Surveillance of vector-borne pathogens under imperfect detection: Lessons from Chagas disease risk (mis)measurement

Abstract

Vector-borne pathogens threaten human health worldwide. Despite their critical role in disease prevention, routine surveillance systems often rely on low-complexity pathogen detection tests of uncertain accuracy. In Chagas disease surveillance, optical microscopy (OM) is routinely used for detecting Trypanosoma cruzi in its vectors. Here, we use replicate T. cruzi detection data and hierarchical site-occupancy models to assess the reliability of OM-based T. cruzi surveillance while explicitly accounting for false-negative and false-positive results. We investigated 841 triatomines with OM slides (1194 fresh, 1192 Giemsa-stained) plus conventional (cPCR, 841 assays) and quantitative PCR (qPCR, 1682 assays). Detections were considered unambiguous only when parasitologists unmistakably identified T. cruzi in Giemsa-stained slides. qPCR was >99% sensitive and specific, whereas cPCR was ~100% specific but only ~55% sensitive. In routine surveillance, examination of a single OM slide per vector missed ~50-75% of infections and wrongly scored as infected ~7% of the bugs. qPCR-based and model-based infection frequency estimates were nearly three times higher, on average, than OM-based indices. We conclude that the risk of vector-borne Chagas disease may be substantially higher than routine surveillance data suggest. The hierarchical modelling approach we illustrate can help enhance vector-borne disease surveillance systems when pathogen detection is imperfect.

Full text

1
SCIEntIFIC REpORTS | (2018) 8:151 | DOI:10.1038/s41598-017-18532-2
www.nature.com/scientificreports
Surveillance of vector-borne
pathogens under imperfect
detection: lessons from Chagas
disease risk (mis)measurement
Thaís Tâmara Castro Minuzzi-Souza1, Nadjar Nitz2, César Augusto Cuba Cuba1, Luciana
Hagström2, Mariana Machado Hecht2, Camila Santana2, Marcelle Ribeiro2, Tamires Emanuele
Vital2, Marcelo Santalucia3, Monique Knox4, Marcos Takashi Obara1, Fernando Abad-Franch5
& Rodrigo Gurgel-Gonçalves1
Vector-borne pathogens threaten human health worldwide. Despite their critical role in disease
prevention, routine surveillance systems often rely on low-complexity pathogen detection tests
of uncertain accuracy. In Chagas disease surveillance, optical microscopy (OM) is routinely used
for detecting Trypanosoma cruzi in its vectors. Here, we use replicate T. cruzi detection data and
hierarchical site-occupancy models to assess the reliability of OM-based T. cruzi surveillance while
explicitly accounting for false-negative and false-positive results. We investigated 841 triatomines with
OM slides (1194 fresh, 1192 Giemsa-stained) plus conventional (cPCR, 841 assays) and quantitative PCR
(qPCR, 1682 assays). Detections were considered unambiguous only when parasitologists unmistakably
identied T. cruzi in Giemsa-stained slides. qPCR was >99% sensitive and specic, whereas cPCR was
~100% specic but only ~55% sensitive. In routine surveillance, examination of a single OM slide per
vector missed ~50–75% of infections and wrongly scored as infected ~7% of the bugs. qPCR-based and
model-based infection frequency estimates were nearly three times higher, on average, than OM-based
indices. We conclude that the risk of vector-borne Chagas disease may be substantially higher than
routine surveillance data suggest. The hierarchical modelling approach we illustrate can help enhance
vector-borne disease surveillance systems when pathogen detection is imperfect.
Vector-borne infectious diseases rank among the most relevant threats to public health globally1. Surveillance of
pathogen presence in vectors allows epidemiologists to track variations of disease transmission risk in time and
space. is, in turn, is crucial for the design, management, and evaluation of strategies for disease prevention2–5.
To correctly interpret surveillance data, however, health ocials need to understand how the tests used to ascer-
tain vector infection actually perform. In particular, they need reliable estimates of each test’s sensitivity and
specicity6,7. Sensitivity is dened in this context as the probability that the target pathogen is detected by a test,
conditional on the vector being infected. Specicity is the probability that the pathogen is not detected by the test,
conditional on the vector being uninfected. ese two quantities are usually unknown and <100% for any given
test or method – including, arguably, the best-studied diagnostic tests, whose nominal sensitivity and specicity
are worked out under extremely articial conditions that may be hard to replicate in eld laboratories8,9.
