Large multi-site followup studies of treatment clients comprise a major source of evidence on the effectiveness of substance abuse treatment in the United States. This report considers challenges to the validity of results in treatment effectiveness studies due to four factors: (1) nonrandom entry to treatment from the population needing treatment; (2) nonrandom selection of treatment providers by researchers for follow-up studies; (3) non-cooperation of the selected treatment providers with the research protocol; and (4) incomplete enrollment and follow-up of clients treated by the cooperating providers. Both kinds of selection/entry and both kinds of non-response are potential sources of bias, so both should be major concerns in the design and analysis of substance abuse treatment follow-up surveys. The response characteristics of four large-scale multi-site studies carried out in the early and mid 1990s are summarized and assessed on a comparative basis. Recommendations are provided for minimizing bias in future studies, and statistical methods are proposed for evaluating biases due to the process of entry into treatment and non-response. An appendix describes the National Treatment Improvement Evaluation Study and Center for Substance Abuse Treatment Demonstrations. (Contains 7 exhibits and 52 references.) (Author/SLD)