2.8. Statistical Analysis

Sample size was calculated on the basis of the results from Rondanelli et al. [9] in which a glucose reduction of −0.61 mmol/L with a 95% confidence interval (−0.89; −0.33) for 8 weeks of intervention was observed. Considering two balanced groups with 1:1 allocation (n1 = n2), an alpha significance level set at 0.01 and 90% of power in detecting differences between the groups, for an effect size d = 0.61, a sample size of 44 subjects in total (22 subjects per arm) was estimated. Differences between groups at baseline were investigated in each continuous variable using t-tests for independent data. To evaluate statistically significant changes over time for primary and secondary endpoints within and between the two groups, we fitted linear mixed models with time, group and the interaction time*group as fixed effect, including a random effect in the form of 1|subject in order to take into account intra-subject correlation produced by the repeated measurements [19]. Normality of residuals was assessed graphically and with Shapiro–Wilk test. All the models were adjusted for age, sex and BMI. Benjamini–Hochberg correction, fixing the false discovery rate (FDR) at α < 0.05, was used to account for multiple comparison [20].

Descriptive statistics are reported as mean ± standard deviation (SD). All analysis was performed on R 3.5.1 statistical software using the nlme and stats packages [21,22].