Many tests are available for detecting pathogens in their hosts and vectors. ey range from direct examina-
tion of samples under the microscope to sophisticated molecular assays that can detect minute amounts of the
pathogen’s genetic material in tissue extracts6,8–10. e performance of these tests may vary substantially. is
1Laboratório de Parasitologia Médica e Biologia de Vetores, Faculdade de Medicina, Universidade de Brasília, Brasília,
72910-900, Brazil. 2Laboratório Interdisciplinar de Biociências, Faculdade de Medicina, Universidade de Brasília,
Brasília, 72910-900, Brazil. 3Laboratório Central de Saúde Pública, Secretaria Estadual de Saúde de Goiás, Goiânia,
74853-120, Brazil. 4Diretoria de Vigilância Ambiental, Secretaria de Saúde do Distrito Federal, Brasília, 70086-900,
Brazil. 5Grupo Triatomíneos, Instituto René Rachou – Fiocruz, Belo Horizonte, 30190-009, Brazil. Fernando Abad-
Franch and Rodrigo Gurgel-Gonçalves contributed equally to this work. Correspondence and requests for materials
should be addressed to F.A.-F. (email: fernando.abad@minas.ocruz.br) or R.G.-G. (email: rgurgel@unb.br)
Received: 3 July 2017
Accepted: 13 December 2017
Published online: 09 January 2018
OPEN
Correction: Author Correction
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
2
SCIEntIFIC REpORTS | (2018) 8:151 | DOI:10.1038/s41598-017-18532-2
mainly depends on the balance between sensitivity and specicity – which as a rule trade-o against one another.
Tests that have both high sensitivity and high specicity tend to be more costly, and oen require higher-level
skills, than simpler, yet not so well-performing, alternatives8,10. is is probably why many routine surveillance
systems, particularly in developing countries, rely on low-complexity pathogen detection tests with suboptimal
performance – they may be just the aordable and technically viable options11,12. Yet a suboptimal test will yield
inaccurate data, and this measurement error can ultimately mislead decision makers8,11. In these cases, knowledge
about test performance will provide particularly crucial insight into the performance of routine surveillance, thus
widening the scope for sounder public health decision making. Still, widespread uncertainty about the true accu-
racy of pathogen detection tests in real-life surveillance settings substantially complicates any such assessment.
Here, we illustrate how a hierarchical modelling approach can help assess the performance of vector-borne
disease surveillance in the face of imperfect pathogen detection. As a case-study, we investigate the detection of a
major human parasite, Trypanosoma cruzi, in its insect vectors13. T. cruzi causes Chagas disease, one of the most
important vector-borne diseases in the Americas14. e parasite is primarily transmitted by blood-sucking bugs
known as triatomines, and entomological-parasitological routine surveillance (EPRS hereaer) is therefore a key
component of Chagas disease control programs13–16. In most such programs, trained technicians identify suspect
insects collected inside or around houses and check them for T. cruzi infection through optical microscopy (OM)
of hindgut contents. OM-based detection of T. cruzi, however, is unlikely to be 100% sensitive or 100% spe-
cic17–23. T. cruzi surveillance data, then, likely contain a certain, yet largely unknown, amount of measurement
error. To quantify this error, we applied multiple detection tests, from routine-surveillance OM to DNA-based
methods, to over 800 eld-caught vectors. We then used multiple detection-state site-occupancy models24 (a
class of hierarchical models) to obtain statistical estimates of each test’s sensitivity and specicity while explicitly
accounting for false-negative and false-positive results. ese models require (i) replicate testing of a subset of
the vectors and (ii) unambiguous identication of a subset of the infections (see Fig.1 and ref.24). is strategy
allowed us to compute corrected estimates of infection frequency in the ve triatomine bug species most oen
found inside and around houses in central Brazil.
Results
Naïve infection indices. We investigated T. cruzi detection in 841 triatomine bugs of ve species collected
inside or around houses in the state of Goiás and the Federal District, Brazil. We used dierent combinations of
OM slide readings (fresh and/or Giemsa-stained slides read in EPRS and/or at the University of Brasília [UnB])
and conventional plus real-time quantitative PCRs (cPCR and qPCR, respectively) (see Tables1 and 2 and
Methods). Overall, 397 bugs were scored as positive in at least one test, for a naïve infection index of 47.2%. Note,
Figure 1. Detecting Trypanosoma cruzi in eld-caught vectors. e gure illustrates our strategy of repeatedly
checking for infection using (i) optical microscopy (OM) including slides read in routine surveillance (fresh,
FS; Giemsa-stained, SS) or at the University of Brasília (fresh, FU; Giemsa-stained, SU), (ii) a conventional PCR
(cPCR), and (ii) a replicate quantitative PCR (qPCR R1 and R2). Blank ‘slides’ represent OM slides that were not
prepared for a given bug (coded ‘−’); in grey, tests that were scored as negative with ambiguity (possible false
negatives, coded ‘0’); in light blue, dark blue, orange, light green, and dark green, tests scored as positive with
ambiguity (possible false positives, coded ‘1’); and, in dark redwith a parasite, a slide scored as positive without
ambiguity (only when a professional parasitologists of the University of Brasília unmistakably identied T.
