quantifying mosquito biting patterns on humans by dna

Am. J. Trop. Med. Hyg., 65(6), 2001, pp. 722–728
Copyright 䉷 2001 by The American Society of Tropical Medicine and Hygiene
Wellcome Trust Centre for the Epidemiology of Infectious Disease, Department of Zoology, University of Oxford, Oxford,
United Kingdom; Vector Control Research Centre (ICMR), Pondicherry, India; Department of Zoology,
University of Cambridge, Cambridge, United Kingdom
Abstract. A major debate in infectious disease epidemiology concerns the relative importance of exposure and
host factors, such as sex and acquired immunity, in determining observed age patterns of parasitic infection in endemic
communities. Nonhomogeneous contact between hosts and vectors is also expected to increase the reproductive rate,
and hence transmission, of mosquito-borne infections. Resolution of these questions for human parasitic diseases has
been frustrated by the lack of a quantitative tool for quantifying the exposure rate of people in communities. Here,
we show that the polymerase chain reaction (PCR) technique for amplifying and fingerprinting human DNA from
mosquito bloodmeals can address this problem for mosquito-borne diseases. Analysis of parallel human and mosquito
(resting Culex quinquefasciatus) samples from the same households in an urban endemic focus for bancroftian filariasis in South India demonstrates that a 9-locus radioactive short-tandem repeat system is able to identify the source
of human DNA within the bloodmeals of nearly 80% of mosquitoes. The results show that a person’s exposure rate,
and hence the age and sex patterns of exposure to bites in an endemic community, can be successfully quantified by
this method. Out of 276 bloodmeal PCR fingerprints, we also found that on average, 27% of the mosquitoes caught
resting within individual households had fed on people outside the household. Additionally, 13% of mosquitoes biting
within households contained blood from at least 2 people, with the rate of multiple feeding depending on the density
of humans in the household. These complex vector feeding behaviors may partly account for the discrepancies in
estimates of the infection rates of mosquito-borne diseases calculated parasitologically and entomologically, and they
underline the potential of this tool for investigating the transmission dynamics of infection.
flicting results regarding the rate of biting on people by mosquitoes in the field. Thus, although some studies have indicated feeding to be random,15,16 others have found adults to
be more attractive than children to mosquitoes.11–14
In contrast, recent developments in molecular ecology
centered on the polymerase chain reaction (PCR)–based fingerprinting of mosquito bloodmeals17 have suggested that
this approach may provide a sensitive method for quantifying the mosquito biting rate of people. Previous applications
of this method have highlighted its potential for addressing
epidemiologically important questions regarding malaria
transmission. Gokool and others18 applied the tool to small
field samples in malaria bed-net trials and demonstrated the
usefulness of the technique for assessing protection provided
by impregnated nets. More recently, Koella and others19 successfully used the method to investigate the ability of Plasmodium falciparum to increase multiple feeding on people
by mosquitoes.
Here, we evaluate for the first time the usefulness of this
method as an epidemiological tool for quantifying the biting
rate on humans in the field by the mosquito species, Culex
quinquefasciatus, an important vector of bancroftian filariasis in many parts of the world. This investigation forms an
integral part of a larger study designed to quantify the impact
and sources of exposure heterogeneity on the immunoepidemiology of filariasis in South India.
