Probabilistic simulations
Probabilistic simulation method is an estimation technique which can flexibly incorporate statistical uncertainties in the simulated parameters. The uncertainties are then taken into account and reflected in the final results. The distributions we used in our models are listed in the table below. The sampling on each parameter was repeated 100,000 times to capture the natural variations in the various metric indices. The 95% confidence intervals for the impact numbers were constructed from the 2.5th and 97.5th percentiles of the simulated values. In this case study, probabilistic simulation method is used to calculate the NNT/NNH, impact numbers (NEPP and PIN-ER-t), BRR and NCB. Parameters are sampled randomly from some appropriate probability distributions based on available data (for details, see full report Rimonabant Wave 1 case study).
In Table 1 below, we show the distributional assumptions of the stochastic parameters in the models we used in the simulations. Due to multiple uncertainties being multiplied together, the final amount of uncertainties in the subsequent analyses may grow to be very large. Therefore, it is important to use robust data available to ensure that the final results do not suffer too much uncertainties, leading to less useful or less actionable conclusions.
Table 1 Distributional assumptions for benefit-risk assessment using impact numbers
| Baseline risk |
These are specific for each criterion and are based on the baseline rates (seeTable 10-12 in the Rimonabant Wave 1 case study report). In general, |
| Relative risk RR | RRs correspond to the criteria, and sampled as follows for each criterion j: |
| Proportion eligible for Rimonabant |
|
| Proportion of overweight and obese |
|
| Size of population |