ULTRAVIOLET RADIATION AND MELANOMA

WITH A SPECIAL FOCUS ON ASSESSING THE RISKS OF STRATOSPHERIC OZONE DEPLETION

EDITED BY

JANICE D. LONGSTRETH

ICF-Clement
9300 Lee Highway
Fairfax, VA 22031-1207

Project Manager and Principal Editor:

Janice D. Longstreth
ICF-Clement
9300 Lee Highway
Fairfax, VA 22031-1207

John S. Hoffman
Office of Air and Radiation
US Environmental Protection Agency
401 M Street SW
Washington, D.C.

Janice Longstreth, Kathleen Knox[1], John S. Hoffman, Patsy H. Lill[2], Sarah A. Foster[3], Audrey F. Saftlas[3], Hugh M. Pitcher[4], Edward De Fabo[5], C. Ralph Buncher[6], David Warshawsky[6], William D. Ward[3]

[1] Statistical Policy Branch, Office of Policy Analysis, U. S. Environmental Protection Agency, 401 M Street, SW, Washington, D.C.

[2] Department of Pathology, University of South Carolina School of Medicine, VA Bldg No. 1, Garnet Ferry Rd, Columbia, SC 29208

[3] ICF-Clement

[4] Benefits and Use Division, Office of Policy Analysis, U. S. Environmental Protection Agency, 401 M St SW, Washington, D.C.

[5] Department of Dermatology, The George Washington University School of Medicine, Room 101 B Ross Hall, 2300 I Street, Washington, D.C., 20037

[6] University of Cincinnati Medical Center, Institute for Environmental Studies, Department of Epidemiology and Biostatistics, Mail Location 183, Cincinnati, OH 45267-0183

CHAPTER 4

TIME-RELATED FACTORS IN THE INCIDENCE AND MORTALITY: AGE, PERIOD, AND BIRTH COHORT EFFECTS

SECULAR TRENDS IN INCIDENCE AND MORTALITY

Sharp rises in incidence and mortality due to cutaneous malignant melanoma have been reported in nearly all Caucasian populations worldwide (see Table 4-1). Magnus (1982) analyzed trends in incidence in Norway, Sweden, Denmark, Finland, and Iceland for the period 1943-1976 and found that, although the absolute levels of incidence varied from country to country, an approximate annual increase of 6 percent was observed in all five countries with no sign of leveling off. Likewise, Connecticut incidence rates increased fivefold from 1935 to 1974, with an annual average increase of 5 percent (Roush et al. 1985a). Osterlind and Jensen (1986) noted that increases in Danish incidence were most pronounced since 1955, and showed no signs of leveling off as late as 1982. However, Danish mortality did not begin to increase until 1965, about 10 years after incidence rates began to rise markedly. Incidence rates in the United States increased steadily until 1983 when a 5 percent decrease was observed; however, this drop is thought to be an artifact due to the advent of DRG (disease related group) legislation, which is resulting in decreased hospitalization of individuals with melanoma in the United States (Sondik et al. 1985).

In general, increases in mortality due to CMM have been less steep than those observed for incidence. From 1943 to 1982, Danish incidence rose fivefold while corresponding mortality rates doubled, with signs of leveling off in women since 1975, but not in men (Osterlind and Jensen 1986). Leveling off of Australian mortality has been evident at least since 1965 to 1969 in both sexes, although incidence continues to increase (Armstrong 1982). In Sweden, overall incidence rose by 7 percent per year from 1959 to 1968, while only negligible changes in mortality occurred in both sexes (Malec and Eklund 1978). Similarly, data from the New Mexico Tumor Registry indicate that incidence rose considerably from 1969 to 1976, but mortality did not change over this period (Pathak et al. 1982). Lee (1982a) reported steady increases in mortality of 3 percent per year over the period 1951-1975 in England and Wales, Canada, and the United States. Lee noted that if diagnosis or treatment had improved over this time period, their influences were sufficiently constant that no change in the mortality trends was observed. More recent data from the United States (1974-1983) indicate that mortality due to CMM has continued to increase significantly both in white males and females (Sondik et al. 1985).

