Table 5.

Pooled odds ratio for outcome reporting bias (fully reported v incompletely reported outcomes) in randomised trials published in December 2000 and cited in PubMed, stratified by study design and sensitivity analyses

	Efficacy outcomes		Harm outcomes
Trial population	No of trials*	Odds ratio (95% CI)†	No of trials*	Odds ratio (95% CI)†
All trials	161	2.0 (1.6 to 2.7)	43	1.9 (1.1 to 3.5)
Parallel group trials	135	1.9 (1.4 to 2.6)	33	3.1 (1.6 to 5.8)
Crossover trials	22	2.7 (1.4 to 5.3)	9	0.32 (0.10 to 1.1)
Excluding survey non-responders	114	2.1 (1.6 to 2.9)	31	1.4 (0.66 to 3.0)
Excluding physiological or pharmacokinetic trials	148	1.9 (1.4 to 2.5)	38	2.4 (1.3 to 4.4)
Fully or partially reported outcomes v qualitatively or unreported outcomes	134	3.2 (2.4 to 4.2)	38	4.2 (2.5 to 7.0)

>1 means that fully or partially reported outcomes have higher odds of being significant (P<0.05) than qualitatively or unreported outcomes.

^*Trials excluded if odds ratio could not be calculated because of empty rows or columns in 2×2 table (see methods for details).

^†Except for the bottom row, odds ratio >1 signifies that fully reported outcomes have higher odds of being significant (P<0.05) than incompletely reported outcomes. In bottom row, odds ratio