Progress in resolving the long-standing epidemiological
debate regarding the relative importance of exposure, acquired immunity, and other factors, such as hormonal effects, in determining observed age and sex patterns of parasitic infection or disease depends on the ability to reliably
quantify the exposure rate of people.1–8 Recent theoretical
work9 has also shown that nonhomogeneous mixing (nonrandom contact) between vectors and hosts has important
implications for the persistence, prevalence, and hence control of mosquito-borne diseases. In particular, when vectors
concentrate on certain hosts, the basic reproductive rate (R)
of the disease (a measure of persistence) and the vectorial
capacity both have the potential to be greater than under
conditions of homogeneous mixing, thus making disease
control more difficult.9,10
Despite these findings and the long-held popular perception by both casual visitors to the tropics and students of
mosquito biting behavior that individual humans differ
markedly in their attractiveness to mosquito bites, there have
been few systematic attempts to investigate the patterns of
human biting by hematophagous insects under natural field
conditions. A major drawback has been the lack of a sensitive tool for quantifying biting rates on people at an epidemiologically relevant scale in the field. Previous methods
of assessing field-biting rates on individual humans by mosquitoes have depended on either making landing catches on
exposed people11,12 or crudely identifying the human source
of mosquito bloodmeals by typing for human blood groups
or other polymorphic blood.13,14 The former, a laborious and
ethically dubious technique, is constrained by biases between catchers; the latter is limited by the degree of polymorphism of the blood markers used. This has led to con-
Study area and samples. Field samples for this study
were collected from 19 randomly selected houses in Muthialpet, a moderately dense neighborhood of Pondicherry,
South India. All household investigations were made with
the consent of the residents, or in the case of children, their
guardians. Ethical approval for the study was obtained from
the institutional ethical committee of the Vector Control Research Center (Indian Council of Medical Research). For all
residents in each house, details of age, sex, and familial relationships were recorded during an initial survey before the
collection of up to 1 mL of blood in heparinized Vacutainers.
All blood samples were frozen within an hour of collection.
Female Cx. quinquefasciatus were collected inside these
houses between 7:30 AM and 9:30 AM. During this time,
mosquitoes are believed to rest on the walls and ceilings
after feeding during the night. Mosquitoes were caught by
means of battery-operated aspirators20; each collector spent
⬃ 10 minutes per house. Mosquitoes from each house were
preserved individually within an hour in 100% isopropanol
in 500 ␮L Eppendorf tubes. Samples were stored at room
temperature until used.
Mosquito bloodmeal dissections. Mosquito body parts
and abdomen membranes were removed from the bloodmeal
in order to remove potential PCR inhibitors. Each bloodmeal
was sized according to one of the following size categories:
one-quarter full, half full, or full.19
DNA extraction. Aliquots (200 ␮L) whole blood and
each mosquito bloodmeal were digested in 200 ␮L 50 mM
Tris-HCl (pH 8.0) containing 1% sodium dodecyl sulfate–
proteinase K. The bloodmeals were crushed in the digest
with micropestles to aid the digestion of clotted blood. All
samples were incubated at 37⬚C overnight. The digested
products were purified by treating with equal volumes of
phenol-chloroform and chloroform–isoamyl alcohol, followed by the addition of an equal volume of LiCl; DNAs
were precipitated by the ethanol precipitation method.21
Multiplex PCR. We developed and optimized a 9-locus
multiplex short-tandem repeat (STR) system derived from
the literature22–24 to undertake the PCR analysis of bloodmeals in this study. The commercially synthesized markers
D21S11, D18S51, D8S1179, HUMF13A1, HUMFES/FPS,
and D3S1358 (MWG, Ebersberg, Germany). For PCR amplification, one of the primers for each marker was end-labeled with [␥33P]-adenosine triphosphate (ICN/FLOW,
Thame, UK) and T4 DNA polynucleotide kinase (NE Biolabs, Hitchin, UK). PCR amplification was performed with
1–50 ng of genomic DNA in a 25-␮L reaction volume. Reactions consisted of 1⫻ PARR buffer (10 mM Tris-HCl, pH
8.3, 50 mM KCl, 1.5 mM MgCl, 1% Triton-X 100 [Cambio
Laboratories, Cambridge, UK]), 1.25 U AmpliTaq Gold
polymerase (Perkin-Elmer, Norwalk, CT), 200 ␮M of each
dNTP (Boehringer, Penzberg, Germany), 0.25 ␮M of each
HUMVWA31/A primer, 0.25 ␮M of each D8S1179 primer,
0.25 ␮M of each D21S11 primer, 0.2 ␮M of each HUMTHO1 primer, 0.125 ␮M of each D3S1358 primer, 0.125 ␮M
of each HUMFES primer, 0.1 ␮M of each HUMF13 primer,
0.06 ␮M of each D18S51 primer and 0.05 ␮M of each
HUMFIBRA primer.