Secular trends in CMM mortality are less steep than those for incidence, probably due to earlier diagnosis rather than increasing diagnosis and registration of benign and semi-malignant cases in recent years (Magnus 1977; Elwood and Lee 1975). A study of cancer registries over the period 1955-1980 in Alabama and New South Wales, Australia (Balch et al. 1983) showed that melanoma skin tumors tended to be detected at progressively earlier stages of development; tumors were thinner, less invasive, and less likely to present with ulcerations. In addition, nodular tumors became less frequent while the incidence of superficial spreading melanomas, which have a better prognosis, increased over the period. Incidence rates of nodular and superficial spreading melanoma in Finland, however, increased to the same extent over the period 1963-1976 (Teppo 1982).

WORLDWIDE INCREASES IN CUTANEOUS MELANOMA: ARTIFACTUAL OR REAL?

Whenever significant changes in incidence or mortality rates are observed, it is necessary to determine whether the trends are genuine. In 1977, Ressuguie suggested that rising trends in CMM incidence were largely a result of improved diagnosis and registration (as cited in Lee 1982a). Histopathological studies, however, indicate that criteria for diagnosis of CMM have been consistent over time in Norway (Magnus 1975), and that the quality of diagnoses in well-run population-based cancer registries has been excellent (Pakkanen 1977; Malec et al. 1977). In addition, reasonable consistency in histologic diagnosis has been demonstrated between pathologists working in the same city (McCarthy et al. 1980) and in different countries (Larsen et al. 1980).

Considerable observational data have also led most investigators to conclude that actual increases in the incidence and mortality of CMM have occurred. Osterlind and Jensen (1986) point out that if substantial changes in histopathological criteria and registration efficiency have occurred, higher rates of diagnosis would most likely have resulted in stepwise increments in incidence, rather than the gradual rise observed. Furthermore, Danish mortality tends to mirror incidence rates with a lag period of 10 years, suggesting that the rise in incidence is real (Osterlind and Jensen 1986). Lee (1982a) has noted that improved diagnosis or registration is not likely to affect mortality since it would mean that the lower mortality rates observed earlier were due to the incorrect registration of large numbers of deaths in countries having very good systems of vital statistics. The fact that proportional increases in incidence have been similar in populations from both high and low incidence areas, such as Queensland and Hawaii (high rates), and Canada and the United Kingdom (low rates), further suggest that the increases are real (Elwood and Hislop 1982) and associated with a universal factor(s). Incidence rates also demonstrate distinct patterns over time by sex, age, and anatomic site, further suggesting that increases in rates are consistent with changes in sunlight exposure habits by sex and birth cohort (see Chapter 5).

There appears to be general agreement that the sharp rise in the incidence of CMM is associated with increasing exposure to ultraviolet radiation through sun exposure as a consequence of changing clothing and leisure habits (Magnus 1981). The etiologic mechanism of solar radiation in the causation of CMM, however, remains controversial (Magnus 1982).

AGE-SPECIFIC TRENDS IN INCIDENCE AND MORTALITY

The age-specific incidence and mortality curves for CMM are unlike those for most other forms of cancer, which tend to increase linearly with increasing age (Cook et al. 1969; Elwood and Lee 1975; Magnus 1982). In contrast, steep increases in CMM incidence begin in adolescence, leveling off through middle age, followed by less steep increases in the older age groups. Distinct changes in the shape of the age-specific curves occur when rates are stratified by sex and anatomical site (see Chapter 5).

Age-specific incidence curves also change in shape according to the decade of diagnosis, as shown in the Danish population for the years 1943-1982 (Figure 4-l). Among individuals diagnosed from 1943 to 1952, incidence increased gradually with increasing age. Over subsequent decades, the age-specific curves showed progressively steeper increases from ages 20 to 50, followed by a more gentle slope or plateau in the older age groups (Osterlind and Jensen 1986). These changes over time in the shape of the cross-sectional age-specific curves suggest the potential influence of birth cohort effects (Lilienfeld and Lilienfeld 1980). Based on Figure 4-1, it is not surprising that the mean age at diagnosis of CMM has tended to decrease over time in Denmark.

