BackgroundIn the Tanzanian city of Dar es Salaam, high coverage of long-lasting insecticidal nets (LLINs), larvicide application (LA) and mosquito-proofed housing, was complemented with improved access to artemisinin-based combination therapy and rapid diagnostic tests by the end of 2012.MethodsThree rounds of city-wide, cluster-sampled cross-sectional surveys of malaria parasite infection status, spanning 2010 to 2012, were complemented by two series of high-resolution, longitudinal surveys of vector density.ResultsLarvicide application using a granule formulation of Bacillus thuringiensis var. israelensis (Bti) had no effect upon either vector density (P = 0.820) or infection prevalence (P = 0.325) when managed by a private-sector contractor. Infection prevalence rebounded back to 13.8 % in 2010, compared with <2 % at the end of a previous Bti LA evaluation in 2008. Following transition to management by the Ministry of Health and Social Welfare (MoHSW), LA consistently reduced vector densities, first using the same Bti granule in early 2011 [odds ratio (OR) (95 % confidence interval (CI)) = 0.31 (0.14, 0.71), P = 0.0053] and then a pre-diluted aqueous suspension formulation from mid 2011 onwards [OR (95 % CI) = 0.15 (0.07, 0.30), P ≪ 0.000001]. While LA by MoHSW with the granule formulation was associated with reduced infection prevalence [OR (95 % CI) = 0.26 (0.12, 0.56), P = 0.00040], subsequent liquid suspension use, following a mass distribution to achieve universal coverage of LLINs that reduced vector density [OR (95 % CI) = 0.72 (0.51, 1.01), P = 0.057] and prevalence [OR (95 % CI) = 0.80 (0.69, 0.91), P = 0.0013], was not associated with further prevalence reduction (P = 0.836). Sleeping inside houses with complete window screens only reduced infection risk [OR (95 % CI) = 0.71 (0.62, 0.82), P = 0.0000036] if the evenings and mornings were also spent indoors. Furthermore, infection risk was only associated with local vector density [OR (95 % CI) = 6.99 (1.12, 43.7) at one vector mosquito per trap per night, P = 0.037] among the minority (14 %) of households lacking screening. Despite attenuation of malaria transmission and immunity, 88 % of infected residents experienced no recent fever, only 0.4 % of these afebrile cases had been treated for malaria, and prevalence remained high (9.9 %) at the end of the study.ConclusionsWhile existing vector control interventions have dramatically attenuated malaria transmission in Dar es Salaam, further scale-up and additional measures to protect against mosquito bites outdoors are desirable. Accelerated elimination of chronic human infections persisting at high prevalence will require active, population-wide campaigns with curative drugs.Electronic supplementary materialThe online version of this article (doi:10.1186/s12936-016-1340-4) contains supplementary material, which is available to authorized users.

BackgroundMore sensitive and scalable entomological surveillance tools are required to monitor low levels of transmission that are increasingly common across the tropics, particularly where vector control has been successful. A large-scale larviciding programme in urban Dar es Salaam, Tanzania is supported by a community-based (CB) system for trapping adult mosquito densities to monitor programme performance.MethodologyAn intensive and extensive CB system for routine, longitudinal, programmatic surveillance of malaria vectors and other mosquitoes using the Ifakara Tent Trap (ITT-C) was developed in Urban Dar es Salaam, Tanzania, and validated by comparison with quality assurance (QA) surveys using either ITT-C or human landing catches (HLC), as well as a cross-sectional survey of malaria parasite prevalence in the same housing compounds.ResultsCommunity-based ITT-C had much lower sensitivity per person-night of sampling than HLC (Relative Rate (RR) [95% Confidence Interval (CI)] = 0.079 [0.051, 0.121], P < 0.001 for Anopheles gambiae s.l. and 0.153 [0.137, 0.171], P < 0.001 for Culicines) but only moderately differed from QA surveys with the same trap (0.536 [0.406,0.617], P = 0.001 and 0.747 [0.677,0.824], P < 0.001, for An. gambiae or Culex respectively). Despite the poor sensitivity of the ITT per night of sampling, when CB-ITT was compared with QA-HLC, it proved at least comparably sensitive in absolute terms (171 versus 169 primary vectors caught) and cost-effective (153US$ versus 187US$ per An. gambiae caught) because it allowed more spatially extensive and temporally intensive sampling (4284 versus 335 trap nights distributed over 615 versus 240 locations with a mean number of samples per year of 143 versus 141). Despite the very low vectors densities (Annual estimate of about 170 An gambiae s.l bites per person per year), CB-ITT was the only entomological predictor of parasite infection risk (Odds Ratio [95% CI] = 4.43[3.027,7. 454] per An. gambiae or Anopheles funestus caught per night, P =0.0373).Discussion and conclusionCB trapping approaches could be improved with more sensitive traps, but already offer a practical, safe and affordable system for routine programmatic mosquito surveillance and clusters could be distributed across entire countries by adapting the sample submission and quality assurance procedures accordingly.

Background: Mosquito feeding behaviour plays a major role in determining malaria transmission intensity and the impact of specific prevention measures. Human Landing Catch (HLC) is currently the only method that can directly and consistently measure the biting rates of anthropophagic mosquitoes, both indoors and outdoors. However, this method exposes the participant to mosquito-borne pathogens, therefore new exposure-free methods are needed to replace it. Methods: Commercially available electrocuting grids (EGs) were evaluated as an alternative to HLC using a Latin Square experimental design in Dar es Salaam, Tanzania. Both HLC and EGs were used to estimate the proportion of human exposure to mosquitoes occurring indoors (π i ), as well as its two underlying parameters: the proportion of mosquitoes caught indoors (P i ) and the proportion of mosquitoes caught between the first and last hour when most people are indoors (P fl ).Results: HLC and EGs methods accounted for 69% and 31% of the total number of female mosquitoes caught respectively and both methods caught more mosquitoes outdoors than indoors. Results from the gold standard HLC suggest that An. gambiae s.s. in Dar es Salaam is neither exophagic nor endophagic (P i ≈ 0.5), whereas An. arabiensis is exophagic (P i < < 0.5). Both species prefer to feed after 10pm when most people are indoors (P fl > > 0.5). EGs yielded estimates of P i for An. gambiae s.s., An. arabiensis and An. coustani, that were approximately equivalent to those with HLC but significantly underestimated P fl for An. gambiae s.s. and An. coustani. The relative sampling sensitivity of EGs declined over the course of the night (p ≤ 0.001) for all mosquito taxa except An. arabiensis.Conclusions: Commercial EGs sample human-seeking mosquitoes with high sensitivity both indoors and outdoors and accurately measure the propensity of Anopheles malaria vectors to bite indoors rather than outdoors. However, further modifications are needed to stabilize sampling sensitivity over a full nocturnal cycle so that they can be used to survey patterns of human exposure to mosquitoes.