Health Tracking & Disease Clusters

The Lack of Data on Chronic Disease Incidence and its Impact on Cluster Investigations

Tony Dutzik

Jeremiah Baumann

U.S. PIRG Education Fund

PennEnvironment Research and Policy Center

September 2002

Acknowledgments The authors would like to acknowledge Alison Cassady of U.S. PIRG for coordinating the research and distribution of this report and Craig W. Trumbo, Ph.D., of the Missouri School of Journalism for his insight and for sharing the results of his research on cancer clusters. Thanks also to Susan Rakov and Brad Heavner for their editorial assistance, to Ellen Montgomery for her research assistance, and to the many state public health officials who provided information for this report. In addition, we would like to thank those who provided editorial review, including Craig Trumbo, Frank Bove, Sc.D., of the Agency for Toxic Substances and Disease Registry, Jerald Fagliano, Ph.D, of the New Jersey Department of Health and Senior Services, and Shelley Hearne, Dr.PH, executive director of the Trust for America’s Health. We express our gratitude to The Pew Charitable Trusts for their financial support of this project. The authors alone are responsible for any factual errors. The recommendations are those of the U.S. PIRG Education Fund and PennEnvironment Research and Policy Center. The views expressed in this report are those of the authors and do not necessarily reflect the views of our funders or those who provided editorial review. Copyright 2002 U.S. PIRG Education Fund and PennEnvironment Research and Policy Center The U.S. PIRG Education Fund is the research and public education center for the U.S. Public Interest Research Group (U.S. PIRG), the national advocacy office of the State PIRGs. The State PIRGs are a nationwide network of nonprofit, nonpartisan, state-based public interest advocacy organizations. Through U.S. PIRG and the U.S. PIRG Education Fund, they promote a national agenda of environmental and consumer protection and good government. For more information about the state PIRGs, visit our Web site at www.pirg.org. PennEnvironment is a statewide, nonprofit, citizen-based environmental advocacy organization. PennEnvironment Research and Policy Center is a 501(c)(3) research and public education center. For more information, visit our Web site at www.pennenvironment.org. For additional copies of this report, send $20 (including shipping) to: U.S. PIRG Education Fund 218 D St. SE Washington, DC 20003 202-546-9707 uspirg@pirg.org

3

Table of Contents

INDEX OF CASE STUDIES.................................................................................................................4

EXECUTIVE SUMMARY...................................................................................................................5

INTRODUCTION.............................................................................................................................8

WHY INVESTIGATE DISEASE CLUSTERS?.........................................................................................10Identifying a Cause of Disease...............................................................................................13Identifying Environmental Hazards......................................................................................13Targeting Public Health Resources.......................................................................................16

HOW DISEASE CLUSTERS ARE INVESTIGATED..................................................................................19A Four-Stage Process..............................................................................................................19Data for Cluster Investigations..............................................................................................21Limitations of Cluster Investigations....................................................................................26

THE CURRENT STATE OF HEALTH TRACKING DATA............................................................................32Cancer Registries....................................................................................................................32Birth Defects Registries......................................................................................................... 34Other Chronic Diseases.........................................................................................................35

THE IMPACT OF HEALTH TRACKING DATA ON CLUSTER INVESTIGATIONS: LESSONS FROM THE CASE STUDIES......................................................................37 Lesson #1: Lack of Data Causes Delays in Cluster Investigations.......................................37 Lesson #2: Lack of Data Deters Pro-active Investigation of Potential Clusters..................37Lesson #3: Lack of Data Deters the Identification of True Clusters..................................38Lesson #4: Lack of Data Leads to Fewer Investigations of Potential Health Threats.......40Lesson #5: Lack of Data Leaves Community Concerns Unaddressed............................... 41A Sixth Lesson: Good Registries Are Not Enough..............................................................43

BEYOND CLUSTERS: INVESTIGATING ENVIRONMENTAL LINKS TO CHRONIC DISEASE.......................................................................................................................47

POLICY RECOMMENDATIONS........................................................................................................50

APPENDIX A: STATUS OF U.S.CHRONIC DISEASE REGISTRIES.............................................................52

APPENDIX B: STATE RESPONSES TO CANCER CLUSTERS....................................................................54

APPENDIX C: GLOSSARY OF ABBREVIATIONS................................................................................. 56

TOMS RIVER, N.J. (CHILDHOOD LEUKEMIA)Lack of Up-to-Date Registry Data SetsBack an Investigation............................ 14

ATKINSON, N.H. (CANCER)Cluster Report Brings Awareness of Water Contamination........................15

LAREDO, TEXAS (NEURAL TUBE DEFECTS)Aggressive Monitoring CatchesEmerging Cluster...................................17

CHARLESTON, S.C. (PLEURAL CANCER)Registry Data Successfully Identifiy anOccupational Cancer Cluster................22

SOUTH BOSTON, MASS. (SCLERODERMA AND LUPUS)Lack of Registry Data Forces CostlyHunt for Cases of Disease.....................23

EL PASO, TEXAS (MULTIPLE SCLEROSIS)Expected Levels of Disease Difficult to Find..................................... 25

SUFFOLK COUNTY, N.Y. (BREAST CANCER)Active Surveillance ProvokesInvestigation...........................................26

Index of Case Studies

FAIRFIELD, MAINE (BRAIN CANCER)Narrow Focus Leads to Questions About Cluster Investigation.................. 28

MARION, OHIO (LEUKEMIA)Cluster Study Leads to Discovery ofContamination, Hints at Links............. 29

FALLON, NEV. (CHILDHOOD LEUKEMIA)When Two-Year-Old Data Aren’t GoodEnough...................................................33

DICKSON COUNTY, TENN. (CLEFT LIP/CLEFT PALATE)Reliance on Birth Records Would HaveShrouded Cluster...................................34

CHESTERFIELD COUNTY, VA. (CANCER)After a Decade of Study, Poor RegistryData Hampers Investigation..................39

HAZLETON, PA. (CANCER)Studies’ Limitations DemonstrateImportance of Exposure Data.............. 42

CALCASIEU PARISH, LA. (CANCER)Inquiry Shows Need for ExposureInformation............................................44

Each year, more than 1,000 calls areplaced to public health officials regard-ing suspected local disease clusters. Inmany of these cases, investigators arecalled upon to determine whether ratesof a disease in a particular communitytruly are excessive – and whether envi-ronmental exposures are to blame.

Requests for disease cluster investigationsare just one sign of the broad public con-cern about the role environmental fac-tors play in the development of chronicdisease. Nearly 90 percent of Americansbelieve that environmental factors suchas pollution cause disease or health prob-lems.

Disease cluster investigations play animportant role in responding to theseconcerns and protecting public health.Cluster investigations can help publichealth officials target resources for dis-ease prevention and treatment, spur thediscovery and cleanup of existing envi-ronmental hazards, and enableresearchers to develop and test hypothe-ses about the possible links between envi-ronmental exposures and chronic dis-ease.

Cluster investigations are notoriously dif-ficult, even under the best of circum-stances and with ample resources.However, in most states, the resourcesavailable for investigating disease clustersare extremely limited. In 1998, 26 statesdevoted less than one half-time person tocancer cluster investigations – a level of staffing virtually unchanged over the pre-vious decade. States also demonstrate

varying degrees of vigor in their response to cluster reports, with some states resolv-ing as many as 99 percent of all clusterinvestigation inquiries during the courseof the initial phone call.

To succeed in cluster investigations,researchers need complete, up-to-dateinformation on the incidence of diseasein a community; data that are typicallycollected in a disease registry or otherhealth tracking system. To be effective,health tracking systems must bestatewide, detailed, up-to-date, utilizemultiple sources of information, andinclude active surveillance by publichealth officials to ensure that all cases ofdisease are recorded.

In many states, however, accurate track-ing systems for chronic disease do notexist. Only three states – California, Iowaand Massachusetts – possess both cancerand birth defects registries that meet thehighest standards for quality and alsoreport having any system at all for thetracking of asthma. And almost no statesconduct systematic tracking of learningdisabilities, neurological disorders suchas Alzheimer’s and Parkinson’s, metabol-ic diseases like diabetes, or auto-immunedisorders such as lupus.

This report details the real-life costs ofthis knowledge gap as it relates to theinvestigation of 14 suspected disease clus-ters over the past decade. A review ofcompleted cluster studies and other liter-ature and interviews with state publichealth officials reveal that lack of accessto high-quality health tracking data:

5

Executive Summary

6

1) Causes long delays in cluster investigations;

2) Prevents public health officials from identifying disease trends;

3) Inhibits the identification of true disease clusters;

4) Reduces the number of cluster investigations carried out by states, meaning that some clusters go uninvestigated; and

5) Deters communities from getting the information and help they need when a suspected cluster arises.

While most states lag behind in their abil-ity to track and investigate disease clus-ters, several states have shown the poten-tial benefits of chronic disease tracking.

• Researchers using data from the California Birth Defects Monitoring System have shown that maternal exposure to air pollution and resi-dence within a quarter-mile of a Superfund site are related to the development of certain birth defects.

• Texas public health officials used their ongoing surveillance of birth defects data to identify a cluster ofneural tube defects in Laredo justmonths after it emerged. The stateis now targeting public health assistance to the area.

• New York state officials are using a first-of-its-kind cancer mapping system to identify potential cancer clusters. The state is now investi-

gating significantly elevated rates of breast cancer in seven Long Island zip codes.

With growing evidence that environmen-tal factors play a significant role in thedevelopment of many chronic diseases –and with growing public concern overthose links – the need to quickly andeffectively identify and investigate diseaseclusters is greater than ever. The creationof a nationwide health tracking networkwould allow public health officials to con-duct quicker, less resource-intensive, andmore accurate investigations of diseaseclusters while providing researchers withthe tools to better assess possible envi-ronmental links to chronic disease. Sucha system would include:

• Tracking of cancers, birth defects, respiratory diseases such as asthma, neurological diseases such as Alzheimer’s and other chronic dis-eases in every state.

• Tracking of environmental expo-sures, such as exposures to PCBs, heavy metals and pesticides.

• An early warning system to alert communities to immediate health crises such as heavy metal and pes-ticide poisonings.

• Federal, state and local rapid response capability to investigate clusters, outbreaks and emerging threats.

In addition to establishing a nationalhealth tracking network, public officialsshould encourage the linkage of healthdata with existing data on environmentalconditions, promote public involvement

in cluster investigations, and take anaggressive stance toward both the pre-vention of new environmental hazardsand the cleanup of existing hazardsnationwide.

7

8

The mother of a young Ohio leukemiapatient calls to offer support to the moth-er of another young leukemia patientwho had graduated from the same highschool as her daughter. During the con-versation, the mother is told that a thirdyoung graduate of the high school hasalso been diagnosed with the disease.The two mothers decide to investigate,and within months find ten graduateswho suffer from various forms of cancer.Environmental testing later uncovers evi-dence of toxic contamination on thegrounds of the school, which was built ontop of a former Army depot.

A woman diagnosed with scleroderma –an autoimmune disease that causes tight-ening of the skin and can attack vitalorgans – learns from a co-worker thatanother woman in her Boston neighbor-hood also has the disease. She begins ask-ing questions around the neighborhood.When the number of local sclerodermapatients hits six, she calls the state healthdepartment. Since that phone call,another 20 scleroderma patients in theneighborhood – along with 60 otherswho suffer from lupus, a related autoim-mune disorder – have been identified.She and others in the neighborhoodwonder if air or water pollution in theneighborhood could be responsible.

Doctors at a clinic in a Texas border towndeliver three babies missing part or all oftheir brains over the course of just a fewhours. They alert public health officials,but without information about otherbirth defects in the region, investigationis difficult. In the years since, clusters ofsimilar birth defects have emerged on

other portions of the U.S.-Mexico border.Public health officials believe vitamindeficiencies are to blame while somecommunity residents suspect that indus-trial pollution could play a role.

Concerned mothers. Curious patients.Alarmed doctors. When it comes to sus-pected clusters of cancer, birth defects,and chronic disease in the U.S., they arealmost always the first to recognize aproblem – and to begin the search foranswers.

Each year, at least 1,000 calls are placedto state public health officials regardingsuspected clusters of disease. Thesereports are just one sign of the broadpublic concern about the role environ-mental factors may play in the develop-ment of chronic disease. Nearly 90 per-cent of Americans believe that environ-mental factors such as pollution causedisease or health problems.

1

In many cases, concerns about a suspect-ed disease cluster are unfounded. But ina significant number of cases, the possi-bility of a cluster is sufficiently real thatpublic health officials feel compelled toinvestigate.

Cluster investigations are notoriously dif-ficult, even under the best of circum-stances. To succeed, researchers musthave access to reliable, up-to-date infor-mation about the extent of disease withina community and solid informationabout potential environmental contami-nants and possible routes of exposure.

Introduction

Most Americans assume that such infor-mation exists. More than half of allAmericans believe that the nation has anational system for monitoring exposuresto environmental threats and trackingchronic disease.

2

We don’t. And the state-level chronic dis-ease tracking programs that do existoften fail to provide researchers with theinformation they need to answer themost basic questions related to diseaseclusters: How many people are sick? Whoare they? And where do they live?

The lack of chronic disease monitoringhas impacts far beyond the investigationof disease clusters. While outbreaks ofinfectious diseases such as Legionnaire’sdisease or E. coli are often almost instant-ly identified (and bring about a promptpublic health response), trends in theincidence of chronic diseases such aslupus or multiple sclerosis can go yearswithout being identified by the publichealth system. Even statistics from reg-istries for the most aggressively trackeddiseases, such as cancer, may be two yearsor more out of date.

The result is a system that leaves citizensand public health officials largely in thedark about trends in chronic disease, andprevents the effective study of the degreeto which environmental factors play arole in the development of those disor-ders. Delaying this understanding meanspublic health officials and policymakerscannot act effectively to prevent disease.

Through case studies, interviews withstate public health officials, and a reviewof existing literature, this report makesthe case that the failure to adequately

track the incidence of chronic diseaseleads to costlier and less conclusive clus-ter investigations that often take longerto complete. In many cases, the lack ofsuch data may result in worthwhile inves-tigations never taking place at all.

To quickly and effectively identify andinvestigate disease clusters, the nation’spublic health system must have betterinformation about where chronic diseaseis occurring in our communities and howindividuals may be exposed to environ-mental contaminants. That informationmust be collected and stored in such away as to be useful to researchers and beeasily linked to other databases of expo-sure or health-related information. Andstate and federal officials must have theresources and the staffing to use thatinformation to thoroughly investigatesuspected disease clusters and the linksbetween environmental contaminationand public health.

9

10

The causes of many cancers, birth defectsand chronic diseases remain shrouded inmystery. While researchers have identi-fied genetic, lifestyle and environmentalfactors that can contribute to the devel-opment of chronic disease, the exactinteractions are poorly understood andother causes may remain unidentified.

Yet, there is growing awareness that envi-ronmental exposures play a significantrole in the development of some chronicdiseases. A 2000 study published in theNew England Journal of Medicine foundthat environmental and lifestyle factorsappear to play a greater role in the devel-opment of most types of cancer thangenetics.

3

Scientists estimate that aboutthree percent of all developmentaldefects can be attributed to toxic expo-sures and, more generally, that environ-mental and lifestyle factors play a role inabout 25 percent of developmentaldefects.

4

The mystery surrounding the causes ofmany chronic diseases leads to concernwhen groups of neighbors, co-workers, orothers who share a common bond con-tract the same disease at roughly thesame time. Naturally, and justifiably,those affected by the disease wonderwhether something more than chance –perhaps something to which they havebeen exposed in their air, water or food –could be responsible. And others in thecommunity wonder whether they will suf-fer the same fate.

Public health officials have a responsibili-ty to the community to respond to these

concerns with sensitivity – and a respon-sibility to the broader public to do so withefficiency. In some cases, this responsewill include a deeper investigation intothe suspected disease cluster. Theseinvestigations can have direct publichealth benefits: the identification andclosing of a contaminated well, thecleanup of a toxic waste site, the targetingof resources for disease prevention andtreatment to those most in need. And insome cases, investigation of suspectedclusters can shed light on the role envi-ronmental exposures play in the develop-ment of chronic disease – discoveries thatcan represent dramatic breakthroughs inpublic health.

