Cancer Incidence Rates in Eddy and Lea Counties

New Mexico, 1970-1994

Issued: April 6, 1998

MATERIALS AND METHODS

This report presents cancer incidence rates, analyzed both temporally and spatially, for Eddy and Lea counties. The cancer data used to calculate incidence rates in this report were collected by the NMTR. The NMTR was established in 1966. In 1973, the NMTR received funding from the National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results (SEER) Program to become one of a network of 11 registries across the US. This network is the primary source of cancer incidence data for the nation. The NMTR serves as a repository and data collection agent for confidential medical abstracts of cancer cases, and reports data on incidence, treatment, mortality, and survival to the NCI. These cases are followed on a yearly basis. The NMTR maintains a 95% follow-up rate of all cancer patients. Each of the SEER registries is a population-based registry and collects data in a standard manner so that it can be compared and combined with the data from other registries in the network. The SEER registries are distributed throughout the US as an accurate representation of the US population, and were chosen for their ability to meet the NCI's standards. The membership of the SEER network has changed slightly over the years, but at the time of this study, data were reported from the states of New Mexico, Connecticut, Iowa, Utah, and Hawaii, and the cities of, and areas surrounding, Detroit, San Francisco, Seattle-Puget Sound, Atlanta, San José, and Los Angeles, representing approximately 14% of the US population.

NMTR tumor registrars and abstractors work with tumor registrars at hospitals and treatment centers to collect cancer data. NMTR's database includes information submitted by hospitals and physician's offices, and data collected during site visits to facilities in New Mexico and other states by NMTR staff. The information used for this report includes data collected at the Guadalupe Medical Center (now Columbia Medical Center of Carlsbad) in Carlsbad, New Mexico, as well as at healthcare facilities and pathology laboratories statewide and in adjacent Texas counties. NMTR staff abstract and code information on the patient (name, maiden name, date of birth, ethnicity, social security number, place of birth, and place of residence); information on the cancer (date diagnosed, location of the cancer, stage, size, and lymph node involvement); information on any surgical treatment, radiation, or chemotherapy the patient receives; and the outcome (survival or death).

The data used for spatial analysis in this report include the NMTR's data set of cases of invasive cancers newly diagnosed in the period between 1970 and 1992, and data used for temporal analysis includes cases newly diagnosed in the period between 1970 and 1994. The data set used for the spatial analysis was extracted from the database on December 8, 1995, and the data set used for the temporal analysis was extracted from the database on January 11, 1996. It is important to note that these data sets reflect cases ascertained at the time the data were analyzed. For more recent years, a small percentage of cases (1 to 2%) may be added or redistributed geographically.

Historically, invasive and in situ cancers have been counted separately (ACS 1995). One reason for this is that in situ cancers are often removed in a physician's office instead of a hospital, and so may be reported differently. Also, after an in situ cancer has been removed, it is not known whether the cancer would have become invasive (i.e., spread). Thus, a reliable way to assess patterns in cancer incidence rates is to compare rates of invasive cancers only. An exception to this is bladder cancer (NCI 1993). Because of biological properties of tumors in the bladder, the line between when a tumor is considered in situ and when it is considered invasive is not clear. The medical community's definition of in situ and invasive bladder cancer may have changed slightly over the years, but the NMTR and SEER definition has remained consistent. Thus, SEER bladder cancer data are comparable through the years, but may not be comparable with international data.

Study Design

A large number of contaminants are present in the waste to be placed at the WIPP (Table 2). Known and suspected carcinogens slated for disposal at the WIPP, and the cancers they may cause, are presented in Table 3. This study addressed cancers of major anatomic sites caused by exposure to radionuclides and substances classified by IARC as Group 1 , indicating that there is sufficient evidence that the substance is carcinogenic to humans (IARC 1987). In addition, to provide a frame of reference, we assessed several cancers of tissues less sensitive to the types of hazardous and radioactive wastes to be disposed of at the WIPP. These sites and the corresponding SEER Program site categories used are presented in Table 4.

