Clinical Trials with Missing Data by O & #39;Kelly Michael Ratitch Bohdana & Bohdana Ratitch

Author:O & #39;Kelly, Michael, Ratitch, Bohdana & Bohdana Ratitch
Language: eng
Format: epub
ISBN: 9781118762530
Publisher: Wiley
Published: 2013-01-23T16:00:00+00:00

We will see in Sections 6.3 and 6.4 how to obtain overall results with multiply-imputed data, but for now, Table 6.2 provides a quick snapshot of how the differences in the imputation methods used above affect the final analysis results at the last time point (Week 8). This table presents least squares mean (LSM) estimates of the difference between experimental and placebo arms for the change from baseline to Week 8 in TST in the insomnia example obtained from an analysis model applied to the imputed data. As would be expected after the examination of the imputation plots in Figure 6.1, the second imputation method produces a slightly smaller estimate for the difference between the experimental and placebo arms for the change from baseline to Week 8 in TST compared to the first method (17.3 vs. 18.7 min). The third method (including MSS assessments as ancillary variables), results in a slightly larger estimate (19.3 min) compared to the other two methods. The differences between the methods in this case are rather insignificant, especially given the fact that the p-values are quite small and are unambiguously below the significance threshold of 0.05 for all three methods. But it illustrates the point that the imputation model affects the results, and thus should be fully pre-specified in advance in order to avoid speculation in case of results that are border-line significant.

Table 6.2 LSM estimates of the difference between experimental and placebo arms for the change from baseline to Week 8 in TST in the insomnia example. Imputation Methods 1, 2 and 3 represent the methods shown in SAS Code Fragments 6.2, 6.3 and 6.4, respectively.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.