Bruce K. Armstrong University of Western Australia Western Australia Australia
The incidence (frequency of occurrence of new cases per unit of population) of malignant melanoma (MM) of the skin is increasing about 5% per year in most white populations (Muir and Nectoux 1982). Lancaster (1956) was the first to suggest that this cancer may be substantially due to ultraviolet radiation from the sun. He observed a threefold variation in mortality from melanoma between the populations of Queensland in the north of Australia (high mortality) and Tasmania in the south (low mortality). The populations of the other states were distributed between them approximately in relation to the latitudes of their main population centers. Similarly, Lancaster noted that European populations residing in the comparatively sunny climates of Australia, South Africa, and California exhibited higher mortality from MM than in the European countries from which they originated.
A number of observations are apparently inconsistent with a simple causal relationship between sun exposure and MM. Other skin cancers, which are generally believed to be caused mainly by sunlight, are more common in men than women (perhaps because men often work outside), increase exponentially in incidence with age (the pattern expected from lifelong exposure to an agent that can initiate cancer), and occur most commonly on the more exposed head, neck, and hands. In contrast, MM occurs as often in women as in men, shows a relative peak in incidence in middle life, and is most common on the back in men and legs in women (Holman et al. 1980). There are also a number of geographical areas in which the incidence of MM does not increase with increasing proximity to the equator. These areas include the Australian states of Western Australia and Queensland, which cover a wide range of latitudes, and Europe where MM incidence decreases with increasing latitude to about 50degN and then increases with increasing latitude (Armstrong 1984). These geographical inconsistencies, however, may be caused by a failure to adequately consider skin pigmentation and climatic factors that modify exposure and sensitivity to sunlight. More significant is the relationship between the incidence of MM and occupation and socioeconomic status. Whereas other skin cancers are more common in outdoor than indoor workers, as would be expected if they are caused by exposure to the sun, the opposite is true for MM. Similarly, incidence of other skin cancers tends to increase with decreasing socioeconomic status (perhaps because low status jobs tend more often to be outside) while the opposite is true for MM (Holman et al. 1980).
These observations led to the "intermittent exposure hypothesis" for the relationship of sunlight to MM. Briefly, this hypothesis states that:
The Western Australian Lions Melanoma Research Project, carried out in Western Australia in 1980 and 1981, was designed to test the intermittent exposure hypothesis.
The methods have been fully described elsewhere (Holman and Armstrong 1984a). In brief, 511 patients with histologically confirmed MM were studied. They were 76% of a total of 670 cases less than 80 years of age and diagnosed in accessible regions of Western Australia in a period of 675 days beginning January 1, 1980. Clinical details were obtained from the doctors who treated the patients and the histopathological diagnosis was reviewed and confirmed by a panel of pathologists.
Five hundred eleven control subjects, each matched to one of the melanoma patients by age, sex, and area of residence, were also studied. They were selected at random from the Australian Commonwealth Electoral Roll (electoral registration is compulsory in Australia) or, if the MM patient was less than 18 years of age, from the student roll of the area public school. The final series of 511 control subjects was 69% of those approached.
The patients with MM and the control subjects were approached in identical fashion and asked to participate in an interview on "environment, lifestyle, and health," which lasted from one to two hours. The interviews were conducted in the subjects' homes (occasionally workplaces) by trained nurse interviewers who were not told whether the person that they interviewed was a MM patient or a control subject. The interview covered demographic, constitutional and genetic factors, sun exposure, hormone use, diet, and some other variables. Objective measurements were made of skin, eye and hair color, weight, height, amount of body hair, number of raised moles (pigmented naevi) on the arms and degree of sun damage to the skin on the back of one hand.
