Table 4.

 Factors associated with prespecification of subgroup hypotheses in subgroup analyses: multivariable logistic regression

Study characteristics	Odds ratio (95% CI)	P value
High impact v lower impact journals	1.82 (0.78 to 4.21)	0.16
Non-surgical v surgical trials	3.50 (1.20 to 10.20)	0.022
Sample size per arm (fourths):
3-32	1 (reference)
33-101	3.10 (0.58 to 16.60)	0.19
102-301	4.73 (0.89 to 25.32)	0.069
≥302	6.53 (1.19 to 35.58)	0.03
No of prespecified primary outcomes	1.04 (0.74 to 1.46)	0.83
Statistically non-significant v significant primary outcome	1.48 (0.80 to 2.76)	0.21
Industry funding v no industry funding	0.49 (0.26 to 0.94)	0.032

Interaction of significance of primary outcome with funding was not significant (P=0.62), suggesting no differential associations in industry funded versus non-industry funded trials. This higher order term was therefore removed from the regression model.