Orsborne et al. Malar J (2018) 17:479 https://doi.org/10.1186/s12936-018-2632-7 Malaria Journal RESEARCH Open Access Using the human blood index to investigate host biting plasticity: a systematic review and meta-regression of the three major African malaria vectors James Orsborne1, Luis Furuya‑Kanamori2, Claire L. Jeffries1, Mojca Kristan1, Abdul Rahim Mohammed3, Yaw A. Afrane3, Kathleen O’Reilly1, Eduardo Massad4, Chris Drakeley5, Thomas Walker1 and Laith Yakob1* Abstract Background: The proportion of mosquito blood‑meals that are of human origin, referred to as the ‘human blood index’ or HBI, is a key determinant of malaria transmission. Methods: A systematic review was conducted followed by meta‑regression of the HBI for the major African malaria vectors. Results: Evidence is presented for higher HBI among Anopheles gambiae (M/S forms and Anopheles coluzzii/An. gam- biae sensu stricto are not distinguished for most studies and, therefore, combined) as well as Anopheles funestus when compared with Anopheles arabiensis (prevalence odds ratio adjusted for collection location [i.e. indoor or outdoor]: 1.62; 95% CI 1.09–2.42; 1.84; 95% CI 1.35–2.52, respectively). This finding is in keeping with the entomological literature which describes An. arabiensis to be more zoophagic than the other major African vectors. However, analysis also revealed that HBI was more associated with location of mosquito captures (R2 = 0.29) than with mosquito (sibling) species (R2 = 0.11). Conclusions: These findings call into question the appropriateness of current methods of assessing host preferences among disease vectors and have important implications for strategizing vector control. Keywords: Blood meal analysis, Host preference, Mosquito, Biting preference, Blood index Background are typically assessed using PCR to identify the host spe- Malaria is transmitted through mosquito bites, making cies from blood-meals in field-caught mosquitoes, and the vectors’ choice of which blood-host species to bite are then quantified according to the human blood index a central component of malaria epidemiology and ecol- (HBI), defined as the proportion of blood-meals that ogy. In Africa, the majority of infections are transmit- are of human origin [1]. Because two mosquito bites on ted by Anopheles gambiae sensu stricto (s.s.), Anopheles a human are required to complete the malaria parasite’s coluzzii, Anopheles funestus and Anopheles arabiensis. life-cycle, HBI has an inflated impact on metrics of trans- Conventional wisdom indicates that the first three vec- mission such as the basic reproduction number, the vec- tors are anthropophagic while the latter sibling species torial capacity and the critical density of mosquitoes for is more zoophagic. Levels of anthropophagy/zoophagy sustained transmission [2]. However, the HBI should not be perceived to have a singular, fixed value; all major African malaria vectors *Correspondence: Laith.yakob@lshtm.ac.uk 1 Department of Disease Control, London School of Hygiene & Tropical have demonstrable plasticity in the host species that they Medicine, London, UK bite [3–5]. It has long been recognized that the same Full list of author information is available at the end of the article © The Author(s) 2018. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/ publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Orsborne et al. Malar J (2018) 17:479 Page 2 of 8 mosquito population will often adjust its biting towards a has never been more important to understand the distri- more locally available host species [1, 6]. This has impor- bution of malaria-vector bites on alternative host species. tant implications for malaria control policy. For example, Here, the current evidence is systematically reviewed and recent studies have observed that increased outdoor bit- a meta-regression conducted to identify the factors asso- ing followed the distribution of insecticide-treated bed ciated with higher HBI in sub-Saharan Africa. nets [7]. In such circumstances, vector control tools that operate effectively outdoors become a critical component Methods for eliminating local malaria transmission. Unfortunately, Findings from the systematic review were reported fol- the huge malaria burden reduction achieved in the years lowing the PRISMA guidelines [16]. The inclusion and since 2000 has relied