Use of Modelling Techniques to Estimate the Number of Additional Events Avoided Through Early Versus Delayed Intervention : The Case of Rotavirus Vaccination

Introduction: Rotavirus (RV) infection is the major cause of diarrhoea in children <5y old. Highest peak in infection is between age 6 months and 2 years. The incidence of RV diarrhoea decreases from age 2 until age 5 after which natural immunity against the virus is well established. Severe cases of RV diarrhoea can result in hospitalisation or even death. To reduce the risk of RV infection, vaccination is a very attractive management strategy especially if the vaccine is administered early in life (within the first 16 weeks after birth). We used a model to compare early (within 16 weeks) versus delayed (within 24 weeks) completed vaccination against RV to estimate the difference in number of diarrhoea events (DE) avoided. Methods: For ethical and logistic reasons it is very difficult to organise a randomised clinical trial to assess early versus delayed vaccination strategies against RV. Hence, modelling can be interesting to assess different vaccination strategies. Based on literature review and database analysis the occurrence of DE in function of age over time (children <5years old) was simulated. Best fit techniques (BestFit from Palisade) were used to select the distribution that optimally compares observed versus estimated numbers of DE over time in order to parameterise the spread. Vaccination at 16 vs. 24 weeks was then applied. The results from this exercise were then compared assuming that vaccination has or an immediate effect or a gradual time-dependent exponential effect in reducing the number of DE. The results are reported as a percentage gain of total number of DE avoided over time due to vaccination. Results: BestFit results to replicate the DE over time in children age <5y yields a Weibull distribution (alpha = 1.5, beta = 24.2) which peaks around the age of 14 months as the most appropriate fit. Assuming 300,000 rotavirus-induced DE over 5 years with a spread over time fitting this Weilbull distribution, one can estimate the extra number of events avoided if a RV vaccination is fully applied at week 16, instead of week 24. With a full vaccination coverage at week 16 the extra number of DE avoided, assuming the vaccine has an immediate effect, is around 15,040 or a 5%-gain. If the vaccine were to have a more gradual effect following an exponential decline, the gain would be less, estimated at 13,000 or 4.3%. Sensitivity analysis over vaccine efficacy and additional different time points will be reported. Conclusion: Modelling techniques allow reliable estimates of the difference in number of RV DE if new interventions are programmed at different time points. In case of rotavirus vaccination different vaccination schedules spread over different time periods may occur. A one month difference in full vaccination protection can be dramatic as we can gain 5% of extra DE avoided with an earlier protection scheme. Given this difference, these results suggest the consequences of this earlier vaccination strategy on cost and QoL of the individual and his direct environment should be assessed