Vaccine - (Page 7) 6162 M. Oviedo et al. / Vaccine 26 (2008) 6157–6164 4.16–5.72), whereas model BN4 estimated 181.59 cases/year, with a reduction of 4.99 cases/year. 5. Discussion The results of this study clearly show that the incidence of hepatitis A has fallen after the introduction of routine hepatitis A vaccination in Catalonia, reaching extremely low incidence rates in the 12–18 years age group (rate of 0.65 per 100,000 persons in 2001 and 0.66 in 2002). The reduction in cases in the 12–18 years age group was dramatic, falling from a mean rate of 8.15 per 100,000 person/years in 1992–1998 to 1.40 in 1999–2005 and most of the reduction was concentrated in 2001–2004 (rate range 0.44–0.67). This range is explained by the interaction of the effect of the year (natural reduction in the incidence) and the effect of vaccination. Although the percentage of vaccination continues to increase, the number of cases have not fallen, as they have been reduced to a very small number (range 2–3 cases in 2001–2004), which can be considered as a base level that would be difficult to reduce in real conditions because, although the programme is offered free to all people, there are some imported cases in non-vaccinated immigrants who have not previously suffered the infection. The prediction model contributes very useful information on the validity, generalizability and possible defects of the model. Model BN4 provides a CI of prediction that includes the observed real incidence and accurately predicts the vaccinated population (12–18 years age group), with a predicted incidence rate for 2006 of 1.77 per 100,000 person/years and an observed rate of 1.75. However, model BN3 (which includes the cycling component) predicts an incidence of 1.48 (95% CI: 0.3–2.7), underestimating the real incidence rate. The prediction model also underestimates the reported cases in the <12 years and ≥40 years age groups. This problem could be corrected by constructing a model with more covariates. There are different reasons why the model does not predict these groups correctly. One may be immigration: a more sophisticated model which includes information on the country of origin or recent travel to endemic areas would be necessary. Another possible reason for the underestimation in the ≥40 years age group is shown by seroprevalence studies which, in 2002, already showed a shift in the seroprevalence curves in Catalonia from an intermediate-high pattern to a much more moderate pattern [19]. Therefore, the forecast is for an increase in the susceptible population, which could imply a greater number of cases in adults. In all the adjusted models, the sensitivity analysis shows very similar estimates of the parameters with the same level of statistical significance. The statistical analysis was crucial in obtaining valid estimates and results, since initial adjustment of the Poisson regression provided heterogenous estimates. Poisson regression has the disadvantage of being sensitive to overdispersion, i.e. violation of the property of equidispersion (observed variance equal to observed mean). The estimate parameters and standard error in the Poisson model were inappropriate and the model provided biased estimates for the 12–18 years age group, with the result that the covariate appeared to contribute more significantly to the model than it does in fact. For this reason, a generalization of the Poisson model, the negative binomial, was proposed. Some studies [24,25] adjust the year as the factor or take a period of years as the pre-vaccination reference level [26]. We also tested these methods, which require the estimation of a larger number of parameters in the model and mean that the effect of the factor year could be confused with vaccination and the results conditioned by the selected reference period. In addition, the aim of this study was not only to perform a retrospective analysis but also to provide a model that contributes a prospective view by introducing the variable year as a continuous variable in the model. The catalytic model [14] is presented as an alternative to the negative binomial regression model used in this study. Although this model permits more accurate estimates and predictions, its main problem is that it requires the calculation of the susceptible population from the seroprevalence data, which are not always available or are not sufficiently up-to-date [25]. The same is true for dynamic models [15], such as SEIR (susceptibleexposed-infectious-recovered) models [16,27], that can also use seroprevalence data, socioeconomic data or under-reporting of information to estimate the reduction in infection rates of HAV in different regions. Our results show that routine hepatitis A vaccination is a preventive intervention, independently of disease cycles. In addition, as shown in Fig. 1, the vaccine benefits not only vaccinated children but also young people <12 years of age and young people aged 19–29 years who have not benefited from the vaccination program. This is consistent with other studies [17]. Samandari et al. [24] also used a mathematical model to distinguish between the reductions in incidence attributable to vaccination and to the natural evolution of the disease and found that a vaccination coverage of 10% in children aged 2–18 years prevented 51% of cases in this age group and 39% of all cases, due to herd immunity. Wasley et al. [25] found that in states which recommended universal paediatric vaccination, incidence rates in 2003 fell by 88% compared with the 1990–1997 pre-vaccination period whereas, in states which did not recommend vaccination, the reduction was only 53%. In Israel [28], after the introduction of universal vaccination of children with doses at 18 and 24 months of age, an annual reduction of 98% was observed in the cohort of vaccinated children and of 90% in non-vaccinated older children and adults in the 2002–2004 period with respect to the 1993–1998 period. The prevented fraction of hepatitis A in preadolescents and adolescents in Catalonia (90%) should be considered as a minimum, as both the coverage and the effectiveness were estimated conservatively [17]. If we assume that vaccine-induced protection is at least 9 years but may be lifelong when an anamnesic response is generated against a new exposure [29,30], universal vaccination in the second year of life, where coverages reach 95–98% [22], would avoid practically all paediatric cases in Catalonia and would continue to protect these children in adolescence and adulthood. This does not rule out the possibility that, in the future, due to migratory phenomena and an increase in trips to endemic regions, the number of cases in some risk groups will increase. Therefore, the possibility of introducing other variables, such as the country of origin, a history of recent travel to endemic regions or mortality [31,32], into the model, should be considered. In conclusion, the adjusted model developed allows the effect of the vaccination programme over time to be estimated and can be used to make predictions and allow new variables of interest to be included. Acknowledgements This study was partially funded by the Instituto de Salud Carlos III Madrid, Spain, and CIBER Epidemiología y Salud Pública (CIBERESP), Spain.
Table of Contents Feed for the Digital Edition of Vaccine Vaccine Vaccine - (Page Cover1) Vaccine - (Page 2) Vaccine - (Page 3) Vaccine - (Page 4) Vaccine - (Page 5) Vaccine - (Page 6) Vaccine - (Page 7) Vaccine - (Page 8) Vaccine - (Page 9)
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.