Amplification reactions were carried out in thin-walled
Gene-Amp reaction tubes (Perkin-Elmer) on a Peltier Thermal Cycler 200 DNA Engine (MJ Research, Inc, Waltham,
MA)—and consisted of an initial denaturation step of 95⬚C
for 9 min and 32 cycles of 93⬚C for 30 sec, 57⬚C for 85 sec,
and 72⬚C for 15 sec, followed by a final 10-min incubation
at 72⬚C. Samples were mixed with 0.4 volume of formamide
dye solution and denatured for 2 min at 95⬚C; then, 2 ␮L
was loaded on a 6% acrylamide gel. After electrophoresis,
the gel was placed onto Whatman 3MM paper, dried, and
exposed to Fuji X-ray film by means of intensifying screens
at ⫺70⬚C. We defined a meal as single if all the identified
alleles came from one person. A meal was defined as multiple if at least one locus had ⬎ 2 alleles or if the alleles at
any of the loci could only have been derived from biting
different people in a single household.19 If none of the identified alleles were found among the inhabitants of a household, the meal was considered to have been taken from people outside the particular household.
Statistical analyses. We used logistic and multiple logistic regressions to test for the effects of covariates on binary
or proportional response data. For testing the effects of age
and sex on the probability that a mosquito took a quartersized meal, we specified a logistic regression for clustered
data25 with each individual person forming the cluster to account for the lack of independence for the data recorded per
individual person. Generalized additive models with binomial errors were used to examine nonlinearity in these data.26
Count data (numbers of mosquitoes and biting density per
person) were analyzed by either quasi-likelihood Poisson
generalized linear models27 or quasi-likelihood Poisson generalized additive models (to uncover nonlinear structure).
Quasi-likelihood estimation was used to account for the significant overdispersion occurring in such data, which may
partly reflect the likely clustering of responses within individual households. Note that we simply account for such
responses here via the quasi-likelihood estimation. Pearson’s
chi-square test was used to assess the goodness of fit of the
negative binomial and Poisson probability distributions to
the observed frequency distribution of estimated bites per
The PCR method uses human microsatellite (STR) primers23 to genotype the bloodmeals of mosquitoes caught while
resting in households and compares the resulting DNA fingerprints with those of the humans residing within the
same.17–19 We developed and optimized a 9-locus ␥33P multiplex STR system23 for this. Figure 1 illustrates an autoradiogram obtained by this system in a pilot study comparing
the DNA profiles of 4 people and 16 Cx. quinquefasciatus
vectors sampled from the same household in Pondicherry,
South India. This pilot investigation shows that the band
pattern of Individual E9 matches those obtained from bloodmeals of Mosquitoes 5, 8, and 31. Similarly, the pattern obtained for Individual E11 matches that of Mosquitos 17 and
22; E12 matches Mosquitoes 9 and 29; and Individual E15
had been bitten by Mosquito 11. By contrast, the fingerprint
obtained from Mosquito 16 contains bands or alleles from
both Individuals E11 and E12, providing evidence for mixed
feeding by this mosquito on these people. The profiles of
only 9 out of the 15 mosquitoes that were amplified (Figure
1) appear to match those obtained from people living in this
household, suggesting that the rest of the mosquitoes may
have fed elsewhere or on visitors.
We used this tool to quantify how biting rate varies with
host age and sex in human communities by collecting par-
FIGURE 1. Autoradiogram derived from applying the 9-locus multiplex single-tandem repeat system to human blood samples and bloodmeals
of the corresponding mosquitoes sampled from a pilot study household in Pondicherry, South India. The DNA banding patterns depicted in
the figure clearly show that polymerase chain reaction fingerprinting of Culex quinquefasciatus bloodmeals provides a reliable tool for identifying and hence estimating the biting rate by these mosquitoes on exposed people. Sample labels are arbitrary and were assigned randomly
to each human and mosquito blood sample. Lane 1 (left) represents results from the negative control (distilled water) used in this experiment.
The bloodmeal of Mosquito 14 failed to amplify.
allel blood and mosquito bloodmeal samples from 19 randomly selected households in an urban locality in Pondicherry (Table 1). A total of 361 mosquitoes were collected
from these households, with the number of blood-fed Culex
quinquefasciatus ranging 2–77 per house per night (Table
More than 75% of the bloodmeal samples from the study
households provided unambiguous PCR results (Table 1).
