Source
Centre for Research on Inner City Health, The Keenan Research Centre in the Li Ka Shing Knowledge, Institute of St. Michael's Hospital, 30 Bond Street, Toronto, Ontario, Canada M5B 1W8. tabackn@smh.toronto.on.ca
Abstract
OBJECTIVES:
To assess what statistical methods are commonly used in high-impact clinical research and how they are presented in abstracts of articles published in high-impact medical journals.
STUDY DESIGN AND SETTING:
A cross-sectional survey of abstracts of original articles published in July 2003 in four high-impact medical journals was conducted. The primary outcome was the distribution of statistical methods used in study results presented in the abstract of articles.
RESULTS:
Seventy articles met inclusion criteria. One hundred twenty-five unique statistical method presentations were analyzed. Sixty-eight percent of statistical methods used summary statistics, and 27.2% used regression analysis. When summary statistics were used, clinical evidence was presented with a P-value or confidence interval (CI) in 51.8% of statistical methods compared to 72.5% when summary statistics were not used (P=0.0282). Clinical evidence was presented verbally in 7.1% of statistical methods when summary statistics were used and in 20.0% when summary statistics were not used (P=0.0323).
CONCLUSIONS:
Summary statistics are the most frequently used statistical method to generate high-impact clinical evidence presented in the abstract of a medical article. Evidence described by summary statistics is significantly associated with less frequent reporting of a P-value or CI, and less frequent verbal presentations.