Association of UV Index, Latitude, and Melanoma Incidence in Nonwhite Populations—US Surveillance, Epidemiology, and End Results (SEER) Program, 1992 to 2001

Objective  To estimate the association between UV index, latitude, and melanoma incidence in different racial and ethnic populations in a high-quality national data set.

Design  Descriptive study.

Setting  Eleven US cancer registries that constitute the Surveillance, Epidemiology, and End Results Program (SEER-11).

Patients  Patients with malignant melanoma of the skin reported between 1992 and 2001.

Main Outcome Measures  Pearson correlation coefficients and regression coefficients were used to estimate the relationship of age-adjusted melanoma incidence rates (2000 US standard population) with the UV index or latitude within racial and ethnic groups.

Results  A higher mean UV index was significantly associated with an increase in melanoma incidence only in non-Hispanic whites (r = 0.85, P = .001), although a nonsignificant association was noted in Native Americans (r = 0.42, P = .20).Negative, but not significant, correlations with incidence were observed in blacks (r = −0.53, P = .10), Hispanics (r = −0.43, P = .19), and Asians (r = −0.28, P = .41).Latitude also had a significant correlation with incidence only in non-Hispanic whites (r = −0.85, P = .001). A substantial portion of the variance in registry incidence in non-Hispanic whites could be explained by the UV index (R² = 0.71, P = .001).

Conclusions  Melanoma incidence is associated with increased UV index and lower latitude only in non-Hispanic whites. No evidence to support the association of UV exposure and melanoma incidence in black or Hispanic populations was found.

Skin cancer is the most prevalent cancer in the United States. In 2005, approximately 59 580 Americans will be diagnosed as having melanoma, 1 of the 3 most common types of skin cancer and the one responsible for the most skin cancer deaths.¹ More than 95% of melanomas are diagnosed in light-skinned populations,² and hence most of the epidemiologic evidence is derived from observation of these populations.

In 1956, Lancaster³ first suggested that the geographic distribution of melanoma mortality (increased mortality at lower latitudes) supported an etiologic role of sun exposure. The lower incidence of melanoma in Hispanics, Asians, and blacks has been attributed to the protective effect of darker skin pigmentation.⁴ After 50 years of epidemiologic investigation, UV radiation (UVR) is the paramount risk factor in the etiology of melanoma, at least in light-skinned populations.^5-9

While it is known that melanoma is diagnosed at more advanced stages and confers a worse prognosis in dark-skinned populations, potential etiologic differences are poorly understood.¹ Nonwhite populations have a larger percentage of their cancers on the distal extremities (especially soles of the feet), including acral lentiginous melanoma, than whites do.¹⁰ Areas such as the soles of the feet receive little sun exposure, and hence this finding suggests an alternative cause.^4,11,12 Yet a case-control study by Green and colleagues¹³ of acral lentiginous melanoma in white populations did suggest a link to sun exposure.

Recent studies have suggested a potential positive association of UVR and melanoma incidence in nonwhites.^14-16 The objective of this study was to estimate the association between the UVR and melanoma incidence in US blacks, Hispanics, Asian/Pacific Islanders, and Native Americans by means of a large, geographically diverse system of cancer registries supported by the National Cancer Institute.

Data on malignant melanoma of the skin were obtained from the Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute for the years 1992 through 2001 (Table 1).¹⁷ The expanded SEER program includes 11 cancer registries (SEER-11), which represent approximately 10% to 14% of the US population. These registries include the states of Connecticut, Hawaii, Iowa, New Mexico, and Utah and the metropolitan areas of Atlanta, Ga; Detroit, Mich; San Francisco–Oakland, Calif; Seattle (Puget Sound), Wash; San Jose, Calif; and Los Angeles, Calif. Cases with unknown race and ethnicity, which could not be standardized by race, were excluded from analysis.

Ultraviolet exposure was operationally defined as the annual mean UV index of the registry and latitude. National Weather Service UV index estimates¹⁸ were used to obtain the annual mean UV index for 1997, consistent with previously described methods.¹⁴ Latitude estimates were obtained from the US Census Bureau’s US Gazetteer.¹⁹ The latitude of the state registries was defined as the latitude of the largest city within the state. The average north-to-south latitude for the state registries was also obtained to allow comparison with previous studies.¹⁴

With the use of SEER*Stat 5.2.2,¹⁷ age-adjusted melanoma incidence rates within racial and ethnic groups (2000 US [19 age group] standard population) were calculated for the individual SEER-11 registries. Statistical analysis was conducted with SAS 8.0 software (SAS Institute Inc, Cary, NC). Pearson correlation coefficients were calculated to estimate the correlation of the age-standardized incidence rate of melanoma with the average UV index or latitude. The effect of the annual UV index and latitude was also examined in univariate linear regression, with melanoma incidence rate as the dependent variable. Analyses were repeated in a limited subset of SEER-11 to minimize the influence of unstable estimates due to low numbers of expected melanomas, including only registries with at least 900 000 black residents per the 2000 US census (6 registries included: San Francisco–Oakland, Connecticut, Detroit metropolitan area, Seattle, Los Angeles, and Atlanta metropolitan area), and with Hispanic populations of at least 500 000 (10 registries examined: Hawaii omitted) or at least 1 million Hispanic residents (8 registries examined: Hawaii, Iowa, and Detroit omitted). Statistical analyses were also repeated in the non-Hispanic white population, omitting the Hawaii registry, which was notably different in latitude and incidence and an apparent outlier with a jackknife residual of 3.7.

