2004 Surgeon General's Report

Those Pesky �Adjustments�


In April, 2003, a couple of doctors in Helena, Mt, released a study on the results of a smoking ban in that community. Helena is a city of 88,140 people. It has a single hospital. According to the study, there was a 40% drop in hospital admissions for heart attacks during the 6 month period after the smoking ban was imposed. After the ban was lifted, supposedly, heart attacks went back to the �normal� level.

This study has been roundly attacked by Jacob Sullum and others because of the small numbers of people involved, the limited time period and the fact that the doctors who did the study were ardent supporters of the ban. During the ban, 26 Helena residents were admitted for myocardial infarction as opposed to an average of 40 during the same period during the same period for the 5 years before the ban. As Sullum has pointed out, however, it is quite possible to find six month periods during those five years in which the number of heart attacks was as low or lower than the average of 26 recorded during the ban.

Sullum, however, has missed another important point, which was mentioned in only a handful of the hundreds of newspaper and other stories on this study. It turns out that, as reported in the Milwaukee Journal Sentinel on April 6, 2003, Stanton Glantz (the anti tobacco crusader) adjusted the figures in the report before it was released, for �seasonal variations�. In other words, we really don�t know the true average number of hospital admissions during the time period covered by the ban. It could have been 20, �seasonally adjusted� to bring it up to 26. But, knowing Glantz, that�s most unlikely! More likely it was 40, �seasonally adjusted� down to 26 to prove his point. We also don�t know the average number of admissions during the control period - the six months prior to the ban - since those, too, were apparently adjusted.

Give me the ability to �adjust� the figures in a study, and I�ll make any study prove my point. Suppose I have a group of smokers and another group of non-smokers and it turns out that the lung cancer rate among the smokers is twice that among the non-smokers. I have data on the age, sex, ethnicity, occupation, educational attainment and various other characteristics of the study subjects. I look over the data and see that a disproportionate number of the smokers with lung cancer were steel workers. I decide that lung cancer must be an occupational hazard of working in steel mils, so I make a little �adjustment�, throwing out the steel workers who are smokers. Now, I have, perhaps, an equal lung cancer rate for the smokers and non-smokers. But that�s not good enough. I want to prove that smoking is good for you. So, I look around for another factor that I can adjust. Maybe, if I take out people of Irish ancestry, that will do the job so I run the data through my computer to see whether it works. Maybe it does, and I have now reached the result that I wanted. If it doesn�t, I can just try another factor.

Of course, if I did that, I�d be justly and roundly condemned as a fraud and a charlatan. But such is the climate of public opinion that anti smoking �scientists� get away with this all the time.

All of which brings me to the matter of the 2004 Surgeon General�s Report. In that Report, the Surgeon General cites a number of �new� studies, which convince him that, in addition to all of the other diseases �caused by smoking�, a whole raft of other diseases must be added to the list, including cancer of the pancreas, cancer of the larynx, bladder cancer, cancer of the oral cavity, etc.

So far as I can tell, all of the new studies cited by the Surgeon General were retrospective, i.e., people were asked for their recollections of their smoking habits and, in many cases, their drinking habits and other lifestyle activities. When a person with a disease like lung cancer or cancer of the larynx is asked about his smoking habits, he�ll invariably think �I must have smoked a lot, for a long time, or I wouldn�t be so sick�. This is called �recall bias� and is the reason why, at the time of the original Surgeon General�s Report in 1964, the authors rejected the retrospective methodology and vowed to rely only upon �longitudinal� studies, i.e., prospective studies in which groups of smokers and non smokers would be followed over a period of time, to determine whether the smokers developed more lung cancer than the non smokers. It�s a sad commentary on the current state of medical �science� that today�s medical �scientists� are willing to accept retrospective studies that would have been considered flawed, 40 years ago.

Most of the studies were also adjusted and some of the adjustments make no sense to me. It�s necessary, of course, to adjust for age, because we can�t validly compare disease rates in 25 year olds with disease rates in 65 year olds. But some of the studies adjusted for religion (of all things), educational attainment, alcohol consumption, geographic residence, and a whole bunch of other things. One study of cancer of the aerodigestive tract adjusted for �cumulative alcohol and tobacco use, race, beverage temperature, religion, wood stove use, and consumption of spicy foods�. What�s especially interesting is that this was a hospital based study, done in Brazil and the cancer patients had already been computer matched with controls for gender, age and quarter of admission, even before the study began. So why were any further adjustments needed?

The adjustment for wood stove use is intriguing. It suggests that if the authors encountered a non-smoker with cancer, they could throw that person out of the study, on the theory that wood smoke caused his disease.

The various studies are also inconsistent. Several suggested that smokers of low tar cigarettes had a lower risk for lung cancer than smokers of high tar brands. But another suggested that �After adjusting for age and total pack years, the difference in risk was insignificant [because] low-tar smokers compensated by smoking almost half a pack more per day�.