cruzi trypomastigotes in a Giemsa-stained slide, coded ‘2’). e last column shows, for each bug, the “detection
history” we used to construct our database, using the codes (‘−’, ‘0’, ‘1’, and ‘2’) dened above. Of the four bugs
in this example, only the rst one was scored as positive without ambiguity (hence its darker colour); the three
light-coloured bugs might or might not have been infected: for the third and fourth, there were some ambiguous
detections; for the last one, the six non-detections could have arisen either because the bug was not infected or
because the tests failed to detect the parasite.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
3
SCIEntIFIC REpORTS | (2018) 8:151 | DOI:10.1038/s41598-017-18532-2
however, that this calculation relies on the assumption that the joint results of the tests applied to each bug ensure
100% sensitivity and 100% specicity. Test-specic naïve indices varied about two-fold, from 17.8% considering
OM results only to 41.5% considering PCR results only – with 23.1% positive by cPCR and 41.3–41.4% by qPCR.
Considering EPRS results only, infection with T. cruzi was reported in 15.1% of the bugs (from 0% to 27.2%,
depending on species; Table2, Fig.2). As above, however, the direct interpretation of test-specic naïve values
hinges on the assumption that each test, including OM-based detection of T. cruzi in EPRS, is 100% sensitive and
100% specic. Further descriptive details are presented in Tables1 and 2 and in Fig.2; the raw data are available
in Supplementary DataS1, and PCR protocols in Supplementary Text S1.
Hierarchical modelling. A site-occupancy model with bug species-specific infection probability and
test-specific sensitivity and specificity (i.e., our focal model; see Methods) had a second-order Akaike’s
Information Criterion (AICc) score25 from 18.5 to >1020 units smaller than models in which those parameters
were held constant across, respectively, bug species and pathogen detection tests. ese simpler models, hence,
had no support from the data25, and we therefore base inference on our focal model. Figures2 and 3 present,
respectively, the estimates of species-specic infection frequency and of test-specic sensitivity and specicity
computed using this model; the values of back-transformed estimates and 95% condence intervals (CIs) are
presented in Supplementary TableS1.
Test sensitivity and specicity. Our focal model suggests that sensitivity varies considerably, and specicity mod-
erately, among tests (Fig.3). e estimated sensitivity of OM-based tests was overall low, both in EPRS (fresh slide,
24.5%, CI 20.2–29.5%; stained slide, 50.7%, CI 42.4–59.0%) and at the UnB (fresh slide, 14.3%, CI 10.2–19.9%;
stained slide, 22.6%, CI 18.3–27.6%). cPCR had low sensitivity (56.0%, CI 50.7–61.1%, comparable to that of
Giemsa-stained OM slides read in EPRS), whereas qPCR had very high sensitivity (99.7%), with a lower CI limit
OM slide combinations
Fresh slides Stained slides
Bugs1
EPRS UnB EPRS UnB
Fresh EPRS + Fresh UnB • • 19
Fresh EPRS + Stained EPRS ••29
Fresh EPRS + Stained UnB • • 67
Fresh UnB + Stained UnB • • 22
Fresh EPRS + Fresh UnB + Stained UnB • • • 334
Fresh EPRS + Stained EPRS + Stained UnB • • • 370
Total (slides/bugs) 819 375 399 793 (2386/841)
Table 1. Optical microscopy (OM) slides examined to detect Trypanosoma cruzi infection in 841 triatomine
bugs, central Brazil, 2012–2014. Fresh and Giemsa-stained slides were examined by entomological-parasitological
routine surveillance (EPRS) and University of Brasília (UnB) sta in dierent combinations. 1All tested also by
one conventional and two real-time quantitative PCRs. •Slide read.