Using cancer registry data from Norway, Magnus (1981) compared the age-specific incidence curves for the two periods, 1955-1970 and 1971-1977, and found that the difference between the two curves was greatest for the age groups 30 to 70 (Figure 4-2). He concluded that this finding was due to birth cohort effects operating primarily on the individuals born between 1900 and 1930. This conclusion will be discussed further in the following section on cohort analyses of CMM incidence and mortality.

COHORT ANALYSES OF CMM INCIDENCE AND MORTALITY

The technique of cohort analysis involves careful study of incidence, mortality, or prevalence rates (i.e., the proportion of individuals with the disease at a specified time period) in individuals born in the same period of time, usually within the same decade. Age-specific rates in cohorts are compared, the major objective being to distinguish the three time-related effects--age, period of diagnosis (i.e., calendar time), and birth cohort--that might explain the changing trends (Kleinbaum et al. 1982). An age effect is present when the disease rate varies by age, regardless of birth cohort; a period effect is present when the disease rate varies by time, regardless of age or birth cohort; a cohort effect is present when the disease rate varies by year of birth, regardless of age (Kleinbaum et al. 1982). Cohort analyses may be conducted graphically, using statistical modelling techniques which attempt to separate the respective contributions of age, period, and birth cohort effects.

As early as 1961, Haenszel suggested on the basis of United States data that "persons born after 1885 have been exposed with increasing intensity to some factor(s) associated with high skin cancer mortality" (Gordon et al. 1961). These increases in skin cancer mortality in successive cohorts were apparently due to malignant melanoma, although CMM was not separately classified until the sixth revision of the International Classification of Diseases in 1950. In 1970, Lee and Carter first associated the long-term trends in total skin cancer mortality with the effects of CMM, and concluded that year-of-birth effects were most likely responsible for the secular increases. Other birth cohort analyses of overall CMM trends have since been conducted by investigators using data from several different countries, all reporting similar findings.

Graphical Analyses of Birth Cohort Effects

Magnus (1981; 1982) plotted age-specific incidence rates for separate birth cohorts in Norway over the period 1955-1977, as shown in Figure 4-3. It can be seen that the risk of malignant melanoma within each cohort rises consistently throughout life, as is true for most other cancers. Graphing CMM rates by birth cohort substantially changed the age-specific curve from that observed in the cross-sectional data; the stable rates in middle age seen cross-sectionally disappeared. Shifting of the birth cohort curves to the left as seen in Figure 4-3 implies that there are consistent increases in incidence for each successive cohort:. The cohort effect in Norway is most marked for individuals born from 1900 to 1930, where distance between the curves is greatest. For example, at: ages 45-49, individuals born in 1920 to 1929 had incidence rates four times higher than individuals born in 1900 to 1909. Cohort curves after 1930 are closer together, suggesting that the differences in incidence rates by cohort are being reduced. One factor postulated as responsible for differences in incidence rates in successive birth cohorts is solar radiation.

Cohort effects are most evident for sites which have shown the greatest increase in incidence over time, such as the trunk in males and lower limbs in females, and are minimal for sites which increased less dramatically, such as the face and neck (Magnus 1981; Houghton et al. 1980; Boyle et al. 1983; Stevens and Moolgavkar 1984).

For CMM of the trunk in Norwegian males 50 to 54 years old, the incidence rate for the 1920-1929 birth cohort was six times that of the 1900-1909 birth cohort (Magnus 1981). As shown in Figure 4-4, among males and females born between 1890 and 1909, the incidence of CMM of the face was greater than that of the trunk among males and lower limbs among females. For cohorts born between 1930 and 1949, however, the highest incidence (per skin surface area) was of the trunk of males and the lower limbs of females. Magnus concluded that the ratio of carcinogenic exposure to the trunk/lower limbs and to the face/neck varied according to year of birth. He suggested that the shift in melanoma distribution by cohort was consistent with changes in clothing and suntanning habits during the first half of this century.

Magnus (1982) notes a slight tendency for the cohort curves to level off with age, particularly in generations born after World War I. The leveling off of the cohort curves is best seen in incidence rates for CMM of the trunk and lower limbs, particularly among Norwegian cohorts born from 1900 to 1929. Magnus notes that this leveling off with age is "rather atypical for most types of cancer," and postulates that it may be due to decreases in sunning with increasing age. This statement implies that sunlight might have a promoting as well as an initiating role in the etiology of CMM.