What Is a Disease Cluster?

A disease cluster is defined as the occur-rence of a greater than expected numberof cases of a particular disease amongmembers of a specific group, residents ofa specific area, or over a specific periodof time.

5

This definition may appearstraightforward, but a deeper readingtells a great deal about what clusters areand how they are viewed by scientists andresearchers.

First, the word “cluster,” under this defin-ition, simply means an excess of disease.Disease can cluster in an area for a varietyof reasons, including demographics,lifestyle choices, occupational exposuresand environmental factors – or for noreason at all. Just because a grouping ofdisease is called a cluster does not meanit has an environmental cause.

Why Investigate Disease Clusters?

Second, the phrase “greater than expect-ed number of cases” is often interpretedto mean a statistically significant excess ofcases when compared to an establishedbaseline. Even a random distribution ofdisease will result in some areas havingmore cases of disease than others. To beconsidered a cluster, an area must havesuch a high rate of disease that the eleva-tion is unlikely to be explained bychance. However, the absence of a statis-tically significant cluster does not neces-sarily mean that there is no environmen-tal link to the development of disease.

Finally, the phrase “particular disease”refers not to overarching categories ofdisease such as “cancer" or “birth defects”but to specific diagnoses within those cat-egories. The term "cancer,” for example,is used to refer to more than 100 separatediseases. An increased rate of variousforms of cancer within a community istherefore unlikely to be deemed a “clus-ter,” while an outbreak of a specific formof cancer is more likely to be given thatdesignation.

The task of defining the existence of acluster, therefore, is different from thetask of assessing the role environmentalfactors may play in the development of adisease in a particular community. Thetask of defining a cluster is primarily a sta-tistical exercise; the task of assessing thecause of a cluster is an epidemiologicalone. In the case studies that accompanythis report, the term “cluster investiga-tion” will be used to include both the sta-tistical efforts made to define and assessthe severity of clusters and the epidemio-logical studies that follow once a clusterhas been identified.

How Common Are Reports of DiseaseClusters?

State public health officials receive some-where between 1,000 and 2,000 reportsof suspected cancer clusters each year – afigure that does not include reportedclusters of other diseases. About two-thirds of these reports come from indi-vidual citizens.

6

The number of cancer cluster investiga-tion requests varies widely from state tostate, ranging from as many as 300requests in California in 1997 to as few astwo in small states such as North Dakotaand Delaware. (See Table 1.)

7

Table 1: Requests for Cancer ClusterInvestigations, 1997 8

Hawaii, Oklahoma, Tennessee and Vermontdid not provide information. Georgia,Nevada and South Carolina reported thatthey do not conduct cluster investigations.Ohio refused to participate in the survey.

11

STATE INVESTIGATIONREQUESTS

CA 300NY 91NJ 71IL 64MA 59MO 43TX 43NC 30WI 29MN 29PA 25MI 24CT 20OR 20LA 18WY 12RI 12AZ 12NH 11ID 11SC 10IN 10

STATE INVESTIGATIONREQUESTS

CO 9MD 9VA 9MS 7AL 6FL 6NM 5UT 5KS 5ME 4AK 3MT 3IA 3WA 3ND 2DE 2NE 2WV 2AR 2KY 2DC 1NV 0

12

Of these reports, most are resolvedimmediately. Often, when citizens aremade aware of the overall prevalence ofcancer and the generally random natureof its distribution, their suspicions of acluster are relieved. Even in cases wherecitizens’ concerns are not relieved, publichealth officials may decide that the situa-tion described is unlikely to constitute atrue cluster and decline to launch aninvestigation.

Between one-quarter and one-third of allcancer cluster inquiries, however,progress beyond the initial phone call.Very few of these inquiries result in iden-tification of a true cluster. State healthdepartments report that less than fivepercent of all inquiries show sufficientevidence of a cluster to warrant a full-scale investigation.

9

Of these investiga-tions, only a few have ever succeeded inlinking a specific outbreak of disease to aspecific environmental cause. (See“Historic Cluster Investigations”)

Why, then, do public health agenciesfocus scarce resources on the investiga-tion of disease clusters? And why are suchinvestigations important for publichealth?

Historic Cluster InvestigationsSeveral investigations of clusters of rare diseaseshave yielded public health breakthroughs.

LEGIONNAIRE’S DISEASE – A mysterious outbreak ofpneumonia at a 1976 American Legion conven-tion in a Philadelphia hotel claimed 34 lives.Researchers traced the cause of the disease to bac-teria in contaminated water used in the hotel’s airconditioning system. The discovery has led to

increased attention to the maintenance of build-ing water systems.

10

RESPIRATORY CANCERS – The first atlas of cancermortality in the U.S., published in 1975, showedclearly elevated rates of respiratory tumors in portcities. Subsequent studies linked the cancers toexposure to asbestos in the shipbuilding industryduring World War II.

11

Asbestos use has dramati-cally declined over the last two decades and it hasbeen removed from many public buildings.

BIRTH DEFECTS – During a two-month period in1961, an Australian doctor delivered three babieswith phocomelia – a rare birth defect in which theupper parts of the arms and legs are missing,resulting in the hands and feet being directlyattached to the body. The mothers of all threebabies had taken thalidomide, a popular tran-quilizer. For much of the previous year, similarcases of the defect had been occurring inGermany, but doctors had failed to establish a pat-tern. Within months, thalidomide was withdrawnfrom the market.

12

VAGINAL CANCER – A 1971 investigation of anunusual number of cases of a rare vaginal canceramong young women traced the cause to a drug,diethylstilbestrol (DES), taken by their mothersduring pregnancy to prevent miscarriage.

13

Physicians were subsequently advised not to pre-scribe DES to pregnant women.

LIVER CANCER – A 1974 study of a rare form of livercancer contracted by four workers in the sameKentucky factory led to the identification of vinylchloride as a potent carcinogen. Regulationsadopted after the discovery reduced the level ofvinyl chloride to which workers could be legally exposed by 99.8 percent.

14

HIV – In the early 1980s, physicians began to learnof outbreaks of a rare form of cancer (Kaposi’ssarcoma) and a rare pneumonia among gay men.The investigation into the cluster eventually ledto the discovery of AIDS and its cause, HIV.

15

13

Benefits of Cluster Investigations

Identifying a Cause of Disease

The greatest potential public health ben-efit that can arise from a cluster investi-gation is the discovery of a previouslyunknown cause of disease or a direct linkbetween a specific environmental expo-sure and the development of disease.Such discoveries are rare, but when theyoccur, the benefits to public health aregreat.

Cluster investigations such as theinquiries into asbestos-related cancersand AIDS have the potential to bringabout sweeping changes in lifestyles andmedical treatment. Asbestos use, forexample, has declined dramatically sincethe 1970s. The identification of AIDS andHIV – the direct result of inquiries intounusual cancer and pneumonia clustersamong gay men in the early 1980s – haveled to public health approaches designedto stem the spread of the disease and tonew treatments that offer improved livingconditions for AIDS patients.

For many reasons, the investigation ofoccupational or medical exposures ismore likely to yield such a dramatic resultthan investigation of environmentalexposures. First, in all but the mostunusual cases, individuals are exposed toenvironmental contaminants at low levelsover long periods of time. The long laten-cy periods of many cancers, for instance,and inadequate understanding of howtoxic exposures at low levels affecthuman health, make it difficult to identi-fy the role such exposures could play inthe development of particular cancers.

Second, people move into and out of

communities frequently, making it diffi-cult to ascertain whether environmentalexposures that took place in one locationled to the development of disease.

Third, people are exposed to a multitudeof toxic substances in a variety of settingsthroughout their lives. Sorting throughthe many possible exposures to deter-mine the specific exposure that caused aparticular disease is an extremely difficulttask. While epidemiologic investigations(and common sense) sometimes suggestthat residence near a toxic waste site orpresence during a specific environmentalrelease could be a cause for a given dis-ease, such suggestions can rarely beproven to the high degree of certaintydemanded by scientists.

While investigations of residential clus-ters are rarely sufficient to prove environ-mental causation, they can suggesthypotheses for further epidemiologicalstudy. By expanding the scope of a studyto include several communities with com-mon environmental exposures (forexample, communities near hazardouswaste sites), some of these concerns canbe addressed, leading to the increasedlikelihood of significant findings.

Identifying Environmental Hazards

In many cases, the intense scrutinycaused by a cluster investigation has ledto the identification of environmentalhazards that may not have been identi-fied otherwise.

Several of the most famous cluster inves-tigations – such as those in Woburn,Mass. and Toms River, N.J. – were able todraw associations between disease out-breaks among certain populations and a

specific exposure pathway; for example,drinking water from certain town wells.From the point of view of identifying thecause of the cluster, these investigationsdid not prove with complete certaintythat the contamination was the cause andthus were “failures” – expensive ones. Butin the Toms River case, the investigationled to the identification of a previouslyunknown contaminant in the problemwell. (See "Case Study: Toms River, N.J." below.)Both the Woburn and Toms River studiescontributed to a scientific body of evi-dence identifying prenatal chemicalexposures as a potential risk factor forchildhood leukemia. And in both cases,the cluster investigations brought newmomentum to efforts to clean up existinghazardous waste sites.

CASE STUDY: TOMS RIVER, N.J. (CHILDHOOD LEUKEMIA)

LACK OF UP-TO-DATE REGISTRY DATASETS BACK AN INVESTIGATION

In the early 1980s, residents of Toms River beganto suspect that environmental contamination couldbe responsible for what they perceived as anincrease in childhood cancer in the community,which is located near a former dye and resin fac-tory listed as an EPA Superfund cleanup site.

Despite the earlier suspicions, however, it wasnot until a Philadelphia nurse who had workedwith many cancer patients from Toms Riverreported her suspicions to the EPA that an inves-tigation took place. The request eventually madeits way to the New Jersey Department of Health

and Senior Services, which conducted an incidencestudy based on registry data. That investigationwas completed in 1995 and found a significantexcess of child brain and central nervous systemcancers in the community and in the surroundingcounty. However, the results did not becomeknown to the public until the publication of anewspaper report the next year.

16

The results of the initial study, while explosive,were far from definitive. The study was based onregistry data that were four years old. As com-munity groups and public officials plotted aresponse to the findings, one of the first orders ofbusiness was to update the cancer registry andre-create the 1995 study based on up-to-datedata. The results of the second study, released in1997, were illuminating: child cancer rates inToms River and its surrounding township wereagain found to be significantly elevated, but thosein the rest of the county were not. The elevationin total cancer cases was most apparent in girlsunder age 5.17

Investigators then began to look at possible envi-ronmental exposures that might have some rela-tionship to the cluster. Based on the excess of dis-ease that had been discovered and the existenceof potential exposures that could be investigated,researchers opted to launch a case-control study.

The study included interviews with nearly 200parents of children in the township, one-fifth ofthem parents of children with cancer. Birth recordsof township children were examined, computermodeling undertaken to reconstruct drinking

14

water conditions at various times over the previ-ous three decades, and environmental monitoringwas carried out of water supplies and localSuperfund sites in order to determine any poten-tial route of exposure the children may have hadto toxic substances.

In December 2001, four years after that studywas initiated, researchers released their results.The study found mothers of the female leukemiacases studied were six times more likely to haveconsumed water from one municipal well thanwere mothers of girls who did not contractleukemia. Female leukemia cases were also morelikely to have been exposed to air emissions fromthe nearby Ciba-Geigy plant. However, investiga-tors did not demonstrate conclusively that anyone contaminant or source of exposure wasresponsible for the cluster.

18

While the Toms River investigation took a total ofsix years and cost millions of dollars, it did breedsome successes. A previously unidentified drinkingwater contaminant was found and new filtrationsystems installed. Measures were taken to limitthe contamination of groundwater from toxicsites.

19

And citizens settled a lawsuit with CibaSpecialty Chemicals, Union Carbide and the localwater company, reportedly for more than $10million.

In addition, the Toms River case highlighted theimportance of access to up-to-date cancer registrydata, bringing about a long-overdue updating ofthe registry. Had those data been available earli-er, researchers may have been able to save time

in their investigation. Also, had there been betterpublic access to the registry and participation inthe inquiry, there may not have been a year’s lagtime between the initial investigations and a pub-lic call for further work. And with consistent sur-veillance of those data, public health officials couldhave turned up evidence of a cluster long before ithappened to be noticed by a nurse in a nearbystate.

Marion, Ohio provides another example.Suspicions of elevated rates of leukemiaamong graduates of a local high schooltriggered environmental testing thatlocated toxic substances on the school’sathletic fields. While no study has conclu-sively linked the contamination with theleukemia cases, the revelations led toclosing of the fields and, eventually, theconstruction of new schools for the dis-trict’s children. (See "Case Study: Marion,Ohio," page 29.) Even when specific expo-sure pathways cannot be identified, clus-ter investigations often result in publicofficials taking added steps to monitorand remediate potential sources of envi-ronmental exposure. (See "Case Study:Atkinson, N.H.," below.)

CASE STUDY: ATKINSON, N.H. (CANCER)

CLUSTER REPORT BRINGS AWARENESS OFWATER CONTAMINATION

In January 2002, residents of Atkinson reportedto state officials their suspicion that a small sec-tion of their town had elevated rates of cancer.Residents identified a total of 32 cases of varioustypes of cancer diagnosed within a half-mile

15

16

radius of Providence Hill Road. Of those cases, atleast 10 had been diagnosed within the past threeyears.20

The New Hampshire Department of Health andHuman Services investigated the claims using1994-1998 data from the state’s cancer registry.In late February, DHHS officials determined thatthe cases – which had been reported over a 22-year period and included several different cancerdiagnoses – did not represent a true cancer clus-ter.

21

However, the state’s inquiry and the public atten-tion focused on environmental conditions did bringabout increased awareness of a potential publichealth concern in the town. The initial cancerreports prompted several area residents to havetheir private wells tested. Those tests revealedlevels of radon in private wells more than 30times higher than the concentrations at which thestate of New Hampshire says well users shouldbe concerned and more than 15 times higher thana proposed federal standard.22 A week after thestate closed its cluster inquiry, test results from apublic water system that serves 850 town resi-dents were released, disclosing levels of radon upto 20 times the level of state concern.23 Resultsof water tests at a local elementary school foundsimilar levels of radon.

The EPA recognizes that radon in drinking waterposes a risk of lung and stomach cancers. But,more than a decade after first proposing a drink-ing water standard for radon in public water sup-plies, the EPA has yet to finalize the rule. As a

result, state officials have no power to compelpublic water suppliers to install equipment toreduce radon levels in drinking water. However,state officials did offer well testing to neighbor-hood residents. In addition, the incident broughtthe lack of a national standard for radon in drink-ing water to the attention of the state’s electedofficials, who would be in a position to press EPAon the matter.

24

The Atkinson cancer cases may not have met thedefinition of a true cancer cluster. But the height-ened community awareness that arose as a resultof the cluster inquiry resulted in the identificationof a potential health hazard, and may help moti-vate public officials to take long-overdue action toprotect the public health.

Critics of cluster investigations some-times suggest that the public healthresources used for those investigationswould be better spent on environmentalcleanup. While citizens should not haveto wait until health problems are docu-mented in order for toxic threats in theirneighborhoods to be cleaned up, historyhas shown that it has frequently onlybeen the discovery of a disease clusterthat has created sufficient urgencyamong public officials to spur cleanupactivity. Moreover, the initiation of a clus-ter investigation has often led to the iden-tification of local environmental threatsthat had previously been unknown orpoorly understood.

Targeting Public Health Resources

The existence of a disease cluster can alsohelp public health officials target preven-tion and treatment resources to the pop-ulations that most need them.

