Abstract
Catches of Anopheles gambiae and An. arabiensis with the Ifakara Tent Trap-model B (ITT-B) correlate better with human landing catches than any other method but fail to reduce the proportion of blood-fed mosquito caught, which indicates that users are exposed to bites during collection. An improved C model (ITT-C) was developed and evaluated by comparing with ITT-B in semi-field and full-field conditions in southern Tanzania. The sensitivity of the ITT-C was approximately two times that of the ITT-B: relative rate (95% confidence interval) = 1.92 (1.52-2.42), 1.90 (1.48-2.43), and 2.30 (1.54-3.30) for field populations of An. arabiensis, Culex spp., and Mansonia spp., respectively. The ITT-C caught 73% less blood-fed An. arabiensis than the ITT-B in open field experiments and none in semi-field experiments, which confirmed that the C design is a safe trapping method. Validation of ITT-C by comparison with human landing catches and parasitologic measures of human infection status may be necessary to confirm that this design produces consistent and epidemiologically meaningful results.