Observations of cancer incidence surveillance in Duluth, Minnesota.

In 1973, amphibole asbestos fibers were discovered in the municipal water supply of Duluth, Minnesota. The entire city population of approximately 100,000 was exposed from the late 1950s through 1976 at levels of 1-65 million fibers per liter of water. Because of previous epidemiologic studies that linked mesothelioma, lung and gastrointestinal cancers to occupational exposure to asbestos, surveillance of cancer incidence in residents of Duluth was initiated to determine the health effect from ingestion of asbestos. The methodology of the Third National Cancer Survey (TNCS) and SEER Program was used. Duluth 1969-1971 rates were compared with TNCS rates for the cities of Minneapolis and St. Paul during 1969-1971; Duluth rates during 1974-1976 are compared with Duluth 1969-1971; Duluth rates during 1979-1980 are compared with Duluth 1969-1971 and with Iowa SEER; and a table of the occurrence of malignant mesothelioma is presented. Statistically significant excesses are observed in several primary sites in Duluth residents. However, lung cancer in Duluth females is the only primary site considered also of biological significance. The mesothelioma incidence rate is no more than expected. This paper also describes the problems of long-term surveillance of exposed populations considered at risk of environment cancer, the need for improved study methodologies and the use of federal records for follow up of exposed individuals.


Introduction
Before presenting recent findings from the surveillance of cancer incidence in Duluth, it seems appropriate to review the history of the situation and the duration and intensity of exposure to amphibole in the city, which at the time of exposure had a population of approximately 100,000.
In 1973, amphibole fibers were discovered in the municipal water supply of Duluth through studies done by the Environmental Protection Agency. A Federal Court ruling indicated that the fibers were a result of a taconite mining company dumping taconite tailings wastes into Lake Superior since 1955 (1). In the late 1950s, the mining company, which is located 50 miles northeast of Duluth on the lake, increased the amount dumped to approximately 67,000 tons per day, which continued into 1980. Taconite is lowgrade iron ore that is mined and processed into pellets of higher grade iron ore and shipped to steel mills on the Great Lakes. This particular *Chronic Disease Epidemiology Section, Minnesota Department of Health, 717 S.E. Delaware Street, Minneapolis, MN 55440.
supply of taconite is mined from amphibole-bearing rock. Cummingtonite-grunerite is the principal amphibole in this deposit.
U.S. EPA data on Duluth water samples in 1939-1940 and 1949-1950 indicated trace amounts of fibers, but samples from 1965 contained large amounts of amphibole. It is not known when amphibole fiber levels increased to those levels. The 1973 tapwater samples collected by EPA contained 1 to 30 million amphibole fibers/L, the level generally dependent on lake weather conditions and length of time the water was in the water distribution system (2). In 1976, water samples were collected from 20 homes in Duluth and a range of 2 to 64 million amphibole fibers/L was found. Storm conditions in the lake have produced levels as high as 100 million fibers/ L. Electron microscope studies at the Minnesota Department of Health have indicated that the physical characteristics of the fibers are: a mean length of 1.13 gm, a mean width of 0.18 gm and thus, an aspect ratio (length/width) of 6.5:1 (unpublished, Minnesota Department of Health, 1978). The aspect ratio of Duluth amphibole is double that of the OSHA arbitrary definition of a fiber with a minimum aspect ratio of 3:1.
A water filtration plant in Duluth became operational in January 1977, removing 99.9% of the fibers. The length of exposure for Duluth residents may be considered to be 17 years, 1960-1976, and the level of exposure to be in the range of 1 to 65 million fibers/L. Until recently, monitoring of the water supply occurred daily at one of several points in the entire water distribution system.
In 1974, because of the known health risk of cancer to those occupationally exposed by inhalation to asbestos (3)(4)(5)(6)(7)(8)(9)(10)(11) and the public health concern regarding the unknown risk of those ingesting asbestos from a public water supply, a study of cancer in Duluth residents was designed to determine cancer incidence during 1969-1981. Generally, 10-20 yr is considered the induction period for cancer of most primary sites and may be as long as 30-50 yr following asbestos exposure (12). Therefore, Duluth cancer incidence rates during 1969-1971, approximately 10 yr following the initiation of exposure, are considered baseline rates with which to compare the results of surveillance through time. The rates beginning in the mid-1970s would be the first rates that could reflect any possible increase in cancer due to exposure, assuming exposure began in approximately 1960.

Methods
The methodology used was that of the Third National Cancer Survey (TNCS) of 1969-1971 (13) and the current SEER Program at the National Cancer Institute (14). Personnel who worked on the Minneapolis-St. Paul component of the TNCS have conducted the study in Duluth. This has been important for the standardization and uniformity of methodology and data collection procedures and for the comparison of data between study areas.
The study requires the identification of all cancer cases in the study population through the review of hospital medical records, pathology and autopsy reports, and death certificates, and the abstracting of patient charts at the three hospitals in Duluth, the Mayo Clinic, and the University of Minnesota and Veterans Administration Hospitals in Minneapolis. The abstracted information includes name and address ofpatient, age, sex, race, date of diagnosis, primary site and histology, and hospital of diagnosis.
Duluth rates during 1969-1971 are compared with TNCS rates for the cities of Minneapolis and St. Paul, by using the Mantel-Haenszel method for determination of statistical significance (15). Minneapolis and St. Paul are considered excellent  (16). Future analysis of the entire study period will include comparison with SEER data from selected cities of Iowa because of similar population characteristics as described above for the cities of Minneapolis and St. Paul. Those Iowa cities are known to be virtually free of amphibole fibers in their municipal water supplies (J. Millette, Health Effects Research Laboratory, U.S. EPA, personal communication).
It is important to note that the methodologies of the cancer incidence studies of Duluth, San Francisco, and Seattle (all presented at this workshop) are virtually identical. A common study methodology makes comparison of those study results quite meaningful among the three cities, all with different exposure factors.