The earlier study by Koella and others19 that used this tool
had a higher failure rate of 33%. However, in contrast with
that study, our results indicated that the lack of success of
the PCR was a direct function of bloodmeal size. Of the 3
bloodmeal categories (fully fed, half fed, and one-quarter
fed) that we assigned to individual mosquitoes, the propor-
Details of human and mosquito samples and summary results from the bloodmeal analysis
House no.
No. of people
in household*
Total (19)
Sex (age in years)
F (15), M (25)
M (13, 19, 21, 23)
F (13, 19), M (7, 12, 25)
F (13, 15, 16, 17, 37, 40), M (18, 65)
F (18, 21, 24), M (14, 20, 29)
F (22, 27, 35), M (20, 60)
F (27, 28), M (13, 25, 39)
F (19), M (25)
F (18, 42), M (20)
F (13, 35), M (17)
F (22, 33, 45), M (20, 48)
F (28)
F (9, 14, 15, 38), M (13, 45)
F (17, 18, 35), M (50)
F (5, 26, 38, 56), M (8)
F (24), M (25)
F (22, 55), M (28)
F (40), M (23)
F (13, 24), M (17)
43 F, 33 M
No. of blood-fed
No. of unclear
bloodmeal samples (%)
No. feeding
households† (%)
No. with multiple
feeds‡ (%)
* Total number of inhabitants in each house, whether bitten or not.
† Out of the bloodmeals successfully genotyped.
‡ Out of mosquitoes feeding within households.
tion of unclear results increased from 16.9% (31 of 183) to
21.8% (26 of 119) to 49.1% (28 of 57) as the bloodmeal
size decreased. This inverse relationship was significant for
the overall data (logistic regression of PCR success on
bloodmeal category: chi-square ⫽ 20.05, degree of freedom
[df] ⫽ 1, P ⬎ 0.0001), but did not differ significantly between the categories of half and fully fed (chi-square ⫽ 1.34,
degrees of freedom [df] ⫽ 1, P ⫽ 0.25). Significant failures
of the PCR system therefore occurred only for insects containing quarter-sized bloodmeals. This is, however, unlikely
to influence the current analysis of age and sex effects because after accounting for clustering of bites on people, multiple logistic regression indicated that the observed proportion of one-quarter–fed mosquitoes did not vary significantly
either with the age (chi-square ⫽ 0.43, df ⫽ 1, P ⫽ 0.51)
or sex (chi-square ⫽ 4.41, df ⫽ 2, P ⫽ 0.11) of the bitten
The results indicate that on average, 27.5% of the mos-
FIGURE 2. Variations in household numbers of Culex quinquefasciatus (␱) found to have fed on people within households as a
function of (a) mosquito density and (b) human density.
quitoes caught within a household did not contain alleles of
the inhabitants of that household (Table 1). Interestingly, the
data in Table 1 suggest that the proportion of such mosquitoes may be high when both the densities of vectors caught
and the number of humans per household are either very
low or very high. The apparent proportional increase in such
mosquitoes at the highest household densities of both variables could, however, represent an anomaly because one
house (H43) contained the highest number of both humans
and mosquitoes (Table 1). When data from this house were
omitted, the number of mosquitoes found biting outside
houses did not differ significantly between the study households in relation to either the density of vectors (quasi-likelihood linear Poisson regression, F ⫽ 1.88, df ⫽ 1,16, P ⫽
0.188) or people per house (quasi-likelihood linear Poisson
regression, F ⫽ 0.22, df ⫽ 1,16, P ⫽ 0.645).
On the other hand, as shown in Figure 2, the number of
mosquitoes biting inhabitants within a house increased significantly and perhaps unsurprisingly with both these variables (quasi-likelihood linear multiple Poisson regression, F
⫽ 68.68, df ⫽ 1,16, P ⬍ 0.0001, and F ⫽ 7.58, df ⫽ 1,15,
P ⫽ 0.014, respectively, for density of mosquitoes and people per household). This implies that the apparent variation
in the proportion of resting mosquitoes found biting people
outside households (Table 1) was mainly the result of between-house differences in the numbers of mosquitoes biting
people inside households.
