April 6 -8, 2004 Cancer Clusters and Environmental Quality Shanghai-California Environmental Health Conference Richard Kreutzer, M.D. California – China Environmental Health Training Program

April 6 -8, 2004 Epidemiology is…  The study of the distribution and determinants of disease in human populations.  Characterizing disease as to person, place and time.

April 6 -8, 2004 “Cluster” “ An unusual aggregation of health events that are grouped together in time and space …” CDC Guidelines for investigating clusters of Health Events, 1990

April 6 -8, 2004 Some Well-Known Cancer Clusters PlaceTimeCancerObsExpO/E Niles, ILL Child Leukemia Sellafield, UK Child Leukemia Woburn, MA Child Leukemia McFarland, CA Child Cancer -several types

April 6 -8, 2004 Clusters  What are they-concept of randomness  5858 census tracts; 80 cancer sites=4686 clusters at 0.01 significance level  Why has DHS studied them  What is the success rate around the world  What has been the approach  The cluster investigation manual  What can be determined by looking at cluster cases

April 6 -8, 2004 Clusters in the United States Represent Fear of environment Distrust of government Frustration with lack of control over one’s surroundings Large degree of ignorance about disease

April 6 -8, 2004 Appendix A There are 440 towns with less than 10,000 population in California, and about 5,000 census tracts with around 5,000 population each. For diseases with an average expectation of five or more cases per time period of interest we can say the following: In 100 such locations, the likelihood that a given disease (e.g..lung cancer) is elevated enough to be statistically significant with a p value of.05 is 5%. So on the average, five out of 100 towns will show an elevation significant at the.05 level, while 95 towns will not. What is the probability of a town escaping both lung cancer and bowel cancer clusters of p value of.05? The probability of - A and B is P(A) x P(B). So the answer is (.95) x'(.95) or.903. The Probability of at least one statistically significant cluster in a census tract home or town Answer:(.95) 80 =.0165 Answer:1.00 minus the probability of no cluster, or 1.00 minus.0165 =.9835 What is the probability of a town escaping a cluster of each and every, one of the 80 major classifications of cancer? What is the probability of at least one type of cancer cluster?

April 6 -8, 2004 So we can expect 8 out of 1,000 towns or census tracts to have at least one of 80 types of cancer elevated at the P = 1/10,000 level of significance. Since there are 5,440 localities in California, that means we can expect about 44 towns will have clusters of that extreme statistical significance. About half the towns will have P = significant clusters. The Probability of at least one statistically significant cluster in a census tract home or town (Cont.) Probability of escapingProbability of at least one of P Value all 80 types of caner 80 cancers being elevated.0l So there is a 98.4% probability that a town will have at least one type of cancer cluster at the p =.05 level. Using your calculator you can verify the following figures:

April 6 -8, 2004 Comparison of Clusters E.coliCancer Cluster Disease Agent Other Causes Latency Rare Can be cultured from a case Few 2-5 days Common Can’t be determined medically Many Years

April 6 -8, 2004 XXX XXXXX XX XXXX XXXX XXX XXX X XXXX XX XXXXX XX XX Taylor & Wilde, “Drawing the line with Leukemia”

April 6 -8, 2004 Limitations of Science for Cluster Questions  Paradox’s of epidemiology  Large numbers –small confidence intervals (clusters disappear in average)  Small numbers-large confidence intervals (insufficient power)  Population vs. individual risk  Can look for known carcinogens  Rarely can identify new carcinogens  What should its role be in a democracy?  Science vs. pseudoscience  Citizen intuition vs. scientific certainty

April 6 -8, 2004 AB CD Categories of Epidemiologic Studies DiseaseNo Disease Total Exposed Total Not Exposed Richard Kreutzer, M.D. Not Exposed Exposed Total with Disease Total without Disease

April 6 -8, 2004 Chemicals in the Environment  Air  Water  Soil Travel Through…….  Breathing  Eating  Touching Get into body by….. Figure 2. Toxicants as Causes of Disease: The General Model  Harmfulness of chemical  Amount of chemical  Length of exposure to chemical HEALTH IMPACT HOW CHEMICALS CAN AFFECT YOUR BODY Chemical effects on your body depend on ……

April 6 -8, 2004 Determinants of Disease Environmental Behavioral Lifestyle Occupational Other Diseases Psychological Genetic Disease

April 6 -8, 2004 Measured LevelRegulatory ActionEHIB Action Below Regulatory number Do nothingProvide public information about regulations and their scientific basis Above regulatory number but below known toxicological threshold Require compliance with regulation. Notify public Provide public information on regulations and toxicology Above toxicological threshold but below level of epidemiolgical detection Require compliance with regulations. Notify public Provide public information on regulations, toxicology and epidemiology Above level of epidemiological detection Require compliance with regulations. Notify public Consider a health study* Figure 4. Regulatory Agency and EHIB Action Regarding Toxicant Levels *Must consider how study would be used and its feasibility. Should obtain community and individual informed consent.

April 6 -8, 2004 (A) Dietary fat and colon cancer (B) DES and vaginal cancer (C) Chernobyl release and thyroid cancer (D) Vinyl chloride and hemangiosarcoma Exposure Common, widespread exposure Rare, unique exposure Type of Health Outcome Common OutcomeRare, Unusual Outcome How would different approaches to looking at clusters perform for these different situations? 1)Respond to inquires – Could pick up (B) and (D) This approach could (and did) confirm the clusters of vaginal cancer and hemangiosarcoma 2)Actively search for clusters – Could pick up (B) and (D) A cluster hunting team scanning registry data could probably have found these rare and unusual clusters, although perhaps later than Approach 1 because of the lag time for registration. 3)Study unusual exposures – Could pick up (C) EHIB has mainly concentrated on being vigilant for new, unusual exposures and their possible consequences (e.g., aerial application of malathion). Conclusions:  The combination of Approaches 1 and 3 could pick up (B), (C), and (D).  None of these approaches would be a good way to detect (A).  Is there a compelling reason for Approach 2? Approaches: