The use of an adequate mathematical model is crucial to evaluate vaccine effectiveness

Abstract

A recent paper by El Khoury et al. (2011) reported estimates of effectiveness of the RotaTeq ® vaccine against rotavirus-related hospitalisations in Brazil of 93%. After carefully examination of the data presented in this paper, it appears that there are important limitations on this study leading to a potential overestimation of the reported vaccine effectiveness. Although the analysis applies a model validated by Rose and Singer in 2008, the conclusions in the Khoury's paper are beyond the original model capabilities. The model used was developed to project expected rotavirus vaccine efficacy in settings where applying unadjusted efficacy data could overestimate the benefits of immunisation. To generate a weighted efficacy estimate for Brazil, El Khoury et al.(2011) used serotype-specific vaccine efficacy estimates against hospitalised rotavirus gastroenteritis (RGE) from the Rotavirus Efficacy and Safety Trial (REST) (Vesikari et al. 2006) and the distribution of rotavirus types observed in four cities of Brazil (Goiânia, Porto Alegre, Salvador and São Paulo) from [2005][2006]. It is difficult to determine whether the strain-specific vaccine efficacy against hospitalised RGE observed in the REST study would be obtained if that study were performed in Brazil. Indeed, differences in access to health care services and differences in health seeking behaviour between locations in Brazil may alter the disease severity presentation of hospitalised children with rotavirus diarrhoea and therefore impact the estimated vaccine efficacy. Thus, it is a daring assumption to conclude that the "true" rotavirus strain distribution in Brazil is equivalent to a weighted average of the strain distributions observed in the four cities studied given the known dynamics of rotavirus strain circulation profile in Brazil (Leite et al. 2008).El Khoury et al. (2011) acknowledged that although the model utilized was developed to estimate projected efficacy, they used the term "projected effectiveness" throughout the paper because their objective was to project the real-life impact of RotaTeq ® vaccine on the reduction of rotavirus-related hospitalisations. This is an important shortcoming of the El Khoury el al. (2011) paper since the two terms are not equivalent and therefore the reported projected effectiveness estimate for Brazil is incorrect. It is well known that rotavirus vaccine effectiveness in a given population is not related to the virus strain type coverage only. Other factors such as malnutrition, transplacental maternal and breast milk antibodies and co-infecting pathogens could interfere with an infants' immune response to the vaccine. Indeed, other live oral vaccines against typhoid, cholera, polio as well as earlier rotavirus vaccines have historically performed less well than expected in developing countries (John 1976, Patriarca et al. 1991, Su-Arehawaratana et al. 1992, Levine 1997, Glass et al. 2006.The per protocol analysis and efficacy subset of the REST study contained only 8% of the participants from Latin Amer...