What is a Medical Study?

As an example of a medical study, lets consider the question of whether 10 miligrams FactStatin (there is no real medication by this name...) taken daily, significantly decreases the probability of getting a Coronary Heart Disease (CHD) in the five years following the start of the investigation.

We start by determining the size of the study - how many people will be tested. This involves a compromise - on the one hand, the more people that are studied, the more statistically significant (more precise) the result will be. On the other hand, the treatment, periodic checkup, data recording and evaluation on a large number of patients is very expensive. Often, the drug manufacturing companies will foot the bill, and this raises a question as to whether the fact that they are paying for it influences the results. We hope that it does not...

Another requirement is to randomly divide the group into two: one group will actually get the FactStatin, whereas the other group will get a placebo. A placebo is an innocuous material fabricated to look like the FactStatin, so that the individual taking the pill doesn't know whether she or he is actually getting the medication. This is because our minds and bodies are inexorably linked: often, if we believe we are taking medication, or just due to the fact that someone is paying attention to us, we feel better, have less significant symptoms, or avoid illness all together. So it is essential to evaluate how much better than a placebo FactStatin performs. After all, if a placebo performs well, why pay for an expensive statin?

Of course, the patient is not told whether s/he is getting a placebo or "the real thing" - this would defeat the purpose of giving a placebo... This is called a "blind study". Generally, the person administering the medication also doesn't know what is being given, to prevent any hint to the patient about the true nature of the medication, or any bias on the part of the reporting physician. This is known as a "double blind" study.

Not all studies fulfill all these requirements. Some studies compare two different types of medications without a placebo. Observational and in particular cohort studies compare populations with a different characteristics - e.g., people who smoke and people who don't, or women who used Hormone Replacement Therapy and women who didn't. Such studies need to account for confounding factors. These are factors that determine the risk of the disease independently from the variable being investigated. For example, in a study to determine the effect of smoking on people's health, it is not enough just compare smokers to non-smokers. For example, it is possible that non-smokers exercise more frequently and/or eat more healthy diets. Therefore, for the study to be meaningful, the study needs to account for all the factors that impact the result, and correct for differences in the factors (e.g., exercise, diet) that are not relevant to the study (determining the effect of smoking).There are times when such observational, non-randomised studies provide inaccurate or misleading conclusions, such as when it was thought that HRT, or taking vitamin E reduced the risk of heart disease. Subsquent double-blind randomized trials disproved this. What could occurr in such a study, is that women who take hormones or people who take vitamin E, also engage in other activities that prevent heart disease. As much as the study tries to correct for all known factors, it is always possible that

  1. The exact correction is not precisely known.

  2. There are unknown confounding factors.

The last type of study we mention is the "Case Study". In this type of study, patients who have the disease are studied in comparison to other similar people who don't have the illness. The comparison evaluates exposure to suspected causes of the disease and compares the two populations. This type of study suffers from the same type of problems as the cohort studies - not all the confounding factors are known, sometimes leading to inaccurate conclusions.

Next

Last Modification - August 14, 2004