Forest plots that illustrate the data are shown for each of the meta-analysis packages, along with the fixed and random effects odds ratios at each stage. The numbers in the graph are slightly different than those found in Table 1 because some data entry errors were discovered at some time points and corrected only after some members reviewed the package. The difference in the numbers between Table 1 and Figure 1 are minor and would not be expected to alter the responses by our reviewers. To remain consistent and avoid confusion, we have provided only the corrected data in Figure 1, and the numbers provided to the reviewers in Table 1. The first error was in Rasmussen 1986 article where the proportions were entered as the raw numbers instead of the frequencies. This was corrected by the next package. A 1988 paper by Rasmussen that was a follow-up analysis based on some criticisms in letters to the editors was included in packages RCT 5 and RCT 10 but then omitted later on. Finally, the numbers for the Thorgersen 1993 paper were correct in the RCT 10 package but were incorrect in the RCT 20 package due to a transcription error when we switched software; this was corrected for the RCT23 package.
Shrier et al. BMC Medical Informatics and Decision Making 2008 8:19 doi:10.1186/1472-6947-8-19