e. the actual acquisition events cannot be directly observed. Moreover, estimation of vaccine efficacy for a colonisation endpoint may need to be adjusted for interactions between the selleck compound multiple strains of the pathogen as they compete in colonising the human hosts. Study subjects may be sampled for colonisation with long sampling intervals or only once. All these aspects should impact the choice of specific colonisation endpoint (e.g. acquisition, duration, or density of colonisation), vaccine efficacy

parameter, and the appropriate methods for estimation. Here and in the accompanying article [14] we discuss the choice of colonisation endpoints for PCV and other pneumococcal vaccine efficacy studies and the associated issues of estimation methods, adjustment for competing non-vaccine type acquisition, control vaccine, timing of colonisation measurements, implications of multiple serotype colonisation, and sample size. We distinguish between vaccine efficacy against acquisition SKI-606 supplier of colonisation (VEacq), vaccine efficacy regarding duration (VEdur) or density of colonisation. A combined efficacy (VET) is defined accounting effects on both acquisition and clearance. For

these and other possible vaccine efficacy parameters, vaccine efficacy against colonisation (VEcol) is used as an umbrella concept. We concentrate on methods that can be used in a cross-sectional study, i.e. based on only one observation of the current colonisation per study subject. The combined efficacy then turns out to be the parameter that requires the smallest set of underlying assumptions. The statistical methodology reviewed here is based on two previous articles ([10] and [11]). These methods are related to the nested case-control design that could be used to estimate vaccine efficacy in a setting with multiple possible endpoints (i.e. colonisation with any of the >90 pneumococcal serotypes), whilst avoiding the need for identifying the actual acquisition events. Related statistical Phosphatidylinositol diacylglycerol-lyase methods for estimation of vaccine efficacy against colonisation or disease in a setting with multiple serotypes include

the indirect cohort method [12] and sieve analysis [13]. Our approach generalises the indirect cohort method to the analysis of transient and recurrent (colonisation) events with appropriate adjustment for replacement carriage within the host. The main difference between our approach and the sieve analysis is that the outcomes in the latter method are non-transient. This work is framed with PCV in mind, however the methods are applicable for newer vaccines such as the protein vaccines. The accompanying article discusses more practical design questions, including the timing of colonisation measurement with respect to the time of vaccination, choice of control vaccine and the statistical power of colonisation endpoint trials [14].