Dear IMPUTE, This is a colleague's question that I recently posted on allstat and s-news. I only received two responses and the question seems appropriate to this group so I will try again. Thanks in advance and apologies to members of the other mail groups.
Hi, I am a statistician for a clinical trial with up to 6.5 years of follow-up with a death rate of approximately 15% per year, and a loss-to-follow-up rate for other reasons of approximately 7% per year. In addition to evaluation of mortality, a key objective of the study is to assess the impact of the randomized treatment interventions on changes in continuous parameters which are measured at regular intervals while the patient remains in follow-up. Existing methods for incorporating informative censoring in mixed effects models appear to be appropriate for dealing with censoring for causes other than death, but seem problematic for high rates of censoring due to death itelf. The problem is that most methods involve estimation of parameters attributed to the full population of patients who started the trial. This seems highly artificial when a high percentage of the patients are dead by several years into the study. Does anyone have experience with methods for longitudinal data which treat censoring by death differently than other types of censoring? -Brett Brett Larive [email protected] Dept of Biostatistics/Wb4 phone: 216-444-9925 Cleveland Clinic Foundation fax: 216-445-2781 9500 Euclid Avenue Cleveland, OH 44195
