model to test for potential confounding: age, BMI, gender, smoking status
and number of BP medicines. Analysis by type of BP medication was not
meaningful because of small patient numbers in the subgroups.
Primary analyses were by intention to treat, including all available data
regardless of protocol deviations, and planned adjusted analysis, excluding
data points owing to BP medication change or participant’s noncompliance.
Tolerability was analysed descriptively and differences between the
groups assessed by w2 test. Differences in acceptability of the treatment
between the groups were assessed by Kruskal–Wallis test. Blinding success
was assessed by Fisher’s exact test for garlic versus placebo, and Kruskal–
Wallis test to ascertain differences between the garlic groups