In Brownsville, Texas, for example, aninvestigation of a cluster of neural tubedefects in the early 1990s did not conclu-sively link the cases to any known envi-ronmental exposure. The identificationof the cluster, however, resulted inrenewed efforts to educate women ofchild-bearing age in the border region ofTexas about the importance of consum-ing multivitamins that include folic acid,which is known to reduce the risk ofneural tube defects in newborns. (See"Case Study: Laredo, Texas," below. )

CASE STUDY: LAREDO, TEXAS (NEURAL TUBE DEFECTS)

AGGRESSIVE MONITORING CATCHES EMERGING CLUSTER

Within a 36-hour period in 1991, three babieswere born with anencephaly (the absence of partor all of the brain) in a single clinic in Brownsville,Texas, on the Mexican border. Over the next fewweeks, several other babies were born with thecondition in Brownsville – highlighting a largerproblem with neural tube defects (NTDs) in com-munities along the border.

The Brownsville cluster – and the lack of informa-tion about the prevalence of NTDs in the state –shocked Texans. In 1993, the state Legislaturecreated the Texas Birth Defects MonitoringDivision (TBDMD), a tracking system for birthdefects in which state public health officialsactively seek out cases of birth defects at localhealth care facilities. The benefits of that systemwere demonstrated in 2001 in another bordercommunity: Laredo.

In early 2001, as field staff from the divisionwere collecting data from health care facilities,they noticed what appeared to be a high rate ofanencephaly cases in Laredo. A subsequent reviewof data collected for the registry found that sevencases of anencephaly had occurred to babies bornof mothers living in Laredo in late 2000 and early2001 – a statistically significant excess of caseseven when compared to other counties along theborder. In May 2001, the division issued a mem-orandum about the high rates and laid out thepublic health response. That response included acampaign by the city of Laredo to encourage theconsumption of multivitamins including folic acid,which has been shown to significantly reduce therisk of NTDs.

25

Meanwhile, state investigators are involved inefforts to determine the cause of the high rate ofNTDs along the Texas-Mexico border. In conjunc-tion with CDC and EPA, Texas health officials areconducting a case-control study, including surveysand laboratory testing of biological samples, in anattempt to narrow down the potential causes ofthe elevated rates.

The active surveillance done by TBDMD has alsoenabled the state to be more pro-active in theinvestigation of suspected birth defects clusters.Of the 76 cluster investigations the division hasundertaken, more than 60 percent have been ini-tiated by public health officials or health careproviders. By way of comparison, less than onethird of all cancer cluster inquiries nationwide areinitiated by public health or health care profes-sionals.

17

18

Moreover, the availability of registry data haseliminated the expense of attempting to trackdown and verify birth defects cases with physi-cians and the uncertainty of relying on birth cer-tificates for investigations. One study mentionedby TBDMD staff found that only 40 percent of allDown’s syndrome cases, for example, were notedon birth certificates.

26

Similar failings with birthcertificates have been noted in other studies ofbirth defects clusters.

A senior epidemiologist at the division reportsthat the registry "has drastically increased thetimeliness and also drastically increased the easeand accuracy with which we can do these investi-gations."27 While the state still faces many chal-lenges in NTD prevention – and in discoveringwhy cases of the disorders tend to occur along theborder – the state’s birth defects registry hasalready proven itself to be a powerful tool ininvestigating community concerns, identifyingproblems as they develop, and helping the statefocus its public health response.

Similarly, the identification of a cluster ofpleural cancer cases in Charleston, SouthCarolina enabled state officials there toprovide medical screening services to for-mer shipyard workers and others at highrisk in the community. (See "Case Study:Charleston, S.C.,” page 22.)

Even when citizen reports of disease clus-ters are clearly incorrect, the ensuinginteraction between the citizen andhealth officials provides the opportunityto spread important information aboutthe prevention of disease. In their initial

response to cluster concerns, most statehealth departments provide generalinformation about the disease in ques-tion and tips for prevention.

In short, while most cluster investigations“fail” to link a specific outbreak of diseaseto a specific environmental cause, the fewthat succeed can have dramatic impactson public health. And those that succeedmore modestly can result in improvedenvironmental conditions and access tohealth resources that provide real healthbenefits to community residents.

A Four-Stage Process

Inquiries into disease clusters are gener-ally conducted at the state level – usuallyby state public health departments, butsometimes by state disease registry pro-grams, state environmental officials orlocal health departments. In some cases,federal agencies, such as the Centers forDisease Control and Prevention (CDC),are called upon to lead or provide assis-tance with cluster inquiries.

The CDC and many states have estab-lished protocols that guide their clusterinvestigations. The CDC protocol, estab-lished in 1990, serves as the basis formany of the state protocols. This protocolbreaks a cluster investigation down intofour stages: initial response, assessment,major feasibility study and etiologic inves-tigation (investigation into cause).

28

Stage 1: Initial Response

This stage begins at the moment publichealth officials receive a report of a sus-pected cluster. Public health officialsgather identifying information and initialdata on the cluster from the caller, obtainnames and contact information for oth-ers with the disease, and share basicinformation about the disease. The CDCprotocol requires that calls be tracked,preferably in a computerized database,and that each call receive a writtenresponse.

At this stage, public health officials mustalso make a determination as to whetherto continue the investigation.Continuation is warranted under the

CDC protocol if the initial report describes a) a single and rare disease, b)plausible exposure to a source of envi-ronmental contamination, or c) plausibleclustering. In the event none of these fac-tors are present, a report is sent to thecaller indicating that no further investi-gation is necessary.

Stage 2: Assessment

Once the decision has been made tomove forward with an investigation, pub-lic health officials then carry out anassessment of whether the incidence ofdisease in the community constitutes atrue cluster. This assessment takes placein several smaller stages moving fromquick and general assessment of inci-dence rates to more detailed investiga-tion, if warranted.

The first step in the assessment is to con-duct a quick study from available data ofwhether a cluster may exist in a commu-nity. In some cases, this may include con-sulting data from a disease registry, wheresuch data are readily available.

At this point, investigators must make acrucial decision; they must define the“community” to be studied – its geo-graphic boundaries and population.Ideally, this decision will be made basedon some hypothesis (or hypotheses) as tothe cause of the cluster – for example, byanalyzing disease rates in sections of atown nearest to a hazardous waste site orthose parts of town receiving water froma suspect well. Investigators may alsochoose to narrow the time period being

19

How Disease Clusters Are Investigated

20

studied to account for an environmentalexposure of limited temporal scope.

In practice, however, decisions as to theboundaries of the population to be stud-ied are often more arbitrary. First, com-munity residents and researchers oftendo not have a good hypothesis when theybegin an investigation. For example, astudy of cancer cluster investigationrequests found that residents did notspecify a suspected cause in 40 percent oftheir initial reports.

29

Second, the waydata are collected and stored in state dis-ease registries often makes it difficult forresearchers to break the data down intothe smaller geographical units of interestin many cluster inquiries. As a result,many preliminary reports on clustersevaluate rates of disease based at a city orcounty level – levels that may not beappropriate for evaluating a suspectedenvironmental exposure.

Once the boundaries of the study havebeen set and initial data on disease inci-dence collected, investigators must deter-mine what rate of disease would beexpected for a similar population andrun a statistical comparison.

If the preliminary evaluation suggeststhat an excess may exist, investigatorsthen must verify the cases by trackingdown medical records or consulting a dis-ease registry.

Finally, if the cases are verified and anexcess confirmed, a thorough assessmentof the potential cluster should take place.Elements of this assessment include: a)determining the most appropriateboundary for the study, b) identifying allcases within that boundary, c) identifying

the data that will be used for the studyand the statistical and epidemiologicalprocedures to be used in analyzing thedata, d) performing an in-depth litera-ture review and considering the plausibil-ity of any association with an environ-mental factor, e) assessing the likelihoodthat an event-exposure relationship mayexist, and f) assessing the community’sperceptions, reactions and needs.

At this point, the investigation will likelycome to one of three conclusions. If anexcess of disease does not exist, the inves-tigation will be terminated. If an excessdoes exist, but there is no plausible linkto an environmental exposure, the inves-tigation will also be terminated, but pub-lic health officials may advise the com-munity of any risks they face and makerecommendations for further disease sur-veillance, environmental monitoring orpublic health follow-up. If both an excessand a possible link to an environmentalexposure exist, the investigation pro-ceeds to the next stage.

Stage 3: Major Feasibility Study

At this stage, investigators reconsider thedefinition of the population being stud-ied, re-evaluate the literature, and mapout a plan – including costs – for con-ducting a major investigation. If adetailed investigation into cause appearswarranted, investigators proceed to thefinal stage.

Stage 4: Investigation into Cause

In this stage, investigators embark on aquest to determine whether a particularenvironmental exposure can be linked tothe disease in question. The CDC notesthat the purpose of this investigation is

not focused on the cluster itself, but onthe public health issues it raises.

Data for Cluster Investigations

One of the most important decisionpoints in a cluster inquiry takes place atthe beginning of Stage 2. It is at this pointthat public health officials must deter-mine if there is significant likelihood of acluster to merit further investigation.Poor decision-making at this stage canlead to the inappropriate expenditure ofpublic health resources on investigationsof unlikely clusters or, alternatively, thefailure to fully investigate conditions thatpose a risk to public health.

Information on the expected and actualincidence of a disease within a definedcommunity is critical to this decision.

Sources of Information on DiseaseIncidence

Researchers generally have four optionsfor assessing the number of cases of aparticular disease in a community.

REGISTRY DATA – The most convenient andgenerally most accurate source of data ondisease incidence is the centralized dis-ease registry. While there are inconsisten-cies in the completeness and accuracy ofsuch data from registry to registry, (see"The Current State of Health Tracking Data" page32.) registry systems typically require sys-tematic reporting of disease incidence byhospitals, labs or other health care pro-fessionals and subject such data to someform of quality assurance. In the bestcases, researchers can, through a relative-ly straightforward query of a database,

derive detailed information about thenumber of cases diagnosed in areas assmall as a census tract, along with demo-graphic information that allows a moreaccurate comparison with similar popula-tions elsewhere.

DEATH RECORDS – Mortality (or death)records typically play a role in investiga-tions of cancer clusters. In areas withgood cancer registries, mortality istracked along with incidence, providingresearchers with a picture both of howmany people in a community have con-tracted the disease and how many havedied from it.

Prior to the advent of cancer registries,death records were the only readily avail-able source of centralized data to trackcancer. Death records have the advantageof being complete – death certificates arefiled in all jurisdictions and have been forgenerations. However, they have numer-ous disadvantages as an analytical tool.One obvious disadvantage is their inabili-ty to track chronic disease that results indisability but not death. Another poten-tial problem is the possibility of inaccu-rate or incomplete descriptions of causeof death (See "Case Study: Charleston, S.C.,"below.). A third problem is that mortalityis not a true gauge of disease incidence.In some areas, people may live longerwith a disease due to demographics oraccess to superior health care. Use ofdeath records in those areas may maskthe degree to which the disease is affect-ing the community.

21

22

CASE STUDY: CHARLESTON, S.C. (PLEURALCANCER)REGISTRY DATA SUCCESSFULLY IDENTIFY AN OCCUPATIONAL CANCER CLUSTER

In December 1998, residents of a Charlestonneighborhood contacted the state of SouthCarolina with their suspicion that a nearby haz-ardous waste site was contributing to a perceivedcancer cluster in their neighborhood.

The South Carolina Central Cancer Registry(SCCCR), which had been established three yearsearlier, took up an investigation. Using cancer inci-dence data for 1996, the registry found signifi-cantly elevated rates of five cancers – colo-rectal,stomach, lung, laryngeal and pleural. The rates ofpleural cancer (a cancer of the lining between thelungs and rib cage) were particularly alarming,with three cases where only a fraction of a casewould have been expected.

SCCCR staff expanded their investigation toinclude a study of mortality statistics from 1969to 1995 and new pleural cancer cases identified in1997. SCCCR also expanded the geographicalscope of the inquiry to a three-county area – iden-tifying a total of 19 cases of pleural cancer, afour-fold increase over what would have beenexpected. Further investigation determined that12 of the 19 individuals had worked in theCharleston naval shipyard, where they were like-ly exposed to asbestos – one of the main risk fac-tors for pleural cancer. A spatial study of the othercancer cases was also undertaken, but found noobvious clustering of the cancers.

30

In May 1999, less than six months after initiatingits study, SCCCR issued a report determining thata cluster of pleural cancer did in fact exist, andrecommending further surveillance of cancer ratesin the area and other public health measures.Screening services have since been offered to localresidents concerned about occupational exposureto asbestos and further studies have been under-taken of cancer rates in the region.

SCCCR staff report that it would have been diffi-cult or impossible to confirm the pleural cancercluster without access to accurate registry data.

31

Without access to a registry, investigators wouldhave had to comb death records, in which pleuralcancer is often confused with lung cancer andother diseases as a cause of mortality.

The existence of the registry has also enabledSCCCR to conduct quicker and more numerous can-cer cluster investigations. Previously, the statehad relied on death records or on self-reportingfrom the community to investigate clusters. Theadvent of the central cancer registry in SouthCarolina has not only aided the discovery of a"true" cancer cluster, but enhances the ability ofstate officials to investigate other suspected clus-ters in the future.

PREVALENCE DATA – Another approach toinvestigating disease clusters is to useprevalence data – data that indicate thenumber of people living with a disease orexperiencing specific symptoms at a par-ticular moment in time. Unlike registrydata, in which each new case of a diseaseis recorded at the time of diagnosis,prevalence is typically ascertained using

23

indirect measures, such as informationon hospital admissions or discharges orphysician caseloads.

Prevalence data – gathered through sur-veys of hospital discharges, emergencyroom visits, and individuals – are oftenused at the national level to ascertaingeneral health trends. Their use at thelocal level, or in cluster inquiries general-ly, can be more problematic. As withdeath records, the number of hospitaldischarges for a certain disease can be asmuch a reflection of access to health careresources or willingness to seek treat-ment as it is an indicator of the degree towhich a disease is affecting a community.

SELF-REPORTING/HEALTH SURVEYS – This ishow nearly all disease cluster inquiriesbegin, with individuals or groups in a par-ticular community reporting their ownexperiences with disease to local or statehealth officials. In cases in which inci-dence data are unavailable from a centralregistry and in which mortality or preva-lence data are not sufficient, it is also theonly way to determine the number ofpeople within a community that sufferfrom a given disease.

Self-reporting can be encouraged byhealth officials through systematic healthsurveys of a community or by putting theword out through newspaper articles,mailings or other forms of public out-reach. However, self-reporting creates anumber of problems for investigators.First, a thorough accounting of diseaseincidence or prevalence requires aggres-sive public outreach, and even with thebest of efforts some people are missed.Second, individuals may be reluctant tocome forward due to concerns about

their privacy or may simply not bother tospeak up, thus reducing the number ofcases identified. Finally, a reliance on self-reporting requires that health officialsexpend the time and effort to verify eachclaim of disease. This process is bothtime- and resource-intensive. (See "CaseStudy: South Boston, Mass.,” below.)

CASE STUDY: SOUTH BOSTON, MASS.(SCLERODERMA AND LUPUS)

LACK OF REGISTRY DATA FORCES COSTLY HUNTFOR CASES OF DISEASE

In 1997, residents of South Boston, an urbanneighborhood just south of downtown Boston,noticed that several women in the community hadcontracted scleroderma – an autoimmune disorderthat can affect the skin or internal organs. Usingword of mouth, residents compiled a list of currentor former residents with the disorder, as well asa larger list of residents with lupus, anotherautoimmune disorder.

In 1998, South Boston residents contacted theMassachusetts Department of Public Health withtheir concerns. Department officials, recognizingthe unusual concentration of cases, initiated aninvestigation. The progress of the investigation,however, when compared to more common cancercluster investigations, has been slow.Massachusetts, like other states, does not sup-port a central registry for autoimmune disease. Asa result, investigators were forced to conductextensive community outreach to identify cases –spreading word of the investigation through the

24

community and asking physicians at local healthcenters that treat autoimmune disease to alerttheir patients from South Boston about the study.Such outreach efforts, taking place over the courseof several years, succeeded in identifying and ver-ifying 26 current or former South Boston residentswith scleroderma and 60 with lupus.

32

Even so,the project manager for the study states that, dueto the limits of self-reporting, "we can never besure we got all the cases."

