Factors associated with early study discontinuation in AACTG studies, DACS 200

Background:When participants in a clinical trial are lost to follow-up, the validity of the results and resource utilization may be seriously affected. Identification of factors associated with loss is essential to inform efforts to minimize early study discontinuation (ESD). Few studies examine subject-specific factors that predict ESD. This large, multi-study, multi-center database allows not only evaluation of the effect of subject-specific factors on ESD, but also of study-specific factors as well as the site of enrollment.Methods:We retrospectively examined the retention of 5605 study entries of 5037 individuals who were enrolled at 30 sites within the National Institute of Allergy and Infectious Disease Adult AIDS Clinical Trials Group (AACTG) from July 1, 1996 through June 30, 1999. Variables examined included: enrollment year, prior enrollment on a study, sex, race, drug use, CD4 count, age, and use of antidepressants, other prescription psychotrophic drugs, or methadone; site location and the proportion of primary care given at a site; and study complexity and type (immunology, complications of HIV, or antiretroviral). Prior antiretroviral experience and baseline viral load were examined in antiretroviral studies.Results:The ESD rate for sites varied considerably from 5.70 to 33.10 per 100 person years of follow-up. Variables associated with higher ESD included complexity of the study, no previous enrollment on a study, black race, female sex, previous intravenous drug use, CD4 < 100, age < 30, sites which did not deliver primary care, no use of tricyclic antidepressants or lithium and use of methadone. After adjustment for all significant variables and their interactions, some sites with hard-to-retain subjects do not lose them, and some sites with populations with low potential for dropout do lose them at a high rate. Follow-up remediation and studies should focus on the sites themselves as well as efforts to retain specific subpopulations.