# A statistical analysis of the latest 'coffee & cholesterol' study.

A statistical analysis of the latest `coffee & cholesterol' study

My college professor introduced us to the art and science of statistics by conjugating the common word for prevaricator: "There are," he proclaimed, "liars, damned liars, and in the superlative, they are statisticians."

This definition is well-exemplified by a recent therapeutic report on coffee consumption. Based upon research at a highly prestigious medical clinic, the media headlined: "Decaffeinated coffee causes rise in cholesterol."

Clinical research is valuable in the study of the effect of various agents on human anatomy and physiology. It is quite different from research in the physical sciences where results are usually clear-cut and reproducible. In clinical research, effects may be 50 percent no change, 25 percent in one direction and 25 percent in the opposite way. This means that months or even years of work may be inconclusive, which is frustrating for the researchers.

In most cases, results for the 50 percent of individuals not affected are discarded and those who are affected are included in the study. If 20 percent go in one direction and 30 percent in the opposite, there is hope for the work. To lend some scientific validity to these possibly questionable results, they are subjected to statistical analysis. Here are techniques to magnify small or inexact differences into respectable mathematical proportions.

Lacking the precision of arithmetic, algebra and geometry, statistics is still one of the most important branches of mathematics because of the way it affects our everyday life. In government, it is the most important basis for text tables, and their periodic modifications, for minimum wages, education allocations, and many other promulgations. In business, it determines expansion of industry, demographics and market research, new products, advertising appropriations, medical and retirement benefits, and many other factors. Quality control, inventories, even certified public accounting are largely based on statistical analysis. All gambling is based upon this discipline, and insurance companies would go bankrupt if based upon unreliable data. In medicine, when validly applied, it leads to discovery and development of valuable vaccines and medications for alleviation and cure of serious ailments. In politics, successful candidates discover voter concerns by accurate polling.

The report on decaffeinated coffee being associated with a rise in blood cholesterol was presented at the annual meeting of the American Heart Association in New Orleans on November 15, 1989 by Dr. Robert Superko, who directed the research at the Stanford Lipid Clinic. The full report had not been published, but an abstract was available in the Medical Journal of Circulation (Vol. 80, Oct. 1989).

Exactly 180 middle-aged, non-smoking men drank a national brand of regular coffee, brewed by a standard procedure, without cream, consuming three to six cups per day. They were then divided into three equal groups and assigned for two months to the same cup per day quota of either the same coffee, decaffeinated coffee, or no coffee.

The researchers found that switching to decaffeinated coffee caused a 4.8 mg. rise in low density lipo-protein, which is currently considered "bad" cholesterol by the medical profession. The group who continued to drink regular coffee and the group that changed to no coffee exhibited an equivalent 4.2 mg. drop in their serum cholesterol blood level.

The good news is that the drop in cholesterol followed consumption of regular coffee for a two month period. The fact that the drop in cholesterol was identical for non-coffee drinkers could lead to the fairly obvious conclusion that caffeinated coffee has no effect on human cholesterol levels and the decrease in both cases was probably due to other factors.

Cholesterol is a key component in several phases of human metabolism. On a completely diet-free cholesterol, the body manufactures its needs from normal food ingredients such as margarine or vegetable oils. Changes in cholesterol content of body fluids can be triggered by dozens of different factors such as variations in total food intake and its components, exercise, sleep, stress and many other conditions. Even though the results of this test indicated ordinary coffee had no effect, the cholesterol problem is so complicated that coffee cannot be ruled out as having a minor, if negligible effect.

The fact that consumers of decaffeinated coffee presented a 4.8 mg. rise, while the other two groups exhibited a 4.2 mg. decrease lends credence to a scientific statistical significant difference between drinkers of decaffeinated coffee and the two groups who partook of regular or no coffee. Mathematically, that is true, but when the underlying facts are examined, the whole report dissolves into nothingness.

The normal cholesterol content of human blood ranges between 130 and 200 mg. It increases with aging and many medical cases have well over 200 mg. levels. The small differences discovered in this clinical test represent a deviation of two to three percent from normal.

Chemical analysis of total cholesterol lacks the precision of other analyses and the error in the analysis of two portions of the same samples is often about 3%.

Before machine analysis, technicians always analyzed two portions of a blood sample and averaged the two for their reports. Machine analyses, though much more rapid, are no more accurate. In addition, the percentage of cholesterol in the same individual may vary up to 5 percent at different times of the same day.

I have not seen the results of the tests on the various individuals involved in this clinical study. In tests of this nature on 60 people each, wide variations will be tabulated. Sometimes there are a few results with wide deviations from the average. These deviations could be caused by one or more factors in a wide range extraneous causes. Some researchers discard them, while others permit them to skew results. In both cases statistical analysis can offer significance where none exist.

Medical reports of this nature must be approved by peers of the author before publication. These include experts in medicine and statistics. Seldom are knowledgeable coffee people consulted.