33

State officials reportthat the number of lupus and scleroderma casesrepresents a roughly two- to four-fold excessover what would be expected. But a lack of accu-rate baseline prevalence data for both diseasesmakes it impossible for state officials to pinpointthe exact degree of excess.

Investigators have extensively interviewed thosepatients and are now identifying a control popula-tion with which they will conduct similar inter-views. The study is designed to determine what,if anything, those with autoimmune disorders inthe neighborhood have in common in order toassess whether environmental exposures may beresponsible for the cluster.

In many ways, the South Boston investigation is apositive model of a state taking aggressive actionto investigate a suspected disease cluster. But italso shows that, absent the existence of readilyavailable registry data, cluster investigations canbe time-consuming and resource-intensive. It isquestionable whether many states would be will-ing to commit four years and $1 million, asMassachusetts has, to such an inquiry.

34

Establishing Expected Rates ofDisease

Of equal importance to establishing ratesof disease incidence is the process ofestablishing the “expected” rate of dis-ease for a given community. For stateswith access to comprehensive disease reg-istries, the task is relatively straightfor-ward – one can calculate the rate of agiven disease statewide or in a particulargeographical area based on demographiccriteria.

States without cancer registries can usethe Surveillance, Epidemiology and EndResults (SEER) database to establishexpected rates of cancer incidencenationally. The SEER database is basedon cancer registry data for a select groupof states. SEER registries cover 14 percentof the U.S. population and are in theprocess of being expanded to cover 26percent.

35

Investigators researching other chronicdiseases typically do not have access tosimilar, comprehensive resources. Often,investigators must rely on published stud-ies of disease incidence that may or maynot reflect local conditions. In somecases, the task of determining the expect-ed rate of a disease for a particular popu-lation can take years or become a majorroadblock to a successful cluster investi-gation. (See "Case Study: El Paso, Texas,"below; "Case Study: Dickson County, Tenn." page34.)

25

CASE STUDY: EL PASO, TEXAS (MULTIPLE SCLEROSIS)

EXPECTED LEVELS OF DISEASE DIFFICULT TO FINDIn 1994, a former El Paso resident with multiplesclerosis contacted the Texas Department ofHealth (TDH) to report an apparent cluster of MScases among people who grew up in one El Pasoneighborhood from the 1940s through the 1960s.Of the 15 people in the neighborhood reported tohave contracted MS, 14 attended a single elemen-tary school. Former residents raised questionsabout the possible role a local metals smelter mayhave played in the outbreak.

In its initial 1996 evaluation, the health depart-ment determined that rates of MS among studentsat the school were approximately four times whatwould be expected, based on national prevalencerates. In 1997, the department received fundingfrom ATSDR to conduct a more detailed investiga-tion.

When completed in 2001, that study confirmedthe existence of an excess risk of MS among stu-dents at the elementary school, but just how largethat risk is remains unknown. Because no registryfor MS exists in Texas, investigators were forcedto undertake a survey of those who had attendedthe school – a survey with limited ability to iden-tify all the cases of MS that arose within the com-munity. The health department reported that theywere unable to locate approximately 70 percentof the students who had attended the elementaryschool and that less than half of those who werelocated participated in the study. Among those

who refused to participate were at least two indi-viduals with suspected cases of MS; those caseswere excluded from the study.

Moreover, the investigation was hampered by alack of data as to the rate of disease that wouldbe expected in such a community. The healthdepartment determined in its final report that"current comparison MS prevalence estimates,appropriate for use with the El Paso schoolcohorts, were not available ..." The studies exam-ined by the department to determine the expect-ed rates of disease for the former students wereeither more than a decade old or based on self-reported, rather than medically confirmed, casesof MS. No prevalence statistics were availablethat were specific to Texas.

Investigators were somewhat successful in sur-mounting another data-related challenge: deter-mining the levels of heavy metals to which stu-dents might have been exposed. Blood test resultsand hair samples from the time showed that stu-dents had been exposed to high levels of lead andother heavy metals.

36

In its final report, the Department of Health rec-ommended that better prevalence data for MS beestablished for Texas, more study of the El Pasocluster be carried out, and a national study on thelinks between heavy metals exposure and MS beconducted. In 2000, the department launched aninitiative to develop better baseline prevalencerates for Texas.

More than six years after the initial report of a

26

suspected cluster – and despite the investment ofsignificant resources – investigators have still notfully identified the severity of the cluster – adetermination that could have been made rela-tively quickly with the existence of good registrydata for MS. However, the study has succeeded infocusing attention on the lack of good data on MSprevalence and has already sparked further inves-tigation into the environmental links to MS.

Limitations of Cluster Investigations

A lack of data is not the only factor thatinhibits the ability to establish the exis-tence of disease clusters. The statisticalmethods often used to evaluate clusters –and the extraneous forces sometimesexperienced by cluster investigators –impose their own limitations.

Methodological Limitations

As noted above, many cluster inquiriesattempt to determine whether a statisti-cally significant excess of disease exists ina community. The purpose of this inquiryis to rule out, to the extent possible, thepossibility that the cluster could be theresult of chance. However, standard sta-tistical tools are generally most useful incluster investigations if the communitybeing studied is large.

In order for a true cluster to be diag-nosed within a small community of a fewhundred residents, the level of diseasemust not just be above the expected level,but several times above the expected level inorder to meet the test of statistical signif-icance. This criterion is rarely met. Inaddition, because common diseases havea high rate of incidence at baseline, it isunusual that elevations of those levels are

so great as to meet the test of statisticalsignificance, unless the population beingstudied is large.

There are some notable exceptions. Onestudy based on registry data has foundthat women in the metropolitan coastalregion of the eastern U.S. are more likelyto contract breast cancer than women inother parts of the country. A New YorkState project to map cancer ratesstatewide has also uncovered the exis-tence of a potential breast cancer clusteron Long Island. (See "Case Study: SuffolkCounty, N.Y.,” below.)

CASE STUDY: SUFFOLK COUNTY, N.Y. (BREAST CANCER)

ACTIVE SURVEILLANCE PROVOKESINVESTIGATION

In 1998, following years of concerns about ele-vated breast cancer rates on Long Island, NewYork State became the first to initiate a cancermapping program highlighting areas of the statewith elevated rates of various cancers.

In May 2002, state officials used the data fromthe mapping project to initiate an investigationinto elevated rates of breast cancer in seven zipcodes in Long Island’s Suffolk County. The areawas the only part of the state to have breast can-cer incidence more than 50 percent in excess ofexpected rates for the 1993 to 1997 period.

37

The investigation will examine factors – includingenvironmental and lifestyle factors and variations

in medical treatment – that might be responsiblefor the excess. Community members will have theopportunity to express environmental concernsand epidemiological studies may take place.

State officials report that the Suffolk Countyinvestigation is the first of many that will resultfrom the state’s Cancer Surveillance ImprovementInitiative. The initiative also includes the mappingof possible risk factors, research on the causes ofcancer and cancer prevention programs. The initia-tive demonstrates the promise that ongoing sur-veillance of disease registry data holds forprompting pro-active investigations of potentialdisease clusters and finding answers to pressingpublic health problems.

Generally, however, to successfully evalu-ate suspected clusters among small popu-lations, researchers must rely on morethan just a statistical comparison of dis-ease incidence rates. They must also havea hypothesis as to the potential environ-mental or behavioral cause of the clusteras well as the means to evaluate thathypothesis – for example, detailed dataon environmental conditions or expo-sures – through an epidemiological inves-tigation. Such investigations, however,are expensive, and may be hindered bythe lack of availability of good data on thedegree to which individuals were exposedto the environmental contaminant inquestion.

Political Influence and Subjectivity

Finally, it is necessary to acknowledge therole of political forces in the conduct anddesign of cluster investigations. Politicalpressures may cut both ways – local citi-zens may clamor for more investigation

of a suspected cluster, while others maywish to avoid an investigation that couldresult in lower property values or harmthe interests of a local industry. Further,public health departments face their owninternal pressures caused by staffing,budgetary and other limitations.

As noted above, investigators must deter-mine, at an early point in their analysis,the spatial and temporal boundaries ofthe “community” they intend to study.Where a hypothesis as to the cause of acluster exists, health officials should usethat hypothesis as the basis of theirboundary setting decision. Yet, such deci-sions are inherently “judgment calls” andopen the possibility for intentional orunwitting distortion. A cluster of diseasethat is limited to the area around a sourceof air pollution, for example, can bemade to statistically “disappear” if publichealth authorities choose to use broadswaths of territory – such as an entirecounty – as the basis for their analysis.Similarly, studying too narrow a popula-tion base may exclude related cases innearby areas or limit the possibility ofdeveloping statistically significant find-ings.

In addition, public health officials oftendetermine which illnesses suspected by acommunity are studied, and whicharen’t. Because many requests for clusterinvestigations include several diseasessuspected to be in excess, limiting thenumber of diseases under study couldlead to some potential clusters beingignored. (See "Case Study: Fairfield, Maine,"page 28.)

27

28

CASE STUDY: FAIRFIELD, MAINE (BRAIN CANCER)

NARROW FOCUS LEADS TO QUESTIONS ABOUTCLUSTER INVESTIGATION

In February 1999, residents of rural FairfieldCenter reported an alarming number of cases ofbrain cancer and other ailments, which they sus-pected could be linked to the illegal disposal oftoxic paper mill waste in a local landfill. TheMaine Cancer Registry (MCR) was called upon toinvestigate the claims, conducting interviews withFairfield residents and learning about casesthrough word of mouth.

At the time of the investigation, the MCR was cur-rent only through 1995. Investigators found threebrain cancer cases that were diagnosed after thatyear – cases that would not have been identifiedhad MCR relied on registry data alone. In all,investigators found six cases of brain cancer thathad been diagnosed in residents between 15 and44 years of age – a rate five times the state aver-age. In October 1999, the MCR confirmed that acluster did in fact exist and the agency soon begana case series investigation to determine what, ifanything, those with brain cancer shared in com-mon.

The results of that investigation were published in2001. It found that more than half the clustercases simultaneously shared residence in Fairfield,spent time near the former paper company land-fill, and were exposed to burning at a municipaldump. But because the study did not include a con-trol group, it was unable to come to a conclusion

as to whether any of these possible routes ofexposure could have been responsible for the clus-ter. Because the number of cases was so small,MCR concluded that such a study would not bepossible, and that even if it were possible, MCRdid not have the resources to conduct it. Thereport recommended that the state continue toconduct surveillance for additional brain cancercases in the area and conduct a case-control studyif more cases arise.

39

The conclusion of MCR’s investigation has notresolved the issue, however. A group of 20 peo-ple or their survivors have filed suit against thoseallegedly responsible for the contamination at thelandfill, claiming that chemicals at the site causeda variety of ailments. While MCR’s investigationfocused on brain cancer, plaintiffs in the case alsoclaim to have suffered from elevated rates oflupus and lymphoma. A former director of theMassachusetts Cancer Registry who submitteddocuments on behalf of the plaintiffs noted thatthere were 10 cases of diagnosed lupus inFairfield in 2000 where less than one case oflupus would be expected.

40

Maine public health officials responded promptlyand aggressively to citizen reports of a brain can-cer cluster in Fairfield. However, had it not beenfor the MCR staff’s active case-finding, the clus-ter may not have come to light, due to the back-log in the state’s cancer registry. Moreover, byfocusing exclusively on brain cancer, public healthofficials may have missed out on other potentialclusters within the community. Three years afterthe initial reports of a brain cancer cluster in

Fairfield, Maine’s cancer registry still does notmeet federal standards for timeliness.

41

Decisions such as these are at the nexuswhere scientific and political concernsmeet. Frequently, communities have sig-nificant concerns or complaints with thescope of a cluster investigation. Allowingcommunity members to participate inthe design of a study at an early phase ofan investigation can mitigate these con-cerns.

In a few instances, cluster investigationshave shown the potential to be overtlymanipulated in response to political con-cerns. In 1998, for example, the individ-ual in charge of the environmentalcleanup of a former Army depot in Ohiolinked to a school-based leukemia clusterwas removed from his position. A federaladministrative law judge later found thathe was removed due to his insistence onan adequate investigation of the site. Thejudge cited in his ruling a 1997 memofrom the man’s superior to Ohio EPAstaff: “The team has been instructed towork closely with ODH (the OhioDepartment of Health) in seeking outinformation which can allow us to con-clude that the existing ‘environmental’conditions in the local community(s) donot pose a threat to human health.” Thememo came two weeks after ODH issuedits first report highlighting potentialincreases in leukemia rates in the town.

42

ODH subsequently concluded that therewas an excess of leukemia among gradu-ates of the high school but did not assigna likely cause. (See "Case Study: Marion, Ohio,"below.)

CASE STUDY: MARION, OHIO (LEUKEMIA)

CLUSTER STUDY LEADS TO DISCOVERY OF CONTAMINATION, HINTS AT LINKS

In 1997, graduates of Marion’s River Valley HighSchool and a nurse at the school contacted Ohiopublic health officials with concerns about apotential leukemia cluster among graduates of theschool. The high school had been built in the early1960s on land purchased from U.S. Army that hadbeen used as a depot for the cleaning and repairof vehicles and heavy machinery. Following thecluster report, the Ohio EPA investigated and dis-covered several carcinogenic substances at andnear the site, some at levels above those thoughtto pose potential dangers to public health.Continued testing by the Army Corps of Engineersdiscovered additional toxic substances. In 1999,officials restricted access to the school’s athleticfields, where toxic substances had been found.

The Ohio Department of Health began its investi-gation into the matter in 1997. First, it studiedleukemia mortality rates in Marion County, dis-covering a sharp increase over time in the deathrate from leukemia within the city of Marion,even as mortality rates in the rest of the countyand the state overall declined.

The health department’s next step was to conducta review of registry data for leukemia for theyears 1992 through 1996. The study ran into sig-nificant delays because Ohio’s cancer registry wasup to date only through 1992. State researcherswere forced to pore through medical records touncover every case of leukemia in Marion over the

29

30

1992 to 1996 period, a process that took nearlya year.

43

When finally completed, the study foundsignificantly elevated rates of diagnosis in the cityand county of Marion only among women 60years of age and older.

44

However, the Department of Health had not yetinitiated the study most wanted by residents – adetailed study of graduates. By 1999, local resi-dents, working on their own, had identified morethan 100 cases of cancer among graduates of thehigh school. They turned their results over to thehealth department, which initiated its own study.

45

That study concluded that rates of leukemia andesophageal cancer among graduates were signifi-cantly elevated. The health department also iden-tified three cases of leukemia among high schoolattendees who did not graduate and one caseamong current students. Only about one-third ofthe graduates contacted responded to the survey.

Finally, in 2001, the Department of Healthreleased a fourth study, this one a case review ofleukemia among River Valley High School gradu-ates and residents of Marion County. The study,which did not include a control group, concludedthat a variety of factors may have resulted in thedevelopment of leukemia among the graduates. Italso concluded that no further study of the clusterwas necessary.

46

However, several of the study’s findings placedthat conclusion in doubt. Six of the nine graduateswith leukemia had had extensive contact with theschool’s athletic fields. The average age of diag-

nosis among graduates (29) was far lower thanthat of Marion County residents as a whole (58),indicating that graduates may develop moreleukemia in the future as they age. Moreover,one-third of Marion County residents (not justgraduates) who had contracted leukemia had hadsome exposure to the depot grounds.

47

The study also failed to include three studentswho had attended River Valley High, but did notgraduate. And it did not expand the pool of thosestudied beyond the eight graduates (and one cur-rent student) identified by the earlier cancer sur-vey – a survey to which less than one-third ofgraduates responded.

The Marion case holds several lessons about clus-ter investigations. First, it demonstrates theimportance of cluster inquiries in focusing atten-tion on local environmental problems. Without theattention brought about by the inquiry, the scaleof contamination on the school’s grounds mightnot have been so quickly discovered. Second, itshows the potential costs of delay – throughoutthe investigation, students continued to attendRiver Valley High School.

