Though it is recognized that the extent of ‘clustering’ of isolates from tuberculosis cases in a given population is related to the amount of disease attributable to recent transmission, the relationship between the two statistics is poorly understood. Given age-dependent risks of disease and the fact that a long study (e.g. spanning several years) is more likely to identify transmission-linked cases than a shorter study, both measures, and thus the relationship between them, probably depend strongly on the ages of the cases ascertained and study duration. The contribution of these factors is explored in this paper using an age-structured model which describes the introduction and transmission of M. tuberculosis strains with different DNA fingerprint patterns in The Netherlands during this century, assuming that the number of individuals contacted by each case varies between cases and that DNA fingerprint patterns change over time through random mutations, as observed in several studies.

Model predictions of clustering in different age groups and over different time periods between 1993 and 1997 compare well against those observed. According to the model, the proportion of young cases with onset in a given time period who were ‘clustered’ underestimated the proportion of disease attributable to recent transmission in this age group (by up to 25% in males); for older individuals, clustering overestimated this proportion. These under- and overestimates decreased and increased respectively as the time period over which the cases were ascertained increased. These results have important implications for the interpretation of estimates of the proportion of disease attributable to recent transmission, based on ‘clustering’ statistics, as are being derived from studies of the molecular epidemiology of tuberculosis in many populations.