A second notable feature regarding the biting behavior of
Cx. quinquefasciatus concerns the number and proportion of
vectors taking multiple feeds (Table 1). Overall, 13.0% of
the mosquitoes biting within households had taken multiple
meals in our study sample (Table 1). This compares remarkably well with the findings of Koella and others,19 who found
14.0% of Anopheles gambiae taking multiple meals in their
African study site. Interestingly, our data show that this pro-
FIGURE 3. Variation in the proportions of household mosquitoes
taking multiple feeds (␱) as a function of the number of inhabitants
per house. Analysis of deviance of nested binomial generalized linear models indicated a significant negative effect for human density
(P ⬍ 0.0001) but not mosquito density (P ⫽ 0.062). Fit of the
generalized linear model (solid line) indicates the effect of household human density; 95% confidence intervals of the fitted model
are also shown (dashed line).
portion may be negatively associated with the number of
inhabitants but not the number of mosquitoes caught per
house (Figure 3). This suggests that interrupted feeding of
Cx. quinquefasciatus may be host density dependent.
Figure 4 depicts the frequency distribution of the number
of bites estimated for each person in this study. The total
number of estimated bites and the mean number of bites
received per person per night in the study population were
233 and 3.1, respectively. The numbers of bites on people
were highly overdispersed (variance to mean ratio ⫽ 5.5)
and were adequately described by the negative binomial
probability distribution (Figure 4). Indeed, ⬎ 55% of the
total bites were received by ⬍ 20% of the sampled people.
Figure 5 examines the variation in the density of mosquito
bites per night among people in relation to age and sex.
There was no significant difference in the overall number of
bites received by males and females (quasi-likelihood Poisson generalized linear model, F ⫽ 0.97, df ⫽ 1,74, P ⫽
0.327), but the age-specific biting rate differed markedly between the sexes (Figure 5). In male inhabitants, biting density showed some tendency to increase in inhabitants aged
between 10–30 years, but overall, the data showed little age
dependency, probably as a result of its high variability (Figure 5a). In the female population, by contrast, there was a
clear nonlinear pattern (Figure 5b): biting density increased
in girls up to the age of ⬃ 15 years before decreasing significantly among women.
Although the numbers of boys and men and girls and
women residing within the study households varied (Table
1), such sex ratio differences were not statistically significant
for either the overall study population (people of all ages)
(chi-square ⫽ 11.92, df ⫽ 18, P ⫽ 0.85) or for people aged
⬍ 25 years (chi-square ⫽ 14.27, df ⫽ 17, P ⫽ 0.65). This
suggests that disproportionate mosquito densities at the
FIGURE 4. Observed frequency distribution of the number of
mosquito bites received per person compared with expected distributions on the basis of the negative binomial and Poisson probability
distribution models. Sample size is 76 people, 233 bites. The distribution is overdispersed, with most people receiving ⬍ 5 bites per
night and a few attracting many bites. The negative binomial probability distribution gave a better fit to the data (chi-square ⫽ 4.64,
degree of freedom ⫽ 6, P ⫽ 0.59) with a k value of 1.03, compared
with the Poisson model (chi-square ⫽ 10.23, degree of freedom ⫽
5, P ⫽ 0.06).
household level are unlikely to bias the observed age-specific biting patterns depicted in Figure 5 for male and female
Our results confirm that DNA fingerprinting of bloodmeals can provide a sensitive method for identifying which
FIGURE 5. Age distribution of estimated Culex quinquefasciatus
biting density per night per person among males and females. Fits
of the respective linear (dotted line) and additive (solid line) quasilikelihood Poisson models to the age-biting intensity data (␱) are
shown for males in (a) and females in (b). The fits indicate some
evidence for a slight age effect in mosquito biting for males but
significant nonlinear age-dependency in biting density for females
(additive model significantly better fitting than the corresponding
linear model (statistic Q ⫽ 8.63, degree of freedom ⫽ 3.5, 44, P ⬍
0.0001).37 The 95% confidence bands of the respective additive models are indicated (dashed lines).
individual humans are bitten by which hematophagous arthropod under natural field conditions.17–19 They also demonstrate for the first time the usefulness of this technique as
an epidemiological tool for quantifying patterns of potential
human exposure to infection. Our analysis of this topic indicates that both variations in mosquito biting and resting
biology at the household level and intrinsic host factors, such
as age and sex, may underlie the natural biting rate of people
in urban endemic localities.
