From the ExpertsJuly 31, 2015 | |
Texas Health Department Report on Cancer Occurrence in East Harris County: An Expert Analysis
Residents near the San Jacinto River Waste Pits (“SJRWP”) Superfund site expressed concern to the Texas Department of State Health Services (“DSHS”) about a possible excess of cancer cases occurring in the flood plain around the San Jacinto River, which includes the SJRWP site. That prompted the DSHS’s Environmental and Injury Epidemiology and Toxicology Unit (“EIET”) and the Texas Cancer Registry (“TCR”) to examine the occurrence of cancer in east Harris County, Texas. Now, the DSHS has issued a report.
On June 19, the Houston Chronicle published a story with the headline, “Texas health officials will further study Harris County cancer clusters,” and the sub-headline, “Disease clusters found near San Jacinto River.” Such headlines are not uncommon. “Cancer clusters” receive a lot of attention – from the residents in the community in which a supposed cluster arises; from local, state, and federal health agencies; and from the media.
According to the Centers for Disease Control and Prevention (“CDC”), over 1,000 suspected cancer clusters are reported to state health departments each year. Data from the CDC also reveals that 75 percent of those suspected cancer clusters are clearly not clusters and are resolved without further intervention. In about five to 15 percent of the reported clusters, “formal” analyses show that the number of observed cases exceeds the number of expected cases in the affected population. But, as noted by the CDC, even in these instances it is rare that an investigation is definitive and chance is usually a plausible explanation for the finding.
What exactly is a cancer cluster? How does a health agency determine if a cluster exists? What are the possible outcomes of a cancer cluster investigation? And how does one interpret the investigation’s findings?
What Is a Cancer Cluster?
The CDC has outlined the criteria for a group of reported cancers to be defined as a “cluster.” Until all the criteria are met, a group of cancer cases often is referred to as a “suspected” cancer cluster. In order to be a cancer cluster:
- Cancer cases: All of the cases must involve the same type of cancer, or types of cancer scientifically proven to have the same cause.
- Population in which the cases occur: The population must be carefully defined by factors such as age and gender for purposes of calculating cancer rates.
- Geographic area: Both the number of cancer cases included in the “cluster” and calculation of the expected number of cases must be carefully defined. It is possible to “create” or “obscure” a cluster by selection or omission of a specific area.
- Time period: The number of cases included in a cluster – and calculation of the expected number of cases – depends on how the time period over which the cases occurred is defined.
- A greater than expected number of cancer cases: A greater than expected number is when the observed number of cases is higher than one would typically observe in a similar setting and in a group with a similar population. This may involve comparison with rates for comparable groups of people over a much larger geographic area – e.g., an entire state.
How Are Suspected “Cancer Clusters” Investigated?
Guidelines for suspected cancer clusters were released in 2013 by the CDC and the Council of State and Territorial Epidemiologists (“Investigating Suspected Cancer Clusters and Responding to Community Concerns”). The guidelines suggest a four step process of investigation:
- An initial contact/response to gather “source” information, including but not limited to the alleged nature of the cluster and the suspected origin of the exposure that is believed to have caused the cluster;
- An assessment of the occurrence of the health event to determine if an excess of cancer cases in fact exist;
- A feasibility study to determine if an epidemiological study can be conducted; and
- An epidemiologic investigation to assess the association between the cancer(s) and environmental causes.
What Are the Possible Outcomes of a Cluster Investigation?
- It may not be a true cluster. The analysis shows that the number of cases may not really be higher than expected when other factors that could explain the increase (such as age, race, and gender) are taken into account.
- It is a cancer cluster. The cause of the cancer or cancers is identified and steps can be initiated to address the issue. In a possible residential cluster, this rarely occurs
- A cancer cluster may be found, but the cause cannot be identified.
How Does One Interpret the Investigation’s Findings?
The last outcome above is the one that occurs most frequently but even then, in absolute numbers, those instances are rare. The possible reasons for not observing a “cause” are also those aspects of the investigation that should be critically examined in critiquing any cancer cluster investigation. Some of those issues include:
- Random patterns: Random patterns can form a cluster and there will be more cases than expected in some geographic areas and fewer cases than expected in others. The areas with more cases than expected are more likely to be noticed. But many of these will be due to the “bull’s-eye effect.” If one took a map of the U.S. and started drawing circles of different sizes in different locations, some of the circles would contain more cancer cases than expected, while other would have less. Some of the circles with more cases might be clusters caused by a single carcinogen, but most would be due to chance. Even if the number of cancer cases in an area is higher than the expected number, the reported increase may not necessarily be related to a single factor or exposure. Random patterns are the most common reason for a cancer cluster with no identified cause.
