Sixteen studies (10 crossover design and 6 parallel design, n=396) were included.
Carbohydrate metabolism.
Compared with high-GI diets, low-GI diets were associated with a borderline statistically significant reduction in fructosamine overall (7 studies, n=171; WMD -10, 95% CI: -20, 0.00, p=0.05). However, a subgroup analysis of diabetic and healthy individuals showed non significant reductions. Low-GI diets were also associated with a statistically significant reduction in HbA1c in people with diabetes (WMD -0.27, 95% CI: -0.49, -0.05, p=0.01), based on 314 participants with diabetes in 8 studies. There was no evidence of statistical heterogeneity (p=0.89; I-squared 0%).
Lipid metabolism.
There was no statistically significant difference between low-GI diets and high-GI diets in HDL cholesterol (11 studies, n=316; WMD -0.04, 95% CI: -0.11, 0.02, p=0.2). There was evidence of moderate statistical heterogeneity (I-squared 37.5%). Subgroup analyses on patients with diabetes (8 studies), patients with CHD (1 study) and healthy individuals (2 studies) showed similar results. There was also no statistically significant difference between the two diets in LDL cholesterol (10 studies, n=308; WMD -0.15, 95% CI: -0.31, 0.00, p=0.06). Subgroup analyses on patients with diabetes (7 studies), patients with CHD (1 study) and healthy individuals (2 studies) showed similar results,
Compared with high-GI diets, low-GI diets were associated with a statistically significant reduction in total cholesterol (13 studies, n=392; WMD -0.33, 95% CI: -0.47, -0.18, p<0.001). There was no evidence of statistical heterogeneity (p=0.40; I-squared 4.8%). Subgroup analyses showed significant reductions in patients with diabetes (9 studies) and healthy individuals (2 studies), but not in patients with CHD (2 studies).
There was no statistically significant difference between low-GI diets and high-GI diets in triacylglycerol (14 studies, n=406; WMD 0.03, 95% CI: -0.12, 0.17, p=0.73). There was no evidence of statistical heterogeneity (p=0.93; I-squared 0%). Subgroup analyses showed similar results for patients with diabetes (10 studies), patients with CHD (2 studies) and healthy individuals (2 studies).