Table 3.

Sensitivity and subgroup analysis based on the incidence of diabetes

	Group	No. of studies	No. of participants	RR	95% CI	Z	P (effect)	I 2	χ2	P (het)
Treatment duration	≥ 12 months	5	967	0.56	[0.45, 0.69]	5.23	< 0.00001	0%	1.62	0.81
	< 12 months	1	60	0.25	[0.03, 2.11]	1.27	0.20	–	–	–
Sample size	≥ 100	4	907	0.57	[0.45, 0.71]	4.85	< 0.00001	0%	1.37	0.71
	< 100	2	120	0.42	[0.21, 0.85]	2.42	0.02	0%	0.31	0.58
Use of placebo	Used	5	967	0.56	[0.44, 0.70]	5.00	< 0.00001	0%	1.94	0.75
	Not used	1	60	0.47	[0.23, 0.97]	2.04	0.04	–	–	–
Publication language	English	1	420	0.64	[0.44, 0.93]	2.33	0.02	–	–	–
	Chinese	5	607	0.50	[0.39, 0.66]	4.99	< 0.00001	0%	1.25	0.87

The incidence of diabetes was analyzed according to different criteria based on treatment duration, sample size, use of placebo, and publication language. RR is risk ratio, CI is confidence interval. Z and P (effect) evaluated the statistics of overall effect; I ² and P (het) were computed to assess heterogeneity