Blood pressure data were analyzed using a linear regression model in which change in BP (mean 6-month value − mean baseline) was regressed on indicators of the 2 behavioral interventions, indicators of clinical center and cohort, and baseline BP. In prespecified subgroup analyses (hypertensive and non hypertensive),
the models also included a main effect for the sub-group indicator and interactions between this indicator and the 2 treatment group indicators.
The effects of the interventions did not differ by clinical center (P=.62 for centertreatment
interaction for systolic BP and P=.54 for diastolic BP).