Between 1992 and 2001, there were 55 353 cases of malignant melanoma in the SEER-11 program: 50 829 cases in white non-Hispanics, 1515 cases in white Hispanics, 293 cases in blacks, 57 cases in Native Americans, 492 cases in Asian/Pacific Islanders, and 2167 cases in persons of unknown race. The age-adjusted incidence rate (per 100 000 population per year) was 21.4 for white non-Hispanics, 4.1 for white Hispanics, 1.0 for blacks, 2.0 for Native Americans, and 1.5 for Asian/Pacific Islanders. Incidence was higher for 1997 to 2001 than for the 5 years earlier for non-Hispanic whites, Hispanic whites, and Asian/Pacific Islanders (Table 2).

Average north-to-south latitude and latitude of the largest city within the state or of the city itself were highly correlated (r = 1.0, P<.001); therefore we report only latitude of the largest city or of the city itself. Latitude was also strongly correlated with the mean annual UV index (r = −0.97, P<.001). The UV index was positively correlated with incidence of melanoma only in non-Hispanic whites (r = 0.85, P = .001) and Native Americans (r = 0.42, P = .20), although the correlation was statistically significant only for the former group (Table 3). Latitude also had a significant correlation with incidence only in non-Hispanic whites (r = −0.85, P<.001). Similar correlations were seen with examination of sex- and race-specific incidence (Table 3).

Scatterplots of age-adjusted melanoma incidence vs the average yearly UV index did not appear to support an association between UV and melanoma for Hispanic whites (Figure 1) or blacks (Figure 2). The incidence-UV scatterplot for non-Hispanic whites did suggest a linear relationship (Figure 3).

Non-Hispanic whites living in Hawaii had the highest age-adjusted incidence: 47.2 cases per 100 000 white population (Figure 3). When data from the 10 continental registries (omitting Hawaii) were analyzed, the Pearson correlation coefficients for UV and latitude with incidence in non-Hispanic whites dropped to 0.41 (P = .24) and −0.49 (P = .15), respectively, and were no longer statistically significant. Analysis of the 6 registries with the highest populations of blacks (r = −0.49, P = .33) or the 8 registries with the highest white Hispanic population (r = −0.15, P = .72) did not change the direction or significance of correlation of UV and incidence.

With univariate linear regression with a dependent variable of the age-standardized incidence in non-Hispanic whites, a substantial portion of the variance could be explained by either UV (R² = 0.71, P = .001) or latitude (R² = 0.74, P = .001). Again, however, when the Hawaii registry was excluded from analysis, substantially less variance in incidence was explained by UV (R² = 0.17, P = .24) or latitude (R² = 0.24, P = .15).

In the SEER program, increased residential UV exposure was significantly associated with increased incidence of melanoma only in non-Hispanic whites. Melanoma incidence was not significantly associated with UVR in blacks, Hispanics, Asian/Pacific Islanders, and Native Americans, with the direction of association for the first 3 groups actually suggesting higher melanoma incidence with less UV exposure. For non-Hispanic whites, a substantial portion of the variance of melanoma incidence explained by UVR was contributed by the Hawaii registry, and when the Hawaii registry was omitted, the association of UVR and melanoma became nonsignificant in the remaining 10 registries in the white non-Hispanic population as well.

There are several limitations of this study. Selection bias due to misclassification of the race and ethnicity of the patient is possible. Although the SEER program has a reputation for quality, Hispanic ethnicity is determined by surname (married as well as maiden names when available), and some Hispanic individuals with indistinguishable surnames may be misclassified as non-Hispanic.⁴ Misclassification by race and ethnicity likely occurs in both directions (ie, missing some Hispanics or including some non-Hispanics). Misclassification of ethnicity is a potential explanation for the absence of cases of melanoma in the white Hispanic population in the Hawaii registry.

Another limitation is that this study examined the association of melanoma and UVR in 5 large race and ethnicity population categories, and hence, significant associations that may exist within smaller subpopulations may be unrecognized.