It is possible, by comparing the data obtained from the patients with MM (often called "cases") and the control subjects, to estimate the extent to which exposure to specific levels or categories of particular exposure variables increases the incidence of MM above the incidence in some arbitrarily chosen reference group (usually those not exposed or those in the lowest exposure category). The statistic calculated is the incidence rate ratio or relative risk (as estimated by the exposure odds ratio), abbreviated hereafter as RR. Because it is a ratio, values of the RR above 1.0 for a particular category of exposure imply that the incidence of MM is increased in that category in comparison with the incidence in the reference category. RRs were calculated by the methods recommended for matched case-control studies by Breslow and Day (1980). For each, a 95% confidence interval (CI) was also calculated. Given that samples of both MM patients and controls were studied, each of the statistics calculated has sampling variability. The CI is the interval in which it is 95% likely that the true value of the RR for the population as a whole lies.
When interpreting an RR for a particular exposure category, it is necessary to consider the possibility that the observed association is influenced by some "confounding" variable that is related to both the exposure variable for which the RR has been calculated and to MM. For example, people with highly sun-sensitive skins may tend to expose themselves less to the sun than those with not-so-sensitive skins. If sun sensitivity is associated with an increased incidence of MM, this could reduce the strength of any association between sun exposure and MM unless the confounding effects of sun sensitivity are controlled when examining the effects of sun exposure. This control was achieved by use of conditional logistic regression analysis (Breslow and Day 1980) and adjusted RRs were calculated, where relevant, free of the effects of specific confounding variables.
Pigmentary Characteristics and Sensitivity of the Skin to the Sun
The RRs for categories of skin color, hair color, and eye color are summarized in Table 1. The skin color measurement was a reflectance measurement; thus low values represent dark skin. It was made on the skin of the upper inner arm to avoid, as far as possible, pigmentation due to sun exposure.
Incidence of MM increased in ordered categories of each of these variables with increase in the characteristics that are usually thought of as being associated with sun sensitivity. Thus the highest incidence of MM was in those with light skin, red hair, and blue eyes. A "P value" of <0.05 means that the probability that the pattern observed in the RRs was due solely to chance (sampling variability) was less than 5% (i.e., 1 in 20). For skin color and hair color it was very much less than 5%.
Sensitivity of the skin to the sun was ascertained by asking two questions:
If your skin was exposed to strong sunlight for the first time in summer for one hour, would you...Relative risks of MM for these categories of sun sensitivity are shown in Table 2. Both acute and chronic skin response to sunlight were strongly related to incidence of MM with the highest incidence being in those with greatest sensitivity to the sun.
After repeated and prolonged exposure to sunlight would your skin become...
- Get a severe sunburn with blistering?
- Have a painful sunburn for a few days followed by peeling?
- Get mildly burnt followed by some degree of tanning?
- Go brown without any sunburn?
- Very brown and deeply tanned?
- Moderately tanned?
- Only mildly tanned due to a tendency to peel?
- Only freckled or no suntan at all?
Of all the pigmentary characteristics and measures of skin response to sunlight, chronic skin response to sunlight was the strongest predictor of the risk of MM. When these variables were included together in a logistic regression model, acute skin response to sunlight and hair color, together with chronic skin response to sunlight, were significantly correlated with incidence of MM. It is at least plausible to suggest that skin response to sunlight is the important predictor of risk of MM and that hair color appeared to be independently predictive only because skin response was measured with some error.
MM is known to be rare in pigmented races (Crombie 1979). Like the association of MM with response of the skin to sunlight in white races, this observation suggests that sunlight may be a cause of MM. In the Western Australian study, subjects were classified by the ethnic origin of their grandparents (if they had two or more grandparents of the same ethnic origin) into one of the following categories: Celtic (Irish, Scottish, or Welsh), English, Australian (mainly Celtic or English, there were no Australian aborigines in the study), Southern European, Northern European, African, or Asian. RRs for MM by ethnic origin are given in Table 3. The RRs have been adjusted for possible confounding effects of age at arrival in Australia (many of the non-Australian subjects were migrants and therefore had a low incidence of melanoma regardless of their ethnic origin; see below) and represent the effects of each ethnic group independently of all the others.
