ways and weighted the amount by their corresponding conditional probability. Summing over the
weighted costs yielded the expected cost of a treatment modality. The same process was undertaken
for both treatment strategies, and the incremental
cost per subject was calculated by taking the difference in expected costs.
We also estimated the incremental number of
bleeding recurrences, surgical procedures, and hospitalization days for the iv PPI strategy versus standard endoscopic treatment in a hypothetical cohort
of 1000 subjects. To estimate these figures, we multiplied the number of specific events experienced
during each of the seven possible pathways in the
decision tree by their corresponding conditional
probability and then summed over each treatment
strategy. The results were multiplied by 1000 to
yield the expected number of events per 1000 subjects. Finally, we calculated the difference between
treatment groups.
One-Way Sensitivity Analyses
To evaluate the impact on the results of uncertainty
surrounding some values in our model, we conducted one-way sensitivity analysis on all variables.
We varied the values over a range drawn from published studies and observed the impact of the
changes on the incremental cost per subject and
number of recurrent bleeds. These analyses also