Page 82 - ESHRE2019
P. 82
Exploring inconsistency
• Detect heterogeneity
– direction of the effect is not a consideration! – variation in effect
– overlap in confidence intervals
– test for heterogeneity (p<0.05), I2 statistic
• Cautious exploration of possible causes –differences in patients?
– interventions?
– outcome definitions? –methodological diversity?
• NOTE: random-effects model does not explain or resolve heterogeneity
Inconsistency
80
77
