Spatial Analysis in Public Health

Did Drinking Water Kill in Woburn?

In 1979 two of eight water wells serving Woburn, MA, USA were found to be polluted (wells G and H):
Even before the wells had been closed, a cancer cluster alarm had been sounded by townspeople, including Anne Anderson - at right - who identified what they perceived to be an excessively high rate of leukemia (see the map below). This map shows "the proximity of the Plaintiffs' [e.g. leukemia victim] homes, wells G & H, and the locations of the Riley Tannery (R), UniFirst Properties (U), and the Cryovac Plant (C) operated by W.R. Grace", and provided the motivation for Anderson and others to file a lawsuit (made famous in the film "A Civil Action") against the owners of the industrial sites.
[Regional Review (1992)]

The town was subsequently found to have an elevated rate of leukemia, and a study was carried out to determine if this elevated rate (and rates of other disorders) could be correlated to these contaminated wells.


[Images from the collection of Professor Scott Bair, Ohio State University.]
Notice that the dots don't quite jive between the two plots directly above. Can you conjecture why? The citizens environmental group For A Cleaner Environment was founded by Anne Anderson.


An Epidemiological Study

Lagakos et al. found that, indeed, the wells were implicated in several diseases in the town, including childhood leukemia. They obtained crucial exposure data from a prior study which characterized the exposure of various neighborhoods to the water from wells G and H by using water system parameters (e.g. pipe diameters, flow rates, etc.).

Lagakos et al. reached their conclusions using proportional hazards modeling. They estimated exposures (to the waters of the wells), calculated observed rates of disease versus expected, and found statistically significant increases of leukemia and several other disorders as a function of increasing exposure to the contaminated water. Their paper generated a significant amount of commentary, including issues of multiple testing, the validity of normal approximations, and confounding (of course!).

Murphy describes the course of events leading up to the study this way: "Various studies ... of the health of the people of Woburn indicate that part of the city had a higher incidence of certain childhood illnesses, including lymphocytic leukemia, than would normally be expected. The affected part of Woburn roughly coincided with the area served by the two contaminated wells. The water distribution system of Woburn was analyzed by Helen A. Waldorf and Robert K. Cleary of the New England Interstate Water Pollution Control Commission. Based on their 1984 report..., a 1986 journal article... by Lagakos, Wessen, and Zelen has reported that the rates of childhood illness have a statistically significant positive association with the use of water from the two contaminated wells. A second water distribution analysis..., made by the author to evaluate and complement the study of Waldorf and Cleary, concurred with their overall assessment."


The Trial

The case (begun 5/14/82) brought by the Plaintiffs was interesting from the legal perspective, as it included issues of science in the courtroom and "Polyfurcation" of complex trials (the case was divided up into sections: in order for the plaintiffs to win, they had to "win" every stage). In fact, "The actual Woburn trial only got through one of three scheduled stages before settlement. The jury ruled that the plaintiffs had no claim against Beatrice Foods, Inc., but could proceed with their case against W.R. Grace.... W.R. Grace then settled for $8 million." (source)

The EPA settled with the polluters in 1990.


Contaminating wells G and H was a collective effort

The NUS Corporation, contractor to the EPA, found that trichloroethylene-tainted soil existed on the Grace property (creating a northeastern plume of groundwater pollution); the Unifirst property, which is adjacent to Grace (creating a northern plume, including trichloroethylene); and the Wildwood (formerly Riley) property (creating a western plume of groundwater contamination).
[Source: EPA Executive Summary on http://research.et.byu.edu/woburn/nus_docs/tocxsum.pdf]


[Image Source: http://www.epa.gov/region01/remed/sfsites/wellsgh/map.gif]


What do hydrology and geology tell us?

Essential arguments were presented by hydrologists and geologists, who helped determine exposure (and create exposure models).

This contour map shows the flow of water into wells G and H after 30 days of sustained pumping. The point is that contamination from WRG or UniFirst will clearly seep into the wells.


[Image from the collection of Professor Scott Bair, Ohio State University.]

Geologists argued for both sides of the case, over such things as the Surficial Geologic Map of the Aberjona River Valley. Types of soil and rock filter water in differing ways, so both sides hoped to show how existing conditions bolster their argument.


The Legacy of Woburn

The EPA touts the Woburn story as one of its success stories. The health of the population of Woburn continues to be monitored, culminating most recently in the Woburn Childhood Leukemia Follow-up Study, which issued its final report (volume I) in 1997.

According to Waller et al., the Woburn case "prompted the Department of Health in the neighboring state of New York in the summer of 1986 to initiate a series of meetings of statisticians, epidemiologists, and public health officials to discuss policy regarding the monitoring of the geographical distribution of cancer cases in their state." In fact, a modern upsurge in interest in spatial statistics in public health may be said to date from this case.


