Confidence Intervals in Procedural Dermatology: An Intuitive Approach to Interpreting Data

Abstract

Significant differences observed in therapeutic trials in procedural dermatology are typically denoted by p values of less than .05. Alternatively, significance can be conveyed by use of confidence intervals. The purpose of this article is to clarify how confidence intervals convey the same information about outcomes as p values, albeit in a slightly different manner. (1) Selective review of textbooks and other relevant literature and (2) presentation of a brief tutorial describing confidence interval determination for therapeutic clinical trials comparing differences between means of two groups. Routine use of confidence intervals is an intuitively satisfying means for conveying the statistical significance of results and can be used in combination with p values for understanding these results. Specifically, confidence intervals are a useful tool for indicating the size, spread, and direction of the observed differences. Unfortunately, dermatologic surgery trials tend to have low sample sizes, which frequently result in outcomes below the threshold of statistical significance (p > .05, or confidence intervals including 1.00). In the absence of statistical significance, neither p values nor confidence intervals yield definitive results. Confidence intervals can complement p values as a means for explaining statistical significant differences. When differences are not statistically significant but are clinically significant and approach statistical significance, neither p values nor confidence intervals can definitively establish whether the observed trends are indicative of an underlying difference. In these cases, common in procedural dermatology, larger, better designed, randomized prospective trials are needed.