Table 2. Annual weight change effect when comparing cohort individuals matching trial participants randomized to either high or low protein intakea.
| Matching variables | Adjusted similar to trialb | Fully adjustedc | ||
| β (range) | p-value d (range) | β (range) | p-value d (range) | |
| Diet | 0.011 (−0.012; 0.034) | 0.807 (0.422; 0.999) | −0.004 (−0.027; 0.020) | 0.934 (0.543; 0.955) |
| Diet and BMI | −0.074 (−0.138; −0.031) | 0.155 (0.008; 0.557) | −0.090 (−0.152; −0.049) | 0.083 (0.003; 0.350) |
| Diet and WC | −0.063 (−0.103; −0.006) | 0.208 (0.042; 0.903) | −0.090 (−0.131; −0.032) | 0.076 (0.010; 0.530) |
| Diet and FMI | −0.074 (−0.125; −0.030) | 0.192 (0.027; 0.594) | −0.079 (−0.132; −0.040) | 0.166 (0.021; 0.482) |
| Diet, BMI and WC | −0.118 (−0.154; −0.051) | 0.025 (0.004; 0.331) | −0.142 (−0.180; −0.078) | 0.008 (<0.001; 0.142) |
Abbreviations: BMI, body mass index; E%, percent of energy intake; FMI, fat mass index; WC, waist circumference.
Five match-combinations: Diet only (Protein E%, carbohydrate E% and glycemic index), diet in combination with BMI, WC or FMI, or diet, BMI and WC. Multiple linear regression analysis was used. Exposure was indicator variables (yes/no) of matched randomization groups: high protein, high glycemic index, control. Outcome was average annual weight change between baseline and follow-up (mean 5.3 years). β = difference in body weight change (kg) between high and low protein group. β and p-values presented as means and summary statistics, respectively, complemented with corresponding ranges across matches of the ten random orders of the trial dataset.
Adjustment for BMI and sex (male/female).
Adjustment for sex (male/female without hormone use/female with hormone use), baseline BMI, age, physical activity (4 groups: inactive, moderately inactive, moderately active, active), education (4 groups: primary school, technical/professional school, secondary school, university degree) and intake of fibers (g/day) and alcohol (E%).
Summary p-values, derived from the means of the β-estimates and of the corresponding standard errors, respectively, over the ten individual matches.