Chapter 6 118 effects of magnesium supplementation were quantified with linear mixed-effect models specifying time, intervention and the interaction between time and intervention as fixed effects. The models were adjusted for age, sex and the baseline value of the outcome of interest. Our primary outcome was c-fPWV. First, we determined effects of magnesium citrate, magnesium oxide and magnesium sulfate versus placebo treatment by means of a 95% confidence interval derived from a linear mixed-effect model. Second, for the non-inferiority testing of magnesium oxide versus magnesium citrate and of magnesium sulfate versus magnesium citrate a 90% confidence interval was reported and a Δ of 0.5 m/s was used to determine non-inferiority 10. For the primary objective, the 4 groups were analyzed in one linear mixed-effect model. Similar models were analyzed for the secondary objectives, but without the placebo group. Secondary outcomes, including blood pressure, plasma magnesium and 24-h urinary magnesium excretion, other plasma and urinary minerals, and body composition were assessed using linear mixed-effect models specifying the same linear predictor. Because data on somatic and GI symptom severity were non-normally distributed, we performed MannWitney U tests to compare between-group differences at baseline, 12-wk and 24-wk. Findings from a recent meta-analysis prompted us to explore effects of magnesium citrate, magnesium oxide and magnesium sulfate supplementation (vs placebo) on c-fPWV in several subgroups, based on their mean baseline values 23. These subgroups included age (<63 y and ≥63 y), sex (m/f), BMI (<28 kg/m2 and ≥28 kg/m2), waist circumference (males: <102 and ≥102 cm; females: <92 and ≥ 92 cm), diuretic use (y/n), dyslipidemia (y/n), hypertension (y/n), and c-fPWV (<9 and ≥9 m/s). Subgroup analyses were also performed based on clinical relevant cutoff points (age <60 y and ≥60 y), BMI (<30 kg/m2 and ≥30 kg/m2), waist circumference (males: <102 cm and ≥102 cm; females: <88 cm and ≥88 cm) and c-fPWV (<10 m/s and ≥10 m/s) as recommended in the Expert Consensus document of 2012 22. Effect modification was investigated by adding interaction terms [intervention (0/1) × subgroup (0/1)]. Subgroup analyses were assessed at 24-wk. Statistical analyses were performed using SPSS 23.0 software for Windows (SPSS Incorporated, Chicago, IL, USA) and Rstudio version 1.1.383 (Vienna, Austria). Results Study subjects A Consolidated Standards of Reporting Trials flow diagram is shown in Figure 1. After screening, 164 individuals were eligible to participate in the study. The first participant was enrolled on March 27, 2018 and the last visit was performed on the June 4, 2020.
RkJQdWJsaXNoZXIy MTk4NDMw