CTSP Grantee Stories, Successes and Struggles
Huntington Breast Cancer Action Coalition (CTSP 1999 Grantee)
Community Breast Cancer Maps
By Roger C. Grimson, Ph.D.
Several community-wide breast cancer surveys have been and are being conducted on Long Island by coalitions of women dedicated to fighting breast cancer. More surveys are being planned in other counties in New York, in other state, and in other countries.
The questionnaires of the survey are brief, typically involving two sides of a page. They are informative and call attention to important matters of breast cancer. Notably, they have prompted many women to obtain mammograms and in several instances, have lead to early diagnoses of breast cancer. In short, the survey questionnaire, and the promotion of it in each community conducting the survey, has had measurable benefits in the community. But there is another objective, the collection of specific information that possibly can lead to informative geographic patterns of breast cancer.
Community surveys contain questions about known breast cancer risk factors including reproductive factors, ethnicity, and age. More important, they contain questions about space and time. The main spatial question pertains to address of residence. The temporal questions pertain to date of birth, year of diagnosis, date that the questionnaire is completed, and year that the respondent moved to her current address. This information provides the ingredients for community-wide breast cancer maps.
Why a Map?
A map is captivating. Other ways of summarizing statistical information, such as charts and tables of numbers, do not generate the same reaction in people that the map does. If I say that 182 out of 5,266 women who live in West Islip have been diagnosed with breast cancer at some time in their lives, I would receive a milder response than I would receive if I displayed a map of 182 homes of women who have been diagnosed with breast cancer and the remainder 5,084 homes of women who have never been so diagnosed. (Henceforth I shall refer to these as "case homes" and "non-case homes" for short.) Any sense of puzzlement I detect when I give the numbers becomes curiosity when I show the map. People closely scrutinize the map for certain details and they back away from the map to get the "bigger picture". Sure, the map contains more information than do the numbers, but even if I present a few numbers giving about as much information as the map, the map still has a greater impact.
The attention-causing effect of breast cancer maps is only one important factor. Another is that the maps may be analyzed for breast cancer patterns on a geographic basis. Here the issue of response is crucial. If the response rate is low, meaning that many women elect not to complete and return the questionnaire, then the opportunity to look for breast cancer patterns is lost; false patterns can arise by characteristics that only the respondents (or only the non-respondents) tend to share, (e.g. older women are more populous in certain regions and older women tend not to respond). Technically, non-response can create biased findings.
What Exactly is Being Mapped?
At first, the answer to this question may seem obvious, but as will be seen, it is not. Thy survey is population based; all women in the community are sent questionnaires, and the non-respondents are sent follow-up questionnaires. Examples of survey questionnaires that are not population based are those that are published in magazines or newspapers. These incur a selective response, only those who read the publication are given the opportunity to reply. The tabulated results of these selective responses then would reflect the responses of the readership and not the entire population. Therefore, any inferences drawn to the population based on the responses of the readership would be biased. (Professional pollsters make the effort to select people at random to complete a questionnaire in order to avoid selective response and bias.) On the other hand, repeated attempts are made in the breast cancer surveys to obtain responses from all women who reside in the community.
In order to protect confidentiality, the maps do not display streets, homes of the non-respondents or other buildings. (If necessary, the maps of colored dots can be transformed in ways that protect confidentiality while preserving any breast cancer pattern.)
The State Cancer Registry had different information. It contains a large percent (ideally, all) of residents who have been diagnosed with breast cancer. It contains no information about non-cases. Some of the cases identified by the registry are deceased or have moved to another community at some time after diagnosis. Thus maps of case homes and non-case homes based on cancer registry data would be maps of different events. But more importantly, residential histories (e.g. length of residency) are not a part of the registry’s database. There are many other differences between registry data and survey data, but it is not necessary to pursue these differences here.
Are the Maps Scientifically Valid?
The community breast cancer maps can be scientifically useful if it is reasonable to speculate that the home environment may be linked with the cause(s) and that possibly some kind of revealing spatial pattern or cluster may arise. The coalition must be willing to accept the notion that the home, the house, the location of the house, or the neighborhood of the house may pertain in some way to the developments of breast cancer. For, without admitting to the possibility that some characteristic or correlate of house (specified or unspecified) is tenably a risk factor for breast cancer, little scientific purpose would be served by mapping the cases and non-cases.
If geographic breast cancer patterns can be discussed using residential maps of cases and non-cases, then it is natural to speculate about differences in the physical environments that may contribute to the patterns. But it is important to realize that such geographic patterns can result from social characteristics of people that tend to be similar among people living within a few blocks or miles from one another and tend to be different among people living farther away – say in the same town or community but a distant neighborhood. So geographic groupings of people according to socio-economic status, for example, should create geographic grouping according to different customs, foods, reproductive factors, occupation, or ethnic groups. A person’s home environment is comprised of a milieu of physical and social circumstances that conceivably are related to breast cancer. These observations regarding home do not preclude the possibilities that place of occupation, transportation, and other environmental factors, as well as genetic factors, play a role. They simply speak to the issue that causes of breast cancer, however complex, may be reflected in residential patterns. But, it cannot be overly emphasized that in order for the map to be unaffected by biases due to non-response, the response rate must be high.
Measures to improve response rates must be given that highest priority in order for the project to be successful and for its conclusions to be valid.
In talking with scientists about the mapping projects, one wants to make clear that the surveys and maps are the results of a planned (proactive) population based survey for which certain preset hypothesis (e.g. case home tend to be near one another) can be tested. This differs from the "reactive" mode of response which investigators occasionally conduct regarding a specific allegation which usually claims large numbers or high rates.
Are the Mapping Projects Useful to the Long Island Breast Cancer Study Project (LIBCSP)?
The LIBCSP consists of ten research studies on breast cancer. These are being conducted at major medical research institutions in the northeast United States. The Columbia case-control study is one of these; the National Cancer Institute has referred to the Columbia study as the cornerstone of the LIBCS. The Long Island mapping projects are not integral to the Columbia Study. The Columbia Study focuses on newly diagnosed cases where-as the surveys for the mapping project focus on breast cancer survivors. There is nothing that the surveys or the maps can contribute to the objectives of the Columbia study. However, presumable the mapping projects would be of interest to the LIBCSP if the LIBCSP involves a Long Island based environmental cause. The National Cancer Institute is planning to fund the development of such a geographic information system (GIS). Most of the maps will be constructed and analyzed by the time the GIS will become a reality, but that should not preclude the information of the surveys and maps from being part of the GIS.
The Long Island breast cancer maps portray the residential locations of breast cancer survivors and of women who have not been diagnosed with breast cancer. Entire community populations, not just subsets of the communities, are targeted for inclusion. From these maps, breast cancer patterns can be sought. (Another issue not discussed here concerns methods of distinguishing patterns that occur randomly from those that occur by non-random processes.)
The mapping projects have a more limited set of possible outcomes than the more comprehensive environmental studies, but geographic breast cancer patterns found on the residential level may be useful scientifically. And yes, if the response rates are high, the projects are scientifically valid.
This is an abridged article. If you would like a complete cop, call HBCAC at 516-547-1518.
All text © by the respective organizations, November 15, 1999
Compilation & web design: Charles Convis, ESRI Conservation Program, November 15, 1999