Table 2.
The type 2 diabetes–glycemic index dose-response risk relation in studies making study-level adjustments for specific nutrients. Studies with CORR > 0.55 were included in this analysis.
| Study-Level Adjustment Made within Studies | Number of Studies | Model a | Mean Combined Relative Risk and (95%CI) | p-Value for RR | Inconsistency I2 |
Heterogeneity τ 2 |
p-Value for τ 2 and I2 | ||
|---|---|---|---|---|---|---|---|---|---|
| Per 10 units GI | (%) | (Per 10 | |||||||
| units GI)2 | |||||||||
| All studies with CORR > 0.55 bar one outlier [40] b | 10 | Random | 1.27 | (1.15–1.40) | <0.001 | 68 | 0.013 | <0.001 | |
| All fiber types c | 7 | Random | 131 | (1.17–1.47) | <0.001 | 76 | 0.0138 | <0.001 | |
| Cereal fiber d | 3 | Random | 1.39 | (1.26–1.53) | <0.001 | 45 | 0.0034 | 0.161 | |
| Vegetable fiber e | 1 | Random | 1.50 | (0.95–2.36) | 0.080 | - | - | - | |
| Magnesium f | 2 | Random | 1.28 | (0.94–1.79) | <0.001 | 69 | 0.0355 | 0.070 | |
| Protein g | 3 | Random | 1.39 | (1.26–1.53) | <0.001 | 45 | 0.0034 | 0.161 | |
| Red meat h | 1 | Random | 1.47 | (1.35–1.60) | <0.001 | - | - | - | |
| Alcohol i | 9 | Random | 1.26 | (1.13–1.40) | <0.001 | 71 | 0.0132 | <0.001 | |
| Energy j | 8 | Random | 1.27 | (1.14–1.40) | <0.001 | 74 | 0.0135 | <0.001 | |
| Saturated fats k | 3 | Random | 1.28 | (1.07–1.48) | 0.007 | 53 | 0.0110 | 0.012 | |
| Trans fats l | 3 | Random | 1.39 | (1.26–1.53) | <0.001 | 45 | 0.0034 | 0.161 | |
a Model procedures: (i) Analysis of doses response, (ii) random effects meta-analysis of the dose-response results. b Values from Table 1. Studies were: Barclay et al. (2007) [68], Bhupathiraju et al. (2014) (2 studies: NHS II, HPFS) [10], Hodge et al. (2004) [69], Mekary et al. (2011) (NHS I) [40], Oba et al. (2013) (men) [44], Sahyoun et al. (2008) [70], Sakurai et al. (2012) [71], van Woudenbergh et al. (2011) [72], and Villegas et al. (2007) [46]. The study of Simila et al. (2012) [41]) was an outlier (p = 0.033) (Table 1, footnote c). c Combined from Villegas et al. (2010) [46], Sakurai et al. (2011) [71], Oba et al. (2013) in men [44], Mekary et al. (2011) [40], Bhupathiraju et al. (2014) (2 studies; NHS I and HPFS) [10], and Barclay et al. (2007) [68]. d Combined from: Mekary et al. (2011) [40] and Bhupathiraju et al. (2014) (2 studies, NHS II and HPS) 10]. e Study was that of Barclay et al. (2007) [68]. f Combined from Oba et al. (2013) for men [44] and Schulze et al. (2004) [45]. g Combined from those studies cited in footnote d. h Mekary et al. (2011) [40]. i Combined studies of footnote b except Barclay et al. [68]. j Combined studies of footnote b except Barclay et al. [68] and Sahyoun et al. (2008) [70]. k Combined studies from Bhupathiraju et al. (2014) (2 studies, NHS II and HPS) [8] and van Woudenbergh et al. (2011) [72]. l Combined from: Mekary et al. (2011) [40] and Bhupathiraju et al. (2014) (2 studies, NHS II and HPS) [8]. Abbreviations: CI confidence interval; HPS, Health Professionals’ Study; NHS, Nurses’ Health Study; RR, relative risk; I2, inconsistency among studies, which is the ratio of the among-studies variance (τ 2) to the sum of the among-studies and within-studies variances expressed as a percentage.