disproportionately on control tools exclusion criteria are listed in Table  1 and advanced operating indoors [8], and there are limited effective search terms were developed following initial manual malaria-vector control options for outdoor use. literature searches and a basic PubMed search (Table 2). One technology that shows promise for targeting The purpose of the initial search was to identify key- mosquitoes regardless of whether they bite indoors or words and synonyms. The authors agreed on the search outdoors involves the use of systemic insecticides— terms and inclusion/exclusion criteria before the system- chemicals applied directly to blood-hosts to kill mosqui- atic search was performed. The Ovid database was used toes that take a blood meal. This technology arose from to search available MEDLINE and EMBASS literature the observation that mosquito mortality was increased from inception to February 2018. Books were excluded following the consumption of sugar-meals [9] or blood- from all searches as well as articles not written in English. meals [10] containing ivermectin—a drug used for Results were retrieved and collated using Mendeley desk- onchocerciasis control. Drugs approved for veterinary top reference manager. use, such as fipronil, have subsequently been demon- After eliminating duplications, abstracts for all publi- strated to have similar impact when livestock are dosed cations retrieved were reviewed for relevance. Full-text orally, or when the chemical is applied topically [11]. reviews were then conducted on all articles to decide More recently, systemic insecticides have had durations on its inclusion in accordance with the pre-specified of their efficacy extended through dosing with higher inclusion and exclusion criteria. If the inclusion criteria concentrations [12], combined dosing with adjuvants were satisfied the estimated human blood index (HBI) [13], and with use of sustained-release devices [14]. The reported was retrieved. Other variables that could have a stage is set for progress in development and evaluation of significant effect on the reported HBI were also retrieved. ivermectin for vector control [15]. Therefore, arguably it These variables included (sibling) species (complex), Table 1 Inclusion and exclusion criteria for systematic review Inclusion criteria Exclusion criteria Studies which used blood meal analysis (PCR, ELISA or precipitin tests) to Semi field studies, studies using baited traps or choice experiments to report the HBI investigate host preference Studies performed in sub‑Saharan Africa Entomological studies not specifically reporting the HBI Studies reporting the HBI for individual mosquito species Studies not reporting total number of mosquitoes caught Reporting HBI for Anopheles gambiae, Anopheles funestus complex or Data points based on less than 50 blood‑fed mosquitoes in total for target Anopheles arabiensis mosquito species species Studies reporting trapping methodology including location of traps (indoors or outdoors) Table 2 Search strategy for systematic review Ovid MEDLINE® Database Human blood index OR HBI OR host preference OR trophic preference OR blood meal preference OR blood host preference OR blood meal OR blood meal analysis OR blood‑meal analysis OR blood meal source OR host blood OR host blood meal OR blood meal identification [multiple posting = MeSH subject heading word, abstract, title, original title, text word (title, abstract), key word heading, name of substance, key word heading word, protocol supplementary concept word, synonym] AND Anopheles OR Anopheles arabiensis OR Anopheles gambiae OR Anopheles funestus [multiple posting = MeSH subject heading word, abstract, title, original title, text word (title, abstract), key word heading, name of substance, key word heading word, protocol supplementary concept word, syno‑ nym] Orsborne et al. Malar J (2018) 17:479 Page 3 of 8 trapping location (indoors, outdoors or both), trap methods had no significant effect (p > 0.05) on the type(s) used and total number of mosquitoes collected. reported HBI when comparing the mean HBI produced The primary effect measure of interest was the HBI. by each collection methodology within its respective col- The double arcsine square root transformed HBI lection areas (indoor and outdoor) for An. gambiae, An. (expressed as a proportion of all