Include only randomized trials

Randomized trials provide the best means of obtaining an unbiased and fair comparison between medical treatments and the problems associated with non-randomized studies are well-documented [30]. As discussed in Chapter 4, prospective allocation by a method that ensures that treatment assignment cannot be known in advance, ensures that individuals cannot be selected to receive a particular intervention, and the process of randomization minimizes the likelihood of systematic differences between patient groups. Therefore, any observed differences should be due to differences in treatment and not to differences in the characteristics of the patients in the two groups. Therefore, most systematic reviews and meta-analyses of treatment interventions in cancer are restricted to 'properly' randomized trials, and any trials allocating treatment by quasi-random methods such as date ofbirth are excluded. Unfortunately, few published reports give adequate descriptions of randomization. For example, an examination of the first twenty RCTs reported in four leading general medical journals during 1987 found that only 40 per cent described the randomization process adequately [31]. Consequently, not all studies described in published reports as randomized can be assumed to be free of bias in the allocation of treatment. On further inquiry, it can turn out that 'randomization' has been done by birth-date, date of clinic visit, by alternate allocation, or by using pre-prepared lists with treatments correctly assigned randomly, but to which clinicians entering patients had free access. All of these could lead to bias. Furthermore, there is evidence that the method of concealment (how difficult it is to know or guess the next treatment to be allocated) is correlated with trial outcome, such that those trials with the most secure methods of allocation tend to give the least positive results [32]. It is therefore important that all systematic reviews ofRCTs should ensure that included trials are indeed randomized.

0 0

Post a comment