Birth cohort analyses of incidence rates in Denmark (1943-1972), Connecticut (1935-1974) (Houghton et al. 1980), and Finland (1953-1973) (Teppo et al. 1978) yielded overall findings consistent with those of Magnus (1981, 1982), i.e., increasing incidence rates were observed in successive birth cohorts. The first sign of changing incidence for all sites was observed among those born in 1892 to 1895 in Connecticut and Denmark. Rising incidence began at different times for different anatomic sites in Denmark, with changes in incidence for those born as early as 1882 for the trunk-neck in males and lower limbs in females (Houghton et al. 1980). The incidence curves for facial CMM changed little in successive cohorts.

Muir and Nectoux (1982) conducted cohort analyses with CMM incidence data from Australia, Czechoslovakia, England and Wales, France, Japan, and the Netherlands. Cohort effects were apparent in all six nations, but were most discernible in individuals born from 1910 to 1930. Even in Japan, where incidence rates are 15 times lower than those in Australia, an upward trend in incidence of comparable magnitude was observed in successive cohorts. Later birth cohorts born from 1940 to 1950 did not differ with respect to incidence. This lack of increase in rates among the youngest cohorts may be due to a small number of cases, or might indicate that later cohorts had similar exposure to the carcinogenic factor(s). Muir and Nectoux concluded that the universality of a cohort effect strongly implicates an environmental factor that is widespread, and affects light-skinned people more and both sexes equally, though on somewhat different parts of the body.

Utilizing recent CMM incidence data from the Connecticut Tumor Registry, (1940-1979) Roush et al. (1985b) plotted age-specific incidence rates for seven age groups by birth cohort (Figure 4-5). The age-specific incidence curves were generally parallel on a semi-log scale, showing that rates within age groups increased similarly with each successive cohort. The slopes of the curves, however, showed a tendency to be less steep for individuals in age groups born in 1925 and later. If rates are actually leveling off in the younger cohorts, this could be interpreted to mean that sunlight exposure has peaked in these individuals, i.e., that the maximum change in behavior patterns with regard to leisure time has occurred. This finding is consistent with those of Muir and Nectoux (1982).

Birth cohort analyses of New Zealand's non-Maori population were conducted for the period 1948-1977, revealing different trends for CMM incidence and mortality (Cooke et al. 1983). While incidence rates continued to increase in recent cohorts, mortality rates stabilized in both sexes by the 1924 birth cohort. These mortality changes are similar to those observed in a number of other countries (Elwood and Lee 1974; Lee et al. 1979; Holman et al. 1980). The New Zealand incidence patterns are also similar to those described in Norway, Finland, Denmark, and Connecticut (Magnus 1981, 1982; Teppo et al. 1978, Houghton et al. 1980). Cooke et al. (1983) concluded that both the incidence and mortality data were correct, and that the stabilization of mortality rates was most likely due to improvements in prognosis.

Modelling Approaches in the Evaluation of Cohort Effects

Lee et al. (1979) studied mortality rates by cohort in the United States (white population), England and Wales, and Canada for the period 1951-1975. They calculated age-specific cohort slopes for each sex within each of the three populations, finding large and consistent differences in mortality rates with each successive cohort. These authors concluded that secular increases in mortality over this time period (approximately 3 percent per year) were caused by cohort effects. Although case-fatality (measured as the proportion of CMM cases who died among CMM cases diagnosed over a specified time period) decreased over this period, slopes of the age-specific cohort curves did not appear changed. This observation suggests that any effects of earlier diagnosis or improved treatment occurred evenly over the study period, thereby failing to alter the trend in the slopes.