It is clear that the power of this method would be enhanced if the technical failure of the PCR for a proportion
of bloodmeals can be overcome. Our study has shown that
this problem may be connected to the low DNA content of
partially completed bloodmeals, although the previous study
by Koella and others19 did not find such a relationship despite the greater failure rate (33%) associated with their 3locus STR system compared with ours (23.5%). Perhaps this
is not only a reflection of the greater sensitivity of our 9
locus STR system to DNA quantity but also differences due
to the vector species (An. gambiae versus Cx. quinquefasciatus) studied,20 such as variations in nutritional and foraging strategies that may affect human-biting propensity (anthropophagism); frequency of feeding; bloodmeal size; and
digestion of intact DNA. We are currently investigating the
use of nested PCR protocols29 as a means to deal with the
low or even possibly degraded DNA content of incomplete
Nonetheless, this study provides the first clear demonstration to date that mosquitoes biting people under field conditions is not random and may be dependent on the person’s
age and sex. This contrasts with the mixed results of studies
conducted on selective feeding in the past.11–16 In part, the
outcome of past studies reflects the biases of previous methods used for assessing an individual host’s biting rate and in
part the fact that previous investigations (largely due to constraints imposed by previous tools) have tended to study relatively small groups of people. The high sensitivity of the
multiplex STR system for identifying bitten people, coupled
with its capacity for dealing relatively speedily with large
samples (which is required to overcome the high variability
of biting data; Figures 4 and 5) means that this newly developed molecular tool may overcome the drawbacks of past
studies and for the first time allow realistic investigations of
the 2 major unresolved questions connected with mosquitotransmitted diseases, namely the factors that underlie differential biting on humans by mosquitoes and the impact of
exposure on the immunoepidemiology of such infections.30,31
The results reported here provide several insights regarding the use of the tool for quantifying and assessing the
impact of these questions for the epidemiology of vectorborne infections. First, the finding that intrinsic host factors
may account for a major proportion of the observed variability in biting by mosquitoes in our study population is of
special relevance to attempts for understanding and modeling the impact of differential host selection by vectors on
the population dynamics of infection.9,10 Specifically, the
present result indicates that in small communities (such as
the present study community, which comprised a few families living on neighboring streets), the effects of variable
biting of hosts by vectors may be adequately addressed by
simply specifying and incorporating age- and sex-dependent
biting in models of infection.32 Whether such intrinsic host
factors may play a similarly large role in the observed variation in host biting rates in larger villages or towns—where
large spatial effects in mosquito biting could be expected to
play a bigger role in host biting heterogeneity—remains to
be seen, but it is clear that the developed PCR tool will
provide an important means to aid the successful investigation of this topic.
The results show that Cx. quinquefasciatus biting and resting behavior may also contribute to the household variation
in bites by this mosquito vector in urban endemic areas. The
analysis presented here shows that such variability may derive in part from the propensities of a significant proportion
of mosquitoes caught resting within households to have either taken bloodmeals elsewhere or to have fed on several
people within households. Although the study was not designed to examine the causes of these findings, it is pertinent
to note that such vector biting and resting behaviors are related to host densities, vector densities, or both at the household level. Such density-dependent vector behavioral patterns may play important roles in the transmission dynamics
of vector-borne diseases.19,33,34 For example, increased multiple feeding by mosquito vectors on different hosts at higher
human densities could clearly increase the transmission rate
of vector-borne infections within endemic communities.19,33
It could also increase the rate at which humans are infected
with multiple parasite clones or clusters, which patently will
have profound implications for the rate of parasite outcrossing and hence for the likely development and spread of drug
By contrast, the propensity of a portion of house-resting
mosquitoes to have taken bloodmeals elsewhere means that
there is unlikely to be a simple correlation between levels
of infection among house-resting mosquito populations and
host infection levels within the same household. This implies
that caution must be exercised when house-resting catches
are used to measure the actual exposure or biting rates of
people and could clearly help explain the discrepancies that
arise when inoculation rates of malaria in people, for example, are estimated entomologically (via assessment of infection levels in house-resting mosquitoes) and parasitologically.13,35,36
These findings and conclusions imply that if the transmission dynamics and control of mosquito-borne infections
are to be better understood and quantified, more thorough
studies of mosquito biting and resting biology at both the
individual and household levels will be required. The present
results suggest that the exquisite sensitivity and highthroughput capability of the newly developed PCR-based
technique for DNA fingerprinting of mosquito bloodmeals
may provide a valuable molecular epidemiological tool in
aiding the successful design and execution of such studies.
