Retrospective, Observational Studies

These begin by observing a recognizable pattern of disease behavior such as—people with high blood pressures have more strokes, those with high cholesterols have more heart attacks, and those with more stress have more bleeding ulcers.  These observations provoke hypotheses or theories that attempt to explain the observed phenomena. Observers might theorize that:

a) high blood pressure—by weakening the arteries—causes strokes,

b) high cholesterol—by dirtying the arteries—causes heart attacks, and

c) stress—by increasing stomach acid—causes bleeding ulcers.

Prospective Studies

Attempting to prove or disprove the proposed theories, initial prospective studies may take a defined number of persons and measure their blood pressures, their cholesterols, and their stress levels at initiation.  Then, by following these persons for a pre-defined period of time, the study collects data about strokes, heart attacks, and bleeding ulcers.  Then the data are plotted against the blood pressures, cholesterols, and stress levels and statistical analyses are applied to detect significant differences.  If the statistical analyses reveal that there are significant differences between those with low and high levels and that the differences cannot be attributed to chance, then the studies can conclude that such high levels are indeed associated with such diseases.  The studies cannot conclude, however, that the high levels are the causes of such diseases because the studies were not pre-designed to show a cause-and-effect.

Controlled, Prospective Studies

To prove cause and effect requires control arms and treatment arms.  The control arms take groups and follow them without treatment.  The treatment arms take similar groups and give them treatments to normalize their blood pressures, cholesterols, or stomach acids.  After a pre-defined period of time, the data are collected and statistical analyses are again applied.  If the control arms are found to have more strokes, heart attacks, and bleeding ulcers than the treatment arms, and if the differences are determined to be statistically significant and not due to chance, then the conclusions would be that high blood pressure indeed causes strokes, that high cholesterol indeed causes heart attacks, and that stress indeed causes bleeding ulcers.

Double Blind, Controlled, Prospective Studies

However, such preliminary, controlled, prospective studies as cited above, which did seem to show cause-and-effect, do suffer from blatant biases.  The biases are caused by the fact that both patients and doctors know who is getting treatment and who is not and it is this pre-knowing or foreknowledge that consciously or unconsciously introduces biases of interpretation and pollutes the integrity of the study.  To avoid the human biases of interpretation, therefore, the studies now have to be redesigned with placebo and treatment arms.  The placebo arms receive pills that look identical to the treatment arms but contain inert powders instead of active medicines against high cholesterol, high blood pressure, and high stomach acid.  Moreover, the doctors and their patients will be blinded, which means that neither the doctors will know what they are giving nor will the patients know what they are receiving.  The codes are only broken after the data are collected at the end of the study and only after that time are the differences subjected to statistical analyses.  It is such, double blind, controlled, prospective studies that are the cornerstone of evidence-based medicine.

Confirmation of Double Blind, Controlled, Prospective Studies 

One study is usually not enough to confirm a new scientific truth because the study may suffer from selection or environmental biases.  Perhaps the study was conducted in Oklahoma where very few international residents live or perhaps it was funded by drug companies whose medicines are being preferentially utilized in Oklahoma.  Consequently, that Oklahoma population may not represent the Washington D.C. population where numerous international delegates reside who might be on other preferred medicines.  It is important to repeat the study in different populations in order to confirm its findings.  When the results of such confirmatory studies are in agreement with the primary study, then the new truth is considered confirmed and worthy of widespread acceptance.  It would require several confirmations over a good stretch of time before a truth becomes solid and indisputable.

Dr. Hanna Saadah

Issue # 97 - August 2010