There were few subjects in the study who belonged to ethnic groups (e.g., born in Africa or Asia) that might reasonably have been expected to have had pigmented skin. The lowest risk, however, was in those of Southern European ethnic origin whose skins are generally darker than those of people originating elsewhere in Europe. The pattern is therefore consistent with a protective effect of ethnically determined skin pigmentation against MM.
Birthplace, Age at Arrival, and Duration of Residence in Australia
Most Australians of European origin who were born outside Australia have migrated to Australia from a region of lower exposure to the sun. Thus, if sunlight is a cause of MM, they would be expected to have lower incidence rates of MM than native-born Australians. This has been observed to be the case in descriptive studies (Armstrong et al. 1982). In the 1980-81 case-control study, incidence of melanoma increased with increasing duration of residence in Australia and fell with increasing age at arrival in Australia. Since age at arrival and duration of residence are correlated one with the other, both were included in a logistic regression analysis to see which, if only one, was independently related to incidence of MM. The results of this analysis are shown in Table 4. After adjustment for duration of residence, incidence still fell with increasing age at arrival while adjustment for age at arrival removed the observed effect of duration of residence. Thus, if sun exposure is responsible for the high incidence of MM in native-born Australians relative to that in migrants to Australia, it appears that exposure early in life is necessary to have this effect.
Mean Annual Hours of Bright Sunlight
To obtain a measure of the potential for exposure to the sun at all their places of residence, Holman and Armstrong (1984b) calculated the mean annual hours of bright sunlight (as given on climatology maps) averaged over all places of residence (as obtained in a residence history) and weighted by the duration of residence. This measure did not account for the time that each subject spent in the sun. The analysis was restricted to native-born Australians to separate the effects of residential sunlight from those of place of birth. The results are shown in Table 5. Incidence of MM nearly doubled between those with less than 2600 mean annual hours of bright sunlight at places of residence and those with more than 2800 hours. Migrants to Australia had about half the incidence of MM as native-born Australians with less than 2600 annual hours of bright sunlight on average. When mean annual hours of bright sunlight were controlled in a logistic regression analysis, the mean latitude of residence showed no association with risk of melanoma. This suggests that the effect of latitude on incidence of MM can be explained by the effect of sunlight.
Other Objective Measures of Total Accumulated Exposure to the Sun
The degree of damage caused to skin on the back of the hand by exposure to the sun was measured by means of cutaneous microtopography (Holman et al. 1984). A silicone mold is made of the skin markings on the back of the hand, examined under a dissecting microscope, and graded on a scale from 1 to 6, 1 indicating the least solar damage and 6 the most. The degree of skin damage was taken to be an indicator of total accumulated exposure to the sun. A history of past non-melanocytic skin cancer was also obtained. Because non-melanocytic skin cancers are believed to be predominately sun-induced they were considered to indicate individuals who had been heavily exposed to the sun.
The relationships of MM incidence with cutaneous microtopograph grade and past history of skin cancer are shown in Table 6. Incidence of MM increased with increasing severity of solar damage to the skin such that the incidence in those with grade 6 damage was nearly three times that in those with only grades 1-3 damage. Similarly, the incidence of MM was higher in those with a past history of non-melanocytic skin cancer than in those without. Adjustment of the latter association for possible confounding effects of acute and chronic reactions of the skin to sunlight, hair color, and numbers of European, African, and Asian grandparents reduced the RR from 3.71 to 2.87 (CI 1.64-5.04).
Pattern of Sun Exposure
The evidence on the associations between incidence of MM and average hours of bright sunlight at places of residence, sun-induced skin damage, and past history of non-melanocytic skin cancer strongly supports the role of sun exposure in the causation of the disease. These measures of sun exposure, however, are essentially measures of total accumulated exposure over a lifetime and reveal nothing about the pattern of exposure. A detailed history of sun-exposure habits was taken from subjects in the Western Australian study to provide evidence relevant to the intermittent exposure hypothesis. They were asked to provide estimates of the time spent outdoors in both summer and winter on typical working and non-working days for all periods of employment throughout their working lifetimes (Mondays through Fridays were treated as typical working days for students, housewives, and retired people). Details were also sought regarding specific outdoor pursuits and clothing habits when outdoors. All the RRs presented are adjusted for the potential confounding effects of acute and chronic skin reaction to sunlight, hair color, ethnic origin, and age at arrival in Australia.
