Racial Variation in Cancer Incidence: Fact or Artifact? Laurence N. Kolonel, * Marc T. Goodman Accuracy in determining cancer incidence rates for ethnic minorities is of fundamental importance to researchers, health service providers, and policy-makers responsible for the allocation of limited health care resources. Thus, the report by Frost et al. (/) in this issue of the Journal raises serious issues for discussion. In their report, the authors show that a very high percentage of Native Americans in the population-based Surveillance, Epidemiology, and End Results (SEER) registry in the Seattle-Puget Sound area are misclassified, largely as White. Vol. 84, No. 12, June 17, 1992
is of paramount importance. Continued intensive efforts in this area will lead to more efficient diagnosis and to more effective forms of therapy for patients with this disease.
References (/) CHODAK GW: Early detection and screening for prostatic cancer. Urology 34 (suppl 4): 10. 1989 (2) MEIKLE AW. SMITH JA: Epidemiology of prostate cancer. In Early Detection and Treatment of Localized Carcinoma of the Prostate (Smith JA. ed). Philadelphia: Saunders. 1990, pp 709-718 (3) STEINBERG GD, CARTER BS. BEATY TH, ET AL: Family history and the
risk of prostate cancer. Prostate 17:337-347. 1990 (4) CARTER BS, BEATY TH, STEINBERG GD. ET AL: Mendelian inheritance of
familial prostate cancer. Proc Natl Acad Sci USA 89:3367-3371. 1992 (5) PEEHL DM, WONG ST. STAMEY TA: Cytostatic effects of suramin on
prostate cancer cells cultured from primary tumors. J Urol 145:624-630. 1991 (6) BERGERHEIM US. KUNIMI K, COLLINS VP. ET AL: Deletion mapping of
chromosomes 8, 10. and 16 in human prostatic carcinoma. Genes Chromosom Cancer 3:215-220, 1991 (7) CARTER BS, EWING CM, WARD WS, ET AL: Allelic loss of chromosomes
16q and lOq in human prostate cancer. Proc Natl Acad Sci USA 87:8751-8755, 1990 (8) BOVA GS, ICHIKAWA T, ISAACS JT. ET AL. Insertion of human chromo-
somes 8 and 11 into human prostate cancer cell lines using microcell transfer technique. J Urol 147 (suppl):2l5A. 1992 (9) THOMPSON TC, SOUTHGATE J, KITCHNER G, ET AL: Multistage car-
cinogenesis induced by ras and myc oncogenes in a reconstituted organ. Cell 56.917-930. 1989 (10) EFFERT PJ. NEUBAUER A. WALTHER PJ. ET AL: Alterations of the p53 gene
are associated with the progression of a human prostate carcinoma. J Urol 147:789-793, 1992 (//) BOOKSTEIN R. Rio P, MADREPERLA SA, ET AL: Promoter deletion and loss of retinoblastoma gene expression in human prostate carcinoma. Proc Natl Acad Sci USA 81:1162-1166, 1990 (12) SHEVRIN DH, KUKREJA SC, GHOSH L. ET AL: Development of skeletal
metastasis by human prostate cancer in athymic nude mice. Clin Exp Metastasis 6:401^*09, 1988 (13) STEPHENSON RA, DINNEY CPN, GOHJI K. ET AL: Metastatic model for
human prostate cancer using orthotopic implantation in nude mice. J Natl Cancer Inst 84:951-957, 1992 (14) Xu H-J. SUMEGI J. Hu S-X. ET AL: Intraocular tumor formation of RB reconstituted retinoblastoma cells. Cancer Res 51:4481^4485. 1991 (15) MUNCASTER MM, COHEN BL, PHILLIPS RA, ET AL: Failure of RBI to
reverse the malignant phenotype of human tumor cells lines. Cancer Res 52:654-661, 1992
Proper designation of race or ethnicity in medical records requires diligent recording by hospital personnel. Increasingly, medical records do not contain such information, which makes the determination of racial differences in rates progressively more difficult. When race is noted in the medical record, one might assume that the information was self-declared by the patient or a family member. As Frost et al. (/) point out, their data suggest that much of the racial information in these records is based on observation of physical appearance by medical personnel. Appearance alone, however, may lead to substantial misclassification, especially among persons of mixed parentage. In one study (2), 32.3% of self-reported Asians/Pacific Islanders and 70% of self-reported Native Americans/Alaska
Received May 14, 1992; accepted May 15, 1992. Cancer Research Center of Hawaii. Epidemiology Program. University of Hawaii at Manoa. Honolulu. 'Correspondence to: Laurence N. Kolonel, M.D, Ph.D. Cancer Research Center of Hawaii, University of Hawaii at Manoa, Rm. 407. 1236 Lauhala St.. Honolulu, HI 96813.