Cancer incidence rates are known to vary by sex and ethnicity. Rates were stratified to examine and control for the differences resulting from this variation. Because Eddy and Lea counties are comprised mainly of non-Hispanic whites and Hispanic whites, the study focused on these two ethnic groups. In addition, the study employed only age-adjusted rates. However, the study did analyze data for childhood (under 15 years old) cancers of all sites and for childhood leukemia. Leukemia is of interest because it is a radiation-sensitive cancer and comprises about one-third of all childhood cancers (NCI 1993).

Several limitations are inherent in the interpretation of cancer incidence data for small populations like those in Eddy and Lea counties. The limitations include the variability of cancer incidence data for small numbers of cancer cases and/or small populations, and the differences in the magnitude of the effects caused by various risk factors. For example, smoking the equivalent of 20 pack years (i.e., one pack per day for 20 years, 2 packs per day for 10 years, etc.) increases lung cancer risk by ten times (CCDPHP 1989). Another additive exposure that increased lung cancer risk by 50% in non-smokers would produce a 10.5-fold increase in the same group of smokers. Thus, even a relatively large increase in risk (50%) would be difficult to detect using standard epidemiologic methods.

Data Analysis

The population estimates for New Mexico used to calculate incidence rates were derived from the censuses of 1960, 1970, 1980, and 1990. Estimates of the Hispanic population in 1960 and 1970 were adjusted for systematic errors in census data processing and for the different techniques used to identify individuals of Hispanic ethnicity (Howard et al. 1983). Age-adjusted incidence rates were calculated by the direct method, using the age distribution of the 1970 US population as a standard to weight age-specific incidence rates in the populations of interest (NCI 1993).

The cancer incidence rates presented are estimates of the average annual number of people who will be diagnosed with cancer per 100,000 people (per 1,000,000 people for childhood cancers). To provide a more accurate picture of the inherent variability in incidence rate estimates, 95% confidence intervals are presented, which are estimates of the range of values within which there is 95% certainty that the true rate lies (Shambaugh et al. 1994). The calculated incidence rates are averages (called means) and approximate a bell-shaped or normal curve when plotted on a graph. Normal curves are highest in the middle at the mean, and fall off symmetrically to the sides. The standard deviation of a normal curve quantifies the dispersion of values to the left and right of the mean. In a normal curve, 95% of the values lie between the center of the curve and 1.96 standard deviations in either direction (Hennekens and Buring 1987). The standard error of the mean measures the variability of the mean as an estimate of its true value. In this study standard errors were calculated for the cancer incidence rate estimates, assuming a Poisson distribution and using the method described by Keyfitz (1966), by dividing each incidence rate by the square-root of the corresponding number of cases (incidence rate/(cases)1/2) (NCI 1993). Confidence intervals are a function of both standard deviation and standard error (Hennekens and Buring 1987). The upper 95% confidence interval was calculated by adding the product of 1.96 and the standard error to the incidence rate (incidence rate + 1.96 * standard error). The lower 95% confidence interval was calculated by subtracting the product of 1.96 and the standard error from the incidence rate (incidence rate - 1.96 * standard error). When the lower 95% confidence interval was negative, the value was truncated at zero.

Data for spatial and temporal analyses were stratified by ethnicity and sex. Spatial trends in incidence rates were assessed by comparing incidence rates from 1970 to 1992 for Eddy and Lea counties with average rates for New Mexico and all other New Mexican counties. Temporal trends in incidence rates were assessed by calculating 3-year moving average incidence rates using data from 1970 to 1994 for Eddy and Lea counties and comparing them with rates for New Mexico and the US. Moving average incidence rates, derived by calculating the mean of rates over a specified time period (3 years, in this report), were used to minimize the sharp fluctuations typical when calculating rates for small populations such as those in Eddy and Lea counties. Confidence bounds for the moving averages were not calculated; thus descriptions of "trends" or other temporal patterns are descriptive rather than statistical interpretations.

Limitations of the Data

Beyond the factors discussed above, a number of potential limitations must be considered in interpreting these data, including random variation, bias, and confounding. These factors are discussed below.

