ABSTRACT
Often in epidemiological and health-service research settings, a variety of different outcome measures and instruments are used. An example is assessing quality of life. Heterogeneity of research instruments and measures poses problems when the quantitative synthesis of a number of individual study results is required (Greenland, 1987; Jones, 1992). The use of standardized effect sizes has been advocated as a means of overcoming this diversity of instruments, especially in a comparative setting. A further complication is the fact that individual studies may report their results in a manner which makes calculation of a standardized effect size problematic, or even impossible. Ideally, individual-patient data could be obtained, so that re-analysis of the studies could be undertaken to calculate the standardized effects in the individual studies (Stewart and Parmar, 1993). However, the process of obtaining individual-patient data is laborious, and frequently infeasible or impractical. Indeed, it may even yield a further complication, in that only individual-patient data for a proportion of the studies in the meta-analysis may be obtained (Sutton et al., 1998; Sutton et al., 1999). We therefore consider a variety
of methods for the synthesis of heterogeneously reported studies, using a variety of outcome measures. In a similar manner, Dominici and Parmigiani (Chap. 3) discuss the synthesis of studies in which the same outcome measure has either been reported on a continuous or a dichotomous scale.