Holman et al. (1980) examined Australian mortality rates due to malignant melanoma over the period 1931-1977, when rates more than quadrupled in both sexes. Estimates for the independent effects of calendar year, birth cohort, and age on mortality were determined using statistical modelling techniques. On the basis of their analysis, Holman et al. concluded that virtually all of the secular trends in mortality rates could be attributed to increases in successive cohorts, beginning with those born from 1865 to 1885. Increases by birth cohort, however, stabilized by the 1925 (women only) and 1935 (men only) birth cohorts. Slowing of mortality rates has also occurred in cohorts born around this period (1926) in England and Wales (Lee and Carter 1970) and Finland (1930-1940) (Teppo et al. 1978). Holman et al. (1980) emphasized that the stabilization of rates was probably not due to the immigration of persons with lower rates of CMM, because migrants were not over represented among the cohorts in which the rates leveled off (i.e., 1925 and after). Instead, cohort trends in CMM mortality were more likely associated with lifestyle changes involving more recreational exposure to the sun over the generations (Holman et al. 1980). The authors, however, were unable to state whether Australian sun exposure habits have stabilized, and if so, whether those born in 1925 and after would have been the first cohorts affected. Improvements prognosis would probably affect all cohorts equally, and therefore, are not likely to account for the stabilization of mortality rates in the later cohorts.

The individual effects of age, birth cohort, and calendar year derived from CMM mortality rates by Holman et al. (1980) are shown on graphs in Figures 4-6, 4-7, and 4-8. The age factor (Figure 4-6) rose sharply between the age groups 10-14 and 30-34, followed by less rapid increases in the subsequent age groups. The time factor (Figure 4-7), after correction for age and birth cohort, demonstrated very little change with year of death. The birth cohort factor (Figure 4-8) increased with successive cohorts from about 1865 to 1935 in males, and 1865 to 1925 in females. Holman et al. concluded that secular increases in CMM mortality will continue for approximately 30 to 40 years, until the cohorts born before the stabilization of rates (i.e., 1925-1935 cohorts) die.

Utilizing a modelling approach similar to that of Holman et al. (1980), Venzon and Moolgavkar (1984) conducted cohort analyses of CMM mortality in five countries: Australia, New Zealand, the United States, Canada, England, and Wales. These countries were selected to represent populations with high, intermediate, and low rates of CMM in order to determine whether similar time-related effects were present in populations varying in CMM risk. Under all models tested, cohort effects were seen to "drive up mortality" within each country. In addition, the relative increases in mortality due to cohort effects were approximately the same in the five countries. Thus, Venzon and Moolgavkar were able to derive an age-specific mortality curve, corrected for birth cohort effects, using combined data from all five populations. A nearly straight-line relationship of CMM mortality and age was observed, the slope being somewhat less in women than men. The authors stated that the lower slope in females might reflect a larger proportion of female deaths in the younger age groups; it could also be interpreted as reflecting an excess of male deaths in the older age groups. Lee and Storer (1980, 1982) discuss a hormonal risk factor in premenopausal women that could be responsible for higher mortality in young women.

When cohort effects derived from statistical modelling were plotted for the five countries, they were seen to be leveling off in recent cohorts (Venzon and Moolgavkar 1984). The authors suggested that this may represent a slowing down of the increase in incidence of melanomas of the trunk and lower limbs (i.e., sites of greatest cohort effects) or possibly an improving prognosis for these sites.

Both, Stevens and Moolgavkar (1984) and Boyle et al. (1983), modelled the independent effects of age (i.e., correcting for birth cohort effects) on site-specific CMM incidence rates using data from Denmark and Connecticut, and Norway, respectively. Both studies noted rapidly increasing risk by birth cohort for all sites. Stevens and Moolgavkar concluded the fit of their model showed a similar age-dependence for all common subsites of CMM, while Boyle et al. found age-dependent relationships differed by site (see Chapter 5). Discrepancies between the findings and conclusions of Stevens and Moolgavkar (1984) and Boyle et al. (1984) may be associated with the application of different statistical models to different sets of data. In addition, these authors grouped the anatomical sites somewhat differently.

