Phenotypic variability in childhood TB: implications for diagnostic endpoints in tuberculosis vaccine trials.

The endpoint definition for infant tuberculosis (TB) vaccine trials should match the TB disease phenotype expected in the control arm of the study population. Our aim was to analyse selected combinations of the clinical, radiological, and microbiological features of pulmonary TB among children investigated under vaccine trial conditions, in order to estimate case frequency for a range of expected TB phenotypes. Two thousand one hundred and eighty five South African children were investigated over a nine-year period (2001-2009). Evidence of TB exposure and classical symptoms were several times more common than chest radiography (CXR) compatible with TB, or positive Mycobacterium tuberculosis culture. Discordance between clinical, radiological, and microbiological features was common in individual children. Up to one third of children with compatible CXR, and up to half the children who were M. tuberculosis culture positive, were asymptomatic. The culture positive rate fell over time, although rates of TB exposure and compatible chest radiography increased. Consequently, the annual incidence of diagnostic combinations that included M. tuberculosis culture fell to <0.2%. However, in this study population (children <2 years of age), annual incidence of the TB disease phenotype that included the triad of TB exposure, symptoms, and compatible CXR, approached 1% (n=848 per 100,000). These findings allow modelling of expected TB case frequency in multi-centre infant TB vaccine trials, based upon benchmarking of diagnostic data against the key indicator variables that constitute the building blocks of a trial endpoint.