Tes t
Triatomine bug speci es
P. megistus T. sordida R. neglectus T. pseudomaculata P. geniculatus
n+n+n+n+n+
Fresh slide EPRS 405 38 273 55 92 24 33 0 16 1
UnB 332 41 0 −1 0 40 0 2 0
Stained slide EPRS 21 3 273 60 91 25 0 −14 0
UnB 395 41(12) 248 28(26) 89 16(16) 45 0(0) 16 0(0)
Microscopy1
EPRS 405 38 273 60 92 25 33 0 16 1
UnB 414 42 248 28 89 16 45 0 16 0
Tot a l 415 64 273 60 92 25 45 0 16 1
cPCR 415 84 273 63 92 22 45 24 16 2
qPCR 1st 415 184 273 89 92 41 45 30 16 3
2nd 415 186 273 88 92 41 45 30 16 3
Molecular1Total 415 186 273 89 92 41 45 30 16 3
Tot a l1415 215 273 100 92 48 45 30 16 4
Table 2. Numbers of triatomine bugs tested and scored as positive for Trypanosoma cruzi infection by optical
microscopy and molecular methods. P., Panstrongylus; T., Triatoma; R., Rhodnius. EPRS, entomological-
parasitological routine surveillance systems (state of Goiás and Federal District, Brazil); UnB, University of
Brasília; cPCR, conventional PCR on the 24Sα subunit of the nuclear ribosomal DNA; qPCR, real-time PCR
on nuclear satellite DNA For each species, the numbers of bugs tested (n) and scored as positive (+) are given;
for Giemsa-stained microscope slides read at UnB, the number of bugs in which infection was unambiguously
determined is also given (in parentheses). 1Bugs scored as positive in at least one test.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
4
SCIEntIFIC REpORTS | (2018) 8:151 | DOI:10.1038/s41598-017-18532-2
just above 98% and an upper limit at 99.9% (Fig.3 and Supplementary TableS1). Specicity was always estimated
at > 90%, although 95% CIs overlapped that value for OM-based tests applied in EPRS (fresh slide, 92.6%, CI
89.8–94.6%; stained slide, 92.8%, CI 89.0–95.4%) and for fresh slides read at UnB (93.1%, CI 88.2–96.0%; stained
slide, 97.1%, CI 95.1–98.3%). e estimate for cPCR specicity was virtually 1.0, with a large standard error
suggesting that the maximum-likelihood solution lies close to the boundary. qPCR again performed very well,
with an estimated specicity close to 100% and a lower CI limit of 95.0% (see Fig.3 and Supplementary TableS1).
Frequency of infection in vectors. Along with the test performance estimates given above, our focal model yielded
a corrected estimate of the frequency of T. cruzi infection (denoted Ψ) in each of the ve vector species in our
sample. As Fig.2 and Table2 and Supplementary TableS1 show, Ψ estimates were very close to naïve indices com-
puted from qPCR data – which reects the excellent performance of this test. Importantly, model-based infection
estimates were consistently larger (about two times, from 1.3 to 3.0) than naïve indices computed from OM or
Figure 2. Trypanosoma cruzi infection in 841 triatomine bugs caught inside or around houses of central
Brazil, 2012–2014. Bars represent the observed proportions of bugs scored as positive with dierent methods,
and circles show infection probabilities as estimated by the focal site-occupancy model. EPRS, entomological-
parasitological routine surveillance; UnB, University of Brasília; cPCR rDNA-24Sα, conventional PCR on the
24Sα subunit of the nuclear ribosomal DNA; qPCR nDNA-sat, quantitative PCR on the nuclear satellite DNA;
CI, condence interval; arrowheads indicate instances in which no bug was tested by a given method (see
Table2).
Figure 3. Performance of microscopy and PCR for the Trypanosoma cruzi detection in insect vectors:
sensitivity and specicity as estimated through site-occupancy modelling. EPRS, entomological-parasitological
routine surveillance; UnB, University of Brasília; cPCR rDNA-24Sα, conventional PCR on the 24Sα subunit
of the nuclear ribosomal DNA; qPCR nDNA-sat, quantitative PCR on the nuclear satellite DNA; CI, 95%
condence interval (an asterisk indicates that the CI could not be estimated for cPCR specicity).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
5
SCIEntIFIC REpORTS | (2018) 8:151 | DOI:10.1038/s41598-017-18532-2
cPCR (Fig.2). Species-specic EPRS-based indices and model-based Ψ values compared as follows: Panstrongylus
megistus 9.4% vs. 44.7%; Triatoma sordida 22.0% vs. 32.6%; Rhodnius neglectus 27.2% vs. 44.6%; Triatoma pseu-
domaculata 0% vs. 66.7%; and Panstrongylus geniculatus 6.3% vs. 18.8% (Fig.2, Table2). Model-based esti-
mates were somewhat smaller than indices based on the combined results from all tests, reecting the fact that
OM-based tests likely yielded some false-positive results. Finally, the size of CIs around model-based Ψ estimates
clearly highlights our uncertainty about infection probabilities when sample sizes are small – particularly for P.
geniculatus (n = 16) but also, albeit to a lesser extent, for T. pseudomaculata (n = 45) and R. neglectus (n = 92) (see
Fig.2 and Supplementary TableS1).