Roush et al. (1985a,b) conducted cohort analyses of CMM incidence data from Connecticut, 1940-1979, using statistical modelling techniques similar to those used by Holman et al. (1980) and Venzon and Moolgavkar (1984). As in similar studies, modelling demonstrated the importance of cohort effects on CMM incidence rates, while period effects were not detected. Roush et al. (1985a) suggested, however, that period effects (i.e., time) could theoretically be present since CMM rates by year of diagnosis show marked fluctuations annually with sunspot activity (Houghton et al. 1978), or seasonally with changes in sunlight exposure (Swerdlow 1979; Holman and Armstrong 1981). The period effects, present as cross-sectional fluctuations in the rates, could be superimposed on the underlying cohort patterns, thus preventing their detection (Roush et al. 1986a). Holman et al. (1983) have suggested that annual or monthly fluctuations in diagnosis of CMM would be consistent with promotional effects of sun on transformed cells in the development of melanoma. Roush et al. (1985a) suggested that period effects and cohort effects may reflect different stages of neoplastic transformation (i.e., promotional and initiating, respectively) in the etiology of CMM.

The dramatically changing public health importance of CMM was emphasized in another recent analysis of Connecticut Tumor Registry data by Roush et al. (1985b). Modelled summary incidence rates (age-adjusted) for cutaneous malignant melanoma were compared with rates of colon cancer within the youngest birth cohorts. The analysis revealed that incidence rates for CMM in the 1955 cohort will rival those for colon cancer, presently the third most common cancer site in Connecticut. Thus, in the coming decades, CMM could easily become one of the most common malignancies in the absence of preventive measures (Roush et al. 1985b).

FINDINGS

The following findings can be drawn from the studies reviewed in this chapter:

4.1 Sharp increases in incidence and mortality have been reported in white but not non-white populations worldwide. Based on observational and analytical evidence, most experts agree that the trends are genuine, and not due to increases in the registration and diagnosis.

4.2 Steeper increases have been reported for incidence versus mortality rates. In addition, there are indications that mortality rates are leveling off in some areas where incidence rates continue to rise annually, such as Australia, Denmark, and New Mexico. Diagnosis at earlier stages of the disease, leading to increased survival, is thought to be a major cause for the leveling off of mortality rates.

4.3 The age-specific incidence curve for CMM is unlike that for most other forms of cancer, which tend to increase linearly with increasing age. Steep increases in CMM incidence begin in adolescence, level off in middle age, and show low rates of increase, if any, in the older age groups. This low slope of age-specific incidence is due to the high lifetime risk of melanoma in younger individuals. The slope of the age-specific incidence curve increases substantially when rates are plotted on a semi-log scale and stratified by birth cohort.

4.4 Most authors who have conducted cohort analyses of CMM incidence and mortality rates conclude that virtually all secular increases in CMM are due to cohort effects. In most countries, the first signs of increasing rates are seen in cohorts born around 1900, although increases in cohorts born as early as 1865 are observed in Australia and New Zealand. In Norway and several other countries, there is a tendency for a slowing of the increase in incidence in cohorts born around and after 1930. Stabilization of mortality rates is also occurring in cohorts born from 1925 to 1935 and later, in countries such as Australia, New Zealand, England and Wales, and Finland.

4.5 On the basis of the Connecticut Tumor Registry data, statistical modelling indicated that the incidence rates of CMM in the 1955 birth cohort will rival those for colon cancer, currently the third most common cancer site in Connecticut.

REFERENCES

Armstrong, B.K., Holman, C.D.J., Ford, J.M., and Wooding, T.L. Trends in melanoma incidence and mortality in Australia. In: Trends in Cancer Incidence, Causes and Practical Implications, Magnus, K. (ed). New York: Hemisphere Publications Corporation, pp. 399-417 (1982).

Balch, C.M., Soong, S., Milton, G.W., Shaw, H.M., McGovern, V.J., McCarthy, W.H., Murad, T.M., and Maddox, W.A. Changing trends in cutaneous melanoma over a quarter century in Alabama, USA, and New South Wales, Australia. Cancer 52: 1748-1753 (1983).

Boyle, P., Day, N.E., and Magnus, K. Mathematical modeling of malignant melanoma trends in Norway, 1953-1978. Am J Epidemiol 118:887-896 (1983).

Boyle, et al. 1984

Cook, P.J., Doll, R., and Fellingham, S.A. A mathematical model for the age distribution of cancer in man. Int J Cancer 4:93-112 (1969).