Questions/Comments:
  1. Notice that the exposure here is not represented simply by space: exposure has been modelled by a complicated network based on the data of the underground water delivery system employed by the city.

    Furthermore, this example shows that an epidemiologist may need to go far outside of her or his expertise (to hydrology, geology, and beyond) to create sensible exposures.

  2. Who identified a frightening pattern of disease? What raised suspicions?

  3. How and when did the EPA get involved?

  4. Could you recreate the two cases maps above if you were given the addresses of the cases? What would you have needed, and could you get it from software or from the net?

  5. The cases maps above have no temporal component: how could you have altered the map to indicate the time at which case was diagnosed? Can you think of any better way of representing the data?

  6. Is there any surveillance system that could have worked better than a network of concerned citizens? Sketch it out, if you can (maybe the EPA will fund you!): what would the data look like? How would it process the data?


References/Acknowledgements/Notes:
  1. Lagakos, S. W., B. J. Wessen, M. Zelen. 1986. An analysis of contaminated well water and health effects in Woburn, Massachusetts. Journal of the American Statistical Association, V81, pp. 583-596. Comments/Replies: pp. 597-614. (local copy)
  2. Kuzmack, A. M. 1987. Comment on Lagakos, Wessen, and Zelen, Journal of the American Statistical Association, V82, p. 703. With Reply.
  3. Waller, L. A., B. W. Turnbull, L. C. Clark, and P. Nasca. 1994. Spatial Pattern Analyses to Detect Rare Disease Clusters. In: Case Studies in Biometry. Eds: Lange, N., L. Ryan, L. Billard, D. Brillinger, L. Conquest, J. Greenhouse. John Wiley & Sons, Inc.
  4. Murphy PJ. Water Distribution in Woburn, Massachusetts. Environmental Institute No. 86-1, University of Massachusetts, Amherst, Massachusetts, 1986.
  5. Murphy PJ. Exposure to Wells G and H in Woburn, Massachusetts. 1990.
  6. Waldorf H, and R. Cleary. Water Distribution System, Woburn, Massachusetts, 1964-1979. New England Interstate Water Pollution Control Commission, Boston, Massachusetts, 1984.
  7. Myette, C. F., J.C. Olimpio, and D. G. Johnson, 1987. Area of influence and zone of contribution to Superfund-Site wells G and H, Woburn, Massachusetts: U.S. Geological Survey Water-Resources Investigations Report 87-4100.
  8. Information made available by the State of Massachusetts (Bureau of Environmental Health Assessment): I especially thank Stephanie Lentz, for her role in tracking down paper copies of information which was not available on-line.
  9. Reports from Charles C. Ryan, the reporter who helped break the story and additional links; especially poignant is the article by Anne Anderson, describing her struggle to get action after her son Jimmy came down with leukemia.
  10. EPA Superfund success story?
  11. Durant, John L., Jia Chen, Harold F. Hemond, and William G. Thilly. 1995. Elevated Incidence of Childhood Leukemia in Woburn, Massachusetts: NIEHS Superfund Basic Research Program Searches for Causes. Environmental Health Perspectives, V103, Supplement 6, September 1995.
  12. Legal lessons of Woburn. If you'd like to go through the case, see what these students have done: it's great!
  13. A picture gallary of the Woburn case. Many of the images used here are from this collection, the collection of Professor Scott Bair, Ohio State University.
  14. Dr. Norm Jones's Woburn Hydrogeology Data site at Brigham Young University. This site also provides documents from the NUS Corporation: "The NUS Corporation performed an extensive remedial investigation of the Woburn site. The report is available from Dr. Jones in its entirety in hard copy (2 copies). The following is the table of contents from the NUS report. Selected portions of the report are available in electronic form. Much of the information is available in the appendix."

    Some sample documents available:

  15. A similar case in CA?
  16. Anyone want to do a spatial analysis of lead?
  17. A Mathematician's Lament: in the movie "A Civil Action", John Travolta (starring as lawyer Jan Schlichtmann, attorney for the victims) intones that "... only 5 cases in 50 will win in appeals court: the odds are as easy to calculate as they are discouraging: they're 10 to 1 against." Since it's not easy for him to calculate the odds correctly, are the odds more or less discouraging?
    [The odds are actually 9 to 1. Think of it this way: if 25 cases in 50 will win, the odds are even that you'll win (i.e. 1 to 1); not 2 to 1, as he would have calculated.]

Website maintained by Andy Long. Comments appreciated.