The use of mosquito trapping techniques to estimate daily vector biting rates experienced by humans requires not only that such approaches are sufficiently sensitive, but also that sampling efficiency is known. The relative sampling efficiency of LT was found to be density-dependent with its efficiency decreasing at high vector densities. This finding supports other reports from areas of low malaria vector density in Kilifi on the coast of Kenya for An. gambiae s.l.  and in Papua New Guinea for An. punctulatus and An. farauti . In other studies, however, the relative sampling efficiency of LT has been found to be density-independent [14, 15, 48, 49]. Unlike the Kilifi study , no zero values were present in the aggregated data and no transformation other than logarithm were necessary. Therefore, this density-dependence cannot be attributed to mathematical artifact  and appears to be a genuine property of the sampling device. Note that our estimate of mean relative sensitivity for the LT differs from previous trials in the same Tanzanian village  and re-analysis of this data reveals the same density-dependence for LT that we have outlined here (Okumu unpublished). The apparently variable trapping efficiency of LT between and within studies, may not necessarily be due to differences in statistical approach  but rather to subtle and intensely heterogeneous factors which inevitably vary through space, time and investigation. Such essentially uncontrollable factors could include the positioning of the paired techniques, use of interventions such as bednets and insecticides, lunar phase, season, weather and house architecture. For instance, one study in Papua New Guinea  reported that the relative sampling efficiency of LT placed indoor was independent of outdoor An. bancroftii density changes and simultaneously density-dependent in relation to indoor vector abundance while the reverse trend was true for An. longirostris. One study in Africa  noted some evidence that the presence of treated nets reduce the relative sampling efficiency of LT but this effect was slight and in other similar studies [16, 51] no effect could be demonstrated. Several previous studies concluded the LT to be free of age-related sampling biases [14, 48, 52] consistent with our observations but not those of other reports [53, 54]. The difference between these studies might be partly explained by variability in both the position of LT relative to the floor , the quality of net , and variability in sleeping behavior of net occupants .
Here the sampling efficiency of Ifakara B traps relative to HLC has been evaluated in two very different eco-epidemiological settings, where malaria transmission intensity ranges from less than one  to over 600  infectious bites per person per year and the vector species in question have clearly distinct feeding behaviors and activity patterns [30, 33]. The sampling efficiency of the Ifakara B design appeared to be independent of vector density in the rural area with high vector abundance but appeared to increase at low densities in the urban setting, possibly reflecting reduced attentiveness of HLC catchers at low mosquito densities. Clearly none of these entomological techniques are precise, accurate or representative of true human biting rates but it is encouraging that the Ifakara B design correlates quite well to the HLC, given its lack of dependence on electricity, access to the inside of houses or intensive effort, plus its increased sampling efficiency at low densities. Also, the modest sampling efficiencies indicated by this crude analysis are sufficiently high to suggest these tent trap designs could be useful for extensive, sustained vector surveillance because of their lower cost and difficulty per trap night of sampling.
The observation that the proportion of fed mosquitoes caught in the tent traps is at least as high as for those caught by HLC implies that either these traps act as resting shelters for freshly fed mosquitoes or that the human bait actually does get bitten while aspirating mosquitoes. We have occasionally observed the latter process occurring in practice and suggest a relatively clear avenue for improvement to develop a design which truly is exposure-free and completely protects the user from exposure to mosquito bites. We conclude that the Ifakara B tent trap may be a valuable tool for large-scale surveillance, particularly in resource-limited settings if concerns about operator exposure while collecting each morning could be overcome through modification or protective measures.
The new trapping method was primarily intended to replace both LTs and HLC for routine monitoring of African malaria vectors in large scale programmes such as the Dar es Salaam Urban Malaria Control Programme (UMCP). As such there is an urgent need to first adapt this tool to prevent operator exposure and then develop a protocol allowing community based volunteers to trap and submit mosquitoes to a central laboratory without the need for intensive, expensive and unsustainable level of support and supervision . We nevertheless caution that all trapping methods described here, including the Ifakara B and HLC have substantial limitations as tools for measuring mosquito biting density and human exposure. Generally speaking, mosquito sampling methods remain poorly characterized and standardized , and as reported here, are difficult to unambiguously relate to human exposure and malaria risk. We therefore suggest that entomological measures of transmission for routine use should ultimately be evaluated in comparison with parasitological indicators of human exposure so that the most representative and epidemiologically relevant tools can be identified.