Linear mixed modeling (Proc MIXED; SAS Institute, 1999) was used to analyze the continuously scaled dependent variables over time as a function of treatment condition. The mixed modeling procedure employs maximum likelihood estimation to calculate parameter estimates, thus taking advantage of all data collected without imputing missing data. In these analyses Treatment condition was treated as a fixed effect. Time (measured in months since baseline) was treated as a fixed, repeated effect, and the intercept was included as a random effect. An autoregressive (ar1) error covariance structure was adopted on the basis of accepted fit criteria (-2RLL, AIC; Judge et al., 1985). Post-hoc contrasts were used to detect between-treatment differences if Treatment X Time interactions appeared.