Cooke, K.R., Skegg, D.C.G., and Fraser, J. Trends in malignant melanoma of skin in New Zealand. Int J Cancer 31:715-718 (1983).

Elwood, J.M., and Lee, J.A.H. Trends in mortality from primary tumours of skin in Canada. Can Med Assoc J 110:913-915 (1974).

Elwood, J.M., and C. Lee, J.A.H. Recent data on the epidemiology of malignant melanoma. Sem Oncol 2:149-154 (1975).

Elwood, J.M., and Hislop, T.G. Solar radiation in the etiology of cutaneous malignant melanoma in Caucasians. NCI Monogr 62:167-171 (1982).

Gordon, T., Crittenden, M., and Haenszel, W. End results and mortality trends in cancer. II. Cancer mortality trends in the United States. NCI Monogr 6:133-350 (1961).

Heston, J.F., Kelly, J.B., Meigs, J.W., Flannery, J.T. Forty-five years of cancer incidence in Connecticut: 1935-79. Cusano, M.M. and Young, J.L., Jr. (eds). NCI Monogr 70:344-349 (1986).

Holman, C.D.J., and Armstrong, B. Skin melanoma and seasonal patterns. Am J Epidemiol 113:202 (1981).

Holman, C.D.J., Armstrong, B.K., and Heenan, P.H. Guest editorial: A theory of the etiology and pathogenesis of human cutaneous malignant melanoma, JNCI 71(4):651-656 (1983).

Holman, C.D.J., Heenan, P.J., Caruso, V., Glancy, R.J., and Armstrong, B.K. Seasonal variation in the junctional component of pigmented neavi. Int J Cancer 31:213-215 (1983).

Holman, C.D.J., James, I.R., Gattey, P.H. and Armstrong, B.K. An analysis of trends in mortality from malignant melanoma of the skin in Australia. Int J Cancer 26:703-709 (1980).

Houghton, A., Munster, E.W., and Viola, M.V. Increased incidence of malignant melanoma after peaks of sunspot activity. Lancet 1:759-760 (1978).

Houghton, A., Flannery, J., and Viola, M.V. Malignant melanoma in Connecticut and Denmark. Am J Cancer 25:95-104 (1980).

Kleinbaum, D.G., Kupper, L.L., and Morgenstern, H. Epidemiologic Research: Principles and Quantitative Methods. Belmont. CA: Lifetime Learning Publications (1982).

Larsen, T.E., Little, J.H., Orell, S.R., and Prade, M. International pathologists' congruence survey on quantitation of malignant melanoma. Pathology 12:245-253 (1980).

Lee, J.A.H. Melanoma. In: Cancer Epidemiology and Prevention. Shottenfeld, D., and Fraumeni, J.F., Jr. (eds). Philadelphia: W.B. Saunders Company (1982a).

Lee, J.A.H. Melanoma and exposure to sunlight. Epidemiol Rev 4:110-136 (1982b).

Lee, J.A.H. Epidemiology of malignant melanoma: 10 years' progress. Pigment Cell 6:1-21 (1983).

Lee, J.A.H., and Carter, A.P. Secular trends in mortality from malignant melanoma. JNCI 45:19-97 (1970).

Lee, J.A.H., Petersen, G.R., Stevens R.G., and Vesanen, K. The influence of age, year of birth, and date on mortality from malignant melanoma in the populations of England and Wales, Canada, and the white population of the United States. Am J Epidemiol 110:734-739 (1979).

Lee, J.A.H., and Storer, B.E., Excess of malignant melanoma in women in the British Isles. Lancet 2:1337-1339 (1980).

Lee, J.A.H., and Storer, B.E. Further studies on skin melanomas apparently dependent on female sex hormones. Int J Epidemiol 11:127-131 (1982).

Lilienfeld, A.M., and Lilienfeld, D.E. Foundations of Epidemiology. New York: Oxford University Press (1980).

Magnus, K. Incidence of malignant melanoma of the skin in Norway, 1955-1970. Variations in time and space and solar radiation. Cancer 32:1275-1286 (1973).

