Page 180 - 15Diarrhoeaandvomiting

Appendix B

Figure B.1 Probabilistic sensitivity analysis of the net costs and effects of ondansetron (n = 10 000)

The PSA results showed that in every simulation ondansetron led to a greater cessation in vomiting
than placebo, providing very strong evidence for the efficacy of the intervention. It also showed
ondansetron to be cost saving in 99.96% of simulations,

Discussion

The baseline result shows that ondansetron is dominant compared with placebo, with increased
cessation of vomiting and a net saving of £51.68. A PSA suggested this cost-effectiveness
conclusion was not sensitive to parameter uncertainty in the model’s probabilities. This is because
the magnitudes of the effect sizes for cessation of vomiting and reduced resources are so large
(and the confidence intervals sufficiently narrow) that they are most unlikely to be due to chance.
As the costs of hospitalisation and IVT are high relative to treatment, a real reduction in the costs
associated with these events is likely to offset the initial ondansetron drug cost.
There are a number of caveats to be considered when interpreting the above results. Firstly, the
PSA assumed independence in the three probabilistic model parameters. In practice, it would be
expected that a reduction in hospitalisation and IVT is dependent on the efficacy of ondansetron.
Therefore, a more sophisticated model would demonstrate an inverse relationship between the
net costs of ondansetron and increased cessation of vomiting. Nevertheless, the importance
of this should not be overstated. As the independent variable, the distribution of cessation of
vomiting rates is not affected by the assumption of independence. Although a more sophisticated
model may show a greater uncertainty around the mean net costs, the efficacy of the intervention
would still be likely to produce some offsetting savings and the probability of a net cost saving
would remain high.
The two studies that reported the outcome of ‘cessation of vomiting’ occurring ‘a few hours’
after ondansetron is taken use different time frames from each other. One study 163 looked at
the proportion of children who vomited while receiving ORT whereas the other 164 measured
the frequency of emesis during the 48 hour period after enrolment. All the studies in the meta-
analysis were undertaken in the USA and treatment practice differs in England and Wales. In the
NHS, a child would be admitted as an inpatient if their vomiting had not stopped within 4 hours
of taking ondansetron in the emergency department, regardless of whether the child ultimately
receives IVT. This inpatient admission would incur costs. Therefore, the time frame used in the

155