Roanoke Times
Copyright (c) 1995, Landmark Communications, Inc.
DATE: SUNDAY, June 19, 1994 TAG: 9406190062
SECTION: NATIONAL/INTERNATIONAL PAGE: A4 EDITION: METRO
SOURCE: Chicago Tribune
DATELINE: CHICAGO LENGTH: Medium
Recently, newspapers and TV anchors made much of a study reported in the Journal of the American Medical Association that found nonsmoking wives of smokers had a 30 percent greater chance of contracting lung cancer than did wives of nonsmokers.
The finding may have caused many women to fear needlessly for their health.
There is no doubt that smoking causes cancer and other serious health problems, and that secondhand smoke is a smelly nuisance and perhaps a health hazard. But the case for environmental smoke as a cancer threat is weak.
And the case illustrates how a statistically unsophisticated media can provide the public with sweeping conclusions based on scientific results that cannot support them.
Anytime the phenomenon under study is rare, as is lung cancer in nonsmokers, small changes can produce large percentage swings, said Temple University mathematician John Allen Paulos, author of the best-selling book "Innumeracy."
"If you look at the new cases of death from AIDS, the fastest growing category could be ladies over the age of 70," Paulos said. "If last year one woman over 70 died from AIDS and this year two do, you get a 100 percent increase in AIDS deaths for that category."
Knowledgeable scientists reading reports in journals can tell at a glance which studies are important and which are marginal, but to many reporters, one study looks like any other. Too often, stories emphasize a study's conclusion without mentioning the limitations that provide necessary context.
The results of the secondhand smoke study are much less compelling than suggested June 7 at an American Medical Association news conference in Washington.
The JAMA study, conducted by Dr. Elizabeth T.H. Fontham of Louisiana State University and colleagues, appears to have been done well and with great attention to detail. For example, researchers required participants to provide urine samples when possible to assure that women who claimed to be nonsmokers weren't sneaking cigarettes.
In the JAMA report, researchers interviewed nearly 17,500 women with lung cancer to find some 650 who had the disease but weren't smokers.
The scientists then selected about 1,250 women at random who didn't have lung cancer and tried to match them with the lung cancer patients according to race, family income and education.
Both the lung cancer patients and their "controls," as the randomly selected women are called, were asked about exposure to smoke as a child and as an adult and other topics such as diet. In some cases, relatives were interviewed because cancer patients had died or were too ill to talk.
The raw study findings aren't difficult to understand.
Among 651 nonsmoking lung cancer patients, researchers found that 433 had been exposed to smoke from their husbands' tobacco use, about 66 percent. Among 1,253 nonsmoking women who didn't have lung cancer, 766 had been exposed to smoke by their husbands, or about 61 percent.
Most people might conclude that the difference between 66 percent and 61 percent isn't much, and they'd be right.
To make the difference between 66 percent and 61 percent meaningful, researchers calculated a risk number - called an odds ratio - of 1.29. An odds ratio of 1.00 indicates there is no added risk; an odds ratio of 1.29 works out to a 29 percent added risk, which was rounded to 30.
This figure suggests a precision the study doesn't have.
Random chance could move the study's odds ratio anywhere from 1.04 to 1.60. That is, any figure in that range is consistent with the study's result, meaning that the risk percentage could have been as low as 4, which suggests no added risk, or as high as 60.
That uncertainty range, which is part of any statistical report, was clearly stated in the study, but it wasn't mentioned in the American Medical Association news release about the finding, nor was it stressed at the news conference.
Even though statistically sophisticated people would find the Journal of the American Medical Association conclusion weak, an AMA spokesman stressed the 30 percent risk figure and called for stricter government regulation of tobacco, giving the impression that this study was clear-cut.
Scientific findings are seldom clear-cut, but that may not deter advocates from making them seem so, and reporters sometimes abandon their innate skepticism when confronted with charts teeming with statistics.
by CNB