EDITORIALS 915
Downloaded from http://jnci.oxfordjournals.org/ at University of Georgia on May 30, 2015
injected into the tail vein during clamping of the inferior vena :ava (12); however, further studies will be required to determine how representative this model will be for study of the metastatic phenotype of prostate cancer. A report by Stephenson et al. (13) in this issue of the Journal describes a new metastatic model for human prostate cancer using orthotopic implantation in nude mice. This innovative approach has been used successfully by other laboratories to establish cell lines from human tumors of other histologies. Of particular relevance to the studies of Stephenson et al., Drthotopic implantation has been shown to result in increased rates of metastasis when compared with ectopic (usually sub:utaneous) implantation. The prostate cancer orthotopic transplantation model may provide a much needed tool not only for studying metastasis but also for evaluating the molecular genetic events associated with initiation and progression of prostate cancer. For example, a number of investigators have shown that introduction of the wild-type RB gene into RB-defective cells resulted in a loss of tumorigenicity when the clones were implanted subcutaneously in nude mice. This result was seen in cell lines derived from several different histologies. However, when retinoblastoma cells were transfected with the wild-type RB gene and then implanted into nude mice either ectopically (subcutaneously) or orthotopically (i.e., intraocularly), the ectopic sites were found to have no tumor, while tumor growth was noted in the implanted eye (14,15). These observations suggest that the specific microenvironment at the injection site may play as important a role as the tumorigenicity of the implanted cells in determining whether an implanted tumor will take. The orthotopic transplantation model may thus be extremely valuable for establishing the role of candidate prostate carcinoma susceptibility genes in tumorigenicity, as well as the role of other tumor suppressor genes in the progression of this malignancy. Increasing our understanding of the molecular mechanisms underlying the diverse clinical behavior of prostate carcinoma
916
is more than three times higher than in the southwest, paralle ing similar differences in smoking-prevalence rates (77). The SEER data from the New Mexico registry, which als identifies cancer cases among Native Americans in Arizona, d not suffer the same problems reported by Frost et al. (7) i Washington State. In a recent extensive review of medical rec ords from the Indian Health Service covering the perio 1983-1991, less than 0.1% of the true Native-American can« cases in the New Mexico SEER registry were misclassifie (Valway S: personal communication). Since census informa tion on Native Americans is also more accurate in rural popuk tions (Horm J: personal communication), it is unlikely that th low overall cancer rates reported for Native Americans in th southwest region misrepresent the true occurrence of the dis ease in that population. If cancer incidence data are to be collected and reported o Native-American populations and on other minority ethni groups, it is critical that the data be of high quality. To obtai high quality data, great care is required on the part of thos responsible for the data collection and, of course, adequat resources are required to do the task well. The registry in Ne\ Mexico, for example, works closely with the Indian Healt Service to check the ethnic classification of all cancer case among Native Americans in its database (Key C: persons communication). Because of the heterogeneity of the U.S. population and th necessity to report cancer rates for special populations so th: their needs are served, efforts to accurately classify the racii or ethnic background of individuals are well placed, eve though they may entail additional costs. The report by Frost t al. (/) serves to remind us of the importance of this issue an of the misrepresentation of disease patterns that may resu: from inadequate data.