Magnus, K. Epidemiology of malignant melanoma of the skin in Norway with special reference to the effect of solar radiation. In: Biological Characterization of Human Tumors. Exerpta Medica, International Congress Series No. 375, Amsterdam, pp 249-259 (1975).

Magnus, K. Incidence of malignant melanoma of the skin in the five Nordic countries: Significance of solar radiation. Int J Cancer 20:477-485 (1977).

Magnus, K. Habits of sun exposure and risk of malignant melanoma: An analysis of incidence rates in Norway 1955-1977 by cohort, sex, age, and primary tumor site. Cancer 48:2329-2335 (1981).

Magnus, K. (ed). Incidence trends in the Nordic countries: Effects of sun exposure. In: Trends in Cancer Incidence, Causes and Practical Implications. New York: Hemisphere Publication Corporation pp 387-392 (1982).

Malec, E., Eklund, G., and Langerlof, B. Reappraisal of malignant melanoma diagnosis in the Swedish Cancer Registry. Acta Pathol Microbiol Scand 85: 707-712 (1977).

Malec, E., and Eklund, G. The changing incidence of malignant melanoma of the skin in Sweden, 1959-1968. Scand J Plast Reconstr Surg 12:19-27 (1978).

McCarthy, W.H., Blach, A.L., and Milton, G.W. Melanoma in New South Wales: An epidemiologic survey 1970-76. Cancer 46:427-432 (1980).

McKay, F.W., Hanson, M.R., and Miller, R.W. Cancer mortality in the United States: 1950-1977. NCI Monogr 59:222-237 (1982).

Muir, C.S., and Nectoux, J. Time trends: Malignant melanoma of the skin. In: Trends in Cancer Incidence, Causes and Practical Implications, Magnus, K. (ed). New York: Hemisphere Publishing Corp. pp 365-385 (1982).

Osterlind, A., and Jensen, O.M. Trends in incidence of malignant melanoma of the skin in Denmark 1943-1982. Recent Results Cancer Res 102:8-17 (1986).

Pakkanen, M. Clinical appearance and treatment of malignant melanoma of the skin. Ann Chir Gynaecol 66:21-30 (1977).

Pathak, D.R., Samet, J.M., Howard, C.A., and Key, C.R. Malignant melanoma of the skin in New Mexico 1969-1977. Cancer 50:1440-46 (1982).

Pondes, S., Hunter, J.A.A., White, H, McIntyre, M.A., and Prescott, R.J. Cutaneous malignant melanoma in south-east Scotland. Quart J Med, New Series L. 197:103-121 (1981).

Roush, G.C., Schymura, M.J., Holford, T.R., White, C., and Flannery, J.T. Time period compared to birth cohort in Connecticut incidence rates for twenty-five malignant neoplasms. JNCI 74:779-788 (1985a).

Roush, G.C., Schymura, M.J., and Holford, T.R. Risk for cutaneous melanoma in recent Connecticut birth cohorts. Am J Pub Hlth 75:679-682 (1985b).

Sondik, E., Young, J.L., Horm, J.W., and Ries, L.A.G. 1985 Annual Cancer Statistics Review. National Cancer Institute (NCI) (1985).

Stevens, R.G., and Moolgavkar, S.H. Malignant melanoma: Dependence of site-specific risk on age. Am J Epidemiol 119:890-895 (1984).

Swerdlow, A.J. Incidence of malignant melanoma of the skin in England and Wales and its relationship to sunshine. Br Med J 2:1324-1327 (1979).

Teppo, L. Incidence trends of melanoma of the skin in Finland by histological type. In: Trends in Cancer Incidence, Causes and Practical Implications, Magnus, K. (ed) Hemisphere Publishing Corp., pp 393-397 (1982).

Teppo, L., Pakkanen, M., and Hakulinen, T. Sunlight as a risk factor of malignant melanoma of the skin. Cancer 41:2018-2027 (1978).

Venzon, D.J., and Moolgavkar, S.H. Cohort analysis of malignant melanoma in five countries. Am J Epidemiol 119:62-70 (1984).

Viola, M.V., and Houghton, A.N. Solar radiation and cutaneous reviews. Hosp Prac 4:110-136 (1982).