Gravar-mail: Impact of subject attrition on sample size determinations for longitudinal cluster randomized clinical trials