BACKGROUND: The full-length (L-) variant of caspase-12 is believed to predispose to sepsis. It has been replaced in the genome of most human populations by the (S-) variant, which leads to premature termination of translation. Strikingly, the L-allele is still widely prevalent in African populations, presumably due to a counterbalancing selective force specific to this continent, for which malaria is a prime candidate. METHODS: We investigated associations between caspase-12 genotype and malarial parameters in three West-African populations, in studies encompassing immunological, clinical and obstetric data. RESULTS: The caspase-12 L-allele was found at frequencies of 11-34%. Plasmodium falciparum-stimulated mononuclear cells from S/L heterozygote donors produced stronger interferon-gamma and interleukin-10 responses than S/S homozygotes (p = 0.011 and p = 0.023 in uninfected and infected donors respectively). Nevertheless, we found no association between caspase-12 genotype and either the presentation of severe malaria or individual clinical parameters in sick children. Amongst pregnant women, the caspase-12 genotype did not influence peripheral or placental malaria infection, or basic obstetric parameters. Interestingly, perinatal mortality was more frequent in children of both S/S and L/L than S/L mothers, independent of placental P. falciparum-infection. CONCLUSION: We find little clinical or epidemiological evidence that malaria has contributed to the persistence of functional caspase-12 in Africa, suggesting either that alternative selective forces are at work or that genetic drift underlies its current global distribution.