Analysis of the Primary Aim
Analyses will be performed under an intent-to-treat design; therefore all participants, including those who drop out of the study will be invited back for assessment evaluations. Methods for analyses of treatment effects in pre-post clinical trials, in the context of missing data, have recently been compared in the statistical literature [49-55]. As is well known [55], complete case analysis and last observation carried forward are not optimal methods. Traditional ANCOVA models require stringent assumptions, e.g., homogeneous variances, and the accompanying software results in exclusion of subjects with missing data. Alternative mixed modeling software (SAS, PROC MIXED), used to compute full-information maximum likelihood (FIML) estimates, permits modeling of such assumptions, and the inclusion in the analysis of participants who do not complete the follow-up assessment (on an intent-to-treat basis). Thus, an ANCOVA model, using SAS PROC MIXED will be used to allow for more flexible modeling of assumptions, treatment of missing observations and inclusion of all subjects with at least one wave of data. Based on prior analytic experience with the outcome variables, it is not expected that transformations will be necessary. The post-treatment values of continuous outcomes will be modeled as functions of baseline values, treatment and the interaction of baseline and treatment.
Prior to analyses, baseline values of all variables from each arm will be examined; however, no p values will be provided, and covariates (other than baseline values) are not proposed for inclusion in the main analyses of treatment effects. Examination of baseline differences on key variables between completers and those lost-to-follow-up will be conducted in order to inform about the nature of the missing data. Methods of examining missing data, e.g., propensity scores, EM algorithm and multiple imputation sensitivity analyses will be considered if necessary. However, thus far missing data are minimal, and may not require modeling. Participation in the program sessions is good, and the majority of respondents attended all sessions. However, if necessary, secondary analyses will be conducted in order to investigate the impact of differential participation, stratifying the participants in the MINT-TLC condition based on their degree of participation and examining differences between strata on the outcome measures at follow-up