blood-meals) was used arabiensis and the An. funestus species complex (Fig. 2). to stabilize the variance across the studies [17] and It should be noted that due to the variety of different then back transformed for ease of interpretation. A lin- methods used and therefore sparsity of data for each ear model was performed on all eligible studies to gain methodology within the “both” categories, a meaningful additional insight into the effect of trapping location and comparison could not be made. Anopheles species on the proportion of HBI. The linear Meta-regression of the data compiled from the 166 model was fit using the HBI (proportion) as the response data points demonstrated a significantly higher propor- variable weighted by the inverse of each study’s vari- tion of blood-meals were of human origin (the human ance to allow the observations with the least variance to blood index, ‘HBI’) among An. funestus (prevalence odds provide the most information to the model, and using ratios [POR] of 1.84 (95% CI 1.35–2.52, p < 0.001) and An. robust error variances. All tests were two-tailed and a gambiae (POR of 1.62, 95% CI 1.09–2.52, p = 0.02) com- p-value < 0.05 was deemed statistically significant. Inverse pared to An. arabiensis. The majority of studies includ- variance weights were obtained using MetaXL (version ing details of An. gambiae did not specify whether they 5.3, EpiGear Int Pty Ltd; Sunrise Beach, Australia) and were M or S forms (or, in more modern nomenclature, the regression models were run using Stata MP (version An. coluzzii or An. gambiae s.s.), so these were com- 14, Stata Corp, College Station, TX, USA). bined. For all three groups, a significantly higher HBI was found from indoor mosquito collections (POR of 2.74, Results 95% CI 2.00–3.75, p < 0.001) as well as combined indoor The search identified 1243 potentially relevant studies. and outdoor collections (POR of 4.20, 95% CI 3.13–5.62, After collating these results and reviewing all abstracts, p < 0.001) versus outdoor only collections. Anopheles spe- 662 studies were deemed relevant. All full text articles cies was not found to be an interaction term for location were retrieved, reviewed for relevance and reviewed collection and HBI, indicating that all species follow a against all inclusion and exclusion criteria. Sixty-one similar trend regarding their preferred location for biting studies resulting in 166 data points fulfilled all criteria humans. The results also revealed that trapping location and where included in the analysis. Reasons for exclusion (R2 of 0.29) had a larger impact on the blood-meal host at full text stage included inadequate number (fewer than species than mosquito species (or species complex) ( R2 of 50) of mosquitoes collected (n = 14) and the use of host- 0.11) and that this difference was statistically significant biased trapping methodologies (n = 4) (Fig. 1). (p < 0.01 resulting from an F-test comparing both univar- Multiple collection methodologies were identified from iable models) (Table 4). the eligible studies. The methodology used was governed by the collection location targeted (indoors or outdoors). Discussion Indoor collections were the most widely used (n = 118) Control of vector-borne diseases is largely, often entirely, with pyrethroid spray catch (PSC) the most commonly dependent on vector control. For malaria, vector control used methodology (n = 78). Other collection methods is achieved primarily through targeting mosquitoes that included manual indoor collections (n = 20) and the use are host-seeking [8]. The major African malaria vectors, of CDC light traps within the household (n = 10). Out- An. gambiae s.s., An. coluzzii and An. funestus, are regu- door collections represented 27 of the total data points larly cited as paragons of anthropophagy, and any non- extracted with manual collection of mosquitoes being human biting exhibited by these species has historically the most common collection method (n = 13). Pit traps been ignored when strategizing control. Here, their bit- (n = 10) and CDC light traps (n = 4) were also an effec- ing behaviour was systematically reviewed and clearly tive collection method. Studies collecting from both demonstrated that the difference in their host choice indoor and outdoor environments consisted of 21 data compared with the zoophilic vector An. arabiensis