However, the Marion investigation also showsthe limitations of statistically based cluster inves-tigations. The Ohio Department of Health’s twoinitial studies, based on registry data, did notattempt to answer the central question ofwhether graduates of the high school – not resi-dents of Marion – had suffered adverse effects.And the decision to cut off investigation after thecase review study – despite the evidence of

potential links between exposure to contamina-tion at the school site and leukemia – demon-strates the subjectivity of the decision to initiateor conclude a cluster inquiry.

In the end, the Marion investigation, while ulti-mately inconclusive, will have one definitiveresult: new schools for the district’s children areunder construction.

The availability of accurate, up-to-dateregistry data, therefore, is almost alwaysnecessary for the swift and effective inves-tigation of disease clusters. But it may notalways be sufficient. Additional informa-tion – particularly data on human expo-sures to toxic substances – appropriatestudy design, and the involvement of thepublic are often also needed to ensuresuccessful cluster inquiries.

31

32

The last decade has seen a substantialincrease in interest in health trackingamong both state and federal officials.Every state now has either an activestatewide cancer registry or one in theprocess of implementation. Most statestrack at least some birth defects. And reg-istries for selected other chronic diseaseshave been established in some locations.

Yet, in many states, the timeliness, accu-racy and completeness of registry datastill leave much to be desired. Even statesthat do track birth defects and cancer doso in ways that make analysis across statelines difficult. Moreover, such databasesare often not linked to available data onsources of potential environmental expo-sures – such as data on drinking watercontamination or air pollution.

These deficiencies can – and have – ham-pered cluster investigations, leading tothe waste of time, money, and opportuni-ties for protecting public health.

Cancer Registries

As mentioned above, every state now has,or is planning, a statewide population-based cancer registry capable of captur-ing all new cancer cases in the state.

The expansion of cancer registriesdemonstrates the positive impact the fed-eral government can make in expandingopportunities for health tracking. WhileConnecticut established the first cancerregistry before World War II, central cancer registries got their

first major boost with the inauguration ofthe SEER program in 1972. SEER, asnoted above, encompasses several stateand regional cancer registries coveringapproximately 14 percent of the U.S.population.

By 1994, 40 states had established centralcancer registries. Most of these registries,however, were underfunded, many wereyears behind in the compilation andreporting of cancer information, andthere was little standardization of registrydata.

By then, however, the situation was begin-ning to improve. In 1992, Congressauthorized the establishment of theNational Program of Cancer Registries(NPCR) through the CDC. NPCR provid-ed funding for states to initiate orimprove central cancer registries in com-pliance with nationwide standards. Thelegislation required that, by 1998, cancercases be reported to central registrieswithin six months.

The federal involvement has led to signif-icant improvements in cancer registries.In 1997, only 14 statewide registries metthe minimum requirements for certifica-tion by the North American Associationof Central Cancer Registries (NAACCR).In 2001, 26 statewide registries receivedNAACCR’s “gold” certification, whileanother 10 received “silver” certification.In Fiscal Year 2002, the federal govern-ment allocated $40 million to NPCR,which supports registries in 45 states.(The other five states are supportedthrough the SEER program.)

48

The Current State Of Health Tracking Data

Among the standards needed to receivecertification are:

• COMPLETENESS OF CASE ASCERTAIN-MENT– Registries receiving gold certi-fication must include 95 percent of expected cancer cases, while those receiving silver certification must be 90 percent complete.

• COMPLETENESS OF RECORDS – Registries must include, with few exceptions, all relevant information including age at diagnosis, sex, race, and county of residence at diagnosis.

• TIMELINESS – All information and corrections must be entered within 23 months from the close of the diagno-sis year.

49

While the increasing number of cancerregistries certified to NAACCR standardsshows improvement in the system, thereis still a long way to go. More than adozen state cancer registries fail to meeteven the NAACCR’s minimum criteriafor certification.

50

The lack of timely or thorough cancerregistry data can have a major impact onthe progress of cluster investigations. (See"Case Study: Toms River, N.J." page 14.) It canalso make it difficult to conduct activesurveillance to detect clusters as they areoccurring. (See "Case Study: Fallon, Nev."below.)

CASE STUDY: FALLON, NEV. (CHILDHOOD LEUKEMIA)

WHEN TWO-YEAR-OLD DATA AREN’T GOOD ENOUGH

In late summer 2000, Nevada public health offi-cials were alerted by local physicians to an unusu-al increase in diagnoses of acute lymphocyticleukemia (ALL) in and around the town of Fallon.Within the span of a few months, five childrenwere diagnosed with the disease among a popu-lation where only one case would be expected tobe diagnosed every five years. Subsequent inves-tigation and diagnoses led to the identification of15 cases of ALL and one case of another form ofleukemia – all but two of them diagnosedbetween 1999 and 2001.

51

Two months after the initial report of the cluster,state health officials began interviewing each ofthe case families to determine possible commonal-ities. Investigations of several potential sourcesof environmental contamination have taken place,environmental samples have been taken, and theCenters for Disease Control and Prevention haveinitiated biological testing of cases and controls.At least $28 million in federal funds have beentargeted to the investigation.

52

The CDC is expect-ed to release its report in the summer of 2002.

The Fallon case, while it was diagnosed as a clus-ter within only a few weeks, drew attention toweaknesses in the state’s cancer registry system.An investigation by a Nevada newspaper foundthat the registry was more than two years behindin the collection of cancer data and does not have

33

34

the resources to examine the data to identifyapparent cancer clusters.

53

However, even a registry that met federal stan-dards (Nevada’s does not) would have beenunlikely to identify the cluster. Those standardsrequire cancer cases to be filed with registrieswithin six months and for registries to make thedata available within two years. In Fallon, wherethe bulk of ALL cases were diagnosed in 2000,researchers may not have recognized the anomalyuntil 2002 at the earliest – if, that is, anyone waslooking for it in the first place.

Birth Defects Registries

While the nation’s cancer registries havetheir share of problems, accurate infor-mation on the incidence of birth defectsis even harder to come by.

At least 34 states now track some birthdefects, but the vast majority of those sys-tems are inadequate. The Trust forAmerica’s Health reports that in 2000,more than 1 million births nationwide –approximately 25 percent of all births –were not covered by a birth defects mon-itoring program.

54

In many states thathave registries, only the most seriousdefects are monitored, surveillance islimited to certain regions of the state, orthe method of data collection is inca-pable of capturing all new cases of birthdefects. In its analysis, the Trust awardedonly eight state birth defects registries its“A” ranking, signifying that the registriesengage in active case surveillance, rely onhigh quality data sources, and meet highstandards for the accuracy of their data.

An additional 14 states received a “B” rat-ing. Yet, even among the eight states thatreceived an “A”, two do not publish datain a timely fashion and two do not coverthe entire state.

55

The lack of access to up-to-date registrydata can make it extremely difficult forpublic health officials to investigate sus-pected birth defects clusters. (See "CaseStudy: Dickson County, Tenn." below.)Conversely, access to such data can allowpublic health officials to conduct activesurveillance of birth defects rates in spe-cific regions of concern (See "Case Study:Laredo, Texas," page 17.) and can supportbroader research into the causes of birthdefects.

CASE STUDY: DICKSON COUNTY, TENN. (CLEFT LIP/CLEFT PALATE)

RELIANCE ON BIRTH RECORDS WOULD HAVE SHROUDED CLUSTER

In June 2000, the Tennessee Department ofHealth (TDH) was alerted by a local early inter-vention center to what appeared to be a cluster ofcases of cleft lip and cleft palate in DicksonCounty. TDH enlisted the help of the Centers forDisease Control and Prevention in conducting acluster study in the area.

Because Tennessee does not have a birth defectsregistry, TDH staff conducted active case-finding,searching birth records and local hospital dis-charge data to identify additional children bornwith cleft lip/cleft palate between 1997 and2000. TDH eventually identified 18 infants born

with clefts in Dickson County. The decision to con-duct active case-finding was essential: only threeof the cleft lip/cleft palate cases between 1997and 1999 were accurately recorded on birth cer-tificates, compared to the 13 identified by TDH.Reliance on birth records alone would not havedemonstrated the existence of a cluster.

While the rates of cleft lip/cleft palate in thecounty were significantly higher than would beexpected, the lack of good baseline data onexpected rates of clefts prevented investigatorsfrom drawing firm conclusions as to the severityof the cluster. Comparing local incidence of cleftlip/cleft palate to regional figures from Atlantaand to national estimates revealed rates approxi-mately five times higher than expected.

Having proven the existence of a statistical clus-ter, CDC researchers surveyed 15 of the 18 moth-ers in an effort to determine whether the cleftshad a common cause. Thirteen of the 15 mothersreported using a municipal water source; no otherenvironmental exposures (with the exception ofsmoking and occupational exposure to chemicals)were included in the survey. The report concludedthat no single factor appeared to be responsiblefor the cluster, but suggested that a more formalcase-control study might be warranted if morechildren are born with the defect.

56

In the year and a half since the investigation wascompleted, the number of children being diagnosedwith cleft lip/cleft palate in the area appears tohave dropped.

57

However, the lack of good datato determine the expected rate for cleft lip/cleft

palate has left researchers and Dickson Countyresidents to wonder whether the cluster of 1997to 2000 was a statistical anomaly, the result ofhigher incidence rates for the disease in Tennesseeas a whole, or the result of some unknown expo-sure that placed children at risk.

Other Chronic Diseases

While there are myriad problems withthe nation’s system of tracking cancerand birth defects, at least they aretracked. The same cannot be said formany other chronic diseases with sus-pected environmental links.

Consider asthma, a disease whose preva-lence has increased dramatically inrecent years (with the number of casesamong children under four increasing160 percent between 1980 and 1994),and which has well-known connectionswith the environment.

58

Yet, only 23 statesand the District of Columbia had any sys-tem at all for tracking asthma cases in2001, based on information reported bythe states to the CDC. Many of those pro-grams track asthma in only limited juris-dictions or among limited populationgroups.

59

The same is true of autoimmune disor-ders such as lupus and scleroderma,developmental disorders, learning dis-abilities, neurological disorders such asAlzheimer’s and a host of other chronicdiseases whose causes and potential linksto environmental exposures are poorlyunderstood or unknown. A 2000 reportby the Pew Environmental HealthCommission found that:

35

36

• Only eight states and the District of Columbia reported tracking at least some developmental disabilities such as autism and mental retardation.

• Only four states reported track- ing autoimmune diseases such as lupus.

• Most states do not systematically track endocrine and metabolic dis- orders such as diabetes or neuro- logical conditions such as migraines and multiple sclerosis.

60

Summary

Appendix A shows the status of cancer,birth defects and asthma tracking in theUnited States.

Of the 50 states, only three – California,Iowa and Massachusetts – possess track-ing systems for both cancer and birthdefects that meet the highest standardsand also report that they possess any sys-tem for tracking asthma cases.

61

In eventhese three states, the ability to trackother chronic diseases with suspectedenvironmental links is limited or non-existent.

The case studies presented throughoutthis report represent a variety of clusterinvestigations and related studies, fromcommon inquiries into local cancer ratesto cutting-edge research into the linksbetween environmental exposures andrates of disease. Each case, however,demonstrates the importance of high-quality data in the completion of prompt,thorough, and effective cluster inquiries.

Five general lessons can be derived froma close examination of the cases:

Lesson #1: Lack of Data Causes Delaysin Cluster Investigations

Time and time again, investigations intodisease clusters have been delayed by alack of access to health tracking data.This is shown most clearly in recent inves-tigations of clusters of non-cancer dis-eases – diseases for which registry dataare completely lacking in most parts ofthe country.

Two investigations of non-cancer clusters– the multiple sclerosis cluster in El Paso,Texas and the scleroderma and lupuscluster in South Boston, Massachusetts –each took more than two years to identi-fy cases of disease within the community;and in neither case are investigators cer-tain that they have identified all cases.

Those studies also face another problem:the lack of good data on the expected

rates of disease in a community. Again,this is a question that could be easilyanswered with access to complete registrydata.

With regard to cancer cluster investiga-tions, the timeliness of cancer registrydata is also critically important. The back-log in entry of new cases into the NewJersey Cancer Registry in the early 1990s,for example, initially forced researchersto use incidence data that were four yearsout-of-date in their investigation of theToms River cluster, and then to re-dotheir analysis two years later once thedata had been updated.

In cases such as the pleural cancer clusterin Charleston, South Carolina, access totimely and complete registry data led torelatively quick diagnosis of a cluster; inthe South Carolina case, within sixmonths of the initial report. Suchprompt investigations of disease clustersand suspected environmental healththreats should be the standard nation-wide. Unfortunately, they are too oftenthe exception.

Lesson #2: Lack of Data Deters Pro-active Investigation of PotentialClusters

Many lament the low rate at which citi-zen-initiated cluster inquiries yield mean-ingful results. Yet few note that relianceon citizen reporting may miss many

37

The Impact of HealthTracking Data on Cluster

Investigations: Lessons fromthe Case Studies

38

potential clusters simply because individ-ual citizens never become aware of otherinstances of disease within their commu-nities. Active surveillance of high-qualityhealth tracking data could catch thesepotential clusters while givingresearchers more tools to analyze thelinks between environmental factors andchronic disease in a pro-active, ratherthan reactive, way.

Of the cluster inquiries studied, onlythree were initiated by public health offi-cials. Most of the rest were initiated byindividual citizens concerned about ratesof disease in their communities. This isnot a scientific sample, but the fact thatmost clusters are reported by citizens iswell established. A 1998 survey of statehealth departments found that 65 per-cent of requests for cluster investigationscame from the public and 10 percentfrom local health officials. The remain-der came from individual physicians, themedia, elected officials and othersources, many of whom no doubt alsoreceived their initial information aboutthe cluster from the public.

62

In many cases – even where tracking datado exist – state public health officials donot have the resources to conduct activesurveillance of disease patterns and initi-ate pro-active cluster inquiries. This rep-resents a crucial missing link in thenation’s public health system.

In recent years, however, several stateshave begun to fill this gap. Texas officialsused ongoing surveillance of birthdefects rates in communities on theMexican border to promptly identify acluster of neural tube defects in Laredo.And New York officials used their exten-

sive new cancer mapping system to initi-ate an investigation into abnormally highrates of breast cancer in an area of LongIsland. However, more states need to fol-low suit in order to ensure that unusualclusters of disease are promptly identifiedand, if warranted, investigated.

Lesson #3: Lack of Data Deters theIdentification of True Clusters

The most damaging potential result of alack of health tracking data is the failureto properly diagnose a “true” cluster ofdisease, or to discover an environmentallink to a particular disease. In at least onecase studied – the South Carolina pleuralcancer cluster – state health officialsreported that previously existing data,such as death records, would not havebeen sufficient to identify a true cluster.

Of similar importance is the role ofhealth tracking data in determining theseverity of suspected clusters. In numer-ous cases, lack of access to – or lack ofconfidence in – registry data has forcedpublic health officials to rely on self-reports of disease, mortality records, orother less-comprehensive sources ofinformation. Reliance on these data setscan inject uncertainty into cluster investi-gations or compel researchers to poursignificant resources into case identifica-tion, as is occurring with the $1 millionSouth Boston lupus/scleroderma study.

Even where case-finding can be done, thelack of registry data can prevent investi-gators from understanding the truedimensions of a cluster due to the lack ofinformation on the rate of disease thatwould be expected in a given community.In the Dickson County, Tennessee case,

39

for example, researchers were unable toconfidently state the degree of excess incleft lip/cleft palate cases due to a lack ofgood baseline data on disease incidence.

In a few cases – such as the Fairfield,Maine and Chesterfield County, Virginiacancer cluster inquiries – self-reporting isactually the preferred alternative for inves-tigating a cluster due to severe weakness-es in state cancer registries. At the time ofthe 1999-2001 investigation of brain can-cer rates in Fairfield, Maine’s cancer reg-istry was only up to date through 1995.Relying on registry data alone wouldhave caused investigators to miss three ofthe reported brain cancer cases. InChesterfield County, health officialsreported as a result of their case-findingthat Virginia’s cancer registry could beunderreporting cancer cases by as muchas 23 percent.