Table 7 summarizes results for average estimated outdoor time per week over the lifespan (since leaving school) and the proportion of the total outdoor time that was recreational (a measure of intermittency of the exposure) between 10 and 24 years of age.
The trend in the RRs for total outdoor time showed the anomalous pattern that has been observed in descriptive data: incidence of MM fell rather than rose with increasing time spent out of doors. This trend, however, could have been due to chance as the P value was quite high (0.13). Both of the components of outdoor time (outdoor time at work and time in outdoor recreation) showed similar downward trends (Holman, Armstrong, and Heenan 1986) but as the proportion of the outdoor time that was recreational increased, the incidence of MM tended to rise (Table 7). The P value for this trend, however, was also rather high (0.25). The recreational exposure proportion was examined for the period of life from 10 to 24 years of age because this was the period in which the negative gradient with total outdoor time was most evident.
Because the intermittent exposure hypothesis is thought to relate most strongly to superficial spreading melanoma (SSM) (Holman, Armstrong, and Heenan 1983), the most common of the four histological types of MM, the analyses of Table 7 were also carried out for this particular type of MM. SSM was more strongly related negatively to average total outdoor time per week during the summer (P=0.09) and positively to the proportion of outdoor time that was recreational (P=0.15) than was MM as a whole.
Incidence of SSM was also strongly related to frequency of participation in some but not all outdoor recreations that involve substantial sun exposure (Table 8). It is particularly interesting that a relationship between frequency of sunbathing and incidence of SSM was evident when only SSM of the trunk was analyzed (the trunk is presumably the body site subject to the most intense intermittent exposure with sunbathing). MM other than SSM did not show strong associations with any of these recreational exposures.
Clothing may modify the relationship between sun exposure and incidence of MM, as Table 9 shows. Incidence of both melanoma and SSM is examined in relation to the person's clothing habit during outdoor work in summer at the primary site of the MM. The incidence of both was substantially higher in those who sometimes or usually exposed the primary site than in those who usually kept it covered. For all MM the RR was highest in those who sometimes exposed the site, an observation that is consistent with the intermittent exposure hypothesis. This was not the case, however, for SSM. SSM of the trunk in women was very strongly related to the type of bathing suit that had been worn in summer between 15 and 24 years of age. Relative to an RR of 1.00 in those who had worn a one-piece suit with a high back-line, the RR was 4.04 (CI 0.65-25.2) in those who had worn a one-piece suit with a low back-line, and 13.0 (CI 1.95-83.9; P value for trend 0.005) in those who had worn a two-piece suit or no bathing suit at all.
The reduction of MM incidence by high natural skin pigmentation and its increased incidence in skin that is highly sensitive to the sun are well known. These observations have been confirmed by the Western Australian study. Indirectly, they implicate sunlight in the etiology of MM. This study has also provided evidence that incidence of MM increases with total accumulated exposure to the sun. Incidence was lower in migrants to Australia than in native-born Australians, was positively associated with mean annual hours of bright sunlight averaged over all places of residence, and increased in those with sun-damaged skin and a past history of non-melanocytic skin cancer (generally accepted as being due to exposure to the sun).
Superimposed on this background, there was some evidence that intermittency of sun exposure may be particularly important in causing MM. Incidence rose slightly with increases in the proportion of total exposure that was recreational, but certain outdoor recreations, often involving intense sun exposure, appeared to be particularly strongly related to it, for example, boating, fishing, and sunbathing for MM of the trunk. When clothing habits were taken into account, there was strong association between MM and unclothed exposure of the primary site of the tumor while at work. Thus the anomalous relationship between MM incidence and outdoor work overall may be due, in part at least, to a tendency of those who work outdoors to be more careful in protecting themselves from the sun than those who expose themselves only recreationally.
