This project conducts a variety of descriptive epidemiologic studies to quantify cancer incidence and mortality; to investigate variations in cancer rates by demographic factors; to examine temporal trends and geographic patterns; and to identify leads for further analytical research. Geographic Variation in Cancer Rates Evaluation of geographic variation in cancer rates may suggest clues to the roles of environmental or cultural influences. Identification of regions at notably high or low risk may indicate areas where more intensive studies might be particularly fruitful. We published a new atlas of cancer mortality in the United States during 1950-94. Data files pertaining to more than 9.5 million whites and 1.1 million blacks who died during 1970-94 and 4.8 million whites who died during 1950-69 from cancer and corresponding estimates of person-years at risk were used by the """"""""Mortality Rate Generator"""""""" (MRG) software program to calculate observed and expected counts, age-adjusted rates, and confidence limits for more than 40 forms of cancer by race, sex, time period, and geographic area. Using special graphics software for Windows on a personal computer, more than 250 maps at the county or State Economic Area (SEA) level were generated and included in the atlas. Summary tables and figures were prepared. The updated maps show that the patterns previously observed for several cancers have persisted, such as the broad stretches of high rates for cancers of the breast, colon, and rectum in the Northeast, although the regional variation has diminished somewhat as rates have risen in many areas of the South. For some tumors, the geographic clustering of areas with elevated rates has become more pronounced in the recent time period, as shown for cancers of the corpus uteri, prostate, bladder, and biliary tract. For lung cancer, there have been remarkable changes in the geographic patterns corresponding to regional/temporal variations in smoking trends by sex and race, with the recent emergence of high mortality rates among white men across the South, among white women in the far western states, and among blacks in northern urban areas. The updated geographic patterns should help in formulating etiologic and other hypotheses, and in targeting high-risk populations for further epidemiologic research and cancer control interventions. The text, maps, rates, and the data used to generate the maps are available on the internet (www.nci.nih.gov/atlasplus). In addition to an online version of the Atlas, there is a component of the Web site that allows the user to create customized maps based on parameters selected by the user from a menu. The user can determine map color, number of data ranges and the percent of deaths to be included in each range, and ranging method (geographic patterns or time trends). The user can also view an entire state at the SEA or county level, as well as view an individual SEA or county. In each case, a table is included with the map and legend showing the mortality rates, number of deaths, statistical significance, and sparseness status of the selected geographic entity, as well as the ones in which it is contained. The Web site database has been expanded to include county data for Blacks, SEA and state data for Other non-whites, data for the nine 5-year periods from 1950-54 to 1990-94, and four age groups (0-19, 20-49, 50-74, and 75+). Multiple maps displaying geographic patterns or time trends are also available, and these can be animated. Furthermore, dynamic charts displaying mortality rates at the state, SEA, or county level are available. Future enhancements will add rates calculated in a variety of ways. A Web-based version of MRG will be added to the site, enabling the user to have more flexibility in choice of cancers, age groups, and sex and race aggregation. Mapping techniques can highlight the spatial or temporal variations in rates of cancer mortality. In mapping geographic patterns of cancer mortality, spatial units are grouped into categories defined by specified rate ranges, and then the units in each category are assigned a particular color in the map. We examined the consequences of using different ranging methods when comparing maps over several time intervals. Two ranging methods were employed: 1) Ranges defined for individual time interval by the deciles of rates in that interval (ranging within intervals), and 2) constant ranges for all time intervals defined by the deciles of rates for the entire 45-year period from 1950 through 1994 (ranging across intervals). Ranging within time intervals displays geographic patterns and changes in geographic patterns, regardless of time trends in rates. Ranging across time intervals shows temporal changes in rates but with some loss of information about geographic variability. We have been analyzing changes in the geographic patterns for those cancers with substantial variation by area, race, and/or gender. Lung cancer mortality rates among white males in the United States were observed to be elevated during 1950-69 in counties with shipbuilding industries during World War II; risk was found to be associated with asbestos exposure. When we evaluated the geographic patterns in more recent years, 1970-94, for whites and compared them with the 1950-69 patterns, we found that rates generally were higher in shipyard counties than in all nonshipyard counties and in coastal nonshipyard counties for both sexes and time periods, rates increased markedly from 1950-69 to 1970-94 in all groups, with the changes more pronounced in females than males, which may be attributed to the combined effects of low asbestos exposures and changes in smoking behavior. From 1950-1954 through 1990-1994, national melanoma mortality rates among whites increased by 191% and 84% among males and females, respectively. The state-specific rates exhibited a strong but diminishing latitudinal gradient, reflecting the complex interplay of UV radiation levels in each