Use Instructions

50
Figure 2 shows that the Bayesian model’s point estimate is very similar to the
estimate from just the FIX-HF-5C study. However, the model further incorporates the
high quality data from the previous randomized, blinded trial which increases the
precision of the estimate. If FIX-HF-5C were a standalone trial, the middle CI would
be appropriate. However, the Bayesian model allows us to incorporate the totality of
the clinical experience which is an increased precision in the effect size estimate and
is shown by the narrower 95% CI with the Bayesian estimate.
Figure 2: Peak VO2 by Study
The improvement in peak VO
2
built up over time, from 3 to 6 months (Figure 3). The
treatment effect can be seen in this graph to be a result of a significant decrease in
VO2 for the control group with relatively little increase in VO2 for the treatment group.
Figure 3: Time Course of Treatment Effect on Peak VO2 (FIX-HF-5C)
Sensitivity analyses involving the primary effectiveness endpoint were conducted in
which missing data were handled with different mechanisms or modifications (Table
5). Method of imputation affected the results and the VO2 estimate varied from 0.48
to 0.84 depending on method. The conclusion of CCM superiority with respect to
mean peak VO
2
was consistent across all sensitivity analyses. In addition, the
primary analysis would achieve statistical significance with any borrowing weight of
0.11 or larger (as noted above, 0.30 was pre-specified in the analysis plan).