James G. Titus
U.S. Environmental Protection Agency
Joseph Scotto National Cancer Institute National Institutes of Health Bethesda, Maryland USASince the early 1970s, the National Cancer Institute, in collaboration with other federal and nonfederal sources, has conducted special epidemiologic surveys of nonmelanoma skin cancer and monitored ground level UV-B measurements at various geographic locations within the United States.
These studies responded to the need for basic data to be used in measuring the extent of certain human health effects that may result from stratospheric ozone depletion.
We focus on solar ultraviolet radiation in the range of 280 to 320 nanometers, called mid-UV or UV-B (see Figure 1). Under experimental conditions, UV-B has been shown to produce skin erythema (sunburn) in man and skin cancer in animals, and it is effective in altering DNA. Cumulative exposure to UV-B is also believed to be partially responsible for the "aging" process of the skin in humans. Except for preventing vitamin D deficiency rickets, which is now confined to populations with inadequate nutrition, UV-B is basically considered biologically harmful.
While no solar radiation below 295 nm ever reaches the earth's surface, a small quantity of UV-B does. Stratospheric ozone depletion may result in increases of UV-B energy reaching the earth and its populations. The physical amplification factor has been put at 2. This means that a 1% decrease in ozone may result in a 2% increase in solar ultraviolet, UV-B.
Basal cell carcinomas and squamous cell carcinomas of the skin, the nonmelanomas, are the most common malignant neoplasms occurring in the white populations of the world. Currently, annual incidence in the United States is estimated at about one-half million patients, and the rates are increasing at about 3% per year. Epidemiologic study has been limited by the fact that most patients are customarily seen and treated in the offices of physicians and not hospitalized. Cure rates are high (about 99%) and only a small percentage of skin cancers are metastatic or result in death.
Because the primary source of data cancer registries is the inpatient hospital file, the statistics routinely collected on skin cancer are usually very incomplete and not comparable with other forms of cancer. Thus, population-based estimates of skin cancer incidence require special surveys to collect data from offices and outpatient files.
Figure 2 shows the geographic locations within the continental United States where skin cancer surveys were conducted and ground level measurements of UV-B were obtained. In 1971-72, there were four areas in the Third National Cancer Survey; in 1977-78, eight areas of NCI Surveillance Epidemiology and End Results Program (SEER) were surveyed.
Two locations, Minneapolis-St. Paul, Minnesota, and San Francisco-Oakland, California, were resurveyed in the late 1970s. Skin cancer incidence data from New Hampshire, Vermont, and San Diego, California, were most recently included. We do not have statistical details for other cancers from these locations, and we are just now receiving UV-B readings from Concord, New Hampshire, and Burlington, Vermont. Patient and general population interview studies were conducted in nine locations. These locations span the United States from Seattle in the north (47.5\'fbN) to New Orleans (30\'a1N). Most of the figures that follow display the epidemiologic details for the eight-area survey conducted in 1977-78. Seven of these are SEER locations where NCI has continuing surveys of all other malignancies. The results are similar for all surveys.
Skin cancer is a disease that rarely occurs in black and pigmented races. The age-adjusted rate of 242 per 100,000 for whites is more than 60 times that for blacks with a rate of under 4 per 100,000. Among Caucasians "Anglos" or non-Hispanics, are at greater risk than Hispanics by about 7-10 to 1.
Figure 3 shows the incidence of nonmelanoma skin cancer and all other cancers combined among whites for each of eight locations plotted according to latitude. For all other malignancy, there is no latitudinal gradient. In contrast, nonmelanoma skin cancer incidence rates were definitely lower at the higher latitudes.
The age-adjusted incidence rates by geographic location for each sex are shown in Figure 4. The rates for males are always greater than those for females. Overall, the male/female ratio is close to 2 to 1 (1.83). Note there are two sets of bars for New Mexico. Because of the high proportion of highly pigmented Hispanics in that State, over one-third, the rates for all Caucasians are lower than that for Anglos only.
Figures 5 and 6 show age-specific rates according to region. The rates increased with age--the highest rates were seen in the oldest age groups. The southern region was clearly at higher risk than the northern region.
Figures 7 and 8 show age-specific rates by specific location, comparing males and females. At each location the male rates were lower or equal to the female rates at early ages. After age 45, in the northern and middle locations, male rates consistently exceeded those of females, and the differences were greatest in the oldest age groups (Figure 7). In the southern locations (Figure 8), the separation between male and female rates began a decade or two earlier, presumably because the UV threshold levels for skin cancer detection were reached sooner.
Tumors appear on the face, head, and neck in over 80% of nonmelanoma skin cancers. Among females, the nose is the most common site while, among males tumors of the nose, cheek, and scalp are equally high. Tumors of the ear occur more frequently among men, compared to women, by a factor of 10 or more, especially in the southern areas. In contrast, tumors of the legs are more common among females who have greater UV exposure and also more melanoma of the lower leg. In both sexes tumors are much more common in the lower lip than in the upper lip.
Age-specific incidence rates according to anatomical site are illustrated in Figure 9. Rates for face, head and neck, and upper extremities progressively increased, while those for trunk and lower extremities reached a plateau or declined at older ages.
With respect to histology, Figure 10 shows a composite of specific geographic locations depicting age-specific incidence patterns for basal cell carcinomas and squamous cell carcinomas of the skin. Overall the incidence rates for BCC was 4 to 5 times higher than SCC. Increases have been observed for each cell type as age increases. The rate of increase may be slightly higher for squamous cell carcinomas. But the squamous cell carcinomas begin at later ages. The ratio of BCC to SCC is greatest in the northern region and lowest in the southern region, ranging from over 12 to 1 in the north to over 2 to 1 in the southern locations.