Discussion
In this study we have demonstrated the use of a hierarchical modelling approach24 to investigate the performance
of routine surveillance in the context of a major vector-borne parasitic disease. e results show that detecting
T. cruzi in its triatomine bug vectors can be dicult. e estimated sensitivity of OM-based tests was below 30%
except for Giemsa-stained slides examined in EPRS – which, in any case, failed to detect about half of the infec-
tions (Figs2 and 3). As a consequence, naïve indices based on EPRS data consistently underestimated the true
frequency of T. cruzi infection in vectors caught inside or around houses in central Brazil. In our hands, cPCR
of ribosomal DNA had low sensitivity (barely above 55%) but very high specicity (Fig.3). e sensitivity and
specicity of the more sophisticated qPCR of nuclear satellite DNA were both >99%; this, therefore, was the only
test yielding nearly unbiased infection indices (Figs2 and 3, Supplementary TableS1).
It is not surprising that detecting microscopic protozoa is a dicult task. Less widely realised, however, is
the fact that imperfect detection can aect any organism including plants, insects, and up to large-sized mam-
mals26–30. It also aects molecules, including antibodies orDNA31–33. e consequences of imperfect detection
can be particularly problematic when the target organism is a human pathogen. In the clinical setting, eorts have
concentrated on the development of better-performing diagnostic tests, but this usually implies higher costs and/
or a need for more sophisticated equipment and skills. In disease surveillance, where large numbers of samples
are oen processed, such high-cost, high-technology tests are usually impractical. is is particularly evident in
developing countries, where decentralised surveillance laboratories oen lack the resources needed to use com-
plex tests in routine practice. It should generally be feasible, however, to blindly re-examine a subset of samples
with high-performance tests, such as qPCR, in central reference laboratories; these data could then be used to
gauge the reliability of imperfect routine tests in an analytical framework similar to that described here23,24,31,34–36.
is would be a fairly straightforward way to enhance disease surveillance when pathogen detection is imperfect.
If, for example, a random subset of just 96 bugs from our samplewere tested using just one high-performance
qPCR assay (99% sensitive and 100% specic), mean EPRS slide-reading sensitivity would be estimated at 71.4%
(CI 63.2–78.4) and specicity at 99.7% (CI 98.4–99.9). Despite a modest, yet evident, upward bias, these estimates
may be seen as a potentially important improvement (particularly regarding sensitivity) over the typical stance of
ignoring the problem by assuming that detection is always perfect (see Text S2).
It should be noted that in some Chagas disease EPRS systems each bug is examined twice – one with a
fresh slide and one with a Giemsa-stained slide. According to the estimates from our focal, top-ranking model,
and assuming independence, the joint sensitivity of a two-slide test would be Sejoint = 1 − (1 − Sefresh) × (1 −
Sestained) = 1 − (1–0.245) × (1–0.507) = 0.628, and the joint specicity Spjoint = 1 − (1 − (Spfresh × Spstained)) = 0.859
(see Supplementary TableS1 for estimate values). is suggests that, on average, such a two-slide–reading tac-
tic would miss about 37% of infections and would mistakenly score as positive about 14% of the bugs. Albeit
somewhat disappointing, these gures do not look entirely hopeless; they suggest that increasing slide-reading
sensitivity might be worth the eort: if, for example, the very low sensitivity of fresh-slide reading could be raised
to about 50%, then joint sensitivity would be ~75% and only ~25% of infections would be missed. is could
perhaps be achieved by simply increasing the number of fresh-slide elds examined in EPRS.
Once reliable estimates of test sensitivity and specicity have been obtained, one can use Bayes’ theorem to
calculate posterior probabilities of infection (or non-infection), given test results37. e posterior probability of
infection, given the bug was scored as positive in a test, is also known as the positive predictive value of that test,
and the posterior probability of non-infection, given a negative test result, as the negative predictive value7. ese
values depend not only on test performance, but also on the true prevalence of infection in the population the
specimen was drawn from7,37. Using test performance estimates from our focal model (including joint perfor-
mance of two OM slides read in EPRS as given above), and assuming each bug was randomly sampled from the
population it belonged to, we calculated the positive and negative predictive values of each test at prior prevalence
ranging from 0 to 100%. Figure4 shows the obviously better performance of qPCR relative to all other tests;
although cPCR also yields very good positive predictive values, a negative result is not very reliable as an indicator
of absence of infection – except, as with the other tests, when the prevalence in the population is very low (Fig.4).
e graph also highlights the relatively poor performance of OM tests as used in EPRS (Fig.4).
