Regression to the Mean: A Potential Source of Error in Clinical Pharmacological Studies

Abstract

The design and interpretation of clinical trials in clinical pharmacology must avoid certain pitfalls to be useful in guiding drug therapy decisions. Here, we discuss one important problem in clinical trial design, the problem of regression to the mean. Because certain disease states tend to wax and wane in severity and because patients tend to present for treatment when disease activity is high, the expected course of the disease is improvement. Therefore, any treatment begun will appear to lessen disease activity. We give four examples of diseases where improper and inadequate recommendations for drug therapy have been made based on clinical trials that ignored the problem of regression to the mean: vasospastic angina, rheumatoid arthritis, chronic congestive heart failure, and renal stones. Studies of drug therapy in diseases such as these should be conducted with concurrent placebo controls rather than having each patient serve as his own control.