Related Experiment Videos
Reporting of numerical and statistical differences in abstracts: improving but not optimal
1Department of Medicine, University of Toronto, Ontario, Canada.
Journal of General Internal Medicine
|April 4, 2002
Summary
Reporting of treatment outcomes in medical abstracts improved significantly between 1986 and 1996. However, abstracts still lack sufficient raw data, P values, and confidence intervals (CIs) for complete understanding of therapeutic efficacy.
Area of Science:
- Medical research methodology
- Scientific communication
- Biostatistics
Background:
- Reporting of relative risk reductions (RRRs) and absolute risk reductions (ARRs) influences perceptions of therapeutic efficacy.
- The use of P values versus confidence intervals (CIs) in reporting statistical significance is debated.
Purpose of the Study:
- To describe how numerical and statistical differences in treatment outcomes are presented in published medical abstracts.
- To compare the reporting of statistical information in abstracts over time.
Main Methods:
- A descriptive study analyzing abstracts from 8 general medical and specialty journals in 1986 and 1996.
- Included controlled intervention trials with binary outcomes.
- Recorded raw data, RRRs, ARRs, number needed to treat, P values, CIs, and verbal significance statements.
Main Results:
- The proportion of abstracts lacking both raw data and a P value or CI decreased from 60% in 1986 to 28% in 1996.
- Reporting of absolute difference measures increased from 0% in 1986 to 12% in 1996.
- Reporting of CIs increased from 2% in 1986 to 35% in 1996.
Conclusions:
- Quantitative outcome and statistical information in abstracts significantly increased between 1986 and 1996.
- Further improvements are needed to enhance the informativeness of abstracts regarding treatment efficacy and statistical significance.