Kathleen Z. Rooney
In order to ensure safety and efficacy of drugs, manufacturers must employ statistics in all phases of production. Drugs need to be safe. They need to have a positive effect on patients. They must be manufactured in a consistent way to ensure accurate dosages. No drug is without unintended or side effects. What matters is to have a positive benefit-to-harm ratio. Benefit must be weighed in terms of symptomatic relief, reduction in disease risk, complications or increased survival. Harm must be measured in terms of unintended deleterious effects on patients. The balance of benefit to harm should be affected by what other drugs are available for the particular condition as well as the seriousness and urgency of the condition.
Prior to clinical studies, companies must do research to identify compounds and create a hypothesis against which they plan to test their drug. They must present meticulously planned studies and any research or prior animal testing in order to apply to do clinical trials in humans. Studies are generally considered Phase 1, 2, 3, or 4. Phase 1 studies are small and look at safety and metabolism in healthy patients. Phase 2 and 3 are clinical trials of patients who are being treated for the targeted conditions. With statistically significant success shown in these trials, a company may apply for a right to market the drug. After approval, Phase 4 consists of follow-up observational studies that look for rare adverse effects.
The FDA requires that a company present a prospectus with a hypothesis before any human testing can begin. An institutional review board must review this proposal for soundness. Not only does stating a hypothesis before beginning trials avoid fishing expeditions to find useful effects of drugs, it is a necessity in order to correctly design an experiment that effectively controls as much variability as possible.
There are several distinct types of study that can be performed for establishing benefits of a particular treatment. The most reliable is considered the double-blind, placebo-controlled, clinical trial. Clinical trials are prospective (forward moving) experiments that can establish cause and effect. Observational studies can show association or a relationship between variables, but are not accurate in demonstrating cause and effect. There must be comparison groups with manipulation of treatments in order to avoid mistaking the effects of confounding variables.
It is of paramount importance to randomize treatment groups of patients. Randomization is critical to establish equality among the treatment groups, to minimize the sources of variability that are not possible to be controlled for. Randomization reduces bias that researchers could intentionally or unconsciously weight the treatment groups. It is the best way to avoid systematic bias, and bias in an experiment can invalidate the results.
Blinding is a second feature of controlled clinical trials. When all parties involved in the treatment of the experiment (subjects, physicians, clinical staff) are not aware of which groups the subjects are assigned to, it is called single-blinding. This can help diminish "Hawthorne-effect"
10
where the knowledge that one is being studied can change the outcome of an experiment. Single-blinding can also mitigate "placebo effect", where patients improve on placebo, simply by feeling that they are being treated. Double-blinding, where all individuals involved in collecting and measuring the results of the treatments (subjects, physicians, clinical staff, researchers) do not know each individual subject's treatment assignment can lessen "ascertainment bias" where the hopes and expectations of subjects and researchers can sway outcomes. A famous National Institute of Health study of vitamin C's effect on the common cold showed that when participants (many of whom were NIH scientists) became curious and unblinded the study, they unconsciously underreported the duration of their symptoms and underestimated the length of their sickness.
11
This study showed that knowledge of treatment could affect both researchers and subjects in the interpretation of symptoms and disease markers.
Sample size is important in statistical theory as well as in practice. Statistics measures variability, and is most effective when given the opportunity to accurately measure the type of variability that exists in a population. Samples that are not sufficiently large cannot introduce the correct amount of actual variability and may under- or overmeasure the effects of a drug.
When clinical trials are not ethical or practical, observational studies or meta-analysis can be used to study safety and efficacy. These can be used to observe teratogenic effects, the effect on a fetus when a pregnant woman is given a drug. Although it would be unethical to test drugs on a fetus, records exist when a woman has been unintentionally treated with a compound during pregnancy, and those records may be used in an observational study.
