A systematic review examines the effect of drug X, a commonly used medication, on all-cause mortality in patients with a chronic pulmonary problem. A systematic search identified 12 retrospective cohort studies. The pooled odds ratio of mortality with drug X use was 0.72 (95% confidence interval: 0.61-0.79). The authors also reported the following funnel plot of the 12 studies, centered around the log of the pooled odds ratio: The graphical findings suggest the presence of which of the following?
A)Lead-time bias
B)Placebo effect
C)Publication bias
D)Small samples in drug X treatment arms

Question

A systematic review examines the effect of drug X, a commonly used medication, on all-cause mortality in patients with a chronic pulmonary problem.  A systematic search identified 12 retrospective cohort studies.  The pooled odds ratio of mortality with drug X use was 0.72 (95% confidence interval: 0.61-0.79).  The authors also reported the following funnel plot of the 12 studies, centered around the log of the pooled odds ratio:  The graphical findings suggest the presence of which of the following?
A)Lead-time bias 
B)Placebo effect 
C)Publication bias 
D)Small samples in drug X treatment arms

Quizplus · Accepted Answer

Explanation: A funnel plot is helpful in assessing publication bias. Each study's treatment effect (on the x-axis) is plotted against a measure of that study's size or precision, usually using the standard error of the treatment effect (on the y-axis). The triangle is centered on a summary estimate of the treatment effect (eg, pooled estimate of odds ratio, usually in log scale) with its sides corresponding to standard errors. If there is no bias, any scatter between the study results should be due to sampling variation, and 95% of studies should lie within the triangle centered on the summary estimate and extending 1.96 standard errors on either side. Because a larger sample size is associated with increased precision, larger studies (more powerful) will be at the top and have a narrow spread whereas small studies will be scattered widely at the base of the triangle. Funnel plots should be symmetric in the absence of study heterogeneity and publication bias. In this figure, however, there are no studies on the right side of the funnel. This asymmetry suggests publication bias (ie, studies showing null/increased effect of drug X on mortality are noticeably absent as they are less likely to have been published). Other possible explanations for asymmetry include heterogeneity, methodological anomalies, artifact, or chance. (Choice A) Lead-time bias occurs when several disease interventions are compared and one of the interventions diagnoses the disease earlier than the others without an effect on the outcome (eg, survival). This would make it appear that the intervention prolonged survival whereas in actuality it only diagnosed the disease sooner. This funnel plot does not help to assess lead-time bias. (Choice B) The placebo effect is the phenomenon whereby a patient's symptoms are alleviated by an otherwise ineffective treatment, most likely as a result of the fact that the individual expects that the treatment will work. This funnel plot does not help to assess placebo effect. (Choice D) A small sample size (low number of patients) in the drug X treatment arms would lead to more studies at the base of the triangle because such studies would have lower power and larger standard errors. Educational objective: Asymmetry in a funnel plot suggests publication bias. References: Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. (http://www.ncbi.nlm.nih.gov/pubmed/21784880) Bias in meta-analysis detected by a simple, graphical test. (http://www.ncbi.nlm.nih.gov/pubmed/9310563)

A Systematic Review Examines the Effect of Drug X, a Commonly