DES JARLAIS ET AL. RESPOND

Abstract

First, we would like to note our appreciation of Dzewaltowski and colleagues’ overall support for the TREND statement. There seems to be strong agreement that standards for the reporting of nonrandomized evaluations are greatly needed and that the TREND statement is an important step in this direction. However, the TREND statement is still clearly a work in progress. Dzewaltowski et al. suggest that the cost of the intervention be included in the TREND reporting items. Malcolm Potts (written communication, March 9, 2002) has also suggested inclusion of cost data. At the time of the TREND journal editors’ meeting, other groups were working on reporting standards for costeffectiveness studies, and the TREND group decided to wait until those had been completed before drafting items for reporting costs. We do agree that cost information is of great importance and should be included in evaluation reports. Most of Dzewaltowski et al.’s comments concern the issue of external validity, including the representativeness of the settings and of the intervention staff. The first version of the TREND statement was developed as a counterpart to the CONSORT statement,1 which mainly focuses on issues of internal validity. Items 3 (description of the settings in which the intervention was studied) and 22 (external validity as a topic in the discussion section) do address external validity, but we agree that much more is needed to develop standards for the reporting of external-validity data. External-validity issues are of great importance for behavioral and public health interventions; an intervention that is highly effective in some settings may be completely ineffective in other settings. It will be critical to extend the TREND ideas to research studies that specifically focus on externalvalidity issues, such as studies that systematically examine the effectiveness of an intervention in different settings. We can envision an additional set of reporting standards for addressing external validity, regardless of whether the original internalvalidity evaluation study used a randomized or a nonrandomized design. We invite all persons interesting in contributing to this discussion to e-mail the TREND group (vog.cdc@DNERT) and to visit the TREND Web site (http://www.TRENDstatement.org). We plan to create linkages from the TREND Web site to the Web sites of other groups interested in these issues, such as Dzewaltowski and colleagues’ RE-AIM (Reach, Efficacy/Effectiveness, Adoption, Implementation, Maintenance) group (http://www.re-aim.org).