In some cases, the estimators obtained in compound tests have better features than the traditional ones, obtained from individual tests, cf. Sobel and Elashoff (1975 ), Garner et al. (1989 ) and Loyer (1983 . The bias, the efficiency and the robustness of these estimators are investigated in several papers, e.g. Chen and Swallow (1990) , Hung and Swallow (1999) and Lancaster and Keller-McNulty (1998) . Thus, the use of estimators based on compound tests not only allows a substantial saving of

more »

... sts, but they also can (in some situations) be more accurate than the estimators based on the individual tests. Nevertheless, each laboratory produces estimates for the prevalence rate of a given infection using different methodologies, such as halving nested 80 J.P. Martins, M. Felgueiras and R. Santos procedures (Sobel and Elashoff, 1975) and square array testing (Kim et al., 2007) . The logistic regression or the weighted least squares regression can be used in order to combine different prevalence rate estimates (Chen and Swallow, 1990) . In this work some meta-analytical techniques are proposed as an alternative approach. This methodology has the advantage of being quite simple and flexible to account for the error source.