Scientific Studies

By: Kavanagh, Terence, Take Heart! Key Porter Books 1998, page 28-30.

In order to understand the studies and trials on which our knowledge is based, an understanding of the terminology is necessary.

Retrospective studies identify patients suffering from a disease and, through medical interviews and examinations, attempt to discover which characteristics they have in common. For example, in the case of coronary disease, the common characteristic might be male gender or cigarette smoking; these are referred to as risk factors for coronary disease.

The predictive accuracy of risk factors established by retrospective studies can be tested by prospective studies. These identify healthy individuals who possess the proposed risk factors, then follow them to see if they eventually develop coronary disease. Both types of studies make no attempt to affect the outcome; they are purely observational in nature. Probably the best- known, and certainly the most-often-quoted prospective study in the field of coronary disease, is the U.S. Department of Health's Framingham Study, which started in 1949 and is still active (as one of its principal investigators, Dr. William Kannel has said " It's a race to see who survives the longest, the investigators or the investigated!").

In contrast, clinical trials test the effect on a disease of some form of intervention - either drug, diet, or even regular exercise. In order to avoid biasing the results by inadvertently preselecting a group of subjects who might respond favorably to the intervention, a process known as randomization is used. Candidates for the trial are randomly allocated, usually by computer, to either a treatment or nontreatment (control) group. In the case of a drug clinical trial, the control group is generally given a placebo, an inert, harmless substance identical in appearance to the drug. This corrects for the placebo effect, a well-known medical phenomenon in which some subjects experience a benefit from any form of treatment merely because they expect one. Since subjects do not know whether they are taking the active drug or the placebo, the trial is described as "blind".

When we come to evaluate the results of a clinical trial, we need to compare the occurrence of a predetermined measure, called the end-point, in the treated and the control groups; for instance, the occurrence of death from a heart attack. However, we have to be sure that any difference in end-points between the two groups is not due to chance. This is solved by subjecting the numbers to statistical significance testing. For a trial to show a statistical significance and a conclusive result, the probability of any difference in end-point between the treated and the control subjects happening purely by chance alone must be 5% of less, expressed as p (probability) = 0.05. A probability of more than 5% (p = 0.06 or greater) means a non-significant or chance result. Obviously a probability of 1% or one in a hundred (p = 0.01), or of 0.1%, one in a thousand (p = 0.001) confers upon the difference in end-point even greater statistical significance, and therefore makes it even more conclusive.

Sometimes a trial may fall just short of achieving a statistically significant result, possibly because of methodological problems, e.g. not enough subjects were recruited, or the time allowed for follow-up was too short. In cases such as this, if the data from similar trials are available, then we can resort to a statistical tool known as meta- analysis. Here we pool all the results and analyze them as if they came from a single trial. In this way we may be able to detect from a series of "almost successful" trials an outcome which may be more conclusive. The use of meta-analysis is rapidly increasing as we come to rely more and more on clinical trials to prove, or disprove, the value of various medical treatments.

More Articles

News Release: Longer Cardiac Rehab Programs Necessary, Says University Of Toronto Study

Although three months are often prescribed for cardiac rehabilitation, it takes nine months for patients to reach peak improvement, say researchers from the University of Toronto.

Learn More