# Table 3

Mean difference in serum glucose (mmol/L) at 2 hours post-sweetener consumption and overnight fast in all participants

Non-

caloric

0.05

0.98

(-1.24,3.25)

τ2 =0.65 (0.35,1.10)

0.16

(-1.46,1.80)

consistent

1.19

(-0.56,2.94)

0.07

(-1.45,1.61)

consistent

-0.37

(-2.07,1.29)

-

Sugar

alcohols

0.38

-0.83

(-2.66,1.03)

0.21

(-1.47,1.84)

-0.93

(-2.56,0.70)

consistent

-1.37

(-2.96,0.18)

consistent

-0.40

(-0.79,-0.01)

N = 1

-

Other

sugars

0.01

1.03

(-0.13,2.20)

-0.09

(-1.00,0.81)

consistent

-0.55

(-1.61,0.50)

consistent

-

-

-

Fructose

0.55

-1.12

(-1.95,-0.27)

consistent

-1.56

(-2.18,-1.02)

consistent

0.30

(-1.99,2.58)

N = 2 I2 = 0

τ2 = 0

0.41

(-2.44,3.26)

N = 1

-0.28

(-1.67,1.11)

N = 7 I2 = 84

τ2 = 1.72

(0.37,1.48)

-0.41

(-1.30,0.47)

N = 9 I2 = 11

τ2 = 0.17

(0.58,2.41)

Sucrose/

HFCS/

Honey

0

-0.45

(-1.15,0.21)

consistent

-

-2.20

(-10.46,6.05)

N = 3 I2 = 85

τ2 = 9.05

(2.94,32.22)

0.10

(-2.46,2.66)

N = 2 I2 = 0

τ2 = 0

-1.40

(-2.05,-0.74)

N = 23 I2 = 77

τ2 = 1.4 (0.68,1.50)

-0.31

(-0.53,-0.08)

N = 15 I2 = 0

τ2 = 0

(0,0.28)

Glucose

0

HFCS, high fructose corn syrup

The mixed evidence of the Bayesian network analysis are in the upper triangle and the direct evidence calculated using the REML estimate of τ2 are in the lower triangle. Sweeteners are reported in the expected order of efficacy[17] (with the exception of other sugars) from the expected lowest to highest 2-hour glucose response, with the estimated probability (or rank) listed in the diagonal. Each table cell contains the mean difference (MD) with the accompanying 95% confidence intervals. In the cells with direct evidence, we also list the number of studies, the I2 (percent of heterogeneity due to between-study heterogeneity) and τ2 (the between-study variance). Blank cells in the lower triangle indicate that no direct evidence was available. In the cells with mixed evidence, we list whether the mixed evidence was consistent with the available direct evidence. Also, in the first cell of the mixed evidence, we list the single τ2 estimate for the mixed evidence. Results are the MD of the expected higher-ranked sweeteners compared to the expected lower-ranked sweeteners (for example, MD of sugar alcohols versus sucrose is 0.41 and is in column 2, row 5 for the direct results, and is -0.93 and is in column 5, row 2 for the network analysis results). MDs less than zero favor the expected higher-ranked sweetener (smaller glucose response). For example, sugar alcohols show an increased serum glucose response by 0.41 mmol/L compared to sucrose using the direct evidence. However, sugar alcohols show a decreased serum glucose response by 0.93 mmol/L using the mixed evidence. However, since both confidence intervals include zero, neither analysis allows a confident judgment about which sweetener is preferable. Pooled evidence significant at P < 0.05 are presented in bold font. All nine mixed and direct results are consistent.

Wiebe et al. BMC Medicine 2011 9:123   doi:10.1186/1741-7015-9-123