[Comparison of annual risk for tuberculosis infection (1994-2001) in school children in Djibouti: methodological limitations and epidemiological value in a hyperendemic context]. / Comparaison du risque annuel d'infection tuberculeuse (1994-2001) auprès d'enfants scolarisés à Djibouti: limites méthodologiques et intérêt épidémiologique dans un contexte hyperendémique.

ABSTRACT

The purpose of this report was to describe a tuberculin survey conducted in 2001 to assess the trend in the annual risk for tuberculosis infection in Djibouti and compare resulting data with those obtained in a previous survey conducted in 1994. In 2001 cluster sampling allowed selection of 5599 school children between the ages of 6 and 10 years including 31.2% (1747/5599) without BCG vaccination scar. In this sample the annual risk of infection (ARI) estimated using cutoff points of 6 mm, 10 mm, and 14 mm corrected by a factor of 1/0.82 and a mode value (18 mm) determined according to the "mirror" method were 4.67%, 3.64%, 3.19% and 2.66% respectively. The distribution of positive tuberculin skin reaction size was significantly different from the normal law. In 1994 a total of 5257 children were selected using the same method. The distribution of positive reactions was not significantly different from the gaussian distribution and 28.6% (1505/5257) did not have a BCG scar. The ARI estimated using cutoff points of 6 mm, 10 mm, and 14 mm corrected by a factor of 1/0.82 and a mode value (17 mm) determined according to the "mirror" method were 2.68%, 2.52%, 2.75% and 3.32 respectively. Tuberculin skin reaction size among positive skin test reactors was correlated with the presence of a BCG scar, and its mean was significantly higher among children with BCG scar. The proportion of positive skin test reactors was also higher in the BCG scar group regardless of the cutoff point selected. Comparison of prevalence rates and ARI values did not allow any clear conclusion to be drawn, mainly because of a drastic difference in the positive reaction distribution profiles between the two studies. The distribution of the skin test reaction's size 1994 study could be modelized by a gaussian distribution while it could not in 2001. A partial explanation for the positive reaction distribution observed in the 2001 study might be the existence of cross-reactions with environmental mycobacteria.