CASE STUDY: CHESTERFIELD COUNTY, VA.(CANCER)

AFTER A DECADE OF STUDY, POOR REGISTRYDATA HAMPERS INVESTIGATION

In 1984, private drinking water wells in theRayon Park residential neighborhood ofChesterfield County, Virginia were found to becontaminated with benzene, trichloroethylene(TCE, a common industrial solvent), and othervolatile organic compounds leaching from the siteof a military supply distribution center. Threeyears later, residents of the area were hooked upto a public water supply.

Residents of the area expressed concern about the

impact the contamination may have had on theirhealth, specifically, the rash of kidney, liver andcentral nervous system disorders that seemed toplague the community. In 1993, ATSDR released astudy of death records that found elevated cancermortality among men in Chesterfield County as awhole between 1950 and 1979 (the last yearanalyzed). However, the study noted that therewere no data available that would allow theanalysis to be localized to the Rayon Park area.Virginia’s cancer registry began mandatoryreporting in 1989 and data were not yet avail-able. Similarly, the state’s birth defects registryhad only one year of data at the time the studywas conducted.

63

In 2001, amid consistent suspicions of high levelsof cancer and other ailments in Rayon Park, stateand local officials initiated a cancer cluster study.The study was based on a survey of cancer casescompared to expected rates for the county derivedfrom Virginia’s cancer registry. Because of thesmall population involved (about 275 individuals),the study includes only an analysis of overall can-cer rates. The study concluded that cancer rates inthe neighborhood were not significantly differentthan those in the county as a whole.

Despite the cluster inquiry and the increasedattention it has brought to Rayon Park, the areastill faces its share of environmental problems.Since the ATSDR’s 1993 study was completed,groundwater contamination from the site has con-tinued to spread beyond Rayon Park. In 2001, anewspaper investigation found TCE-tainted waterpouring from a pipe at the site into a stream.

64

40

And in early 2002, a contractor cleaning up thesite was indicted for illegally discharging contam-inated groundwater into a local creek.

65

While local health officials conducting the clusterstudy report that the state’s cancer registry wasuseful in determining expected rates of disease,the final report on the cluster investigation notesthat, of 26 confirmed cases of cancer found inRayon Park, six had not been reported to theVirginia Cancer Registry. As a result, the reportestimates that registry may be underreportingcancer cases by as much as 23 percent.

66

Further,the cluster inquiry was limited to residents wholived in Rayon Park after 1985 – despite evidenceof groundwater contamination prior to that date –because earlier cancer registry data were consid-ered unreliable. The Virginia Cancer Registry doesnot meet national certification standards.

The inability of ATSDR to access reliable local reg-istry data prevented the agency from investigat-ing local health problems in 1993. A decade later,registry data continue to shape – and limit – howpublic health officials can investigate health condi-tions in Rayon Park.

Access to up-to-date, high-quality diseaseincidence data would eliminate or allevi-ate many of these problems, allowingresearchers to more confidently deter-mine when a cluster really is a cluster.

Lesson #4: Lack of Data Leads to FewerInvestigations of Potential HealthThreats

Because the cases studied here largely

represent completed cluster inquiries,they do not answer the question ofwhether a lack of health tracking dataresults in some clusters not being investi-gated at all.

Research and common sense, however,indicate that the more investigations costand the fewer resources states have topursue them, the fewer investigations willbe done. Cluster investigation protocolsgive states wide latitude to determinewhen a cluster is worth investigating andwhen it is not. Studies show wide varia-tions in the percentage of investigationrequests that are pursued beyond the firstphone call, with some states satisfying asmany as 99 percent of all inquiries at firstcontact and others satisfying as few as 10percent.

67

(See Appendix B.)

A 1991 telephone survey of public healthofficials responsible for responding tocluster investigation requests document-ed the wide variety of approaches takenby state public health officials in responseto cluster inquiries.

That study found that:

• Two states had no process for responding to cluster inquiries. When citizens with concerns about clusters called, “the officials assumed that they were routed around the agency until they gave up.”

• Several states actively discouraged callers from pursuing cluster inquiries, either by emphasizing the lifestyle causes of cancer or requiring citizens to fill out cum-bersome forms with 10-20 pieces

of information for every cancer case. “Some respondents acknowl-edged that while the forms were imposing, the state did not have the resources to do the epidemiol-ogy themselves. In other words, if citizens did not do the work, the study would not be done.”

Some state officials contended “they would not follow-up unless people showed sufficient interest to return the forms or had their elected repre-sentative write or call.”

• Only five states clearly encouraged callers to pursue their request for cluster investigations.

68

It is unclear the degree to which this situ-ation has changed in the last decade, butthe results of a 1998 survey conducted bya Missouri School of Journalismresearcher, when contrasted with the1991 results, suggest that not much haschanged at all. The 1991 study found that27 states had less than one half-time per-son committed to cluster inquiries; the1998 survey found 26 states with less thanone half-time person. The 1991 studyfound that only two states had formallyevaluated their communications with thecommunity about cancer clusters; the1998 survey found the same results. Andonly two-thirds of the states respondingto the 1998 survey had standing proto-cols for handling disease clusters, nearlya decade after the CDC issued its ownprotocol.

69

The 1991 study concluded that stateswith the strongest overall commitment toenvironmental protection and publichealth were more likely to have a well-

developed capacity to respond to clusterinquiries.

A review of the case studies suggests simi-lar conclusions. In several cases, an initialassessment was not begun until manyyears after citizens began to report theirsuspicions about the clustering of healthproblems in their community. InChesterfield County, Virginia, for exam-ple, it took 17 years after the first detec-tion of contamination in ground water tolaunch a study of Rayon Park residents.In other cases, public health officialsresponded promptly to citizen requestsfor investigations, often consulting reg-istry data to make quick determinationsof whether further study would be war-ranted.

Clearly, access to registry data can makeinitial investigations into clusters far easi-er and less resource-intensive, and thusmore attractive to public health officials.The existence of registry data can alsomake it possible for citizens and inde-pendent researchers to conduct theirown inquiries when public officials fail totake action.

Lesson #5: Lack of Data LeavesCommunity Concerns Unaddressed

Communities that suspect they are expe-riencing a disease cluster are often des-perate for information and help. It isimperative that the public health systemprovide communities with as much infor-mation as possible promptly and withsensitivity, and that environmental andpublic health measures be taken quicklywhere they are needed.

41

42

CASE STUDY: HAZLETON, PA. (CANCER)

STUDIES’ LIMITATIONS DEMONSTRATEIMPORTANCE OF EXPOSURE DATA

In the early 1990s, residents of Hazleton andneighboring Hazle Township began to complain ofgasoline odors infiltrating their homes.Investigation by environmental officials discov-ered that approximately 50,000 gallons of gaso-line had leaked from a series of underground stor-age tanks into the land underneath the homes,many of which sit atop abandoned mines. Severalhomes were evacuated and others monitored forlevels of benzene and other pollutants.

In 1996, the EPA took over cleanup of the area,known as the Tranguch site, at the request ofPennsylvania environmental officials. EPA con-ducted its own residential air quality testing anddetermined that benzene levels did not warrantfurther action. However, EPA failed to communi-cate the results of the testing to residents.

70

Fouryears later, residents again began to express con-cerns about exposure to gasoline vapors. HazleTownship contracted with the University ofPittsburgh to conduct a health study of the area,while the Pennsylvania Department of Healthlaunched its own study of the entire spill area.

The University of Pittsburgh study found statisti-cally elevated rates of leukemia, prostate andstomach cancers among Hazle Township residentsin the spill area. The study relied primarily on self-reporting, rather than registry data, to determinecancer incidence.

71

Four months later, the stateDepartment of Health released its own study,

based on registry data, which found statisticallyelevated levels of stomach cancer only (levels ofleukemia were also elevated, but not statisticallysignificant). The study suggested that the increasein stomach cancer could be linked primarily to thearea’s history as a coal mining center and not toany ongoing environmental contamination.

72

University of Pittsburgh researchers are continu-ing their study of the area, this time focusing onaffected properties in Hazleton.

The Tranguch studies hold several importantlessons. First, they demonstrate the importance ofstudy definition. Because three of the fourleukemia cases in the spill area took place in HazleTownship, the rate of leukemia there was judgedstatistically significant in the University ofPittsburgh study (which included only the town-ship) but not in the state health department study(which included the entire spill area).

Second, both studies demonstrate the difficulty ofrelying on statistical methods to assess diseaseclusters in small geographical areas. The larger ofthe two studies – the Pennsylvania Department ofHealth investigation – included a base populationof only 900 residents. The University ofPittsburgh study included only 207. Because ofthe small sample population, levels of leukemiaincidence more than three times the expected rate(in the state study) still fail to meet the test ofstatistical significance.

Clearly, registry data have proven helpful inassessing the health problems affecting residents

in the Tranguch spill area. However, more infor-mation – such as historical information on benzenelevels in the area or blood testing for evidence ofbenzene exposure – would be needed to draw astrong conclusion about the impact of the spill onresidents’ health.

The consequences of failing to do so canbe severe. When a lack of data delays acluster inquiry or results in unnecessarilyvague or uncertain conclusions, theresult can be the intensification – ratherthan the easing – of a community’s fearand apprehension. In case after case,poorly conducted cluster inquiries haveresulted in divided communities, erodedproperty values, and increased levels ofstress, fear and frustration among resi-dents – results that can have their owneffects on public health and well-being.

In both Hazleton, Pa. and Toms River,N.J., the failure of public officials to shareinformation with residents at an earlystage of an investigation led to strainedrelations between public health officialsand the community. In the Toms Rivercase, public health officials responded byworking to include local citizens indesigning the public health response tothe cancer cluster. However, in theHazleton case, relations remain strainedto the point where the communities ofHazleton and Hazle Township commis-sioned their own health studies, even asthe Pennsylvania Department of Healthconducted its cluster inquiry.

In addition, the inability to quickly iden-tify clusters can lead to delays in theimplementation of important publichealth measures – for example, folic acideducational efforts in areas plagued by

high rates of neural tube defects or envi-ronmental cleanups where suspectedclusters led to the identification of conta-minated sites.

While even the best cluster inquiries willnot resolve every question in the mindsof community residents, the existence ofhealth tracking data can allow publichealth officials to quickly reassure resi-dents of communities where clusters donot exist and speed the delivery of infor-mation and assistance to all communitieswith public health concerns.

A Sixth Lesson: Good Registries AreNot Enough

While access to complete, up-to-dateinformation on disease incidence isessential to cluster investigations, theavailability of such data does not guaran-tee that such investigations will be suc-cessful or that public health concerns willbe addressed.

The Need for Tracking ofEnvironmental Exposures

Investigations of suspected clusters of dis-eases that are common or occur in asmall geographical area will not often besuccessful in identifying statistically sig-nificant levels of disease. Where environ-mental exposures can be plausibly linkedto these cases, researchers must employother methods – such as biological orenvironmental monitoring or epidemio-logical studies – to determine whetherenvironmental factors can be associatedwith the outbreak of disease.

43

44

CASE STUDY: CALCASIEU PARISH, LA. (CANCER)

INQUIRY SHOWS NEED FOR EXPOSUREINFORMATION

Mossville, in Calcasieu Parish, is a small, predom-inantly African-American residential enclave in theshadow of a giant Condea Vista Co. chemicalplant. Refineries and chemical plants have operat-ed in the area since the 1940s.

In the 1980s, Mossville residents began toexpress concern about possible links betweenemissions from the chemical plant and theirhealth. Ethylene dichloride, a suspected carcino-gen, was found to have migrated into the soilbeneath Mossville. Following a protracted legalbattle, Condea Vista and the plant’s formerowner, Conoco, agreed to settlements totalingapproximately $47 million, including a voluntarybuyout of some homeowners.

73

Around the time in 1997 that Conoco was settlingits lawsuit, lawyers gathering information for apotential class-action suit tested the blood of 11people around Calcasieu Parish. Three of the sam-ples came back with unusually high levels of diox-in – one of the most highly toxic families of sub-stances known to science. To confirm the results,the lawyers tested a pooled blood sample takenfrom a local hospital. That sample showed levelsof dioxin that were at the high end of the nation-al average and levels of one type of dioxin, TCDD,that were higher than average. The results of thetests were sent to Louisiana public health offi-cials, who were asked to conduct an investigation.The Louisiana Department of Health and Hospitalsdeclined.

74

Federal officials at the Agency for ToxicSubstances and Disease Registry (ATSDR) didtake up the investigation, however. In late 1998,they confirmed the earlier test results, and in late1999, ATSDR released the results of follow-upblood testing that found average levels of dioxinthree times the national average.

75

The levelswere among the highest ever reported in theUnited States for a non-occupational exposure.

Louisiana state officials then embarked on theirown study of cancer patterns in Calcasieu Parish.Based on data from the Louisiana Tumor Registryfrom 1988-1997, the study found that overallcancer rates in the parish were similar to those inother parts of Louisiana. But the study noted sig-nificantly elevated levels of lung cancer and softtissue cancers, both of which have been linked todioxin exposure.

76

While the Louisiana study hinted at the possibilitythat the elevated rates of some cancers could bethe result of dioxin exposure, it stopped wellshort of asserting causation, largely due to thedearth of information about the nature of theexposures. Federal officials are currently conduct-ing air monitoring to attempt to pinpoint thesource of any ongoing dioxin releases. They havealso launched another round of blood testing ofarea residents.

The existence of a tracking system to monitordioxin exposures could have tipped offresearchers to the problems in Calcasieu Parishyears before they were discovered by the commu-nity. Four years after that discovery, the commu-

45

nity and state and federal officials are still work-ing to ascertain how widespread the dioxin cont-amination problem is in the parish, whether expo-sure to the chemical is ongoing, and whether thereis any link between the chemical and health prob-lems among parish residents.

Biological monitoring of environmentalexposures provides the most accurateinformation on the level of exposures towhich residents of a community havebeen subjected. Yet the current availabili-ty of data on environmental exposures isextremely limited. A 2000 GeneralAccounting Office (GAO) study foundthat federal exposure surveys measure inthe general population only 6 percent ofthe more than 1,400 toxic chemicalsthought to pose potential health prob-lems. Even when exposure data are avail-able, public health officials lack a clearunderstanding of the levels of exposurethat would or should be expected in thegeneral population.

About 90 percent of public health offi-cials surveyed by the GAO reported thathuman exposure data from tissue sam-ples was extremely or very important foraddressing environmental health con-cerns. Yet almost two-thirds of those offi-cials reported that they could includesuch data in less than half of the studiesin which they deemed it important. Lessthan one-tenth always or almost alwayscould include such data. Key barriers tothe use of such data included a lack oflaboratory capacity and a lack of knowl-edge for how to set the results of suchtesting in context.

77

In contrast to the scarcity of biological

exposure data, there is a significantamount of data collected by state and fed-eral governments from environmentalmonitoring efforts. Most states regularlycollect, track, and compile at least someinformation on the levels of various cont-aminants in drinking water, the applica-tion of pesticides, the discharge of toxi-cants from industrial facilities, or the lev-els of various pollutants in the air. Severalstudies have connected these databaseswith information from disease registriesto achieve new insights into the environ-mental links to chronic disease. (See"Beyond Clusters," page 47.)

Better information on exposures andpotential exposures, better training onhow to use those data, and better link-ages between health outcome and envi-ronmental databases are all of vitalimportance in investigating the linksbetween environmental exposures andchronic disease.

The Need for Public Access,Involvement and Education

While the primary responsibility forinvestigating disease clusters remainswith public health officials, individualresearchers and citizens should have thetools to conduct their own assessments.Information collected through chronicdisease tracking systems should be madereadily available to researchers and thegeneral public – provided that effectiveprocesses are in place to preserve patientconfidentiality.

In addition, as noted above, decisions onthe geographic and temporal scope ofcluster investigations and the types of dis-ease covered can have a direct impact onthe results of a cluster investigation.