Random Variation

One way to assess the precision of a rate estimate for a given population is by calculating a confidence interval for that rate (Hennekens and Buring 1987). The narrower the confidence interval, the more precise the estimate of the rate, and the less likely it is that the rate is due to random variation (i.e., chance). Cancer cases often seem to occur in groups or clusters, and because of the large number of people who will eventually develop some type of cancer, clusters may be expected to occur for no other reason than chance (NCCC 1994). Clusters can be defined spatially, geographically, or by some other aspect that individuals may have in common, such as age, sex, or occupation. This concept is especially important when analyzing small samples (such as the populations of Eddy and Lea counties) because a few cases of cancer may result in a comparatively large and misleading incidence rate.

Bias

A number of biases are possible in surveillance systems such as cancer registries, including biases due to incomplete case ascertainment, abstracting or coding inconsistencies, problems with assignment of ethnicity, and changes in population estimates resulting in changes in cancer incidence rates. The NMTR and SEER registries work to minimize these biases, producing the most comprehensive cancer data available nationally (Brownson et al. 1993) and making a valuable contribution worldwide (IARC 1993). The NMTR's efforts to avoid bias are discussed below.

Case Ascertainment: The NMTR uses a variety of methods to collect all diagnosed cases of malignant cancer. Within one month, 85% of reported cases are identified. Within 6 months, 99% of reported cases are identified and completely abstracted. Although it is almost certain that all histologically confirmed cancers have been collected by this time, a few cases may not be documented in routine resources such as hospital records, outpatient clinic records, and pathology laboratory records. To identify any missing cancer cases, the NMTR uses death information from New Mexico and other states which accounts for about 3% of the total yearly reported cases. In addition, physician's offices and treatment facilities are visited on a regular basis by NMTR staff to identify potential cases. Quality assurance measures include repeating case finding procedures using hospital, clinic, or laboratory records, for a specified time period.

Abstracting/Coding: Abstracted and coded data include information about patient demographics, cancer site and histology, extent of the disease, treatment, and survival (NCI 1993). Although errors are rare, the NMTR implements a rigorous quality control and quality assurance process to correct any abstracting or coding errors. Quality control measures include yearly audits by NCI personnel, monthly audits of randomly selected cases by NMTR personnel, utilization of computer programs that check for implausible data combinations, and a focus on education of NMTR abstractors and tumor registrars.

By Ethnicity: Assignment of ethnicity is an important issue in New Mexico because of the state's comparatively large Hispanic and Native American populations. To ensure correct assignment of ethnicity, NMTR has initiated routine verification procedures that provide a high percentage of accuracy for ethnic identification. The NMTR uses multiple sources to assign or verify ethnicity. The sources used to help assign ethnicity include hospital charts, death records, Health Care Financing Administration records, Indian Health Service records, information collected during SEER special studies, the 1980 Census list of Spanish surnames, and the GUESS program (a computer program designed by a former NMTR staff member, that assigns ethnicity based on parts of, and whole, last names). Sources used to verify ethnicity include many of the same, and some additional sources, including data supplied by the next of kin when a patient has died, records from the Health Care Financing Administration (only for people over 65 years old, and only for the race distinctions of white, black, or other), Indian Health Service records (only for Native Americans), follow-up letters to doctors asking for confirmation of ethnicity, and information gathered during SEER special studies. The GUESS program alone successfully assigned ethnicity for 93.3% of Hispanic males residing in New Mexico who died during the period between 1985 and 1993, and 89.4% of Hispanic females who died during this period (Darling 1996).

Population Estimates: Population estimates are periodically adjusted, resulting in a change in cancer incidence rates (calculated by dividing the number of reported cases for a specified time period by the population at risk). In an effort to minimize variation, the NMTR uses Bureau of Census population data, divided into ethnic groups by the University of New Mexico's Bureau of Business and Economic Research (BBER), to estimate population (including for the years between censuses). The BBER publishes yearly population estimates for New Mexico. The NMTR uses this information to calculate changes in population estimates for standard ethnicities used by the NMTR and SEER program. Population estimates may be readjusted for past years after the decennial census is completed.

Confounding

The data sets analyzed in this report were age-adjusted and stratified for sex and ethnicity. Associations found between cancer incidence rates and sex or ethnicity may actually be due, in whole or in part, to other factors that were not measured such as tobacco use, lifestyle, occupation, eating habits, or socioeconomic status.