Downloaded from http://jnci.oxfordjournals.org/ at University of Georgia on May 30, 2015
Natives were classified as White or Black on the basis of appearance alone. Thus, hospital personnel should be required to obtain data on ethnicity by direct inquiry at the time of admission. Variations in racial or ethnic classification also occur in other vital records, such as birth and death certificates. Furthermore, the definitions used to classify individuals differ from one geographic area to another (3). Since vital records are often used by cancer registries to obtain missing information on race, consistency in the procedures for designating race in all such records would be of considerable value. Proper racial assignment of cancer cases is only one component of the problem. Census counts of minorities are also notoriously inaccurate (4). Thus, for the computation of cancer rates, a major issue is the comparability of racial designations in the data used for the numerators and denominators. If the same misclassification occurs in both sources (i.e., individuals are categorized uniformly in hospital records and in the census), the resulting rates may nevertheless closely represent the true occurrence of disease. Accurate cancer rates also depend on the completeness of recording of the cases. If case ascertainment for a racial group is less complete than the census counts, the resulting rates will be underestimated and vice versa. American Indians appeared to be substantially overcounted by the 1980 census, primarily in states with very small Native-American populations not living in Indian areas such as reservations (4). This overestimate would compound the problem of underreported cancer cases noted in the paper by Frost et al. (/) and could contribute further to artifactually lower cancer incidence rates. Another problem related to ethnic differences in selfidentification is illustrated by the Native-Hawaiian population. Historically, persons of mixed Hawaiian ancestry were separately identified from those who were ethnically pure. Accordingly, many Native Hawaiians of mixed parentage report themselves as other races in the census, even though they are classified as Hawaiian in medical records and in other vital records. To deal with this misclassification problem, more precise racial data are collected on a random sample of the population of Hawaii, and this information is used to correct the census estimates (5). Because of the many potential problems in developing cancer rates for minorities and the important uses to which these data are put, such rates should be reported only when a reasonably high degree of accuracy can be assured. The SEER Program is well known for its attention to data quality. Because of limited population sizes and other concerns about data accuracy, regional incidence rates for Native Americans (including Native Hawaiians) have been presented in official SEER publications for the areas covered by the New Mexico and Hawaii registries but not the Seattle-Puget Sound or other SEER registries (6,7). Nevertheless, in a few publications (8,9), a combined data set on Native Americans from several SEER registries has been used. While this approach increases the population base and the stability of the calculated rates, it is meaningful only if the rates are similar in different regions of the country. In fact, site-specific cancer mortality data show substantial differences among Native-American populations (10). For example, lung cancer mortality in the northern plains
References
(/) FROST F, TAYLOR V, FRIES E: Racial misclassification of Native Amcr cans in a Surveillance, Epidemiology, and End Results cancer registry. Natl Cancer Inst 84:957-962, 1992 (2) MASSEY JT: A comparison of interviewer observed race and responder reported race in the National Health Interview Survey. Proc Am Ste Assoc: Soc Stat Section, 1980, pp 425-428 (i) HAHN RA: The state of federal health statistics on racial and ethni groups. JAMA 267:268-271, 1992 (4) PASSEL JS, BERMAN PA: Quality of 1980 census data for America Indians. Soc Biol 33:163-182, 1986 (J) OYAMA N, JOHNSON DB: Hawaii Health Surveillance Program Surve Methods and Procedures. Research & Statistics Report No. 54. Honoluli Research and Statistics Office. Hawaii State Department of Health. 198 (6) YOUNG JL JR, PERCY CL, ASIRE AJ, EDS: Surveillance. Epidemiology
and End Results: Incidence and Mortality Data 1973-1977. Natl Cance Insl Monogr 57:1-1082. 1981 (7) MUIR CS, WATERHOUSE J, MACK T. ET AL, EDS: Cancer Incidence in Fiv
Continents. Vol V. IARC Sci Publ No. 88. Lyon: IARC. 1987 (8) BAQUET CR. RINGEN K. POLLACK ES. ET AL. EDS: Cancer Among Black
and Other Minorities: Statistical Profiles. DHHS Publ No. (NIH 86-2785. Bethesda. Md: NCI. 1986 (9) YOUNG JL JR, RIES LG, POLLACK ES: Cancer patient survival amon ethnic groups in the United States. JNCI 73:341-352. 1984 (10) VALWAY S, ED: Cancer Mortality Among Native Americans in the Unite States: Regional Differences in Indian Health. 1984-1988 and Trend Over Time, 1968-1987. Washington. DC: Indian Health Services. Publi Health Service. US Department of Health and Human Services. 1991. (//) Indian Health Service facilities become smoke-free. MMWR Morb Moi tal Wkly Rep 36:348-350. 1987
Journal of the National Cancer Institut
is of paramount importance. Continued intensive efforts in this area will lead to more efficient diagnosis and to more effective forms of therapy for patients with this disease.
