pp.162-163 Junk Science Judo    Steven J. Milloy

Study Finds High Exposure to Dioxin Increases Cancer Risk

reported the Associated Press on May 5, 1999. The article continued:

"Chemical workers exposed to high levels of dioxin have a 60 percent increased chance of dying of cancer, but the chemical poses no added cancer risk to the general population, a study says. Kyle Steenland, co-author of the study appearing today in the Journal of the National Cancer Institute, said the research suggests most people typically are not exposed to dangerous levels of dioxin and that cancer appears to result only among those who had extremely high and long-term exposures to the chemical 15 to 20 years ago.
"There is not a significant increase in cancers until you get to the upper exposure levels,” Steenland said Tuesday. "This is not raising a red flag by saying that low level exposure is causing a lot of  cancer."

The researcher's conclusion is based on a pooling of the weak statistical associations for specific cancers into a category called "all cancers." Check out the study's results for "all cancers" and  specific cancer sites in Table 11.1
You'll notice that other than the statistical associations for bladder and larynx cancers, none of the reported statistical associations is statistically significant by virtue of its confidence intervals. (We don't know about the p-values because they weren't reported—or do we?)
The researchers try to circumvent this problem by combining data for the specific cancer sites into an "all cancer" category. But as Bruce Charlton pointed out earlier:

"The root of most instances of statistical malpractice is the breaking of mathematical neutrality and the introduction of causal assumptions into the analysis without scientific grounds. This amounts to  performing science by sleight-of-hand: the quickness of statistics deceives the mind."

In this case, the researchers assumed that it was biologically possible for dioxin to be a cancer-causing agent at all sites in the body—even though there is really no persuasive evidence that it acts as a cancer-causing agent even at one site. The "all cancers" category is a statistical trick to enable the conclusion that dioxin is a cancer-causing substance.
A general rule that may be taken from the "all cancers" trick is, in epidemiologic studies, beware of statistical pooling intended to overcome data that provide the "wrong" answer.

Table 11.1: Dioxin and Cancer

Type of Cancer                     No. of Deaths             SMR (95% CI)
All cancers                                 377                            1.13 (1.02–1.25)
Esophagus                                   13                             1.46 (0.77–2.49)
Stomach                                      13                             1.04 (0.55–1.78)
Small intestine, colon                     34                            1.16 (0.80–1.61)
Rectum                                           6                             0.85 (0.31–1.85)
Liver and biliary                               7                            0.88 (0.44–1.57)
Pancreas                                       16                             0.96 (0.55–1.56)
Peritoneum and unspecified             3                              2.19 (0.45–6.41)
Larynx                                         10                              2.22 (1.06–4.08)
Lung                                           125                             1.06 (0.88–1.26)
Prostate                                        28                             1.17 (0.78–1.69)
Kidney                                          13                             1.56 (0.82–2.66)
Bladder                                         16                            1.99 (1.13–3.23)
Lymphatic and hematopoietic         35                             1.11 (0.78–1.54)
Hodgkin's disease                             3                             1.09 (0.22–3.19)
Non-Hodgkin's lymphoma             12                              1.10 (0.56–1.91)
Multiple myeloma                           10                              2.07 (0.99–3.80)
Leukemia and aleukemia                10                               0.81 (0.38–1.48)
Brain and nervous system                 8                               0.81 (0.35–1.60)
Connective tissue and soft tissue       4                               2.32 (0.63–5.93)

Note from RJB--Two out of 19 tests are significantly different from 1 at the 5% level, according to the confidence limits shown. The probability of this occurring by chance, according to our studies on sampling statistics and the Central Limit Theorem, is almost 20%.  Pooling of all the data gave a larger sample size, but one could question whether this is legitimate. Five of the categories give results, albeit insignificant, “suggesting” a protective effect of dioxin (relative risk<1), and even with pooling, the relative cancer risk from high level exposure to dioxin is barely significant at the 5% level. Moreover, who knows what other characteristics the subjects might have had in common?

Dioxins are a family of some 75 compounds that occur naturally from a number of processes, including forest fires. There is some evidence that higher levels resulting from human activities may be detrimental to human health. Do a google search to see some of the controversy. Here is a good overview: http://www.heartland.org/Article.cfm?artId=12374

Ben & Jerry assert in a brochure that "the only safe level of dioxin is zero,"  which is impossible to achieve. Michael Gough and Steven Milloy actually had dioxin levels in Ben & Jerry’s ice cream measured by two independent labs and found one serving gave a dioxin exposure 200 times the “EPA safe level.” Nobody has questioned the findings. Instead, various environmental groups have subjected Milloy to bitter personal attacks for being "pro-industry."
here are a few references:



back to 3300 page