As with any vaccine-preventable disease, hepatitis B vaccination programmes should be reviewed according to the evolution of the disease, and surveillance data are crucial to enable correct assessments of the situation [1].
The results of our study are consistent with those for hepatitis A, which showed that the estimates of GLM using negative binomial distribution were better [19]. In addition, GAM permitted estimation of non-linear trends of continuous variables as year, vac and immigrant. The incidence of hepatitis B fell from 1992 to 2000, and has since increased non-linearly, being much higher in groups with a high percentage of immigration and falling when vaccination coverages are higher.
The results also coincide in the case of categorical variables in the GLM models. Poisson or negative binomial regression can cause problems in estimating the coefficients due to the low incidence of cases. This is a limitation of the model that was solved by aggregation into age groups, and not introducing it as a continuous variable. An alternative would have been to adjust using models for count data with Poisson distribution with many zeros (ZIP) or negative binomial (ZINB) [20].
With respect to the predictions, future studies could simulate the predictions of future incidence rates with vaccination strategies other than those proposed in this study. This could allow evaluation of these strategies and the detection of atypical incidence rates other than the estimated pattern.
The use of two data sets (1992-2007) and (2000-2007, which included the category male/female, has advantages and disadvantages. Using the (1992-2007) database has the advantage of a 16-year historical record in which the estimated coefficient of the variable year is significant. This is one possible explanation of why the estimates are better using this database. In the second case, the use of the (2000-2007) database supposes a historical record of only eight years. The positive aspect is that the clearly-differentiated incidence rates in males and females can be estimated. However, the estimated coefficient of the variable year is not significant in this period in the models with the lower AIC (glm.nb3 y gam2). As the number of observations is reduced, the number of degrees of freedom and, therefore, the estimates, are worse.
La Torre et al. [12], applied jointpoint regression and estimated that the incidence decreased in all groups, and highlighted the importance of analyzing the changes in disease incidence in the evaluation of vaccination policies.
The increased incidence in men aged 19-49 years from 2001 onwards obtained in the present study with the overall rate increasing from 2.16 in 2000 to 5.74 in 2007(x10-5 person-years) respectively, coincides with the mass influx of immigrants to Catalonia with rose from 5.6% in 2000 to 22.7% in 2007 [15]; in this same period, the incidence in men aged 19-34 years increased from 2.9 in 2000 to 5.3 in 2007 (x10-5 person-years) and in men aged 35-49 years from 1.4 in 2000 to 7.5 in 2007 (x10-5 person-years). These results differ from those obtained in the USA [21, 22], where the incidence rate of hepatitis B has decreased steadily in all age groups, from 6.3 x 10-5 person-years in 1992 to 1.5 x10-5 person-years in 2007. In that country, although progress has been made in reducing disparities in incidence of new infections rates among non-Hispanic blacks have declined, the incidence rates remain more than twofold higher than those among other ethnic population and the rate in elderly immigrants is much lower than in young adults, although the number of elderly immigrants is very low. Several authors have already analyzed the relationship between the incidence of hepatitis B and immigration, showing that there is a positive relationship between both [23, 24].
In 1992, the rate in females in Catalonia was 2 x10-5 person-years, which fell to 0.45 x10-5 person-years in 2000 and remained fairly stable in the remainder of the study period. In the USA [22] the rate fell from 7.34 in males and 5.01 in females (x10-5 person-years) in 1992 to 3.6 and 2.09 x 10-5 person-years, in males and females, respectively, in 2000 and 1.85 in males and 1.15 in females (x10-5 person-years) in 2007, while in Germany, rates fell from 1.96 x10-5 person-years in 2001 to 0.89 x10-5 person-years in 2006 and 0.6 x10-5 person-years in 2007.
In people aged 12-18 years, the incidence was zero in 2000 and less than one case per year, later. No gender differences in the incidence were detected in indigenous children of vaccination age or in children aged < 12 years, with very low or zero rates. The rates in indigenous women and people aged < 18 years were less than 1x10-5 person-years, as a result of the vaccination programs implemented. In contrast, in immigrants and other age groups, the reported incidence was more than two-fold higher for men than for women.