A notable limitation of this study is that it is a descriptive, observational study and thus (1) is able to demonstrate only association as opposed to causality and (2) is limited by the data available. Most notably, this necessitated using the average geographic UVR of place of residence as opposed to an individual’s exposure measure. This is an important limitation for many reasons. First, individuals may have migrated once or several times during their lifetimes, living in areas of varying UVR. Using the location at diagnosis to estimate UVR may not accurately reflect their cumulative UVR or UVR during critical exposure periods. This limitation may be particularly relevant if youth exposure is indeed a critical factor.^20-22 By using the average UV index as well as latitude in our analysis, we have reduced some of the limitations of using only latitude as a proxy for UVR; however, there may be other confounders that we did not examine, including ambient temperature.²³

Other sources of geographic confounding are possible, and there may be regional behaviors (including recreational and occupational sun exposure), attitudes, or genetic factors that are unaccounted for in this study.²¹ The likelihood of developing melanoma may be a combination of inherited and environmental exposure, and this study explores only the association of race and average residential UVR.²³

It is unclear whether these findings are generalizable outside of the study population. The SEER program encompasses a large and geographically diverse group of registries and hence likely represents the United States in general. These findings may not be generalizable outside of the United States, especially since a single nation, even one as large as the United States, may not provide sufficient latitude or UV gradients. Furthermore, despite the large population analyzed, there were only 11 cancer registries, and one of these (Hawaii) was an outlier in terms of latitude and UVR.

This study has several strengths. The SEER program represents approximately 14% of the US population and contains data on more than 3 million cases of cancer.²⁴ This population-based registry gathers data from large university medical centers, small community hospitals, pathology laboratories, and other sources, and accumulates cancer cases from cities, suburbs, and rural areas of America. Because of its size, scope, and central quality control, SEER-11 provides important cancer data on a diverse group of dark-skinned populations, allowing us to examine melanoma in several low-incidence groups. Furthermore, SEER data are precise and validated, including complex variables such as consistency of histologic diagnosis.²⁵ During the last 30 years, the SEER program has led in guiding US public policy and stimulating progress in cancer research.²⁴

The evidence in the published literature provides meager support at best for a direct association between melanoma incidence and UV flux in brown-skinned populations.^14-16 Following the report by Hu et al,¹⁴ which claimed that solar radiation exposure plays a role in melanoma development in Hispanic and black populations in the United States, we sought to investigate this potential association in SEER program data. Hu and colleagues examined the melanoma incidence from 6 state cancer registries during the late 1990s. While Hu et al found a positive correlation of UV index with melanoma incidence in whites, Hispanics, and blacks, this correlation was significant only in white men and women and in black men (r = 0.93, P = .01). They did not observe significant association in black women, Hispanic men, or Hispanic women.

Krishnamurthy¹⁵ observed patterns of melanoma incidence in India similar to those in whites and suggested that this might be due to UV light exposure. Using data from the National Cancer Registry Project, Krishnamurthy analyzed melanoma from 1982 to 1985 in 7 regions of India.¹⁵ However, the association with latitude was of borderline significance (P = .07) and was nonsignificant for UV-B flux or altitude.

Pennello and colleagues¹⁶ suggested from their findings that sunlight exposure increased skin cancer risk in blacks. However, their examination of SEER data from 1973 through 1994 failed to demonstrate significantly increased incidence rates with increasing UV-B radiation in blacks. They did, however, find an association with melanoma mortality in black men but not in black women.

Although they did not directly investigate the relationship of melanoma incidence and UVR, Bergfelt and colleagues¹² reported that they found a higher incidence of melanoma in Hispanic residents of New Mexico compared with Puerto Rico. Since San Juan, Puerto Rico, is at 18.5°N while Albuquerque, NM, is at 35.1°N,¹⁹ this suggests that UV does not lead to higher melanoma risk.

The populations investigated in this study are diverse in a variety of genetic and other factors that may be relevant to melanoma risk. Hence, the diversity of rates for the studied populations and lack of simple geographic pattern is not entirely surprising. The most important etiologic factors for melanoma in dark-skinned populations remain to be documented.

This investigation failed to find substantial evidence suggestive of a role of UVR in the etiology of melanoma in the US black, Hispanic, Asian/Pacific Islander, and Native American populations. Until better evidence is available regarding primary prevention of melanoma in these populations, public health messages for those with dark skin are best focused on early detection.

Correspondence: Melody J. Eide, MD, MPH, Providence VA Medical Center, 830 Chalkstone Ave (111D), Providence, RI 02908 (Melody_Eide@brown.edu). (After July 1, 2005: Melody J. Eide, MD, MPH, Department of Dermatology, Henry Ford Hospital, 2799 W Grand Blvd, K-16, Detroit, MI 48202-2689 [mje@alum.dartmouth.org]).

Accepted for Publication: December 21, 2004.

Previous Presentation: This study was presented in part at the 14th Annual Meeting of the Photomedicine Society; February 17, 2005; New Orleans, La.

Acknowledgment: This study was supported by a fellowship award from the Dermatology Foundation, Evanston, Ill (Dr Eide), and by grants CSP 402 from the Department of Veterans Affairs, Office of Research and Development, Washington, DC, and CA 106592 from the National Cancer Institute, Bethesda, Md (Dr Weinstock).

Financial Disclosure: None.