Cross design synthesis : a new strategy for medical effectiveness research : report to Congressional requesters / United States General Accounting Office.

  • United States. General Accounting Office.
Date:
[1992]
Catalogue details

Licence: Public Domain Mark

Credit: Cross design synthesis : a new strategy for medical effectiveness research : report to Congressional requesters / United States General Accounting Office. Source: Wellcome Collection.

    20/128 (page 18)
    Randomized Studies Chapter 1 Introduction and Review of Study Designs for Evaluating Treatment Effectiveness treatments in use today (Miké, 1982). Logically, however, the difference between the average outcomes cannot be attributed to the treatments received unless the patients in the current study and the patients who historically received the alternative treatment are equivalent. A historical control trial typically cannot ensure the equivalence of these groups. One reason, among others, is that patients who are selected for the current study (and who agree to “try out” the new treatment) may represent a special group; these patients, then, would not be comparable to the more typical patients used as historical controls (see Pocock, 1983; Gehan, 1984; Sacks et al., 1983). Numerous other evaluation designs, termed “quasi-experiments” (see Cook and Campbell, 1979), include the one-group pretest-posttest design, the regression-discontinuity design, and many, many others. Despite differences in these designs (and in their rigor), they share with historical trials an inability to ensure the isolation of the treatment effect. The 1920s brought a significant breakthrough in controlled trials: the randomized design (R.A. Fisher, 1925, 1935° ). Relying on a chance process (randomization) for assigning current patients to two alternative treatments ensured that the only source of differences between the two groups, at baseline, would be chance.® Thus, one could assess the likelihood that an observed difference in outcomes occurred because of the difference in treatments rather than chance. Given an objective outcome measure, one could be relatively sure of the validity of this assessment. Strictly random procedures preclude the possibility that investigators would (even unconsciously) assign the healthier patients to receive the new treatment or usual care. In addition, there is a statistical expectation of equivalence in the two groups. Probability theory allows investigators to estimate the size of the average difference in outcomes that might occur by chance alone—provided that a random process was used to assign treatments to individual patients. Thus, only a difference in outcomes that is larger than would be expected on the basis of chance alone will be interpreted as a statistically significant indicator of a treatment effect. ®The development of the randomized design has also been linked to earlier work by Neyman (1923), given a newly translated version of Neyman’s 1923 theoretical work (see Speed, 1990; Neyman, 1990; Rubin, 1990b). *Randomization of treatment assignment can be accomplished by flipping a coin, drawing lots, using a table of random numbers or using an appropriate computer program.