Results
Tables 2 and 3 contain absolute numbers of Duluth cancer cases and average annual ageadjusted incidence rates of selected primary sites for males and females, respectively, in the cities  In the 1969-1971 comparisons, excesses in Duluth that are statistically significant are observed, by sex and primary site, for male peritoneum, retroperitoneum and intra-abdominal; male other peritoneum and digestive; female uterine corpus; and male prostate. In the comparison between the two Duluth time periods, statistically significant increases are observed for female peritoneum, retroperitoneum, and intra-abdominal; female lung cancer; female uterine corpus; and female multiple myeloma. A decrease in male prostate is statistically significant. Data from some of these sites involve very small numbers and must be interpreted with caution even though statistically significant. Table 4 provides the most recent data for Duluth, for 1979-1980 for males and females separately, again for selected sites. Absolute numbers of cases, average annual crude incidence rates, and SMRs compare Duluth 1979-1980 rates with rates for Duluth 1969Duluth -1971 (16). An increase in diagnostic surveillance during the mid-1970s, generated from increased concern in this community relative to malignant mesothelioma, may partially account for these differences. Based on the observed frequency of pleural malignant mesothelioma, one would expect approximately one newly diagnosed case per year. In fact, six cases were ascertained during the 3-yr period of heightened concern, 1974-1976, which is double the expected frequency of three.

Discussion
We are unable to explain the variations in rates and the differences that are statistically significant. However, it is clear that rate differences between geographic areas and through time do occur without any apparent biologic reason. Also, in the determination of statistical significance at the p = 0.05 level, for every 100 comparisons, 5%  Female  1969  1970  1971  1972  -1973  1  1974  2  1  1975  1  1976  2  1  1977  -1978  1  1979  1980  1  Total  6  3  1 will be statistically significant by chance alone. In our opinion, the only statistically significant result with clear biological significance is the increase in female lung cancer in Duluth, undoubtedly a reflection of the increase in cigarette smoking over the past few decades. This increasing trend of lung cancer in women is seen nationally (14). The marginally statistically significant excess of stomach cancer in Duluth males, compared with Iowa, is consistent with historical observations of stomach cancer mortality for St. Louis County, the location of Duluth (17). The lack of a statistically significant excess in Duluth females, compared with Iowa, is probably due to the smaller number of female cases. Historically, the rates have been high for stomach cancer in males and females long before the occurrence of asbestos exposure. However, the impact of asbestos exposure on the risk of stomach cancer in excess of the risk attributed to known risk factors (i.e., Scandinavian dietary practices) on current cases, cannot be determined (18). Other differences of Duluth cancer incidence rates may be important, and a longer time of surveillance coupled with evaluation of mobility and follow-up of the exposed population will need to be conducted before these issues can be resolved.

Needs of Further Research
In my opinion, it will be important to complete the analysis of study period 1969-1981 so that three solid years of data during [1979][1980][1981] around the 1980 census year can be analyzed. At that time, thorough review and analyses of all the data should be conducted. In addition, plans should be made to continue the surveillance, possibly using a modified methodology to accommodate the entire length of exposure of 17 yr and the induction periods of 30-50 yr. Furthermore, the intent of the study design was to incorporate methods for determining length of residency of Duluth cancer cases as a measure of length of exposure and also for determining migration patterns of the exposed population more scientifically than the Census Bureau data currently can permit. A basic problem in study methodology arises in such long-term surveillance when one questions to what degree the observed cancer incidence is measuring the occurrence of disease in the actual exposed population, even for the city of Duluth, which is considered to be a very stable population.
Similar questions also arise for this and other diseases perceived to be environmental threats to human health and yet having very long latent periods. These questions require new methodologies and procedures such as rapidly registering exposed individuals at time of exposure, taking biologic specimens, and incorporating methods of long-term follow-up. Such follow-up would be accomplished much more easily, and scientific study would probably be much more valid if it were possible to use some IRS records and Census Bureau records of individuals in large, exposed populations. Use of those records would greatly enhance and enable long-term follow-up in epidemiologic and biomedical research and would re-duce the expense of more difficult and time-consuming follow-up procedures, which are now required (19).
It is apparent that there is another major area of needed epidemiologic research regarding the health effects from the ingestion of asbestos in drinking water-the need for development of methods for clinical laboratory studies applied within an epidemiologic design, which would address questions related to low-level dose-response effects, host sensitivity and reactivity for large populations, and the effects of three variables known to affect carcinogenicity of asbestos fibers: the structure, geometry, and surface adsorption characteristics of fibers. Currently, it is difficult to predict the carcinogenic behavior of fibers in vivo in both animals and humans.