Acknowledgments: We thank the expert entomology team at the
Vector Control Research Centre for carrying out the mosquito collections. We also acknowledge the support and encouragement of
the Indian Council for Medical Research in the implementation of
this study.
Financial support: The study was financially supported through a
Medical Research Council (UK) Career Development Fellowship
awarded to E.M. D.A.P.B. and B.T.G. acknowledge the support of
the Wellcome Trust.
Authors’ addresses: E. Michael, G. Barker, M. R. Paul, and D. A.
P. Bundy, Wellcome Trust Centre for the Epidemiology of Infectious
Disease, Department of Zoology, University of Oxford, South Parks
Road, Oxford OX1 3PS, United Kingdom. K. D. Ramaiah, S. L.
Hoti, J. Yuvaraj, and P. K. Das, Vector Control Research Centre,
Indian Council for Medical Research, Medical Complex, Indira Nagar, Pondicherry 605 006, India. B. T. Grenfell, Department of Zoology, University of Cambridge, Downing Street, Cambridge CB2
3EJ, United Kingdom.
Reprint requests: E. Michael, Department of Infectious Disease Epidemiology, Imperial College Medicine School, Norfolk Place, London W2 1PG, United Kingdom. Telephone: ⫹44-0-207-594-3294,
Fax: ⫹44-0-207-402-2150 (e-mail: [email protected]).
1. Warren KS, 1973. Regulation of the prevalence and intensity of
schistosomiasis in man: immunity or ecology? J Infect Dis
127: 595–609.
2. Anderson RM, May RM, 1985. Herd immunity to helminth infection and implications for parasite control. Nature 315:
3. Bundy DAP, 1988. Gender-dependent patterns of infection and
disease. Parasitol Today 4: 186–189.
4. Alexander J, Stimson WH, 1988. Sex hormones and the course
of parasitic infection. Parasitol Today 4: 189–193.
5. Brabin L, 1990. Sex differences in susceptibility to lymphatic
filariasis and implications for maternal child immunity. J Epidemiol Infect 105: 335–353.
6. Woolhouse MEJ, Taylor P, Matanhire D, Chandiwana SK, 1991.
Acquired immunity and epidemiology of Schistosoma haematobium. Nature 351: 757–759.
7. Fulford AJC, Butterworth AE, Sturrock RF, Ouma JH, 1992. On
the use of age-intensity data to detect immunity to parasitic
infections, with special reference to Schistosoma mansoni in
Kenya. Parasitology 105: 219–227.
8. Michael E, Bundy DAP, 1998. Herd immunity to filarial infection is a function of vector biting rate. Proc R Soc Lond B
265: 855–860.
9. Dye C, Hasibeder G, 1986. Population dynamics of mosquitoborne disease: effects of flies which bite some people more
frequently than others. Trans R Soc Trop Med Hyg 80: 69–
10. Burkot TR, 1988. Non-random host selection by Anopheline
mosquitoes. Parasitol Today 4: 156–161.
11. Muirhead-Thomson RC, 1951. The distribution of anopheline
mosquito bites among different age groups. A new factor in
malaria epidemiology. Br Med J 1: 1114–1117.
12. Carnevale P, Frezil JL, Bosseno MF, Le Pont F, Lancien J, 1978.
Étude de l’agressivite d’Anopheles gambiae A en function de
l’age et du sexe des sujets humains. Bull World Health Organ
56: 147–154.
13. Bryan JH, Smalley ME, 1978. The use of ABO blood groups
as markers for mosquito biting studies. Trans R Soc Trop Med
Hyg 72: 357–360.
14. Port GR, Boreham PFL, 1980. The relationship of host size to
feeding by mosquitoes of the Anopheles gambiae Giles complex (Diptera: Culicidae). Bull Entomol Res 70: 133–144.
15. Smith A, 1956. The attractiveness of an adult and child to A.
gambiae. E Afr Med J 33: 409–410.