A reliable surveillance system for any vector-borne infection, including Chagas disease, has two main compo-
nents. Entomological-parasitological surveillance aims at providing data on how oen the vectors, and in particu-
lar infected vectors, are found in close proximity to humans2–5,14,15. Epidemiological surveillance aims at detecting
human disease cases – and, in particular, new cases, which provide insight on incidence3,4,38. Here we have shown
that EPRS data are likely to substantially underestimate natural T. cruzi infection in Chagas disease vectors. We
know, in addition, that detecting infestation by those vectors can also be dicult, with sensitivity estimates oen
below 50%27,28,39. Finally, Chagas disease patients typically present with few signs and symptoms, if any, in the
acute phase of infection, and it is estimated that less than 5% of new cases are diagnosed14. Chagas disease routine
surveillance data therefore contain three types of (downward) measurement error – infestation by triatomines,
vector infection, and human disease all are almost surely more frequent than reported39. is suggests that the
risk of Chagas disease transmission to humans is almost surely higher than what crude surveillance data would
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
6
SCIEntIFIC REpORTS | (2018) 8:151 | DOI:10.1038/s41598-017-18532-2
seem to imply. If unaccounted for when interpreting such data, this composite error can generate a false sense of
security that may ultimately mislead expert advisors and decision makers into, respectively, the wrong advice and
the wrong decisions39.
Our results come with some important caveats; most of them relate to model assumptions, as outlined in the
Methods section. We caution, in particular, that the conditional independence assumption common to most
diagnostic test evaluations (e.g., those based on standard latent class analysis23,34,40,41) may not fully hold in our
data. is means that our estimates of test performance metrics may be somewhat optimistic, which would lead
to some bias in infection frequency estimates – likely a downward bias for the higher estimates and an upward
bias for the lower estimates40. Our approach, on the other hand, let us relax the assumption of perfect detection
and hence circumvent the need for internal amplication controls in PCR assays – whose absence is a wide-
spread, major limitation of studies that assume perfect detection42. We also stress that OM examination of fresh
slides is seldom meant to identify T. cruzi parasites; instead, the examiner usually records the detection of motile
microorganisms and labels them as ‘trypanosomatids’. is is typically regarded as an initial, supposedly more
sensitive test whose results need to be conrmed by the supposedly more specic examination of a stained slide14.
In a sense, then, our estimates of fresh-slide specicity are somewhat ‘unfair’. We found, however, very small
dierences in specicity for fresh and stained slides (Fig.3), suggesting that a similar, small fraction of fresh and
stained slides scored as positive did not contain viable T. cruzi parasites. Further analyses (OM- and PCR-based)
at UnB revealed Trypanosoma rangeli in seven and Blastocrithidia triatomae in 85 of the bugs in our sample43;
these results will be presented elsewhere. Finally, we note that we formally considered only a few among the many
potential sources of heterogeneity in test performance. In particular for OM, the skills of sta preparing and read-
ing slides, their adherence to written protocols, the quality of reagents and microscopes, or bug-specic parasite
loads may all vary and have an eect on performance. We did not have the means to measure or model all these
plausible candidate predictors; instead, we present average test performance estimates for what may be considered
typical circumstances of EPRS.
In summary, we have demonstrated how hierarchical site-occupancy models can help us develop a more
realistic understanding of the performance of T. cruzi detection tests – from the OM-based methods widely used
in routine surveillance to sophisticated molecular assays. is quantitative knowledge about test performance
allowed us to compute corrected estimates of the frequency of T. cruzi infection in the ve vector species most
oen found in houses across central Brazil. Our analyses revealed a considerable downward bias in T. cruzi
infection indices generated by routine surveillance; since vector occurrence indices are also probably biased low,
we conclude that the frequency at which infected triatomines occur inside or around houses, and hence the risk
of Chagas disease transmission, may be substantially higher than surveillance data suggest. It is rather likely that
similar biases aect routine surveillance systems aimed at other pathogens and other vectors. Any such bias must
be singled out, quantied, and explicitly taken into account if we are to draw sound, epidemiologically meaning-
ful conclusions from imperfect surveillance data.
Methods
The bugs. From August 2012 to December 2014, the parasitology laboratory at UnB received 841 triatomine
bugs caught inside or around houses during EPRS in Goiás and the Federal District of Brazil. All bugs were iden-
tied to species at UnB using the keys by Lent & Wygodzinsky13.