In regard to our estimates of ground level UV-B measurements, we have been monitoring and editing data in collaboration with researchers at the National Oceanic and Atmospheric Administration and their network of weather stations (see Berger in this volume for details of the Robertson-Berger meter).
Figure 11 shows, as expected, that the annual amounts of sunburn-producing UV-B correlated with latitude. To put these numbers in perspective, it is estimated that a count of 440 R-B units may produce a perceptible sunburn. It is possible to receive such a dose within 20 minutes on a midsummer day around noontime. It is important to note that these measurements are affected by altitude, cloud cover or water vapor, and other meteorological factors. At altitudes of over a mile high Albuquerque and Salt Lake City received greater amounts of solar UV-B than expected; while Tallahassee with relatively high humidity received less than the expected dose for that latitude. I should like to point out that current observations indicate a general downward trend in meter readings at several locations. The relative positions for Detroit and Minneapolis have changed, with current figures now lower at Detroit.
The monthly averages of R-B counts for Albuquerque, New Mexico, and Seattle, Washington, respectively, are shown in Figure 12. These were the highest and lowest exposure areas included in NCI's nonmelanoma skin cancer surveys. The ratio of UV-B exposure is about 2 to 1 for these locations.
Next my discussion focuses on our correlation studies of population-based skin cancer incidence and estimated UV-B dose.
The annual UV-B levels and age-adjusted incidence rates for nonmelanoma skin cancer in white males and females were determined in the two special surveys conducted in 1971-72 and 1977-78 (see Figures 13 and 14). The incidence data are plotted on a log scale so that a straight line with a positive slope represents a constant percentage increase in incidence. Mathematical models were used to describe dose-response relationships and do not reflect the mechanism by which UV causes skin cancer. Using an exponential model previously applied to the 1971-72 data, estimates of the biological amplification factor (that is, the relative change in skin cancer incidence due to a relative change in UV-B radiation) were derived. Assuming a common slope, the exponential model may be written as a logarithmic expression as shown:
In the regression analyses, the logarithms of the age-adjusted incidence rates were weighted by the inverse of their estimated variance. Assuming that the annual UV-B counts were to increase by 1% at each location, the relative effects on skin cancer incidence were found to vary from a low of 1.19% to a high of 2.88%. The estimates were lowest for females residing in areas of low UV-B exposure levels. Overall, the biological amplification factors were estimated to be roughly twofold, but steeper for squamous cell carcinomas.
An update of the incidence and UV-B correlations for males and females is shown in Figure 15. There are ten locations plotted, which include only the most recent surveys of 1977-80. We have tentatively estimated the UV-B index for NH/VT at 96 (x10,000 SU), and we use average annual counts for the years 1977 through 1981 for all other locations. In general, the average UV values are lower than those previously estimated for a one-year period.
Biological amplification factors (using exponential model) show no substantial changes; however, the range is now between 1.03 and 2.5. While these estimates are a little lower than those previously calculated, the degree of uncertainty has been reduced and the estimates for the 95% lower limits have in fact increased. Figures 16 and 17 depict these correlations with respect to cell type. The slopes are steeper for SCC compared to BCC. Correlations according to anatomical site are shown in Figures 18 and 19. The relationship of skin incidence and ground level UV-B exposure is consistent for each site group. However, there appear to be stronger associations and steeper slopes for the face, head, and neck, and upper extremities among white males and females. Keep in mind that over 87% of all nonmelanoma skin lesions occur on these relatively more exposed anatomical sites.
As many researchers have suggested, and as the results of our studies show, there are demographic factors that may reflect increased or decreased skin cancer risk in certain population groups similar to the differences we have noted for Anglos and Hispanics. So, in addition to the incidence surveys, telephone interview sample surveys of skin cancer patients and general population controls between the ages of 20 to 75 were conducted at nine locations. Information was sought on several host and environmental factors that may affect the risk of skin cancer. Examples follow showing the correlation of skin cancer incidence with UV-B radiation according to the presence or absence of certain constitutional factors.
In Figures 20 and 21 we see familiar patterns. Skin cancer incidence increases as UV-B radiation increases for white males with and without freckles. We also note that at each location the risk is greater for those with freckles. The estimated relative risk adjusting for age and location is 1.8 and 1.7 for men and women, respectively, compared to those without freckles.
The correlation for fair skinned complexion is illustrated in Figure 22. Again, the UV-B gradients are observed. Overall relative risk estimates were 2.6 for men and 1.6 for women, respectively, compared to those without fair skinned complexions.
Figure 23 shows the patterns for Celtics or those of Irish or Scottish ancestry. Estimates of relative risk were the same as those observed for individuals with freckles.
Other high-risk groups include individuals with blond or red hair color, blue or green eye color, and those who sunburn easily; individuals treated for acne, moles, warts or psoriasis; individuals exposed to radiation or radiation therapy, coal tar or pitch, and arsenic. Individuals at low relative risk include those of Mexican or Spanish ancestry and those who are never outdoors on their principal occupation and those who develop deep tans.
With respect to the consistent excesses in skin cancer risk observed for men compared to women, we noted that the average amount of time spent outdoors was greater for males by a factor of 1.5 to 2.
I am reminded of a statement made in a review article by our Chairman (Emmett, CRC 1973). It said:
...in the USSR where the occupations of women nearly parallel those of men, the incidence rates of skin cancer are the same for males as for females except in the elderly females who may retain more domestic occupations. On this basis, it ma y be unlikely that the male human has a biological predisposition to solar skin cancer.In conclusion, I would add that this disease represents a major health and economic problem in the United States and other parts of the world. Twelve to thirty percent of the Caucasians in this country may expect to develop skin cancer in their lifetime if current rates and trends persist.