Original Analyses Sample Clauses

Original Analyses. Due to the small sample size of the pilot study, original analyses included ▇▇▇▇▇▇’▇ exact test (for ▇▇▇▇ scores) and Wilcoxon’s rank sum test (for DRS scores). All analyses were stratified by the initial severity of the injury (moderate vs. severe). At 30 days 40% of the severely injured patients randomized to placebo were deceased whereas only 13.2% of those randomized to treatment were deceased. Mortality rates among the moderately injured were similar across the treatment arms (14.3% among the control group versus 16.7% among the treated). Among the survivors with severe initial injury severity, 26.7% of patients in the control group had a ▇▇▇▇ score at 30 days indicative of moderate or good recovery versus 21.2% of the treated. Among the moderately injured survivors, none in the control group attained moderate to good recovery versus 55.6% in the treated. Additionally, although the confidence intervals overlap, the mean estimated DRS score among severely injured patients was lower (indicating more improvement) in the placebo group, whereas DRS results among It is worth noting that repeating the original study’s Wilcoxon rank sum test on the one year follow-up data fails to find a statistically significant difference in DRS scores between the groups (p-value = 0.5, see Table 4.3 below), although patients in the treatment group do have a lower mean DRS (lower is better). If the worst DRS score (29 - extreme vegetative state) is substituted for all patients who died before the one year follow-up (a common alternative to survivor-only analyses), Table 4.1 results, Table 4.1: Mean DRS Assuming DRS = 29 for Deceased Mean DRS (SD) Progesterone 9.1 (12.3) Placebo 13.0 (13.7) which still indicates improvement among the treated patients, but much worse recovery overall. Additionally, this is clearly misleading since it adds back potentially inaccurate outcome values for 30% of the control patients and 18% of the treated patients. However, even this amount of data replacement fails to produce a statistically significant difference between groups using a Wilcoxon rank sum test (p-value = 0.2). Traditional methods are hampered by the significant amount of missing data resulting from the two different sources - by the one year follow-up 50% of outcome data are missing, either due to loss to follow-up or truncation due to death. Both survivor-only analyses and the alternative of substituting the worst possible outcome for all deceased patients are likely to provide ...