was points. These studies used a variety of different methods; dwarfed by the difference found when comparing indoor many used a combination of the most effective indoor with outdoor collections. In other words, where the mos- and outdoor collection methods. CDC light traps were quito was collected was substantially and significantly the most common (n = 12) followed by other combina- more influential on host choice than which mosquito spe- tions of indoor and outdoor methods; CDC light trap cies was collected. plus PSC (n = 2) and pit traps and manual indoor collec- This raises an important question: where should vectors tions (n = 2) (Table  3 and Additional file  1). Collection be collected from in order to provide the most useful HBI Orsborne et al. Malar J (2018) 17:479 Page 4 of 8 Fig. 1 PRISMA flow diagram of search phases with numbers of studies included/excluded at each stage estimates? Results indicate that a single HBI for a given Longitudinal assessments initiated before rolling out location risks presenting quite a biased estimate for local control tools, and followed up over the time course of the vector biting behaviour. A standardized HBI accounting programme would provide a valuable source of informa- for both indoor and outdoor behaviours would probably tion. For example, these would determine the timeframe constitute an invalid metric because of the increased dif- across which LLIN-derived exophagy [7], as well as zoo- ficulty posed by collecting blood-fed mosquitoes out- phagy [18] occurs, as well as provide unbiased estimates doors i.e., tools are lacking for the estimation of indoor of the magnitude of effect. These entomological data versus outdoor mosquito numbers with any confidence. would also be able to inform on whether there is a rever- Therefore, current best practice should be to present sion to behavioural norm after a certain period post-dis- both estimates for an indoor HBI and an outdoor HBI. tribution, and the rate at which this occurred. Orsborne et al. Malar J (2018) 17:479 Page 5 of 8 Table 3 Data points extracted from eligible studies for each collection location and trapping methodology Species Collection location Methodology Indoor Outdoor Both (indoor + outdoor) Indoor Outdoor Both Pyrethroid spray Manual CDC light Manual CDC light Pit traps CDC light CDC light trap Others catch (PSC) aspiration trap aspiration trap trap + PSC Anopheles gambiae 37 7 3 22 10 2 7 0 0 1 0 2 Anopheles arabiensis 50 14 10 39 2 7 1 4 9 0 10 0 Anopheles funestus s.l. 31 6 8 17 8 1 5 0 1 1 2 5 The top three methodologies based on number of data points extracted are displayed here Orsborne et al. Malar J (2018) 17:479 Page 6 of 8 Fig. 2 Reported mean HBI (+ 95% CIs) for individual collection methods when sampling from indoor, outdoor or both (indoor and outdoor) environments for An. gambiae, An. arabiensis and An. funestus Better data on this behaviour and its temporality will availability of alternative blood meals increases mos- do much more than inform a fundamental aspect of quito survival more than counters the impact of divert- mosquito ecology: it will have considerable ramifica- ing feeds. This risk could be reduced or eliminated with tions pertaining to malaria control. For example, if sig- systemic insecticidal dosing that is judiciously timed with nificantly reduced HBI is detected immediately following LLIN roll-out. Mathematical models already exist for the distribution of LLINs, this may present an excellent optimal systemic insecticide deployment [20] including opportunity to synergize bed nets with systemic insecti- its integration with LLINs [21]. These could immediately cide-treated livestock. Saul [19] described the potential be capitalized upon once the temporal HBI data became for zooprophylaxis to switch into zoopotentiation if the available. Orsborne et al. Malar J (2018) 17:479 Page 7 of 8 Table 4 Predictors of  human blood index: univariable Additional file and multivariable regression models Predictors Univariable Multivariable Additional file 1. All eligible studies (and corresponding data points) retrieved from systematic search. POR (95% CI) R2 POR (95% CI) R2 Anopheles species 0.11 0.40 Authors’ contributions An. arabiensis 1.00 1.00 LY conceived the study. JO, LFK and LY performed the systematic review and meta‑regression. All authors contributed to results interpretation and manu‑ An. gambiaea 1.50 (0.95–2.36) 1.62 (1.09–2.42) script drafting. All authors read and approved the final manuscript. An. funestus 1.95 (1.34–2.85) 1.84 (1.35–2.52) Location 0.29 Author details1 Department of Disease Control, London School of Hygiene & Tropical Medi‑ Outdoor only 1.00 1.00 cine, London, UK. 2 Department of Population Medicine, College of Medicine, Indoor only 2.83 (2.04–3.93) 2.74 (2.00–3.75) Qatar University, Doha, Qatar. 