Figure 4. Publication bias was assessed by inspection of funnel plots for mortality.

Dots was basically symmetrical distribution on both sides of dashed line, indicating that there was no obvious evidence of significant publication bias.