ABSTRACT
I. Introduction Consider the following statements, which were included as part of an abstract to
an article reporting the results from a randomized trial comparing stent placements to balloon angiography in obstructed coronary bypass grafts:
“As compared with the patients assigned to angioplasty, those assigned to stenting had a higher rate o f procedural efficacy...(92% vs. 69%, p < 0.001), but they had more frequent hemorrhagic complications (17% vs. 5%, p < 0.01).... The outcome in terms of freedom from death, myo cardial infarction, repeated bypass surgery, or revascularization of the tar get lesion was significantly better in the stent group (73% vs. 58%, p = 0.03)”1
Proper interpretation of the above results, and of similar reports from much of the modern clinical literature, depend in large part on the understanding of statisti cal terms. In this case, terms such as “significant” and “/^values” were given, and in other reports one may see terms like “confidence intervals”, “f tests”, “Chi-squared tests”, “power”, “type I and type II errors”, and so on. Clearly, surgeons and other clinicians who wish to keep pace with new techniques and technologies must at least have a basic understanding of statistical language. This is true not only if they desire to plan and carry out their own research, but also if they simply want to read the medical literature with a keen critical eye, or if they want to make informed deci sions about which new treatments they may wish to apply to their own patients, and under which circumstances.