3 Department of Medical Microbiology, College of Health Sciences, University of Ghana, Korle Bu, Accra, Ghana. 4 School Both 3.98 (3.14–5.06) 4.20 (3.13–5.62) of Applied Mathematics, Fundacao Getulio Vargas, Rio de Janeiro, Brazil. 5 POR prevalence odds ratio Department of Immunology & Infection, London School of Hygiene & Tropi‑ a Most studies did not specify M/S form of An. gambiae (and pre-dated the cal Medicine, London, UK. renaming of these forms as An. coluzzii and An. gambiae s.s., respectively) Acknowledgements Not applicable. One further, important unknown pertaining to HBI Competing interests is the spatial scale across which within-mosquito pop- The authors declare that they have no competing interests. ulation plasticity occurs. Over 50  years ago, Garrett- Availability of data and materials Jones described differing HBI estimates for mosquitoes All data generated or analysed during this study are included in this published collected from proximal locations [1]. Given the cur- article and its additional file. rent concerns over altered biting behaviour poten- Consent for publication tially compromising recent gains in malaria burden Not applicable. reduction [22], a fuller comprehension of the scale and magnitude of this variability is timely. A recent study Ethics approval and consent to participateNot applicable. conducted in southern Ghana describes the success- ful piloting of a novel experimental design to address Funding exactly this phenomenon [23]. It demonstrated that JO has an MRC London Intercollegiate Doctoral Training Partnership Student‑ship. TW and CLJ are funded through a Wellcome Trust/Royal Society Sir Henry statistically significant alteration in host choice for Dale Fellowship (101285/Z/13/Z) awarded to TW. LY received funds from a An. coluzzii was detectable over a range of 250 m [23]. Royal Society Research Project (RSG\R1\180203). Funding bodies had no role Heterogeneity in mosquito biting rates has been dem- in the design of the study and collection, analysis and interpretation of data nor in writing the manuscript. onstrated to be key to malaria transmission, first by theoretical work [24], but more recently with empirical Publisher’s Note studies using genotyping of blood-meals [25]. Future Springer Nature remains neutral with regard to jurisdictional claims in pub‑ modelling frameworks will need to account for this lished maps and institutional affiliations. additional form of village-level heterogeneity in biting Received: 4 October 2018 Accepted: 14 December 2018 behaviour. Conclusion References Results demonstrate that where mosquitoes are col- 1. Garrett‑Jones C. The human blood index of malaria vectors in lected from (indoors versus outdoors) is significantly relation to epidemiological assessment. Bull World Health Organ. 1964;30:241–61. more associated with the HBI than which of the major 2. Waite H. Mosquitoes and malaria. A study of the relation between the African malaria-vector mosquito (sibling) species is number of mosquitoes in a locality and the malaria rate. Biometrika. collected. Some of the more important consequences 1910;7:421–36. 3. Bogh C, Pedersen EM, Mukoko DA, Ouma JH. Permethrin‑impregnated to disease control of this behaviour are described. bednet effects on resting and feeding behaviour of lymphatic filariasis Some new theoretical and empirical developments vector mosquitoes in Kenya. Med Vet Entomol. 1998;12:52–9. that may improve both HBI assessment and how this 4. Takken W, Verhulst NO. Host preferences of blood‑feeding mosquitoes. Annu Rev Entomol. 