Census Tract Epidemiology Strikes [out] Again

BrooklynDodger blogged previously about census tracting for vermiculate in New Jersey. Here's an uncontroverted attempt to quell concerns about explosives residues and nuclear processing in Missouri. BrooklynDodger questions whether enough stuff from the feared facility could get far enough to measurably perturb mortality in the target zip codes or county. So a null finding would likely be uninformative whether the leakage, which could still be, in EPAese, an "unreasonable risk to public health" and an association would likely be with something other than the targeted leakage.

Nevertheless, BrooklynDodger was bored enough to search for the actual report. Findings were variously headlined as "no cancer cluster", "are not higher," are "not higher than normal."

Somewhat, this turns on combining races in all analyses, and combining all races and genders in some of the key analyses. BrooklynDodger questions this increasingly common practice. Gender is a "real" thing, race is a reification of real social and hypothetical genetic ecological differences between populations. But population rates by race widely diverge. Transparent public health practice would identify the most concerning evidence in a race-gender group in addition to stating the investigators' strongest beliefs on the correct data set on which to base a conclusion.

Which brings BrooklynDodger to that strongest number. Leukemia mortality among males in St. Charles country was 99 observed, 82 expected. BrooklynDodger's statistical website calculates an SMR of 1.21, CI = 0.99-1.47. Ratios in females were weaker but still greater than 1.0.

So what's a risk communicator to do? [U of M is running a big symposium on risk communication. What's the MSDS for soft soap when the ingredients include John Graham and Dennis Paustenbach?]

"Not higher" without qualifiers [weasel words?] is sketchy. "Not higher than NORMAL" is still pretty sketchy. "Not signficantly different" would convey more information if it were not significantly "higher," but obviously would draw a follow up question from the great unwashed.

Fact is, the study maybe ruled out an excess greater than 47% among the men.

Persons concerned about radiation would be better served by measuring radiation, not that hard to do and with a substantial data base for assessing risk. Those concerned with past exposure to explosives residues would have to put their faith in toxicology.

PS: The Missouri public health web page leads with "Alternatives to Abortion."

>>>>>>>>>>>

For Immediate Release:July 28, 2005
Contact:Nanci GonderOffice of Public Information573-751-6062
State health department study finds leukemia rates are not higher inWeldon Spring area
An investigation by the Missouri Department of Health and SeniorServices has determined the number of leukemia cases in St. CharlesCounty is not higher than normal. The investigation was conducted inresponse to concerns expressed by citizens in the Weldon Springcommunity about cancer – primarily childhood leukemia – and pastchemical contamination in the area.

>>>>>>>>>>>>>>>>>>>>>>>>

http://www.dhss.mo.gov/DataAndStatisticalReports/weldonspringcancerreport.pdf

Analysis of Leukemia Incidence (1996-2000) and Mortality (1994-2002) Data
in St. Charles County, and Weldon Spring and Surrounding Areas

Summary: In this study, the observed number of incident cases of leukemia in St. Charles County, Weldon Spring and its surrounding ZIP codes during 1996-2000 were compared with the expected number of cases based on the incidence rates for the State of Missouri during the same period; the observed number of leukemia deaths during 1994-2002 were also compared with the expected numbers of death based on the leukemia mortality rates for the State of Missouri during the same period. This study found that leukemia incidence rates during 1996-2000 and mortality rates during 1994-2002 in St. Charles County, Weldon Spring and its surrounding ZIP codes were not significantly different from that of the State of Missouri; the incidence rates for cancers that may be associated with exposure to radiation sources (such as sunlight, x-rays, gamma rays and other naturally occurring or man made sources of radiation) or result from a number of other genetic, lifestyle or unknown risk factors during 1996-2000 and mortality rates during 1994-2002 in St. Charles County were not significantly higher than that in the State of Missouri during the same periods.