February 10, 1997

Methodological Studies of Systematic Reviews: Is There Publication Bias?-Reply

Author Affiliations

Hamilton, Ontario

Arch Intern Med. 1997;157(3):357-358. doi:10.1001/archinte.1997.00440240123019

This article is only available in the PDF format. Download the PDF to view the article, as well as its associated figures and tables.


We would like to respond to the questions raised by Laupacis regarding our article on the importance of the quality of primary studies in producing unbiased systematic reviews. We have now reanalyzed our data using logistic regression, with pregnancy as the dependent variable, to assess the impact of study quality on treatment effect. The value of the β coefficient for the interaction between the variables "study quality" and "treatment allocation" provided an estimate of the amount of bias and was tested formally against the null hypothesis that the β coefficient is 0 or its exponent (OR) is 1.0. We observed that for trials of low quality the treatment effect was significantly higher than 1.0 (OR, 4.22; 95% CI, 1.42-12.50; P=.008), indicating that the treatment effect was overestimated by such trials. This relationship between studies of poor quality and exaggeration of treatment effect estimation has been reported by several other investigators

First Page Preview View Large
First page PDF preview
First page PDF preview