geographic region, the sun-protection behaviors of each generation of males and females in childhood and adulthood, the geographic mobility of the population, and the risk awareness and early detection. Ultraviolet B (UVB) radiation exposure increases the risk of skin cancer in whites. Motivated by indications that United States geographic variation of relative skin cancer risk in blacks approaches that in whites, we used Poisson regression to estimate the risk of skin cancer in blacks as a function of average annual surface-levels of UVB radiation, measured by Robertson-Berger meters. For black males, the age-adjusted relative risk of mortality for a 50% increase in UVB radiation was significantly above one for malignant melanoma, 1970-1994 (1.16; 95% confidence interval, 1.02-1.32) and nearly so for nonmelanoma skin cancer, 1970-1981 (1.18; 95% confidence interval, 1.00-1.39), for which the time period was chosen to avoid AIDS-related deaths from Kaposi's sarcoma. Although the public health implication is uncertain because of the much lower absolute risk of skin cancer in blacks compared with whites, the findings suggest that sunlight exposure increases skin cancer risk in blacks. Research into the geographic patterns and their determinants is ongoing. The geographic patterns for breast cancer mortality have remained remarkably static, but are more pronounced for women older than age 50 years than for younger women. We are examining the age-specific patterns in greater detail over time and comparing them among white and black women. We are using a spatial scan statistic using county-based mortality data from 1970 through 1989 to test whether the elevated regional rates of prostate cancer mortality shown in the recent atlas of cancer mortality were statistically different from rates in the rest of the country and whether they could be explained by known or hypothesized risk factors. Generating New Hypotheses based on Cancer Trends and Demographic Patterns We analyzed the descriptive epidemiology for cancers of the esophagus, bladder and kidney, the leukemias, as well as for cancers among children, among the elderly, and among women, using national and international incidence and mortality data. Our analysis of the recent U.S. trends in lung cancer mortality found an unexpected worsening of the birth cohort trend in risk for people born after 1950; we also found a significant decrease in the slope of the calendar period trend in 1990 for both whites and blacks, observed primarily in people 55 years of age and older. Prostate cancer is the most commonly diagnosed cancer in western men, and incidence is rising rapidly in most countries, including low-risk populations. Age-adjusted incidence and mortality rates from 15 and 13 countries between 1973-77 and 1988-92, respectively, were compared to provide leads for future analytic studies. Large increases in both incidence and mortality rates of prostate cancer were seen for all countries. Increasing incidence rates in the United States and Canada are likely to be due in part to the widespread use of transurethral resection of the prostate and prostate-specific antigen testing, while increases in the Asian countries are probably related to westernization in these low-risk populations. The large disparities in incidence between high- and low-risk countries may be due to a combination of genetic and environmental factors. Future studies are needed to examine gene-gene and gene-environment interactions in various countries concurrently to shed light on the etiology of prostate cancer and to help elucidate reasons for the large differences in risk between populations. In contrast to mortality data, which are limited to specifying the form of cancer, incidence data include information on histologic type of the tumor and in many instances, the subsite of origin. We have used incidence data from the Surveillance, Epidemiology, and End Results (SEER) program to investigate further the demographic patterns to discern subgroups that may be of etiologic significance. Clinical investigations have shown prognostic heterogeneity within the non-Hodgkin's lymphomas (NHLs) according to histology, but few descriptive studies have considered NHLs by subgroup. We assessed the demographic patterns and trends in population-based rates of different histologic subgroups of NHL. Overall, the broad categories of small lymphocytic, follicular, diffuse, high-grade, and peripheral T-cell NHL emerged as distinct entities with specific age, sex, racial, temporal, and geographic variations in rates. Findings from our large, population-based study reveal differing demographic patterns and incidence trends according to histologic group. Future descriptive and analytic investigations should evaluate NHL risks according to subtype, as defined by histology and new classification criteria. Oral cavity and pharynx (OCP) cancers include a variety of histologic types and anatomical sites. We evaluated age-adjusted incidence rates based on data for cases diagnosed in nine SEER registries for OCP cancers by histologic type, anatomical site, race, and sex to identify subgroups that may be etiologically distinct. During the 1975-1996 period, 86% of OCP cancers were squamous cell carcinomas (SCC), 7% adenocarcinomas (AC), 2% Kaposi's sarcoma (KS) and 5% other or non-specified histologic types. The male/female rate ratios ranged from less than 1.2 for AC to 3.8 for SCC and more than 10.0 for KS. Male black/white rate ratios exceeded two for cancers of the palate, tonsil and other pharyngeal subsites, except the nasopharynx. In contrast, rates were higher among white males than black males for cancers of the lip and salivary glands, and very similar for gum (gingival) cancer. Among females, rates by race were similar for SCC and AC and all oral sites with the exception of lip, whereas rates for most subsites in the pharynx were higher among blacks than whites. The data suggest that OCP cancers may be separated into five