I read with great interest the meta-analysis by Parise et al comparing intravascular ultrasound–guided and angiography-guided percutaneous coronary intervention. The investigators concluded in the text of the abstract that intravascular ultrasound “guidance for bare metal stent implantation improved the acute procedural results (angiographic minimum lumen diameter) and thereby reduced angiographic restenosis and repeat revascularization and major adverse cardiac events.” However, visual inspection of the funnel plots for restenosis, repeat revascularization, and major adverse cardiac events (Figure 7) reveals asymmetry. In fact, the investigators stated in the text of the “Results” section that “the Begg rank correlation test was statistically significant for [major adverse cardiac events] (p = 0.02) and restenosis (p = 0.04) and was borderline for repeat revascularization (p = 0.10), indicating some potential bias for these variables.”

Therefore, the trim-and-fill method was used to obtain symmetry in the funnel plot and to determine the impact of hypothetical or imputed studies on the pooled estimate (R version 2.12.2). In the analysis, the asymmetric studies from the left-hand side of the funnel plot are trimmed to locate the unbiased effect, and then the plot is filled by reinserting the trimmed studies on the left as well as their imputed counterparts to the right of the mean effect, producing a symmetric funnel plot. The analysis incorporating 3 hypothetical or imputed studies showed that any pooled estimates were no longer significant (for restenosis: odds ratio 0.82, 95% confidence interval 0.39 to 1.73; for revascularization: odds ratio 0.64, 95% confidence interval 0.41 to 1.02; for major adverse cardiac events: odds ratio 0.79, 95% confidence interval 0.46 to 1.33). Publication bias could explain, in part, the potential benefit of intravascular ultrasound guidance.