Summary. Two recently published metaanalyses of controlled trials of a wide variety of progestational agents, used in pregnancy (Daya 1989; Goldstein et al. 1989), prompted this third metaanalysis of placebocontrolled trials involving the prophylactic use of a single agent, 17αhydroxyprogesterone caproate. Of seven relevant published reports of controlled trials, six had involved women considered to be at high risk of miscarriage or preterm birth. This analysis provides no support for the view that 17ahydroxyprogesterone caproate protects against miscarriage, but suggests that it does reduce the occurrence of preterm birth. The latter effect was reflected in a reduced rate of low birthweight babies, but not in a statistically significant reduction in perinatal mortality and morbidity. The difference between this metaanalysis and the two earlier metaanalyses illustrates the problems both of selective subgrouping and of comprehensive pooling of data from small trials.