Table 2.

Primary pooled analyses of the effect of fructose exchange for carbohydrate on blood lipids

Outcome	No. (studies)	No. (subjects)	Effect estimate		Heterogeneity
			SMD (95% CI)	P	χ2	I2 (%)	P
TG
Any diabetes	16	236	0.01 (−0.19–0.21)	0.9	36.75	59	0.001
Type 1 diabetes	4	54	−0.03 (−0.51–0.46)	0.91	10.96	73	0.01*
Type 2 diabetes	11	156	0.06 (−0.17–0.30)	0.61	23.22	57	0.01†
Total cholesterol
Any diabetes	14	172	−0.02 (−0.18–0.14)	0.79	40.96	71	<0.0001
Type 1 diabetes	2	14	0.17 (−0.32–0.67)	0.49	5.45	82	0.02
Type 2 diabetes	11	132	−0.08 (−0.26–0.10)	0.37	32.29	72	0.0002
LDL cholesterol
Any diabetes	7	99	0.02 (−0.07–0.11)	0.69	7.00	14	0.32
Type 1 diabetes	1	6	0.25 (−0.03–0.53)	0.08	—	—	—
Type 2 diabetes	6	93	−0.01 (−0.10–0.09)	0.86	3.97	0	0.55
HDL cholesterol
Any diabetes	12	164	0.02 (−0.05–0.10)	0.51	48.01	77	<0.00001‡
Type 1 diabetes	1	6	−0.01 (−0.46–0.44)	0.96	—	—	—
Type 2 diabetes	10	132	0.02 (−0.05–0.10)	0.53	47.95	81	<0.00001‡

Analyses were performed by the generic inverse variance method using fixed-effects or random-effects (if heterogeneity was significant at P < 0.10) models with paired analyses applied for crossover trials (12). —, no available data for subgroup analyses or, if there are no data for heterogeneity in the presence of a corresponding effect estimate, data only available from one study, precluding calculation of I² and the corresponding P value.

*Removal of Osei et al. (20) during sensitivity analyses explained heterogeneity (I² = 0%, P = 0.59).

†Removal of Osei and Bossetti (24) during sensitivity analyses explained heterogeneity (I² = 0%, P = 0.88 for all diabetes; I² = 0%, P = 0.76 for type 2 diabetes).

‡Removal of Pelkonen et al. (16) during sensitivity analyses explained heterogeneity (I² = 10%, P = 0.33).