This study proposes to use the capture-recapture method to estimate the number of new HIV diagnoses among children in mainland France during the 2003–2006 period. Three data sources were used: the mandatory HIV reporting system (DOVIH), the French Perinatal Survey (EPF), and laboratory screening surveillance (LaboVIH). Estimates were generated using log-linear models that accounted for dependencies between the sources and heterogeneity variables. Two hundred thirteen cases were identified by at least one of the three sources. The log-linear model used accounted for dependencies between the DOVIH and EPF sources, EPF and LaboVIH, and interactions between EPF and the region of diagnosis, EPF and the country of birth, and DOVIH and the year of diagnosis. The number of new HIV diagnoses among children under 13 was thus estimated at 364 for the 2003–2006 period, 62% of whom were born abroad. The completeness of the DOVIH registry was estimated at 28%. Despite a possible overestimation—which would require comparing the results with field data—this study made it possible to estimate, for the first time, an average annual number of 90 new HIV diagnoses among children in metropolitan France. Recommendations for screening pregnant women, repeated if necessary at the end of pregnancy, are essential for optimizing the prevention of mother-to-child transmission. The high prevalence in certain regions of the world, particularly in sub-Saharan Africa, would justify offering HIV screening to children arriving in France, so that they can receive appropriate care. The involvement of laboratory professionals since 2007 and increased awareness among pediatricians should improve the completeness of the mandatory reporting system for HIV in children. (R.A.)