April 29-30, 2003
National Institute on Drug Abuse
A 1998 study by the Institute of Medicine determined that despite the availability of efficacious behavioral treatments, established research-based treatments have not been adopted widely in clinical practice. The gap between research and practice is especially evident with regard to group therapies for substance use disorders. Most substance abuse treatment in community settings is delivered in a group format, yet there is little empirical evidence to inform the development or delivery of group therapies. The unique challenges of conducting group therapy research may account for the observed disconnect between clinical science and practice. Some of the challenges inherent in group work are constituting one or more groups in the context of a research project, managing fluctuations in group membership over the course of treatment, accounting for heterogeneity in group members responses to treatment, choosing appropriate levels of analysis for group treatment data, and identifying key mechanisms of action of group treatment.
As part of NIDAs ongoing efforts to develop behavioral treatments for substance abuse that are both efficacious and community-friendly, a science meeting on group therapy research was convened. The purposes of the meeting were to clarify the state of the science regarding group behavioral treatment, and to discuss the challenges of conducting group therapy and potential solutions to these challenges. These discussions were intended to inform future initiatives on group therapies, and to enhance technical assistance provided to investigators proposing group therapy research studies.
On April 29 and 30, 2003, the National Institute on Drug Abuse convened a meeting to discuss the state of the science of group therapy research for drug abuse and dependence. The meeting brought together experts on group behavioral treatment research, both for substance-related disorders, and for other psychiatric disorders with potential relevance to the treatment of substance abuse and dependence. Over the course of the 1 _-day meeting, the participants evaluated the existing empirical literature regarding group treatment for substance abuse and dependence, highlighted a number of cutting-edge group therapy research projects, identified the major challenges posed by group treatment research, and discussed potential solutions to these challenges. Through funding initiatives such as targeted RFAs and NIDAs ongoing Behavioral Therapies Development Program Announcement (see: http://grants1.nih.gov/grants/guide/pa-files/PA-03-126.html), and via dissemination of the outcomes of this meeting, NIDA hopes to advance further a program of research on group behavioral treatment.
Overview of Research on Group Therapy for Substance Abuse and Dependence
A context-setting presentation on published studies of group treatments for substance abuse and dependence acknowledged that, while group therapy is the most commonly-used treatment modality in community drug treatment settings, fewer than 20 controlled studies of group treatment for substance abuse had been published as of this meeting. Those studies that had been published addressed such a wide range of populations, target drugs of abuse, and types and intensities of group treatment, that it is difficult to draw conclusions about the efficacy of group treatment for drug abusers.
Cutting-Edge Group Therapy Research Projects
Following a brief review of existing literature on group therapy for substance abuse and dependence, presenters described recent or ongoing research projects focused on group therapy. These presentations highlighted the wealth of lessons learned through conducting group therapy research in areas outside the substance abuse field. Additionally, they generated discussion of a very basic question: When should a group therapy project (as opposed to a different treatment modality) be conducted?
When should a group therapy project be conducted? The participants offered two answers to this question. The first emphasizes a theoretical or clinical reason to expect that a group format would be beneficial. Examples in which a group format might be beneficial are group therapy for social phobia, in which group interaction in itself may address the presenting problem (e.g., research presented by Dr. Heimberg), and treatment for adolescents, a developmental stage in which the influence of peers is especially important (e.g., research presented by Dr. Winters).
The second answer offered to this question emphasized the aim of developing treatments that are community-friendly, because the majority of community treatment occurs in a group format, and the assumption is that group therapy is cost-effective. While a group format seems intuitively less costly, there may be tradeoffs in effectiveness or other expenditures such as therapist training, supervision, and crisis management. It seems clear that more work needs to be done in establishing that group treatment is indeed cost-effective. Also, although group treatment appears to be the default cost-effective alternative to individual or family treatment, further research on other potentially cost-effective modalities (e.g., web-based interventions, self-paced instruction, etc.) may provide other options. Finally, if one goal of research is to develop streamlined, cost-effective treatments, it is vitally important to begin to identify and test the mechanisms of action of group therapies. Identifying therapy mechanisms of action allows for preserving those components of treatment that contribute most to outcomes when moving the treatment into community settings.
