Conclusion
• Meta-analyses don’t replace RCTs
• Good data more important than tests
• Beware of zoom effect
• Hypotheses need confirmation