46

These decisions should be made with theparticipation of the affected communi-ties. CDC and state investigation proto-cols allow for the inclusion of citizen advi-sory boards in the decision-makingprocess for cluster investigations. Theseadvisory groups should be constituted asearly as possible in the process in order tobuild confidence with local residents andallow for the creation of studies that bestserve the community’s public healthneeds.

Finally, state disease registries should pro-vide detailed annual information on can-cer incidence in their states. Cancer-map-ping projects such as the one in New YorkState have the potential to broaden pub-lic understanding of cancer trends intheir communities.

The history of cluster investigationsshows that trust and channels of commu-nication between citizens and publichealth officials can easily break down.Where these breakdowns have occurred,the results have been damaging: reducedlegitimacy for public health agencies;wasted time and money on suspect clus-ter inquiries and environmental remedia-tion measures; anger, emotional stressand reduced property values for commu-nities.

History also shows that, generally, theearlier and more completely informationis shared with the public, the less likely itis that the crucial bonds of trust and com-munication will be broken. Involving thepublic in the design and conduct of clus-ter inquiries can be a useful tool forensuring that such investigations lead topositive outcomes for public health andunderstanding.

The Need to Prevent and CleanUp Environmental Health Threats

While the investigation of suspected dis-ease clusters is a matter of importance topublic health, it should never interferewith or delay the cleanup of known orpotential environmental threats. In manyof the case studies presented in thisreport, local citizens have been forced towage simultaneous battles to have theirhealth concerns taken seriously and toclean up long-standing environmentalproblems in their communities.

In such cases, environmental and publichealth officials must take a precautionaryapproach by moving quickly to clean upsources of suspected environmental cont-amination and prevent future contamina-tion – even in the absence of “proof” thatsuch contamination has caused a specificdisease.

In this regard, current proposals toreduce the number of Superfund haz-ardous waste cleanups conducted annual-ly and ease New Source Review require-ments for large industrial facilities underthe Clean Air Act must be viewed withcaution. Evidence from many studies sug-gests that exposure to toxic chemicals, airpollutants and drinking water contami-nants play a role in the development ofserious, chronic disease. Research intothese links must continue, but it must bematched with aggressive action by thepublic health community and govern-ment to prevent pollution and clean upalready contaminated communities.

As noted above, the investigation of local-ized disease clusters rarely results, in andof itself, in the discovery of a direct linkbetween an environmental pollutant andthe development of disease. Among thetypical limitations of community-basedcluster inquiries are small sample sizeand the presence of other, confoundingfactors that could be responsible for theincreased incidence of disease.

Epidemiological studies that includemultiple communities with similar envi-ronmental exposures can reduce theimpact of these limitations and allowresearchers to develop more informedhypotheses about the potential environ-mental links to chronic disease.

Disease registry data and information onenvironmental exposures are critical tosuch studies. Detailed and reliable dataabout who suffers from a disease, whencombined with data on biological expo-sures or environmental conditions, canenable researchers to probe the connec-tions between the environment and pub-lic health with greater precision than canoften be obtained through inquiries intothe cause of localized clusters.

Research activities in several states showthe promise of this kind of research fordocumenting the environmental links tochronic disease.

The California Birth DefectsMonitoring Program

Since its founding in 1982, the CaliforniaBirth Defects Monitoring Program hasplayed an important role in collectinginformation on birth defects in the stateand in supporting research into the caus-es of birth defects. In several cases, caseinformation drawn from the monitoringprogram has been used – in concert withinterviews or environmental monitoringdata – to probe the environmental linksto birth defects.

• PROXIMITY TO HAZARDOUS WASTE SITES – A 1997 study based on inter-views with more than 2,000 women found that women who lived within a quarter-mile of a Superfund site dur-ing the first three months of pregnan-cy had a greater risk of having babies with heart and neural tube defects. Even with the large sample size, however, the study was limited by the small number of women who live in such close proximity to hazardous waste sites, reducing the statistical signifi-cance of the finding. The researchers recommended further studies based on measurement of actual environ-mental exposures.

78

• EXPOSURE TO PESTICIDES – A 1999 study, again based on interviews with more than 2,000 women, found that women living within a quarter-mile of agricultural crops and those who engaged in household gardening

47

Beyond Clusters: InvestigatingEnvironmental Links to

Chronic Disease

48

were more likely to give birth to babies with certain birth defects. Interestingly, no association was found between birth defects and self-application of pesticides inside the home or occupational exposure to pesticides. The researchers suggested further study based on actual expo-sure information.

79

• AIR POLLUTION – A 2002 study eval-uated rates of certain birth defects along with air pollution levels detect-ed by air monitors in Southern California in an attempt to determine whether exposure to air pollutants at key points during pregnancy was asso-ciated with birth defects. The study found links to various defects: expo-sure to carbon monoxide during the second month of pregnancy was asso-ciated with the development of ven-tricular septal defects and second-month exposure to ozone was associ-ated with aortic artery and valve defects, pulmonary artery and valve anomalies, and conotruncal defects. No similar effects were found for other months during pregnancy or other air pollutants studied. The study was the first of its kind in the United States and researchers cautioned that the findings need to be confirmed by other studies.

80

The existence of complete, timely anddetailed birth defects registry data wasintegral to all three studies, demonstrat-ing the power such data can have forinvestigating the environmental links tochronic disease.

New Jersey Drinking Water Studies

Between 1979 and 1984, New Jersey cre-ated registries for both cancer and birthdefects and began monitoring for thepresence of 14 volatile organic com-pounds (VOCs) in public communitywater systems. In the 1990s, researcherscombined the health tracking databaseswith information on drinking water cont-amination to explore the links betweendrinking water and health.

In the late 1980s, researchers with thestate Department of Health compareddata on leukemia incidence taken fromthe state’s cancer registry with water test-ing results in a part of the state with abroad range of contamination. The studyshowed a significant association betweenconcentrations of trichloroethylene(TCE) and perchloroethylene (PCE) andthe overall leukemia rate among femalesin 27 towns.

In 1993, the state expanded its investiga-tion to 75 towns with a total population of1.5 million and also included non-Hodgkin’s lymphomas as well asleukemias. The study also found an asso-ciation between TCE and PCE in drink-ing water and certain kinds of leukemiasand non-Hodgkin’s lymphomas.

81

Three years later, researchers examinedthe potential links between a variety ofdrinking water contaminants and adversebirth outcomes, including birth defects.The study – like the earlier studies of can-cer – was based on a combination ofwater monitoring data and informationfrom the state’s birth defects registry.The researchers found associationsbetween six types of drinking water cont-

aminants and defects, low birth weights,and small-for-gestational-age births.

82

The New Jersey studies, while not thefinal word on the link between drinkingwater contamination and cancer andbirth defects, are part of a growing bodyof literature expressing concern aboutexposure to certain VOCs. The studiesalso demonstrate the potential benefitsof linking environmental monitoringinformation with disease registry data.

Other States

Several other states have embarked onefforts to link existing information onenvironmental conditions to disease reg-istries. Since 1987, for instance, Iowa hasbeen regularly compiling information ondrinking water contamination and inves-tigating the potential health threats itmay pose. In 2001, University of Iowaresearchers compared data on contami-nants in public drinking water supplieswith information on cancer incidencefrom the state’s cancer registry to explorethe links between nitrate levels and can-cer. Their research showed an associationbetween long-term exposure to low levelsof nitrates in water and the developmentof bladder cancer in women. Researcherssuggested that the EPA’s standard fornitrate in drinking water may not be ade-quately protective of human health butcautioned that more follow-up research isnecessary to confirm the link.

83

The studies noted above demonstrate thepotential of epidemiologic studies thatspan wider geographic areas in attempt-ing to ascertain the links between envi-ronmental exposures and chronic dis-ease. Clearly, more such research is need-ed to fill in the gaps in medical knowl-

edge about the environmental causes ofcancer, birth defects and other condi-tions with suspected environmental links.However, to make that research possible,states must begin to compile accurateinformation on disease incidence, assem-ble good information on human expo-sure to environmental pollutants, andallow for the linkage of registry data toexisting information on environmentalconditions.

49

50

Create a National Health Tracking Network

Regardless of what state they call home,citizens deserve swift and thorough eval-uation of suspected environmentalhealth threats in their communities.Researchers conducting those evalua-tions need timely, complete and accuratedata to help make their work effectiveand affordable. And all of us deserve apublic health system that works aggres-sively to probe the potential linksbetween environmental exposures andchronic disease and protect us fromexposures that may harm our health.

Creating such a public health systemwould require the commitment of newresources to tracking the incidence ofchronic disease in our communities andour exposure to harmful substances inour environment. Public health organiza-tions across the country have ralliedbehind proposals to create just such anationwide health tracking network.

Such a network would include:

• Systems in all 50 states to track chronic diseases including: asthma and chronic respiratory diseases birth defects, developmental and other neurological disorders, cancers, neu-rological diseases such as Alzheimer’s and Parkinson’s, and other chronic diseases.

• Systems in all 50 states to track human exposures to environmentalhazards, beginning with such priority substances as PCBs, dioxin, heavy met-

als such as mercury and lead, pesti-cide, and water and air contami-nants.

• An early warning system to alert communities to immediate health crises such as heavy metal and pesti-cide poisonings. This system would be similar to current systems to alert communities to the outbreak of infec-tious diseases such as West Nile Virus.

• Up to 20 pilot programs to investi-gate clusters of disease and local health priorities not covered by the network. These programs could serve as models for later inclusion in the network and would allow states the capacity to track problems of particu-lar concern. For example, Massachusetts could track autoim-mune disorders such as scleroderma and lupus; lessons learned there could lead to nationwide tracking for these conditions.

• The creation of rapid response teams of federal, state and local offi-cials to investigate clusters, outbreaks and emerging threats. The teams would be well-trained and have access to high-quality equipment and lab facilities that, in many cases, do not currently exist.

• The involvement of academic cen-ters and local communities in envi-ronmental health research through the sharing of data and perspectives.

Creation of such a nationwide healthtracking network would resolve several

Policy Recommendations

problems that typically hamper diseasecluster investigations. First, the networkwould expand tracking for cancer, birthdefects, and other diseases into jurisdic-tions where existing registries are eithernonexistent, incomplete, or of insuffi-cient quality. The availability of such datawould allow local and state public healthofficials to make determinations as to thevalidity of reports of disease clustersquickly and with little expenditure ofadditional public health resources. Thelikely result would be greater efficiency inthe handling of cluster investigationrequests and the provision of better infor-mation to the public.

Second, by creating nationwide trackingof chronic diseases – and by amassingdata on human exposures to toxic sub-stances in the environment – the networkwould provide researchers with impor-tant information with which to explorethe environmental links to chronic dis-ease. The availability of such data wouldallow researchers to compare rates of dis-ease in communities with similar envi-ronmental exposures, conduct multi-community studies, and conduct moreauthoritative cluster studies in small com-munities where statistical methods aloneare unlikely to be effective.

Third, by increasing the amount ofresources devoted to investigating envi-ronmental links to chronic health prob-lems, the network would reduce the pres-sure on state and local officials to mini-mize cluster inquiries due to staff andresource limitations. The result wouldlikely be fewer delays in the investigationof potentially significant disease clusters.

51

52

Appendix A: Status Of U.S. Chronic Disease Registries

CANCER (a,b) BIRTH DEFECTS (c,d) ASTHMA (e)

First Year of Population-Based Data Collection NAACCR First Year of TFAH (1) Certification Available Data Statewide? Grade Tracking?

AK DNR Gold 1996 Yes B NoAL DNR None 1998 No B NoAR DNR None 1980 Yes A YesAZ 1995 None 1986 Yes B YesCA 1988 Gold 1983 No (3) A YesCO 1988 Gold 1989 Yes B NoCT 1935 Gold 1993 Yes C YesDC DNR Gold None F YesDE 1972 Gold Collecting data Yes C NoFL 1981 Silver 1996 Yes B YesGA DNR Gold (2) 1968 No A NoHI 1960 Silver 1986 Yes A YesIA 1973 Gold 1983 Yes A YesID 1970 Gold None F NoIL 1986 Gold 1988 Yes B YesIN 1987 None None D NoKS DNR Silver 1985 Yes F NoKY 1991 Gold 1997 Yes B NoLA 1988 Gold 2003 (est.) Yes D NoMA 1982 Gold 1998 Yes A YesMD 1982 Gold 1984 Yes C NoME 1983 None No data yet Yes D NoMI 1985 Gold 1992 Yes C YesMN 1988 Gold 2002 Yes D NoMO DNR Silver 1980 Yes B YesMS DNR None 2001 Yes D NoMT 1979 Silver No data yet Yes D NoNC 1990 Silver 1989 Yes C YesND DNR Gold No data yet Yes F YesNE 1987 Gold 1973 Yes B Yes

53

NH 1987 Silver No data yet Yes D NoNJ 1979 Gold 1985 Yes B NoNM 1973 Silver 1995 Yes B NoNV 1979 None Collecting data No C NoNY 1976 Gold 1997 Yes B YesOH DNR Silver None F NoOK DNR None 1992 Yes A YesOR DNR Silver None F YesPA 1985 Gold Collecting data Yes D NoRI 1986 Gold 1999 Yes D YesSC DNR Gold 1993 Yes B YesSD DNR None None F NoTN 1989 None Collecting data No C NoTX 1992 None 1995 (4) Yes A NoUT 1966 Gold 1994 Yes C YesVA 1990 None 1989 Yes B NoVT DNR None None F YesWA 1992 Gold 1987 Yes D YesWI 1978 Gold 2003 Yes D YesWV 1993 Gold 1989 Yes C NoWY 1962 Gold None F No

DNR= Did not respond.

(1) In many locations, hospital-based tumor registries had collected limited information on cancer prior to this date.(2) Gold certification is for metropolitan Atlanta registry.(3) California’s registry, while not statewide, is designed to be representative of births in the state as a whole.(4) Texas had collected statewide data for a time but has since discontinued collection in some areas due to budget cuts.

SOURCES:(a) First year of population-based registry from Maria Hewett and Joseph V. Simone, eds.,Enhancing Data Systems to Improve the Quality of Cancer Care, National Academy Press, 2000.(b) Certification data from North American Association of Central Cancer Registries,“Cancer Registries Certified for High Quality 1999 Incidence Data,” downloaded fromhttp://www.naaccr.org/Certification/1999Certification.html, 24 April 2002.(c) List of states and first year of data collection from Centers for Disease Control andPrevention, National Center on Birth Defects and Developmental Disabilities, “State BirthDefects Surveillance Programs Directory,” Teratology, Vol. 64, S47-S116, 2001.(d) Grade from Trust for America’s Health, Birth Defects Tracking and Prevention: Too ManyStates Not Making the Grade, 2002.(e) Trust for America’s Health, Short of Breath: Our Lack of Response to the Growing AsthmaEpidemic and the Need for Nationwide Tracking, July 2001. Asthma tracking data is based oninformation reported by the states to the CDC.

54

Appendix B: State Responses To Cancer Clusters

CONDUCTINVESTIGATIONS?

HAVE STANDINGRESPONSE PROTOCOL?

NUMBER OF STAFF AVAILABLE FOR CLUSTERS

ESTIMATED SATISFACTION OF COMPLAINTS(1=SATISFIED,7=UNSATISFIED)

ESTIMATED % OF COMPLAINTS SATISFIED AT FIRST CONTACT (VIA PHONE)

INVESTIGATIONREQUESTS IN 1997

AK Y N 0.25 2 75% 3AL Y Y 0.5 4 50% 6AR Y N 0.25 4 50% 2AZ Y Y 0.5 5 70% 12CA Y Y 1 4 DNR 300CO Y Y 0.25 2 95% 9CT Y Y 0.5 4 75% 20DC Y N 0.25 4 50% 1DE Y Y 0.25 2 75% 2FL Y Y 0.75 2 10% 6GA N N NA NA NA NAHI Y N 1 6 0% DNRIA Y Y 0.25 2 20% 3ID Y Y 0.25 3 75% 11IL Y Y 0.25 3 50% 64IN Y N 0.25 4 50% 10KS Y Y 0.25 5 80% 5KY Y N 1 DNR DNR 2LA Y N 1 4 50% 18MA Y Y >2 5 40% 59MD Y N 0.25 4 55% 9ME Y Y 0.25 2 99% 4MI Y Y 0.75 2 75% 24MN Y Y 1 DNR 95% 29MO Y Y 1.5 DNR 60% 43MS Y N 0.25 2 90% 7MT Y Y 0.25 3 80% 3NC Y Y 0.25 2 95% 30ND Y Y 0.25 5 85% 2

55

NA= Not ApplicableDNR = Did Not Respond

SOURCE: C.W. Trumbo, “Public Requests for Cancer Cluster Investigations: A Survey ofState Health Departments,” American Journal of Public Health, Vol. 90, No. 8, 1300-1303.