Challenges in Conducting Group Therapy Research and Potential Solutions
Closed vs. rolling groups. The meeting participants identified several major logistical and methodological challenges in conducting group therapy research. Among the most challenging logistical issues discussed was whether to constitute closed or rolling groups. In closed groups, members are enrolled before treatment begins, and group membership stays fairly constant. In contrast, for rolling groups, membership changes as new members join and previous members graduate. In constituting closed groups, subject flow must be adequate to form a group in a reasonable time frame. Because risk of drop-out increases with increasing wait-times for treatment, it may be necessary to implement strategies to retain subjects. In addition, some disorders may require treatment as soon as possible, so that waiting for a closed group to fill before beginning treatment may not be clinically appropriate. In such cases, researchers might consider a baseline treatment during the waiting period. Of course, introducing additional treatment has the potential of confounding the effects of the treatment under study. One way to minimize confounding may be to offer a baseline treatment that is unrelated to the treatment under study. Also for closed groups, if too many members drop-out, this may have a deleterious effect on the remaining group members. The question of how many group members are required for effective treatment is an important one.
While constituting rolling groups avoids some of the logistical problems with wait-listing patients, it generates other problems related to group heterogeneity. Group heterogeneity is a problem to the extent that the composition of the group affects treatment process and/or outcome. For instance, if group norms are an important mediator of therapeutic outcomes, and group norms change with changing group membership, outcome for any single group member probably depends on some combination of the group norms that member experienced. Statistical strategies may help to manage such complexity. One such strategy is to treat each change in group membership as a new group, so that the number of groups becomes another variable in analyses. Another strategy is to model the hypothesized key elements of group composition over time, and account for these in the analyses. For example, if group norms are thought to mediate outcome, assess group norms over time, and add group norms (or change in group norms) to equations predicting outcome.
The non-independence of group members. Another methodological challenge of group therapy research is accounting for the non-independence of group members. By definition, group members in the same group participate in the same therapy, which implies some level of interdependence. However, the experiences of individual group members in these shared group sessions are probably varied, so that the level of interdependence may be small. In general, interdependence is less of a problem in small groups, i.e., N < 6. In cases where individual group members are thought to have a great influence on other members, it may be possible to quantify the effect of those individual(s), so that change in the group can be tracked by change in that individuals behavior. For instance, the catalyzing effect of a charismatic group organizer, or the intimidating effect of a so-called predator, can be quantified and modeled in predictive equations. Hierarchical linear models (HLM) or multi-level analytic strategies can account for both within-group and between-group variables, making it possible to model both those factors that make group members interdependent, and factors that distinguish one group from another. Also, statistically, the non-independence of group members influences the significance of any effect found, rather than the size of the effect. This implies that interdependence is likely to lead to premature acceptance of findings as significant, and that appropriate cautions could be taken in setting a compelling statistical significance level.
The unit of analysis in group therapy research. Another related issue has to do with the unit of analysis in group therapy research. Data collected from individual group members can be transformed into sums, averages, a measure of group variability, and in other ways. In cases where group members are highly-interdependent, it may make sense to consider the group, rather than the individual, as the unit of analysis. (But preserving the individual level of analysis may be necessary for answering some research questions, such as identifying moderators of treatment effect.)
Randomization in group therapy research. Another methodological issue discussed by the meeting participants related to appropriate strategies for randomization procedures. Individuals can be randomly-assigned to different group treatments, or groups can be constituted and then randomized to treatment conditions. The latter may be especially appropriate if there are clinical indications for constituting a group of a particular composition (e.g., single-gender, day vs. evening schedules, presenting problems, etc.).
Controlled group therapy experiments. The meeting participants also discussed the possibility of conducting group therapy research through controlled, pseudo-laboratory studies similar to traditional social psychology experiments. As in some social psychology studies, confederates acting as group members might increase the degree of control in a group therapy study.
Recommendations for Future Group Therapy Research
The meeting concluded with a lively discussion about recommendations for future research in group therapy. The following are some of those recommendations:
- Encourage new investigators to develop group therapy projects
- Support translational research, e.g., from social psychology, business schools looking at work teams, behavior analysts, etc.
- Conduct secondary data analyses using new statistical strategies that account for group data
- Brainstorm about methodologies for answering good questions with good methodologists
- Include in group therapy research hypotheses directly related to the group aspect of the project
- Include questions in research that address the unique challenges of group therapy, such as how one makes a rolling group work, and how to aggregate group data
- When proposing to develop or manualize a group therapy (i.e., Stage I research), be clear about when iterations to the manual will be made
- For review committees, demonstrate the feasibility of doing group therapy research, given the many challenges inherent in the work
- Study the mechanisms of action of group therapy
- Let the research question guide the study design and analyses, not the other way around