2013;58:433–53. metric can inform malaria control optimisation are 5. Lefèvre T, Gouagna L‑C, Dabiré KR, Elguero E, Fontenille D, Renaud F, discussed. et al. Beyond nature and nurture: phenotypic plasticity in blood‑feeding behavior of Anopheles gambiae s.s. when humans are not readily acces‑ sible. Am J Trop Med Hyg. 2009;81:1023–9. Orsborne et al. Malar J (2018) 17:479 Page 8 of 8 6. Chandler JA, Boreham PF, Highton RB, Hill MN. A study of the host selec‑ 15. Rabinovich NR. Ivermectin: repurposing an old drug to complement tion patterns of the mosquitoes of the Kisumu area of Kenya. Trans R Soc malaria vector control. Lancet Infect Dis. 2018;18:584–5. Trop Med Hyg. 1975;69:415–25. 16. Moher D, Liberati A, Tetzlaff J, Altman DG, The PG. Preferred reporting 7. Russell TL, Govella NJ, Azizi S, Drakeley CJ, Kachur SP, Killeen GF. Increased items for systematic reviews and meta‑analyses: the PRISMA statement. proportions of outdoor feeding among residual malaria vector popula‑ PLoS Med. 2009;6:e1000097. tions following increased use of insecticide‑treated nets in rural Tanzania. 17. Barendregt JJ, Doi SA, Lee YY, Norman RE, Vos T. Meta‑analysis of preva‑ Malar J. 2011;10:80. lence. J Epidemiol Comm Health. 2013;67:974–8. 8. Bhatt S, Weiss DJ, Cameron E, Bisanzio D, Mappin B, Dalrymple U, et al. 18. Pombi M, Calzetta M, Guelbeogo WM, Manica M, Perugini E, Pichler V, The effect of malaria control on Plasmodium falciparum in Africa between et al. Unexpectedly high Plasmodium sporozoite rate associated with low 2000 and 2015. Nature. 2015;526:207–11. human blood index in Anopheles coluzzii from a LLIN‑protected village in 9. Pampiglione S, Majori G, Petrangeli G, Romi R. Avermectins, MK‑933 and Burkina Faso. Sci Rep. 2018;8:12806. MK‑936, for mosquito control. Trans R Soc Trop Med Hyg. 1985;79:797–9. 19. Saul A. Zooprophylaxis or zoopotentiation: the outcome of introducing 10. Chaccour C, Lines J, Whitty CJM. Effect of ivermectin on Anopheles animals on vector transmission is highly dependent on the mosquito gambiae mosquitoes fed on humans: the potential of oral insecticides in mortality while searching. Malar J. 2003;2:32. malaria control. J Infect Dis. 2010;202:113–6. 20. Yakob L. Endectocide‑treated cattle for malaria control: a coupled ento‑ 11. Poche R, Burruss D, Polyakova L, Poche D, Garlapati R. Treatment of mological–epidemiological model. Parasit Epidemiol Control. 2016;1:2–9. livestock with systemic insecticides for control of Anopheles arabiensis in 21. Yakob L, Cameron M, Lines J. Combining indoor and outdoor methods western Kenya. Malar J. 2015;14:351. for controlling malaria vectors: an ecological model of endectocide‑ 12. Smit MR, Ochomo EO, Aljayyoussi G, Kwambai TK, Abong’o BO, Chen T, treated livestock and insecticidal bed nets. Malar J. 2017;16:114. et al. Safety and mosquitocidal efficacy of high‑dose ivermectin when 22. Sokhna C, Ndiath MO, Rogier C. The changes in mosquito vector co‑administered with dihydroartemisinin‑piperaquine in Kenyan adults behaviour and the emerging resistance to insecticides will challenge the with uncomplicated malaria (IVERMAL): a randomised, double‑blind, decline of malaria. Clin Microbiol Infect. 2013;19:902–7. placebo‑controlled trial. Lancet Infect Dis. 2018;18:615–26. 23. Orsborne J, Furuya‑Kanamori L, Jeffries CL, Kristan M, Mohammed AR, 13. Chaccour CJ, Hammann F, Alustiza M, Castejon S, Tarimo BB, Abizanda Afrane YA, et al. Assessing the blood‑host plasticity and dispersal rate of G, et al. Cytochrome P450/ABC transporter inhibition simultaneously the malaria vector Anopheles coluzzii. bioRxiv. 2018;26:49–52. enhances ivermectin pharmacokinetics in the mammal host and phar‑ 24. Dye C, Hasibeder G. Population dynamics of mosquito‑borne disease: macodynamics in Anopheles gambiae. Sci Rep. 2017;7:8535. effects of flies which bite some people more frequently than others. Trans 14. Chaccour CJ, Ngha’bi K, Abizanda G, Irigoyen Barrio A, Aldaz A, Okumu F, R Soc Trop Med Hyg. 1986;80:69–77. et al. Targeting cattle for malaria elimination: marked reduction of Anoph- 25. Guelbéogo WM, Gonçalves BP, Grignard L, Bradley J, Serme SS, Hellewell eles arabiensis survival for over 6 months using a slow‑release ivermectin J, et al. Variation in natural exposure to anopheles mosquitoes and its implant formulation. Parasit Vectors. 2018;11:287. effects on malaria transmission. eLife. 2018;7:e32625. Ready to submit your research ? Choose BMC and benefit from: • fast, convenient online submission • thorough peer review by experienced rese archers in your field • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations • maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions