38] In addition, UV radiation is known to activate multiple signaling cascades, leading to inflammatory reactions. [39] Despite all of these possible pathways, the mechanisms of the association between actinic skin damage and mortality remain unclear. Further study is needed to elucidate this association.

Study implications

Previous studies have found that optimal vitamin D status may not only be important for bone health but also may protect people from some cancers and other chronic diseases. [3] It has been estimated that approximately one billion people worldwide have vitamin D deficiency or insufficiency.[3] In the United States, at the end of the winter, 42% of 15-49 year-old black girls and women had a 25-hydroxyvitamin D level below 20 ng/mL. [3], [40] Exposure to sunlight is the main source of vitamin D for most people. [41] For all of these reasons, although the American Academy of Dermatology (AAD) recommends that an adequate amount of vitamin D should be obtained from a healthy diet and/or vitamin D supplement,[42] people are encouraged to increase their sun exposure to prevent vitamin D deficiency by many studies, [3], [43], [44] and the WHO is now addressing the issue of whether current sun protection messages are too strong. [45] However, the prevalence of hypovitaminosis D was only 4.2% for white women of reproductive age, [40]. Hence to better understand the risk and benefit of sun exposure to human health is needed. It is well known that over exposure to sun increase risk for skin cancer and skin damage. If the finding from our research about the positive association between skin damage and mortality can be confirmed by other studies, an evaluation on the sun protective message should be warranted.

Children are at high-risk for sun-induced skin damage. An estimated 50% to 80% of the skin's lifetime sun damage is thought to occur in childhood and adolescence. [46] Among youths aged 11 to 18 years in the United States, 72% reported having had at least one summer sunburn, 30% reported at least 3, and 12% reported at least 5 sunburns. [47] Sunbathers are another group at high risk for actinic skin damage. [47] In a national survey, 59% of white adults reported sunbathing during the past year, and 25% reported frequent sunbathing. Of the subsample who reported sunbathing during the month before the interview, only 47% routinely used sunscreen. [48] Other high risk populations include tourists, outdoor workers, and fair skinned people living in lower latitudes. [4]

Skin damage caused by over exposure to sunlight is preventable. By running programs such as education, provision of shade in open spaces, rescheduling of work and sports activities, removal of sales taxes from approved sunscreens, and provision of cheap sunscreens, there have been substantial changes in knowledge, attitudes, and beliefs about sunlight and suntans, as well as an increase in sun protection behavior in Australia. [49]

Sunlight exposure plays an important role in both vitamin D synthesis and actinic skin damage pathology, which means that whether increased sun exposure is a protective or risk factor for chronic disease may depend on the balance between the UV induced production of vitamin D and UV induced skin damage. Throughout the United States, the estimated daily solar exposure to maintain a serum 25(OH)D level of 30 ng/mL is 15 minutes in early fall or later spring, from 11:00 AM to 2:00 PM under clear skies, assuming exposure of arms, shoulders, and back. [50], [51] Ultraviolet exposure beyond the minimal erythemal dose does not increase vitamin D production further. [51] The concentration of previtamin D in the skin reaches an equilibrium in white skin within 20 min of ultraviolet exposure. [51] However, the optimal amount of sunlight exposure a person should receive is not known but is likely to depend on many factors related to skin type, latitude, and sunscreen application. [4]

Study limitations

A major limitation of this study is that the measurements of actinic skin damage were subject to errors of classification. [18], [52] Although the 101 examiners received training, the written instructions for this component of the examination were limited, and there were no diagnostic criteria employed in the definition of actinic skin damage severity. [52] In addition, the actinic skin damage severity evaluation was done subjectively by dermatologists. [18] However, the impact of any particular dermatologist's judgments on the overall results was minimized by the large number of examiners used in the survey. [18] In addition, the follow-up survey did not specifically collect dermatologic data [17] and we thus had to assume that the subjects' actinic skin damage did not vary over time, a factor that might result in biased hazard ratios. Another study limitation is the lack of information in potentially important characteristics such as skin type, outdoor/indoor workers or sun seeking behavior. In the present study, we used other surrogates, such as eye color, occupation and family income, to minimize these affects. Furthermore, baseline actinic skin damage data were collected almost 40 years ago, hence the study findings may not be generalizable to the current US population.

Despite these limitations, our study used a large population-based cohort with an average of 16 years of follow-up, focusing on whites only. The fact that follow-up information was available for more than 97% of the study participants and that an extensive number of covariates were available for the NHANES I subjects further enhanced the validity of our findings. Moreover, to the best our knowledge, this is the first study to investigate the relation between sun-induced skin damage and mortality.

Age could be the most important confounder in this study, because age has a very strong relationship with both actinic skin damage and all-cause mortality. [20] To adjust adequately for chronological age, we used age as the time scale and all analyses were stratified by birth cohort to control for calendar period and cohort effector. [26] Furthermore, subgroup analyses by age provided similar results. After further controlling for potential confounding factors such as smoking, occupation, region of residency, season and social-economic status, the association of over-all actinic skin damage and its components with all-cause mortality persisted, and the results remained similar when we used cancer and cardiovascular disease as the outcome.

Conclusions

In conclusion, our study detected an association of actinic skin damage with cardiovascular disease, cancer and all-cause mortality in whites. However, this result should be interpreted with caution since unmeasured or incompletely measured confounding factors cannot be ruled out in this study and since mechanisms for these associations are lacking. More independent studies are needed before any practical recommendations can be made.