- Case definition: Although a cancer cluster may involve a very rare and histologically unique cancer, most suspected cancer clusters involve cancer of different organs that are not known to have a common etiology. Attempting to identify a common underlying cause in such an instance likely will produce misleading results. Even cancers of the same organ systems should not be viewed as a “single group of cancers” because they may be of different molecular types and have different etiological mechanisms.
- Latency: There is also the question of latency – the interval between exposure and the diagnosis of cancer that can be, depending on the cancer, anywhere from a few years to many decades.
- Exposures: Given the latency period, it is difficult to ascertain which of a myriad of exposures might be the “cause.” If there is a potential exposure of interest, investigators must determine if, when, and how this group of people was exposed. Was it a single event or has it been ongoing? Was it five years ago? Ten years? Twenty? And what did the people with cancer have in common during that time? What alternative cause or causes (i.e.,other factors for cancer induction) might this group have in common? Determining to what substance or substances the group may have been exposed, in what manner, and for how long is a long and involved process.
- Mobility: The mobility of the group (both of the cases and those in the comparison group) needs to be taken into account. Some people will have moved into the community, while others may have moved away. Should the cases of people who moved into town in recent years be included? What about the people who moved away?
- The geographic cluster area: Defining the geographic cluster area is not necessarily an easy process. We go back to the “bull’s eye” effect – how large or small should the circle be? Should the cluster be investigated in one area of the community (the local neighborhood); the entire community; one or more census tracts; or the county or the state? Areas outside the perceived “cluster area” may in fact have cases that may or may not be related to those in the area of interest.
No matter how thorough and well conducted an investigation may be, questions may still remain.
The San Jacinto Cancer Cluster Investigation
As noted, the study referenced in the newspaper headline at the beginning of this article was begun in response to citizen concerns about a possible excess of cancers occurring in the flood plain around the San Jacinto River, which included the SJRWP Superfund site. One of the primary materials of concern was dioxin. The DSHS undertook an investigation of the occurrence of cancer in East Harris County, Texas. The DSHS investigation was the second step (assess the occurrence of the health event to determine if an excess of cancer cases exist) as described in the CDC and Council of State and Territorial Epidemiologists guidelines mentioned above.
The report, released in June 2015, concluded, “This assessment identified a number of statistically significant results that warrant further discussion.” The report specifically noted the elevated (greater than expected) standardized incidence ratios (“SIRs”) for childhood glioma in one census tract; childhood melanoma in one census tract; and childhood retinoblastoma in two census tracts.
(An SIR is the number of observed cases compared to (divided by) the number of expected cases for each cancer type. An SIR greater than 1.00 indicates that the observed number of cases of a specific cancer type is higher than expected and an SIR less than 1.00 indicates that the observed number of cases of a specific cancer type is lower than expected. Expected numbers of cancers are obtained by multiplying the number of person-years (people at risk times time of observation) times the cancer rates for a given group of people.)
The assessment also found more cases than expected of childhood lymphoma and melanoma, more cases than expected of brain and cervical cancer, and fewer cases than expected of thyroid cancer in the area as a whole.
Based on these findings, the DSHS announced that it will consult with a group of internal and external experts in cancer, epidemiology, toxicology, and environmental issues to evaluate whether a follow-up epidemiologic study is both feasible and recommended. According to the Houston Chronicle and the ABC News affiliate in Houston, the DSHS never has recommended the next step of seeing if an epidemiologic study is feasible. The ABC News story noted that the DSHS has conducted 260 cancer cluster investigations since 2004.
What Does the Texas DSHS Report Actually Say?
Overall there is not much that is unusual in the DSHS report. If one looks at Tables 1 and 2 (pages 8 and 9), the two cancers (out of 17 analyzed) that appear possibly elevated are childhood lymphoma and childhood melanoma. It is interesting to note that of the two childhood cancers, the elevation for lymphoma is over the entire study area, but no single census tract shows a significant elevation (Table 2). For childhood melanoma, 13 total cases were observed and only one census tract out of 38 showed a “significant” elevation, which was based on fewer than five observed cases. Overall, a review of Tables 1 and 2 do not show any extraordinary trends.
