Effect of study duration on the interpretation of tuberculosis molecular epidemiology investigations

Many molecular epidemiological investigations of M. tuberculosis are reported using data collected over relatively short timeframes. We postulated that such studies would tend to under-estimate the amount of disease in a community attributable to ongoing transmission. To test this hypothesis we used 12-year datasets of both real and simulated epidemics with the latter being based on two possible models of transmission. We analysed the effect of viewing the datasets through time windows of varying sizes on the measured degree of strain clustering as an indicator of ongoing transmission. We found that shorter windows significantly under-estimated transmission and that this effect was inversely correlated with the size of a cluster. Accordingly, we recommend that molecular epidemiological studies of M. tuberculosis, for the purposes of estimating transmission, be conducted over a minimum of 3-4 years and that the distribution of cluster size be taken into account in the interpretation of such data.