The male-female ratio doubled over the study period, unlike the results in the USA [22] and Germany [25]. In the USA, the male-female ratio increased slightly, from 1.5 in 1992 to 1.8 in 2006, and in Germany, it rose from 1.91 in 2001 to 2.2 in 2008. This may be explained by the increased incidence in immigrants, a group containing a large proportion of men of working age. The statistical models used show that the incidence of cases increased due to immigration, especially in groups with > 15% of immigrants.
Although the impact of disease prevention measures and the maintenance of high vaccination coverages are important, this impact may be offset by an increase in cases in adult immigrants. The proportion of immigrants aged >50 years is still low, which could explain the low incidence of acute hepatitis B in this population compared to younger ages. Therefore, as suggested by the distribution of cases in other countries [22, 25], the incidence may increase in these groups. For this reason, vaccination strategies for risk groups, including travellers to countries with high or intermediate prevalence of chronic Hepatitis B virus infection should be reinforced [26, 27]. This recommendations should also be applicable to susceptible immigrants (children and adolescents) coming from countries with high or intermediate prevalence where hepatitis B vaccination programmes have still not been launched or where coverages are still very low [28].
In countries like the USA, Germany and Italy, incidence rates have declined in all age groups. In Germany, the rate for males decreased from 2.8 x10-5 person-years in 2001 to 1.2 x10-5 person-years in 2007, while in the USA [22] the rate fell from 3.6 and 2.09 (x 10-5 person-years) in 2000 to 1.85 and 1.15 (x10-5 person-years in 2007 in males and females, respectively. In Catalonia, the rate increased from 1.3 x 10-5 person-years in 2001 to 2.9 x10-5 person-years in 2007. However, it should be noted that, in Germany, the percentage of immigrants has remained constant at around 9% since 1995 [25]. In the USA the percentage of immigrants increased from 10.4% in 2000 to 12.6% in 2007 [29]. However, in Catalonia, the rise was much higher, from 4.1% in 2000 to 15.7% in 2007.
This study was conducted using routine surveillance data. It would probably be useful to focus more-closely on immigrants and risk groups to obtain a better understanding of the situation of hepatitis B virus infection in order to design strategies to increase vaccination coverages [21, 30].
In addition to the protection afforded to individuals in avoiding the risk of chronic liver diseases such as cirrhosis and hepatocellular carcinoma [31], a strategy focused not only on universal vaccination of infants or preadolescents but on risk groups, and immigrants should be considered because it takes into account the substantial indirect effects of vaccination, as avoiding new infections avoid the cases transmitted by them [32].
A possible limitation of this study is the underdetection of cases in the immigrant population. The attendance of health services is free in Catalonia for all people (indigenous and immigrants), but we do not know if there are differences in the attendance to medical services for acute hepatitis B. So, we cannot rule out some underdetection of cases in immigrant population. In order to improve our estimates we would need to improve the surveillance of acute hepatitis B disease and to know what level of completeness of reporting we have for immigrant and indigenous population.
With respect to the quality of the proposed forecasts, the randomness in the incidence pattern of the disease is much lower than in males, making it much easier to predict the temporal disease evolution. However, the randomness in the incidence is greater in males, depending, amongst other factors, on age and disease outbreaks, and therefore it is more difficult to predict its future evolution. The observed incidence was lower than the confidence intervals of the predictions, indicating a change in the pattern of the evolution of the disease.
Future research might concentrate on extending the proposed models to include the spatial-temporal distribution of the disease, as we have done for hepatitis A [33]. Likewise, Pearce and Dorling [34] studied a period of rapid social, economic and political changes in which differences in life expectancy between men and women had a clear geographical dependence. It is essential that future studies have access to more reliable information on immigration and that they can recover missing data whenever possible, in order to detect disease outbreaks earlier and adjust the vaccination strategy to these situations.