[The capture-recapture applied to epidemiology: principles, limits and application].

By matching several sources of information coming from the same population, the capture-recapture method allows, if matches between sources can be identified, to estimate the number of cases not identified by any of the sources, the total number (N) of cases in the population and the sensitivity of each source. Used first in zoology and demography, this technique has been increasingly applied to epidemiology in the last two decades. However, six conditions need to be satisfied: all cases identified by each source are true cases; occurred during the time and the geographical area studied; the study population is closed; all matches are true matches; sources are independent and there is no heterogeneous catchability within sources. It is possible to assess and take into account dependence between sources when 3 or more sources are used, in particular by log linear modelling. However interactions of maximal order cannot be adjusted for. In case of a biased estimation because of the non respect of the conditions, it is recommended to apply a method of correction. The complete application of the methods is illustrated in the study of estimating the number of Salmonella infectious foodborne outbreak. As capture-recapture studies often use data collected for other purposes, this technique should be regarded as exploratory or be used to validate an estimate of a number of cases obtained by another method. It should not replace population sampling methods only because of its low cost. Otherwise it is a very informative approach to evaluate the sensitivity and representativeness of surveillance systems and registries.