References (/) CHODAK GW: Early detection and screening for prostatic cancer. Urology 34 (suppl 4): 10. 1989 (2) MEIKLE AW. SMITH JA: Epidemiology of prostate cancer. In Early Detection and Treatment of Localized Carcinoma of the Prostate (Smith JA. ed). Philadelphia: Saunders. 1990, pp 709-718 (3) STEINBERG GD, CARTER BS. BEATY TH, ET AL: Family history and the
risk of prostate cancer. Prostate 17:337-347. 1990 (4) CARTER BS, BEATY TH, STEINBERG GD. ET AL: Mendelian inheritance of
familial prostate cancer. Proc Natl Acad Sci USA 89:3367-3371. 1992 (5) PEEHL DM, WONG ST. STAMEY TA: Cytostatic effects of suramin on
prostate cancer cells cultured from primary tumors. J Urol 145:624-630. 1991 (6) BERGERHEIM US. KUNIMI K, COLLINS VP. ET AL: Deletion mapping of
chromosomes 8, 10. and 16 in human prostatic carcinoma. Genes Chromosom Cancer 3:215-220, 1991 (7) CARTER BS, EWING CM, WARD WS, ET AL: Allelic loss of chromosomes
16q and lOq in human prostate cancer. Proc Natl Acad Sci USA 87:8751-8755, 1990 (8) BOVA GS, ICHIKAWA T, ISAACS JT. ET AL. Insertion of human chromo-
somes 8 and 11 into human prostate cancer cell lines using microcell transfer technique. J Urol 147 (suppl):2l5A. 1992 (9) THOMPSON TC, SOUTHGATE J, KITCHNER G, ET AL: Multistage car-
cinogenesis induced by ras and myc oncogenes in a reconstituted organ. Cell 56.917-930. 1989 (10) EFFERT PJ. NEUBAUER A. WALTHER PJ. ET AL: Alterations of the p53 gene
are associated with the progression of a human prostate carcinoma. J Urol 147:789-793, 1992 (//) BOOKSTEIN R. Rio P, MADREPERLA SA, ET AL: Promoter deletion and loss of retinoblastoma gene expression in human prostate carcinoma. Proc Natl Acad Sci USA 81:1162-1166, 1990 (12) SHEVRIN DH, KUKREJA SC, GHOSH L. ET AL: Development of skeletal
metastasis by human prostate cancer in athymic nude mice. Clin Exp Metastasis 6:401^*09, 1988 (13) STEPHENSON RA, DINNEY CPN, GOHJI K. ET AL: Metastatic model for
human prostate cancer using orthotopic implantation in nude mice. J Natl Cancer Inst 84:951-957, 1992 (14) Xu H-J. SUMEGI J. Hu S-X. ET AL: Intraocular tumor formation of RB reconstituted retinoblastoma cells. Cancer Res 51:4481^4485. 1991 (15) MUNCASTER MM, COHEN BL, PHILLIPS RA, ET AL: Failure of RBI to
reverse the malignant phenotype of human tumor cells lines. Cancer Res 52:654-661, 1992
Proper designation of race or ethnicity in medical records requires diligent recording by hospital personnel. Increasingly, medical records do not contain such information, which makes the determination of racial differences in rates progressively more difficult. When race is noted in the medical record, one might assume that the information was self-declared by the patient or a family member. As Frost et al. (/) point out, their data suggest that much of the racial information in these records is based on observation of physical appearance by medical personnel. Appearance alone, however, may lead to substantial misclassification, especially among persons of mixed parentage. In one study (2), 32.3% of self-reported Asians/Pacific Islanders and 70% of self-reported Native Americans/Alaska
Received May 14, 1992; accepted May 15, 1992. Cancer Research Center of Hawaii. Epidemiology Program. University of Hawaii at Manoa. Honolulu. 'Correspondence to: Laurence N. Kolonel, M.D, Ph.D. Cancer Research Center of Hawaii, University of Hawaii at Manoa, Rm. 407. 1236 Lauhala St.. Honolulu, HI 96813.