Figure 4. Posterior probabilities of vector infection (or non-infection) with Trypanosoma cruzi, given test
results: positive and negative predictive values. EPRS, entomological-parasitological routine surveillance; UnB,
University of Brasília; cPCR, conventional PCR on the 24Sα subunit of the nuclear ribosomal DNA; qPCR,
quantitative PCR on the nuclear satellite DNA. Note that positive predictive values of cPCR (black line) and
qPCR (green broken line) largely overlap.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
7
SCIEntIFIC REpORTS | (2018) 8:151 | DOI:10.1038/s41598-017-18532-2
Optical microscopy. EPRS sta checked 420 bugs for T. cruzi infection with one fresh OM slide only and
399 bugs with two slides – one fresh and one Giemsa-stained (see below). e results of each OM examination
carried out in EPRS, as well as stained slides, were sent together with the bugs to the UnB, where we examined
456 bugs with one OM slide only (19 fresh only, 437 stained only) and 356 bugs with two OM slides (one fresh,
one stained). At the UnB, each slide was read for up to 5 minutes; if slim trypomastigotes with a large, round,
subterminal kinetoplast44,45 were seen in a slide, then it was scored as unambiguously positive (see below) and the
examination stopped. Table1 shows the number of bugs examined with dierent combinations of OM slides both
in EPRS and at UnB. Each fresh slide was prepared by homogenizing one droplet of bug hindgut contents (faeces
and possibly urine) in one drop (~50 μL) of buered saline solution on a microscope slide and covering it with
a cover slip; the slide was examined under a light microscope at 400× magnication17,22,44. For Giemsa staining,
thin smears of hindgut contents homogenised in saline solution were xed with methanol and stained with a
buered 10% Giemsa stain solution; these slides were examined at 1000× magnication44,45. Fresh slides were
scored as positive whenever motile forms suggestive of T. cruzi infection were observed; fresh slide scoring was
always considered ambiguous. Stained slides were scored as unambiguously positive only when parasites were
unmistakably identied as T. cruzi by UnB researchers; in all other cases, the results were considered ambiguous
(see Identifying and coding ambiguity below).
DNA-based methods. We stored all bugs at −20 °C at UnB. e bugs were thawed and dissected on steri-
lised glass slides in a laminar ow safety cabinet. e hindgut was removed with watchmaker forceps and stored
in sterile phosphate-buered saline at −20 °C until DNA extraction. e forceps were thoroughly washed twice
(with HCl 0.1 M and with 70% ethanol) and amed before re-use. DNA was extracted with the Illustra Tissue and
Cells Genomic kit (GE Healthcare, Piscataway, NJ), the QIAamp DNA Mini Kit (Qiagen, Valencia, CA), or the
Biopur Mini Spin Plus kit (Biometrix, Curitiba, Brazil) according to each manufacturer’s instructions. DNA was
quantied with a NanoVue Plus spectrophotometer (GE). We used the QIAamp DNA Mini Kit to extract DNA
from (i) one T. cruzi culture (Berenice strain, 5 × 105 epimastigotes/mL) as the positive control and (ii) the hindguts
of laboratory-reared, uninfected triatomines (Dipetalogaster maxima) as the negative control. Positive, negative,
and blank controls (with no DNA) were included in each PCR round. To ensure that all DNA extracts from bug
hindguts contained good-quality DNA, we PCR-amplied ~414 bp of the bugs’ mitochondrial cytochrome b gene46,
including blank controls in each PCR run (Supplementary Text S1). Each bug was tested for T. cruzi DNA using two
dierent PCRs. First, we ran a cPCR that targets the variable D7 domain of the 24Sα gene of the nuclear ribosomal
DNA47. A bug was scored as cPCR-positive (with ambiguity; see below) when the assay yielded a 270–290-bp band47.
Second, we ran a qPCR targeting the nuclear repetitive satellite region of T. cruzi48 and using SYBR Green technol-
ogy49. For each bug, we ran two independent qPCR assays. To score detection/non-detection of the parasite’s DNA
by qPCR, we built a standard curve50 with serial dilutions of T. cruzi DNA extracted from an epimastigote culture
(see above) and ranging from 10−1 to 104 parasite equivalents/mL. As per standard curve results, a bug was scored
as qPCR-positive (with ambiguity; see below) when the assay yielded a signal corresponding to ≥ 0.1 parasites/mL
(see Supplementary Text S1). We note that all samples contained good-quality DNA (as indicated by amplication
of the ~414-bp cytochrome b fragment) and all positive, negative, and blank controls yielded the expected results.
Modelling. To assess how dierent tests, including those used in EPRS, perform at detecting T. cruzi infec-
tion in triatomine bugs, we used the hierarchical modelling approach developed by Miller et al.24. ese ‘multi-
ple detection-state site-occupancy models’ explicitly accommodate both false-negative and false-positive results.
The models make use of repeated detection/non-detection data (possibly with missing results) to compute
maximum-likelihood estimates of (i) the probability that an organism (here, T. cruzi) is detected in a sampling
unit where it actually occurs (here, a T. cruzi-infected bug), and (ii) the probability that the organism is detected
in a sampling unit where, in reality, it does not occur, i.e., the false-positive error rate24. With this information, the
latent, unobserved probability of each sampling unit being ‘occupied’ by the target organism (here, probability of
infection) is also estimated24. ese probabilities can in addition be modelled as a function of covariates in a gen-
eralised linear modelling framework24. e models require that at least a subset of sampling units is checked for
the presence of the target organism more than once, and that a subset of the detections can be considered unam-
biguous24. Below we give details on the modelling approach, which we implemented in P 11.8 (ref.51).
