Statistical modelling methodologies to predict the impact of reduced-risk products (with the potential to reduce the public health risks associated with smoking, “RRP”), namely e-cigarettes and heated tobacco products, require many input parameters; i) demographic projections (e.g. births, deaths, migrations), ii) status and evolution of tobacco consumers’ prevalence (e.g. initiations, cessations, product transitions), and iii) estimates of the potential reduced-risk of new products (as compared to the risks associated with smoking conventional cigarettes).

Generally, due to the large volume of data and heavy computation processing time, a small range of input parameters is carefully selected (e.g. low, medium, high scenarios) and is evaluated for sensitivity purposes.

The objective of this work is to build a statistical platform to compute valid and reliable predictions integrating exhaustive variability by inputting not only point estimates but the full parameter distributions.

A standard SAS/IML® script using resampling techniques simulations has been created to mimic the density of the reduced-risk profile and smoking status transition rates probabilities coming either from assumptions defined by the users of the platform or from clinical biomarker and observational studies.

The script is producing automatic outputs showing the predictive demographics distribution over time including life-year gain in a population after introduction of an RRP. Sensitivity and robustness of the model can then be checked by looking at percentiles of all simulations (e.g. 5 % lowest percentile showing positive life-year gain after introduction).

This method has been applied using official Japanese demographic projections, vital statistics and smoking prevalence data to better describe what would be the impact of model parameters variability on public health predictions after introduction of a new RRP.