EDITORIALS 915
Downloaded from http://jnci.oxfordjournals.org/ at University of Georgia on May 30, 2015
injected into the tail vein during clamping of the inferior vena :ava (12); however, further studies will be required to determine how representative this model will be for study of the metastatic phenotype of prostate cancer. A report by Stephenson et al. (13) in this issue of the Journal describes a new metastatic model for human prostate cancer using orthotopic implantation in nude mice. This innovative approach has been used successfully by other laboratories to establish cell lines from human tumors of other histologies. Of particular relevance to the studies of Stephenson et al., Drthotopic implantation has been shown to result in increased rates of metastasis when compared with ectopic (usually sub:utaneous) implantation. The prostate cancer orthotopic transplantation model may provide a much needed tool not only for studying metastasis but also for evaluating the molecular genetic events associated with initiation and progression of prostate cancer. For example, a number of investigators have shown that introduction of the wild-type RB gene into RB-defective cells resulted in a loss of tumorigenicity when the clones were implanted subcutaneously in nude mice. This result was seen in cell lines derived from several different histologies. However, when retinoblastoma cells were transfected with the wild-type RB gene and then implanted into nude mice either ectopically (subcutaneously) or orthotopically (i.e., intraocularly), the ectopic sites were found to have no tumor, while tumor growth was noted in the implanted eye (14,15). These observations suggest that the specific microenvironment at the injection site may play as important a role as the tumorigenicity of the implanted cells in determining whether an implanted tumor will take. The orthotopic transplantation model may thus be extremely valuable for establishing the role of candidate prostate carcinoma susceptibility genes in tumorigenicity, as well as the role of other tumor suppressor genes in the progression of this malignancy. Increasing our understanding of the molecular mechanisms underlying the diverse clinical behavior of prostate carcinoma
916
is more than three times higher than in the southwest, paralle ing similar differences in smoking-prevalence rates (77). The SEER data from the New Mexico registry, which als identifies cancer cases among Native Americans in Arizona, d not suffer the same problems reported by Frost et al. (7) i Washington State. In a recent extensive review of medical rec ords from the Indian Health Service covering the perio 1983-1991, less than 0.1% of the true Native-American can« cases in the New Mexico SEER registry were misclassifie (Valway S: personal communication). Since census informa tion on Native Americans is also more accurate in rural popuk tions (Horm J: personal communication), it is unlikely that th low overall cancer rates reported for Native Americans in th southwest region misrepresent the true occurrence of the dis ease in that population. If cancer incidence data are to be collected and reported o Native-American populations and on other minority ethni groups, it is critical that the data be of high quality. To obtai high quality data, great care is required on the part of thos responsible for the data collection and, of course, adequat resources are required to do the task well. The registry in Ne\ Mexico, for example, works closely with the Indian Healt Service to check the ethnic classification of all cancer case among Native Americans in its database (Key C: persons communication). Because of the heterogeneity of the U.S. population and th necessity to report cancer rates for special populations so th: their needs are served, efforts to accurately classify the racii or ethnic background of individuals are well placed, eve though they may entail additional costs. The report by Frost t al. (/) serves to remind us of the importance of this issue an of the misrepresentation of disease patterns that may resu: from inadequate data.