subgroups: SCC of the lip, SCC of the oral cavity, SCC of the pharynx, AC, and KS. Ongoing analyses are investigating the international patterns and trends in liver cancer rates and prostate cancer rates according to ethnic group and place of residence. Using the most recent data, we are examining the U.S. age-specific incidence and mortality patterns among whites and blacks by sex and cohort year of birth for lung cancer and for kidney cancer. Using the SEER incidence data, we are investigating the ovarian cancer incidence patterns by histologic type. Based on the findings of our recent paper on non-Hodgkin's lymphoma, we are comparing the patterns of several subtypes that may be etiologically distinct. Using Descriptive Data to Assess Consistency with Hypotheses Cancer incidence and mortality rates may be used to assess consistency with hypotheses regarding cancer etiology suggested by other scientific studies. To further our understanding of the rising rates of cutaneous melanoma and to determine whether they are diagnostic artifact or real, we examined the most recent incidence patterns among whites by sex, age, tumor stage and thickness using data from the SEER program. We found elevated risks for females born since the 1960s. In the more recent time period 1990-91 through 1996-97, in general, age-specific rates for distant stage tumors stabilized or declined more among females than males. In addition, rates for thick tumors (> 4 mm) increased significantly (P <.05) only in males aged 60 years and older. Thus, the recent trends reflect increased sunlight exposure more than diagnosis. Ovarian cancer incidence and mortality rates have declined among U.S. women age 35-59 years during the period 1970-1995. Epidemiologic studies have shown that ovarian cancer risk decreases with increasing parity and increasing duration of oral contraceptive use. During this period, parity declined while oral contraceptive use increased. We compared temporal trends in observed ovarian cancer incidence rates with rates predicted by changes in parity and duration of oral contraceptive use to determine whether the changes in these characteristics could explain the declining rates in younger women. In addition, we wished to examine whether oral contraceptive use continues to be protective to postmenopausal women. We found that the predicted rates agreed well with observed rates in young women (age 30-49) but were substantially lower than observed rates in older women (age 50-64). The data indicated that the relative decrease in incidence rates due to the protective effect of oral contraceptive use declines with age. To evaluate positive findings from an earlier report, we studied the relation between retinoblastoma incidence and ultraviolet (UV-B) radiation levels in the US using weighted regression, as well as in international data after adjusting for race, economic development, and climate. The findings suggested that environmental factors other than UV-B may be responsible for the geographic patterns of retinoblastoma. International Collaborations We are continuing our analyses of data from Shanghai with our Chinese collaborators. We updated the incidence trends using population-based data from the Shanghai Cancer Registry for 1972-1994. During 1993-1994, cancers of the lung, stomach, and liver were the 3 leading forms among men, followed by cancers of the colon and esophagus. Among women, cancers of the breast, stomach, and lung were the most common tumors, followed by cancers of the colon and liver. Over the 23-year period, the rate for all cancers combined, excluding non-melanoma skin cancer, decreased from 247.5 to 215.2 per 100,000 person-years among men and from 173.6 to 154.0 among women. However, trends for individual forms of cancer varied considerably. Rates doubled for cancers of the colon and biliary tract in both sexes, and they increased substantially for cancers of the brain and nervous system, kidney, pancreas, prostate, corpus uteri, female breast, and ovary, and for non-Hodgkin's lymphoma. Rates for cancers of the lung and rectum changed little. Rates declined by at least one-half for cancers of the esophagus and cervix, with notable decreases also for cancers of the stomach and liver. Some of these trends may reflect variations in diagnostic or screening practices, although changes in lifestyle and other environmental exposures are likely to play important roles. Further epidemiologic research in China is needed to identify risk factors influencing the cancer incidence trends. We analyzed in greater detail the rapidly declining incidence rates for cervical cancer in Shanghai. Based on more than 7000 cases of cervical cancer diagnosed during 1972-94 among female residents of the ten urban areas, rates declined from 26.7/100,000 in 1972-74 to 2.4/100,000 in 1993-94, dropping in rank from 1st to 15th among all cancers among females. Decreases occurred in all age groups and approached 90% among women aged 35-64. A cervical cytology program was implemented in Shanghai in the late 1950s. The wide availability of Pap smears and treatment for cervical precursor lesions and improved personal hygiene are responsible for the large reductions in incidence. Continued detection of early-stage cases by systemic cytology screening are needed to lower incidence further. We are continuing our collaboration using the Shanghai data to compare the colorectal cancer incidence trends with estimates of consumption of various types of food and to investigate in greater detail the trends in lung cancer incidence. With our collaborators in Beijing, China, we mounted a pilot study, using data from five provinces, to investigate the feasibility of comparing cancer mortality rates from the sample survey of 1990-92 with rates from the full survey of 1973-75. Preliminary results suggest that we can proceed, and we recently received data for the full 27 provinces for both time periods. We are editing the data and preparing for our full analysis.