16. Clyde DF, Shute GT, 1958. Selective feeding habits of anophelines amongst Africans of different ages. Am J Trop Med
Hyg 7: 543–545.
17. Coulson RMR, Curtis CF, Ready PD, Hill N, Smith DF, 1990.
Amplification and analysis of human DNA present in mosquito bloodmeals. Med Vet Entomol 4: 357–366.
Gokool S, Smith DF, Curtis CF, 1992. The use of PCR to help
quantify the protection provided by impregnated bednets.
Parasitol Today 8: 347–350.
Koella JC, Sørensen FL, Anderson RA, 1998. The malaria parasite, Plasmodium falciparum, increases the frequency of
multiple feeding of its mosquito vector, Anopheles gambiae.
Proc R Soc Lond B 265: 763–768.
Service MW, 1993. Mosquito Ecology. Field Sampling Methods.
Second edition. London: Chapman and Hall.
Sambrook J, Fritsch EF, Maniatis T, 1989. Molecular Cloning.
A Laboratory Manual. Second edition. Cold Spring Harbor,
NY: Cold Spring Harbor Laboratory Press.
Kimpton CP, Gill P, Walton A, Urquhart A, Millican ES, Adams
M, 1993. Automated DNA profiling employing multiplex amplification of short tandem repeat loci. PCR Methods Appl 3:
Kimpton CP, Oldroyd NJ, Watson SK, Frazier RR, Johnson PE,
Millican ES, Urquhart A, Sparkes BL, Gill P, 1996. Validation
of highly discriminating multiplex short tandem repeat amplification systems for individual identification. Electrophoresis 17: 1283–1293.
Li H, Schmidt L, Wei MH, Hustad T, Lerman MI, Zbar B, Tory
K, 1993. Three tetranucleotide polymorphisms for loci:
D3S1352; D3S1358; D3S1359. Hum Mol Genet 8: 1327.
StataCorp, 1997. Stata Statistical Software. Release 5.0. Volume
2. G-O. College Station, TX: Stata Press.
Hastie TJ, Tibshirani RJ, 1996. Generalized Additive Models.
London: Chapman and Hall.
McCullagh P, Nelder JA, 1991. Generalized Linear Models.
Second edition. London: Chapman and Hall.
Young LJ, Young JH, 1998. Statistical Ecology. A Population
Perspective. Boston: Kluwer.
Messmer TO, Skelton SK, Moroney JF, Daugharty H, Field, BS,
1997. Application of a nested, multiplex PCR to psittacosis
outbreaks. J Clin Microbiol 35: 2043–2046.
Grenfell BT, Das PK, Rajagopalan PK, Bundy DAP, 1990. Frequency distribution of lymphatic filariasis microfilariae in human populations: population processes and statistical estimation. Parasitology 101: 417–427.
Michael E, 1998 The population dynamics and epidemiology of
lymphatic filariasis. Hoffman S, Pasvol G, eds. Tropical Medicine: Science and Practice. New York: Imperial College
Press, 41–81.
Chan MS, Srividya A, Norman RA, Pani SP, Ramaiah KD, Vanamail P, Michael E, Das PK, Bundy DAP, 1998. Epifil: a
dynamic model of infection and disease in lymphatic filariasis. Am J Trop Med Hyg 59: 606–614.
Boreham PFL, 1975. Some applications of bloodmeal identifications in relation to the epidemiology of vector-borne tropical diseases. J Trop Med Hyg 78: 83–91.
Dye C, Williams BG, 1995. Non-linearities in the dynamics of
indirectly-transmitted infections (or does having a vector
make a difference?). Grenfell BT, Dobson AP, ed. Ecology of
Infectious Diseases in Natural Populations. Cambridge: Cambridge University Press, 260–279.
Davidson G, Draper CC, 1953. Field studies on some of the
basic factors concerned in the transmission of malaria. Trans
R Soc Trop Med Hyg 47: 522–535.
Pull JH, Grab B, 1974. A simple epidemiological model for
evaluating the malaria inoculation rate and the risk of infection in infants. Bull World Health Organ 51: 507–516.
Davison AC, Hinkley DV, 1997. Bootstrap Methods and Their
Application. Cambridge: Cambridge University Press.