The report has a number of problematic aspects; here are five of them:
- There is a large “multiple comparisons” problem – when you perform a large number of statistical tests, some of the results will be statistically significant based on chance alone. The report considered a total of 17 cancer types over 38 census tracts. Thus, there are potentially 646 statistical comparisons (17 x 38) which at a five percent significance level would be expected to yield 32 “significances.”
- The study area is huge. A quick graphical analysis of Figure 1 suggests that it is around 700 square miles. The area is also part of Harris County Texas, which includes Houston. Since the total area of the county is 1,778 square miles, the study area is, in terms of area, about 40 percent of the county.
- Since Houston is a less than pristine environment, it seems quite likely that such a large area would include a variety of toxic exposure opportunities.
- The data on cancer cases was obtained from the Texas Cancer Registry, which was established in 1995. The report covers the period 1995-2012. New cancer registries typically experience initial startup problems that lead to an undercount of cases (they fail to record some cases because they were not notified of the cases). This in turn will result in initial underestimates of cancer rates, which will tend to bias the analysis toward higher numbers. The report does not provide a breakdown of the number of cases per year. Even for later time periods, case ascertainment can take up to two years due to reporting laws and the complexity of the data.
- Population estimates for 1995 to 2012 were calculated using linear interpolation based on population counts obtained from the U. S. decennial census for the years 1990, 2000, and 2010. This is probably the best the DSHS could do but it is not very accurate. Moreover, the population estimates are for an unspecified number of age and race specific strata. This makes interpolation even more problematic (e.g., the total population count may be fairly accurate but particular age groups will be much less precise).
Conducting a “cancer cluster” investigation in a large urban area of 700 square miles when both the size of the at-risk population and the applicable “baseline” cancer rates are poorly known seems to an impossible task. On the other hand, doing nothing was not an option and inaction on the part of the DSHS could lead to even less correct studies being conducted by “concerned citizens.”
Reports of cancer clusters should not be ignored but neither should they be taken at face value. When confronted with claims of cancer clusters, one needs to determine the details of the claims and also must conduct a technical review, along the lines of the one we have just presented, of the analyses underlying such claims. Although such a review is not guaranteed to make claimed cancer clusters “go away,” it can be invaluable in informing next steps for responding to such claims.
* * *
About the Experts
Dr. Michael E. Ginevan is the President of M.E. Ginevan and Associates, LLC (http://www.ginevan.com/). Dr. Ginevan has more than 35 years of experience in the application of statistics and computer modeling to problems in public health and the environment and in the conduct of environmental, epidemiologic, and risk assessment studies. He is the author of “Statistical Tools for Environmental Quality Measurement” and over 60 other publications in the areas of statistics, computer modeling, epidemiology, and environmental studies.
Dr. Ginevan is a founder and past Secretary of the American Statistical Association (“ASA”) Section on Statistics and the Environment and a recipient of its Distinguished Achievement Medal. He is a charter member of the Society for Risk Analysis. Dr. Ginevan has served on numerous review and program committees for the ASA, the U.S. Department of Energy, the U.S. Nuclear Regulatory Commission, the National Institute for Occupational Safety and Health, the National Cancer Institute, and the U.S. Environmental Protection Agency. He also served as a member of the National Academy of Sciences Committee on Health Risks of the Ground Wave Emergency Network. He is currently a member of the Editorial Board of Regulatory Toxicology and Pharmacology and has served as a peer reviewer for many other professional journals.
Deborah Watkins, Proprietor of Watkins Consulting, is an epidemiologist with over 30 years of experience in public health. She formerly was a Research Assistant Professor and Deputy Director of the Division of Occupational Health Studies in the Department of Family Medicine, Georgetown University School of Medicine; she remains a member of the adjunct faculty.
Prior to her work at Georgetown University, Ms. Watkins served concurrently as Director of Legislative Affairs at the Pennsylvania Environmental Council and Director for the Hazardous Waste Management Project at the Pennsylvania Environmental Research Foundation. She was the Executive Director of the Society for Occupational and Environmental Health in Washington, D.C.
* * *
For further information on litigation over the San Jacinto Waste Pits, see Texas Governor Greg Abbott Signs Bill that Limits Pollution Lawsuits; Texas County Appeals in San Jacinto River Case.