Reported efficacies from vaccine trials may depend heavily on the clinical case definition used in the trial. The dependence may be particularly striking for diseases such as malaria, in which no single case definition is appropriate. We used logistic regression modeling of the relationship between parasitemia and fever in data sets from Ghanaian children to determine the fraction of fevers attributable to malaria and to model how the choice of a threshold parasitemia in the clinical case definition affects the measured efficacy of malaria vaccines. Calculated clinical attack rates varied 10-fold as a function of the threshold parasitemia. Strikingly, measured vaccine efficacies in reducing clinical malaria depended heavily on the threshold parasitemia, varying between 20% and 80% as the threshold varied between 1 and 5000 parasites/ microL. We suggest that clinical case definitions of malaria that incorporate a threshold parasitemia are arbitrary and do not yield stable estimates of vaccine trial end points.