Downloaded from http://jnci.oxfordjournals.org/ at University of Georgia on May 30, 2015
Natives were classified as White or Black on the basis of appearance alone. Thus, hospital personnel should be required to obtain data on ethnicity by direct inquiry at the time of admission. Variations in racial or ethnic classification also occur in other vital records, such as birth and death certificates. Furthermore, the definitions used to classify individuals differ from one geographic area to another (3). Since vital records are often used by cancer registries to obtain missing information on race, consistency in the procedures for designating race in all such records would be of considerable value. Proper racial assignment of cancer cases is only one component of the problem. Census counts of minorities are also notoriously inaccurate (4). Thus, for the computation of cancer rates, a major issue is the comparability of racial designations in the data used for the numerators and denominators. If the same misclassification occurs in both sources (i.e., individuals are categorized uniformly in hospital records and in the census), the resulting rates may nevertheless closely represent the true occurrence of disease. Accurate cancer rates also depend on the completeness of recording of the cases. If case ascertainment for a racial group is less complete than the census counts, the resulting rates will be underestimated and vice versa. American Indians appeared to be substantially overcounted by the 1980 census, primarily in states with very small Native-American populations not living in Indian areas such as reservations (4). This overestimate would compound the problem of underreported cancer cases noted in the paper by Frost et al. (/) and could contribute further to artifactually lower cancer incidence rates. Another problem related to ethnic differences in selfidentification is illustrated by the Native-Hawaiian population. Historically, persons of mixed Hawaiian ancestry were separately identified from those who were ethnically pure. Accordingly, many Native Hawaiians of mixed parentage report themselves as other races in the census, even though they are classified as Hawaiian in medical records and in other vital records. To deal with this misclassification problem, more precise racial data are collected on a random sample of the population of Hawaii, and this information is used to correct the census estimates (5). Because of the many potential problems in developing cancer rates for minorities and the important uses to which these data are put, such rates should be reported only when a reasonably high degree of accuracy can be assured. The SEER Program is well known for its attention to data quality. Because of limited population sizes and other concerns about data accuracy, regional incidence rates for Native Americans (including Native Hawaiians) have been presented in official SEER publications for the areas covered by the New Mexico and Hawaii registries but not the Seattle-Puget Sound or other SEER registries (6,7). Nevertheless, in a few publications (8,9), a combined data set on Native Americans from several SEER registries has been used. While this approach increases the population base and the stability of the calculated rates, it is meaningful only if the rates are similar in different regions of the country. In fact, site-specific cancer mortality data show substantial differences among Native-American populations (10). For example, lung cancer mortality in the northern plains
References
(/) FROST F, TAYLOR V, FRIES E: Racial misclassification of Native Amcr cans in a Surveillance, Epidemiology, and End Results cancer registry. Natl Cancer Inst 84:957-962, 1992 (2) MASSEY JT: A comparison of interviewer observed race and responder reported race in the National Health Interview Survey. Proc Am Ste Assoc: Soc Stat Section, 1980, pp 425-428 (i) HAHN RA: The state of federal health statistics on racial and ethni groups. JAMA 267:268-271, 1992 (4) PASSEL JS, BERMAN PA: Quality of 1980 census data for America Indians. Soc Biol 33:163-182, 1986 (J) OYAMA N, JOHNSON DB: Hawaii Health Surveillance Program Surve Methods and Procedures. Research & Statistics Report No. 54. Honoluli Research and Statistics Office. Hawaii State Department of Health. 198 (6) YOUNG JL JR, PERCY CL, ASIRE AJ, EDS: Surveillance. Epidemiology
and End Results: Incidence and Mortality Data 1973-1977. Natl Cance Insl Monogr 57:1-1082. 1981 (7) MUIR CS, WATERHOUSE J, MACK T. ET AL, EDS: Cancer Incidence in Fiv
Continents. Vol V. IARC Sci Publ No. 88. Lyon: IARC. 1987 (8) BAQUET CR. RINGEN K. POLLACK ES. ET AL. EDS: Cancer Among Black
and Other Minorities: Statistical Profiles. DHHS Publ No. (NIH 86-2785. Bethesda. Md: NCI. 1986 (9) YOUNG JL JR, RIES LG, POLLACK ES: Cancer patient survival amon ethnic groups in the United States. JNCI 73:341-352. 1984 (10) VALWAY S, ED: Cancer Mortality Among Native Americans in the Unite States: Regional Differences in Indian Health. 1984-1988 and Trend Over Time, 1968-1987. Washington. DC: Indian Health Services. Publi Health Service. US Department of Health and Human Services. 1991. (//) Indian Health Service facilities become smoke-free. MMWR Morb Moi tal Wkly Rep 36:348-350. 1987
Journal of the National Cancer Institut
