You Call This Science?

• Author: Orient, Jane M.
• Position: UNDER THE MICROSCOPE

SCIENCE has undergone a radical metamorphosis. People with M.D. or Ph.D. degrees who have published hundreds of papers in the scientific literature suddenly have become "anti-science"--as judged by media personalities, politicians, bureaucrats, and fact-checkers. What does this mean? One clue: chief White House medical advisor Anthony Fauci, M.D., said that anyone who attacked him was attacking "science." Anyone skeptical of the accepted COVID narrative may be called "anti-science." The articles of faith are that it is an existential threat, and the only solution is universal vaccination and surrender of our freedom to work, assemble, worship, travel, or even go outdoors without the approval of public health authorities.

One key feature of science that we should have learned in grade school seems to have been forgotten, albeit selectively: experiments need to have a control group. Ancient or medieval physicians may have said that bleeding and purging cured disease. The only time these cures did not work was when the patient was too far gone to be saved. The science was settled; no control group needed.

Doctors may like to imagine that we have god-like powers but, in fact, we never can know what would have happened had we done something different. That is why studies of treatment must have a control group. The "gold standard" is a double-blind randomized controlled study (RCT) to correct for observer bias and the placebo effect. Giving a fake medicine (placebo) sometimes helps.

Any experiment can have two types of error. A type I error "finds" a difference between the treatment and control group just because of random variation. A type II error fails to find a sufficiently large difference to reach "statistical significance" even though one exists. Studies need to be adequately "powered"--have an adequate number of subjects--to minimize type II errors. A power of less than 80% generally is considered unacceptable. A study that is too small is called "underpowered."

Ideally, there should be a 50-50 allocation of treated and control subjects. If only 30% of subjects are in the control group, the study loses significant power. If only 10% are in the control group, the power of the study is only 40% to 60%, writes data scientist Mark H. White II. At the moment, about 30% of Americans have not taken the COVID vaccine, and government keeps trying to reduce that percentage to as close to zero as possible.

There was a 50-50 allocation in the...