Statistical analysis
Data were analyzed using R (Version 3.6.2). Continuous variables were
presented as mean ± SE and baseline differences among two exposure
groups were evaluated using a t-test. Frequencies were presented as
percentages and 95% confidence intervals (CI) and exposure groups
compared using a Z-test. Multiple linear regression analysis was used to
investigate the relationship between exposure groups and changes for
each outcome between Q4 and Q1 for weight, BMI, SBP, DBP, CMRF, and
between last and first measure for A1C measurements. Other confounding
variables were included in the model, the confounders were determined
based on the 10% rule [11]. To
account for the effect modification, a variable was also included if a
significant interaction term was observed in a model consisting of that
variable and the exposure group. To explore exposure effects on
clinically significant changes in cardiometabolic outcomes, a binary
logistic regression analysis (reduction of at least vs no reduction of
weight [5%], BMI [1 kg/m2], SBP [2 mmHg],
of A1C [0.3%]) was performed. Confounders and effect modifiers were
included, following the same rule as the multiple linear regression. A
p-value of <0.05 was considered statistically significant.