Systematic review of methods for individual patient data meta- analysis with binary outcomes

Table 3 Statistical analysis method categorized by overall strategy among 26 IPD meta-analyses of binary outcomes

Analytic approach¹		n/N (%)
One stage approach (n = 19)
Ignored clustering by study	Logistic regression	5/19 (26%)
Fixed effects	Logistic regression	4/19 (21%)
Random effects	Logistic regression	10/19 (52%)
	Fixed study effect with random treatment effect²	1/10 (10%)
	Random study effect with fixed treatment effect²	2/10 (20%)
	Random study effect with random treatment effect²	2/10 (20%)
	Unclear¹	5/10 (50%)
Two stage approach (n = 6)
Fixed effects	Unspecified	2/6 (33%)
Fixed effects	Cochrane-Mantel-Haenszel	1/6 (17%)
Random effects	Der Simonian Laird	2/6 (33%)
Random effects	Unspecified	1/6 (17%)

¹It was unclear from one article [38] which approach was used.
²Among studies that used random effects logistic regression, where the intercepts and/or treatment effects allowed to vary across studies.

ISSN: 1471-2288