In or out? Methodological considerations for including and excluding findings from a meta-analysis of predictors of antiretroviral adherence in HIV-positive women.

AIM: This paper is a discussion detailing the decisions concerning whether to include or exclude findings from a meta-analysis of report of quantitative studies of antiretroviral adherence in HIV-positive women. BACKGROUND: Publication constraints and the absence of reflexivity as a criterion for validity in, and reporting of, quantitative research preclude detailing the many judgements made in the course of a meta-analysis. Yet, such an accounting would assist researchers better to address the unique challenges to meta-analysis presented by the bodies of research they have targeted for review, and to show the subjectivity, albeit disciplined, that characterizes the meta-analytic process. DATA SOURCES: Data were 29 published and unpublished studies on antiretroviral adherence in HIV-positive women of any race/ethnicity, class, or nationality living in the United States of America. The studies were retrieved between June 2005 and January 2006 using 40 databases. REVIEW METHODS: Findings were included if they met the statistical assumptions of meta-analysis, including: (1) normal distribution of observations; (2) homogeneity of variances; and (3) independence of observations. RESULTS: Relevant studies and findings were excluded because of issues related to differences in study design, different operationalizations of dependent and independent variables, multiple cuts from common longitudinal data sets, and presentation of unadjusted and adjusted findings. These reasons led to the exclusion of 73% of unadjusted relationships and 87% of adjusted relationships from our data set, leaving few findings to synthesize. CONCLUSION: Decisions made during research synthesis studies may result in more information losses than gains, thereby obliging researchers to find ways to preserve findings that are potentially valuable for practice.

Duke Scholars

Author Victor Hasselblad Biostatistics & Bioinformatics, Division of Biostatistics