In February, 1995, Stacey Peterson, Director, Cancer Data Registry of Idaho (CDRI) was contacted by Colleen LaMay, Idaho Statesman, regarding an investigation of cancer incidence in 17 selected counties. The counties of interest to Ms. LaMay and which, therefore, comprised the study area were: Bannock, Bingham, Blaine, Bonneville, Butte, Caribou, Cassia, Clark, Custer, Fremont, Jefferson, Jerome, Lincoln, Madison, Minidoka, Power, and Twin Falls. The remaining 27 counties in Idaho comprised the area from which comparison cancer incidence rates were calculated. Ms. LaMay was referred to Christopher Johnson, Center for Vital Statistics and Health Policy, Division of Health, who provided a listing of sites and types of cancer with established relationships to ionizing radiation sources (from Biological Effects of Ionizing Radiations V, 1990). Although the timeframe of interest for this study was since 1969, the beginning of CDRI data collection, it was limited to 1971 to 1992, the period for which CDRI possesses a complete, population-based registry of cancer cases in Idaho.
Because the Rocky Mountain Cancer Data System software utilized by CDRI is set up to utilize the state of Idaho as a comparison area for studies of cancer incidence, and not groupings of counties such as comprise the 27 county comparison area, analysis was requested from and provided by the Center for Vital Statistics and Health Policy. Prior to the completion of this report, the methodology and results were reviewed by CDRI staff and the Cancer Cluster Analysis Work Group (Supervisor and Toxicologist, Office of Environmental Health; Supervisor and Health Policy Analyst, Health Statistics Section, Center for Vital Statistics and Health Policy; State Epidemiologist, Bureau of Communicable Disease Prevention; Director, Cancer Data Registry of Idaho).
Using data from the Cancer Data Registry of Idaho (CDRI), the Centers for Disease Control and Prevention (CDC), the Bureau of the Census, and the Center for Vital Statistics and Health Policy, cancer incidence for the 17 counties in the study area was compared to the remainder of the state for the years 1971 to 1992. The cancer sites and types for which incidence comparisons were made were: (1) parotid/salivary, (2) esophagus, (3) stomach, (4) colon, (5) rectosigmoid/rectum, (6) liver, (7) sinus/paranasal, (8) lung/bronchus, (9) bone, (10) female breast, (11) ovary, (12) bladder, (13) kidney/renal pelvis, (14) brain/central nervous system (CNS), (15) thyroid, (16) parathyroid, (17) non-Hodgkin's lymphoma, (18) multiple myeloma, and (19) leukemia. Because of the small number of cases statewide for cancer of the parathyroid, comparisons were not conducted for this site. Cancer cases were abstracted by the CDRI for Idaho residents only. The data elements used were primary cancer site code, histology code, age at diagnosis, sex, and county of residence at diagnosis. Person-years (a measure of the length of time a person is at risk for disease), the denominators for incidence rates, were calculated by adding 1991 and 1992 Center for Vital Statistics and Health Policy population estimates by age, sex, and county to 1971-1990 person-year estimates from the CDC.
Standardized morbidity ratios were calculated using the indirect method. Comparison incidence rates for each site and type of cancer by age and sex were calculated by aggregating cases and person-years for the 27 counties in the comparison area. These rates were multiplied by age and sex specific person-year estimates for each of the counties in the study area to determine the expected number of cases for each site and type of cancer in each study county. Probabilities of the observed numbers of cases exceeding those expected were calculated based upon the Poisson probability distribution, using a one-tailed test. For those cases where expected exceeded observed, no p-values were calculated. All statistical tests were performed using SAS software.
RESULTS AND DISCUSSION
The tables on pages 4 to 13 list the number of observed cases, expected cases, the standardized morbidity ratio, and the unadjusted p-value for each cancer site/type by county. A total of 324 observed to expected comparisons were performed. Based upon chance alone, it was expected that 16 of the 324 comparisons would have shown statistically significant elevations at p=.05. Fourteen statistically significant elevations were found. While adjustment for multiple tests is warranted, the p-values shown do not adjust for the number of tests performed. Using the Bonferroni (Jones and Rushton, 1982) or Tippett's (1952) adjustment, none of the cancer sites or types were significantly elevated at the .05 level in any of the 17 counties studied or the conglomerate study area.
The table on page 14 shows cancer incidence rates for Idaho, 1983 to 1992, as compared with Surveillance and Epidemiology End Results (SEER) rates for 1987 to 1991. SEER is a program of the National Cancer Institute that collects cancer data from 9 states representing 13.9% of the United States population. The states were chosen such that SEER rates are representative of the nation. This provides a frame of reference for interpreting the cancer incidence in Idaho and counties within Idaho. Some of the site/type categories listed on page 14 differ from those presented in the previous tables on pages 4 to 13 due to the availability and comparability of SEER data. Overall, cancer incidence in Idaho is significantly lower than SEER rates.
Several behavioral risks and other factors related to cancer were not addressed in this ecologic study. While all of the cancer sites/types selected for inclusion in this study are sensitive to ionizing radiation (BEIR V: ionizing radiation has been associated with an increase in cancer incidence), the cancer sites/types are not specific to radiation. (Other agents have been linked to increases in incidence, such as smoking tobacco and lung cancer.) Although some studies of irradiated cohorts have provided dose-incidence data for skin cancer, such studies have been complicated by the fact that skin cancer carries a low mortality and is grossly underreported (BEIR V). The Cancer Data Registry of Idaho collects data on basal and squamous cell carcinoma only for mucous membranes. For these reasons, skin cancer was not included in this study. Migration patterns were not taken into account. Persons may have resided in counties other than those in the study area or outside of the state of Idaho for most of their lives, but have been diagnosed with cancer after moving to one of these areas. In such circumstances, conditions in these areas would be unrelated to cancer etiology. Other risk factors not investigated in the study include diet, drinking, smoking habits, and familial history, all of which have been linked with several cancer sites/types.
A fundamental limitation of ecologic studies such as this is the inability to link cancer rates to causes. Because no person-specific exposure data are available, there is no evidence to conclude that elevated environmental contaminants or other etiologic sources caused cancer. Consequently, cause and effect relationships are not discernable for any excess cancer incidence identified by ecologic studies.County Comparisons