If the intermittent exposure hypothesis for the relationship of MM to sunlight is correct, it has important implications for prevention. Figure 1 shows a concept of the exposure-response relationship for SSM and sunlight under the intermittent exposure hypothesis. While this concept applies to lesions that have their origin in SSM, as they are probably the majority of MM, it may be considered to apply to MM as a whole. On average (curve C) incidence of SSM first rises as frequency of exposure increases and then falls as some critical exposure frequency is passed. Thus, on average, whether or not reduction in sun exposure will be a "good thing" (in terms of prevention of MM) will depend on where a population lies on this exposure-response curve. The position will probably depend both on the genetic composition of the population and the available sunlight (expected absorbed dose per unit of time of exposure to the sun).
At the individual level, the shape of the exposure response curve will probably be determined by the pigmentary response of the skin to sunlight. Those with a poor pigment response (curve B) may experience progressively increasing incidence of SSM, whatever the frequency (and dose) of sun exposure, because a protective tan is never obtained. Others who tan readily (curve A) may show little rise in incidence of SSM at all with increasing sun exposure before incidence falls back to background levels with further sun exposure. While the former may be best advised to protect themselves against the sun at all times; the latter, if exposed more than minimally, should perhaps continue sun exposure to ensure that they stay as far as possible to the right of the peak. Until these concepts are clarified or discarded, caution should be exercised in making recommendations about sun exposure--at least with reference to prevention of MM.
Armstrong, B.K. 1984. Melanoma of the skin. Brit. Med. Bull. 40:346-50.
Armstrong, B.K., C.D.J. Holman, J. M. Ford, and T. L. Woodings. 1982. Trends in melanoma incidence and mortality in Australia. In Trends in cancer incidence. Causes and practical implications, ed. K. Magnus, 399-417. Washington: Hemisphere.
Breslow, N.E., and N.E. Day 1980. Statistical Methods in Cancer Research. Lyon: International Agency for Research on Cancer.
Crombie, I.K. 1979. Racial differences in melanoma incidence. Brit. J. Cancer 40:185-93.
Holman, C.D.J., and B.K. Armstrong. 1984a. Pigmentary traits, ethnic origin, benign nevi, and family history as risk factors for cutaneous malignant melanoma. J. Nat. Cancer Inst. 72:257-66.
Holman, C.D.J., and B.K. Armstrong. 1984b. Cutaneous malignant melanoma and indicators of total accumulated exposure to the sun: An analysis separating histogenetic types. J. Nat. Cancer Inst. 73:75-82.
Holman, C.D.J., B.K. Armstrong, P.R. Evans, G.J. Lumsden, K.J. Dallimore, C. J. Meehan, J. Beagley, and I.M. Gibson. 1984. Relationship of solar keratosis and history of skin cancer to objective measures of actinic skin damage. Brit. J. Dermatol. 110:129-38.
Holman, C.D.J., B.K. Armstrong, and P.J. Heenan. 1983. A theory of the etiology and pathogenesis of human cutaneous malignant melanoma. J. Nat. Cancer Inst. 71:651-6.
Holman, C.D.J., B.K. Armstrong, and P.J. Heenan. 1986. Relationship of cutaneous malignant melanoma to individual sunlight exposure habits. J. Nat. Cancer Inst. 76:403-14.
Holman, C.D.J., C.D. Mulroney, and B.K. Armstrong. 1980. Epidemiology of preinvasive and invasive malignant melanoma in Western Australia. Int. J. Cancer 25:317-23.
Lancaster, H.O. 1956. Some geographical aspects of mortality from melanoma in Europeans. Med. J. Aust. 1:1082-87.
Muir, C.S., and J. Nectoux. 1982. In Trends in cancer incidence. Causes and practical implications, ed. K. Magnus. Washington: Hemisphere.