NE Y Y 0.25 2 90% 2NH Y Y 1.5 4 75% 11NJ Y Y 1.5 5 75% 71NM Y Y 0.25 3 50% 5NV N N NA NA NA 0NY Y Y >2 2 90% 91OK Y N 0.25 4 90% DNROR Y N 0.25 2 80% 20PA Y Y 0.5 4 10% 25RI Y Y 0.25 6 95% 12SC Y Y 0.25 DNR DNR 10SD N N NA NA NA NATN Y Y 0.5 3 80% DNRTX Y Y 0.5 4 20% 43UT Y Y 0.5 4 50% 5VA Y Y 0.25 3 65% 9VT Y N 0.25 DNR DNR DNRWA Y Y 0.25 2 90% 3WI Y Y 1 DNR DNR 29WV Y N 0.25 5 0% 2WY Y Y 1 DNR DNR 12

56

Appendix C: Glossary Of Abbreviations

ALL: Acute lymphocytic leukemia

ATSDR: U.S. Agency for Toxic Substances and Disease Registry

CDC: Centers for Disease Control and Prevention

DES: Diethylstilbestrol

DHHS: Department of Health and Human Services (New Hampshire)

EPA: Environmental Protection Agency

GAO: U.S. General Accounting Office

MCR: Maine Cancer Registry

MS: Multiple sclerosis

NAACCR: North American Association of Central Cancer Registries

NPCR: National Program of Cancer Registries

NTD: Neural tube defect

ODH: Ohio Department of Health

PCBS: Polychlorinated biphenyls

PCE: Perchloroethylene

SCCCR: South Carolina Central Cancer Registry

SEER: Surveillance, Epidemiology and End Results program

TBDMD: Texas Birth Defects Monitoring Division

TCDD: 2,3,7,8-tetrachlorodibenzo-p-dioxin

TCE: Trichloroethylene

TDH: Tennessee Department of Health

TDH: Texas Department of Health

VOC: Volatile organic compound

57

Notes

1

The Pew Charitable Trusts, Public Opinion Research on Public Health, Environmental Health, and the Country’s Public Health

Capacity to Adequately Address Environmental Health Problems, May 1999. Survey conducted March 1999.2

Ibid.3

“Environment More Important than Heredity to Cancer Risk, Study Suggests,” CNN.com, 13 July 2000.4

National Academy of Sciences, Scientific Frontiers in Developmental Toxicology and Risk Assessment, June 2000, 1.5

Centers for Disease Control and Prevention, “Cancer Clusters,” downloaded from

http://www.cdc.gov/cancer/npcr/clusters.htm, 29 April 2002.6

C.W. Trumbo, “Public Requests for Cancer Cluster Investigations: A Survey of State Health Departments,” American

Journal of Public Health, Vol. 90, No. 8, 1300-1303.7

Ibid.8

Ibid.9

National Cancer Institute, “Cancer Facts: Cancer Clusters,” downloaded from http://cis.nci.nih.gov/fact/3_58.htm,

15 May 2002.10

Centers for Disease Control, Division of Bacterial and Mycotic Diseases, “Legionellosis,” downloaded from

http://www.cdc.gov/ncidod/dbmd/diseaseinfo/legionellosis_g.htm, 9 May 2002; National Institute of

Environmental Health Sciences, “Legionnaire’s Disease,” downloaded from

http://www.niehs.nih.gov/external/faq/legion.htm, 9 May 200211

North American Association of Central Cancer Registries GIS Work Group, Using Geographic Information Systems

Technology in the Collection, Analysis and Presentation of Cancer Registry Data: Introduction to Basic Practices, draft, April 2002.12

Trent Stephens, “Reinventing Thalidomide,” Chembytes E-zine, Royal Society of Chemistry (Great Britain), November

2001.13

Darrell E. Ward, “Cancer Clusters,” Frontiers, Ohio State University, Spring-Summer 1999.14

Centers for Disease Control, “Epidemiologic Notes and Reports Angiosarcoma of the Liver Among Polyvinyl

Chloride Workers – Kentucky,” MMWR Weekly, 7 February 1997 / 46(05);97-101.15

Darrell E. Ward, “Cancer Clusters,” Frontiers, Ohio State University, Spring-Summer 1999.16

Roberta D. Calhoun, “Q&A: Cancer Clusters,” The Brain Tumor Society, downloaded from

http://www.tbts.org/cancerclusters.htm, 17 May 2002.17

New Jersey Department of Health and Senior Services, Summary: Childhood Cancer Incidence Health Consultation, A

Review of Cancer Registry Data, 1979-1995 for Dover Township (Ocean County) New Jersey, September 1997.18

New Jersey Department of Health and Senior Services, Case Control Study of Childhood Cancers in Dover Township

(Ocean County) New Jersey, Public comment draft, December 2001.19

New Jersey Department of Health and Senior Services, Dover Township Childhood Cancer Investigation: Progress Report

#3, March 2001.20

Margot LeSage, “State Ending Probe, Not Me,” Eagle-Tribune, 13 February 2002.21

Melinda Carpenter, New Hampshire Assistant State Epidemiologist, personal communication, 10 June 2002.

58

22

Margot LeSage, “State Ending Probe, Not Me,” Eagle-Tribune, 13 February 2002.23

Margot LeSage, “Radon Found in Atkinson’s Public Water,” Eagle-Tribune, 1 March 2002.24

Dennis Pinski, New Hampshire Bureau of Health Risk Assessment, personal communication, 7 June 2002.25

Memorandum from Mark Canfield, Director, Texas Birth Defects Monitoring Division, “Subject: Recent High Rates

of Anencephaly in Laredo,” 15 May 2001.26

Peter Langlois, Senior Epidemiologist, Texas Birth Defects Monitoring Division, personal communication, 12 June

2002.27

Ibid.28

Centers for Disease Control and Prevention, “Guidelines for Investigating Clusters of Health Events,” Morbidity and

Mortality Weekly Report, 27 June 1990.29

C.W. Trumbo, “Public Requests for Cancer Cluster Investigations: A Survey of State Health Departments,” American

Journal of Public Health, Vol. 90, No. 8, 1300-1303.30

Tim E. Aldrich, Susan W. Bolick and Laura A. Cave, A Report of Cancer in Zip Code 29405, South Carolina

Department of Health and Environmental Control, 13 May 1999.31

Laura C. Sanders, Special Projects Manager, South Carolina Central Cancer Registry, personal communication, 7

May 2002.32

Michael Lasalandra, “Southie Study of Lupus, Scleroderma Enters Final Phase,” Boston Herald, 18 April 200233

Molly Jacobs, Project Manager, South Boston Scleroderma and Lupus Health Study, Massachusetts Department of

Public Health, Bureau of Environmental Health Assessment, personal communication, 14 May 2002.34

Michael Lasalandra, “Southie Study of Lupus, Scleroderma Enters Final Phase,” Boston Herald, 18 April 2002.35

National Cancer Institute, “About SEER,” downloaded from http://seer.cancer.gov/about/, 15 May 2002.36

Judy Henry, et al, El Paso Multiple Sclerosis Cluster Investigation: Public Comment Draft, Final Report, Texas Department

of Health, U.S. Agency for Toxic Substances and Disease Registry, August 2001.37

New York State Department of Health, “State Health Department Announces Cancer Mapping ‘Follow-Up

Investigation’ in Suffolk County,” press release, 20 May 2002.38

New York State Department of Health, “Cancer Surveillance Improvement Initiative: Unusual Disease Patterns

Investigation,” downloaded from http://www.health.state.ny.us/nysdoh/cancer/sublevel/unusual.htm. 39

Margaret A. Parsons, Judith Graber, Nancy Sonnenfeld, Case Series Investigation of Reportable Brain Tumors Among Young

Adults in Fairfield, Maine, Summary of Preliminary Findings, Maine Bureau of Health, 22 January 2001.40

Doug Harlow, “Expert Ties Chemicals to Illnesses,” Central Maine Morning Sentinel, 31 August 2001.41

Margaret A. Parsons, Maine Cancer Registry, personal communication, 14 May 2002.42

Paul Jayko v. Ohio Environmental Protection Agency, U.S. Department of Labor Office of Administrative Law Judges,

Case 1999-CAA-5.43

Randall Edwards and Jill Riepenhoff, “Health Questions Linger,” Columbus Dispatch, 4 October 1998.44

Ohio Department of Health, Community Health Assessments Section, Case Review of Leukemia Among Residents of

Marion County, Ohio, 1992-1999 and Graduates of River Valley High School, 1963-2000, July 2001.45

Hal Karp, “A Town Divided: How a Cancer Threat is Tearing a Community Apart,” Family Circle, August 2001.46

Ohio Department of Health, Community Health Assessments Section, Case Review of Leukemia Among Residents of

59

Marion County, Ohio, 1992-1999 and Graduates of River Valley High School, 1963-2000, July 2001.47

Bruce Molholt, “RE: Case Review of Leukemia Among Residents of Marion County, Ohio, 1992-1999 and Graduates

of River Valley High School, 1963-2000,” memo to J.R. Kolmer, 17 September 2001, downloaded from

http://www.envirodocsrivervalley.org/frpg29, 16 May 2002.48

Chronic Disease Directors, “National Program of Cancer Registries Appropriations Fact Sheet,” February 2002;

Centers for Disease Control and Prevention, “National Program of Cancer Registries: About the Program,” down-

loaded from http://www.cdc.gov/cancer/npcr/register.htm, 11 June 2002.49

Maria Hewett and Joseph V. Simone, eds., Enhancing Data Systems to Improve the Quality of Cancer Care, National

Academy Press, 2000.50

North American Association of Central Cancer Registries, “Cancer Registries Certified for High Quality 1999

Incidence Data,” downloaded from http://www.naaccr.org/Certification/1999Certification.html, 24 April 2002.51

Nevada Health Division, “Churchill County (Fallon) Childhood Leukemia Update,” downloaded from

http://health2k.state.nv.us/healthofficer/leukemia/fallon.htm, 17 May 2002; Associated Press, “Sixteenth Child

Added to Cancer Cluster in Northern Nevada Town,” 29 July 2002.52

Associated Press, “Rep. Gibbons Seeks Funds for Fallon Leukemia Cluster Study,” 17 April 2002.53

Associated Press, “Doctors, Families Criticize Nevada Cancer Registry,” 30 December 2001.54

Trust for America’s Health, Birth Defects Tracking and Prevention: Too Many States Not Making the Grade, 2002.55

Ibid.56

Dana C. Crawford, “Memorandum: Epi-Aid Trip Report: Possible Cluster of Orofacial Clefts,” Centers for Disease

Control and Prevention, 24 January 2000 (sic).57

Bonnie Bashor, Tennessee Department of Health, personal communication, 17 May 2002.58

Pew Environmental Health Commission, Attack Asthma: Why America Needs a Public Health Defense System to Battle

Environmental Threats, May 2000.59

Trust for America’s Health, Short of Breath: Our Lack of Response to the Growing Asthma Epidemic and the Need for

Nationwide Tracking, July 2001.60

Pew Environmental Health Commission, America’s Environmental Health Gap: Why the Country Needs a Nationwide

Health Tracking Network, September 2000.61

Based on NAACCR gold registration for cancer registry, Trust for America’s Health “A” grade for birth defects

tracking and possession of an asthma registry as reported to CDC, per Pew Environmental Health Commission.62

C.W. Trumbo, “Public Requests for Cancer Cluster Investigations: A Survey of State Health Departments,” American

Journal of Public Health, Vol. 90, No. 8, 1300-1303.63

U.S. Agency for Toxic Substances and Disease Registry, Public Health Assessment: U.S. Defense General Supply Center,

Richmond, Chesterfield County, Virginia, 21 April 1993.64

Meredith Fischer and Paige Akin, “Contamination Spreading: Chemical Found South of Site,” Richmond Times-

Dispatch, 7 March 2002.65

Meredith Fischer and Paige Akin, “Leaks Lead to Indictments; Engineering Firm is Charged in DSCR Creek

Contamination,” Richmond Times-Dispatch, 10 April 2002.

60

66

Chesterfield County Health Department, Evaluation of Cancer Incidence in Rayon Park, 14 June 2002; Dr. William

Nelson, Chesterfield Health District Director, personal communication, 22 May 2002.67

C.W. Trumbo, “Public Requests for Cancer Cluster Investigations: A Survey of State Health Departments,” American

Journal of Public Health, Vol. 90, No. 8, 1300-1303.68

Michael Greenberg, Daniel Wartenberg, “Communicating to an Alarmed Community About Cancer Clusters: A

Fifty State Survey,” Journal of Community Health, Vol. 16, No. 2, April 1991.69

Michael Greenberg, Daniel Wartenberg, ”Communicating to an Alarmed Community About Cancer Clusters: A

Fifty State Survey,” Journal of Community Health, Vol. 16, No. 2, April 1991; C.W. Trumbo, “Public Requests for Cancer

Cluster Investigations: A Survey of State Health Departments,” American Journal of Public Health, Vol. 90, No. 8, 1300-

1303.70

U.S. Environmental Protection Agency, Office of Inspector General, EPA Region III’s Management of Tranguch

Gasoline Site, Hazleton, Pennsylvania, 29 August 2001.71

Evelyn O. Talbott et al, Hazle Township Health Effects Study 1990-2000, Preliminary Findings, University of Pittsburgh

Graduate School of Public Health, 29 August 2001.72

Pennsylvania Department of Health, Tranguch Gasoline Spill Report, Hazleton, Pennsylvania, December 2001.73

Joe Mathews, “Paying Neighbors to Move,” Baltimore Sun, 6 December 1998.74

John McQuaid, “A Health Risk,” New Orleans Times-Picayune, 23 May 2000.75

U.S. Agency for Toxic Substances and Disease Registry, Health Consultation: Exposure Investigation Report: Calcasieu

Estuary, Lake Charles, Calcasieu Parish, Louisiana, 19 November 1999.76

Kimberly Gallo, Jessica King, Joseph Sejud, Cancer in Calcasieu Parish, Louisiana: 1988-1997, Louisiana Department

of Health and Hospitals, January 2002.77

U.S. General Accounting Office, Toxic Chemicals: Long-Term Coordinated Strategy Needed to Measure Exposure in Humans,

May 2000.78

California Birth Defects Monitoring Program, Birth Defects & Hazardous Waste Sites, April 1999.79

California Birth Defects Monitoring Program, Pesticides & Birth Defects, April 1999.80

Boate Ritz, Fei Yu, Scott Fruin, Guadalupe Chapa, Gary M. Shaw and John A. Harris, “Ambient Air Pollution and

Risk of Birth Defects in Southern California,” American Journal of Epidemiology, vol. 155, no. 1 (2002), 17.81

Perry Cohn, Frank Bove, Judith Klotz, Marian Berkowitz, Jerald Fagliano, Drinking Water Contamination and the

Incidence of Leukemia and Non-Hodgkin’s Lymphoma, New Jersey Department of Health, May 1993, Executive Summary.82

F.J. Bove, M.C. Fulcomer, J.B. Klotz, J. Esmart, E.M. Dufficy, J.E. Savrin, “Public Drinking Water Contamination and

Birth Outcomes [Abstract],” American Journal of Epidemiology (1995), Vol. 141: 850-862. 83

University of Iowa Health Care, “Information Consistent with Previous Data,” week of 30 April 2001, downloaded

from http://www.uihealthcare.com/news/news/2001/04/30drinkingwater.html, 12 July 2002.