Table 2.

Sensitivity analysis.

Sensitivity analyses	k	N	Hedges' g	95% CI	p-value	Heterogeneity
Primary meta-analysis	12	431	−0.47	−0.71 to −0.24	p < 0.0001	Q = 15.06, df = 11, p = 0.18, T2 = 0.0313, I2 = 27%
• Random allocation: low-risk of bias	10	379	−0.46	−0.74 to −0.17	p = 0.0018	Q = 14.9, df = 9, p = 0.09, T2 = 0.0754, I2 = 39.6%
• Concealed allocation: low-risk of biasa	2	105	−0.10	−0.49 to −0.28	–	–
• Blinding of assessors: low-risk of biasb	1	26	−0.62	−0.1.41 to 0.17	–	–
• Completeness of follow-up: low-risk of bias	6	237	−0.50	−0.83 to −0.18	p = 0.0022	Q = 8.28, df = 5, p = 0.14, T2 = 0.0293, I2= 39.6%
• Intention-to treat analysis: low risk of bias	5	257	−0.45	−0.77 to −0.13	p = 0.0063	Q = 7.7, df = 4, p = 0.1, T2 = 0.0347, I2 = 48.1%
• Only peer-reviewed articles	7	308	−0.56	−0.90 to −0.23	p = 0.0011	Q = 11.48, df = 6, p = 0.07, T2 = 0.0839, I2 = 47.7%

k number of effect size estimates, N number of participants, Hedges' g pooled effect size estimate, CI confidence interval.

^aThe two included studies rated low-risk of bias for “concealed allocation” are Carter et al. (2015) and Jeong et al. (2005).

^bThe only included study rated low-risk of bias for “blinding of assessors” is Hughes et al. (2013).