Identifying and coding ambiguity. We considered detection of T. cruzi as unambiguous (coded ‘2’) only when
T. cruzi parasites with normal morphology (slim trypomastigotes with a large, round, subterminal kinetoplast
– which were judged viable) were unmistakably identied in Giemsa-stained OM slides examined at UnB45,52–54.
For the rest of tests and trials, all detections and non-detections were considered ambiguous and coded ‘1’ and ‘0’,
respectively. We thus explicitly acknowledge the possibility that some positive results may be false positives and
some negative results false negatives24. Microscope slide examination oen yields false-negative results17–23. False
positives may occur because T. cruzi and other triatomine-infecting trypanosomatids, such as Blastocrithidia
spp. or T. rangeli, can be mistaken for one another, especially in fresh OM slides52,54. In the case of PCRs,
cross-amplication of heterospecic genomic targets may produce false-positive detections. In addition, a sample
containing T. cruzi DNA but no viable parasites may yield a positive, yet epidemiologically irrelevant, PCR result.
Cross-sample spillover of amplicons (i.e., contamination), another source of false-positive PCRs, is routinely
minimised with appropriate protocols for handling samples and can be detected using negative controls. Finally,
polymerase inhibition or variation in primer-binding sequences may yield false-negative PCR results. To account
for this pervasive potential for ambiguity, when specifying our models we xed at 0 the probability (denoted b)
that a detection event was classied as unambiguous, given the bug was infected and a detection occurred, for all
tests except stained OM slides read at UnB – for which b was estimated.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
8
SCIEntIFIC REpORTS | (2018) 8:151 | DOI:10.1038/s41598-017-18532-2
Dening outcomes and modelling eects. We dened sensitivity (denoted Se) as the probability of detecting
viable T. cruzi parasites, conditioned on their occurrence, in a given bug sample. We let Se vary as a function
of detection test. Each test was also allowed to have its own probability of false-positive detections, denoted
Pfalse (with 1 − Pfalse estimating the specicity Sp of each test). As noted above, b values were only estimated
for stained OM slides read at UnB. In our research setting, ‘site-occupancy’ corresponds to the frequency of T.
cruzi infection in vectors – the probability Ψ that a bug is ‘occupied by’ (i.e., infected with) viable T. cruzi para-
sites. To get species-specic estimates of this probability, we let Ψ vary among triatomine bug species. We then
computed maximum-likelihood estimates of the parameters and eects of interest (and their variances) using
Miller et al.’s 24 models and P 11.851. We compared our focal model (estimating test-specic Se and Sp
and species-specic Ψ) with simpler models assuming constant Ψ across triatomine species or constant Se and Sp
across tests. We used Akaike’s Information Criterion scores corrected for nite sample size (AICc, with N = 841)
to evaluate relative model performance25.
Model assumptions. As in previous applications of the site-occupancy approach24,31–33,35,36 and other forms of
latent class analysis23,34, our models assume conditional independence of diagnostic test results. is means that,
conditional on the true infection status of a bug, whether one test yields the right (or wrong) result does not aect
the probability that another test will also yield the right (or wrong) result40. If there is positive covariance between
test results, then false-positive and false-negative error rate estimates will be too low and sensitivity and speci-
city will be overestimated; infection frequency estimates might also be somewhat biased high for bug species
with lower infection frequency and somewhat biased low for bug species with higher infection frequency40. e
models also assume independence of individual bugs in relation to infection status, equal test performance across
bug species, and that the infection status of the bugs does not change between OM and DNA extraction24. Because
we apply multiple tests with two distinct biological underpinnings to a large sample of bugs in several species, we
believe that our estimates are unlikely to be badly biased and their condence intervals unlikely to be exceedingly
narrow. We stress, however, that the possible eects of violation of the assumptions above must be kept in mind
when interpreting our results.
References
1. Institute of Medicine. Vector-borne Diseases: Understanding the Environmental, Human Health, and Ecological Connections (e
National Academies Press, 2008).
2. Gu, W., Unnasch, T. ., atholi, C. ., Lampman, . & Nova, . J. Fundamental issues in mosquito surveillance for arboviral
transmission. Trans. . Soc. Trop. Med. Hyg. 102, 817–822, https://doi.org/10.1016/j.trstmh.2008.03.019 (2008).
3. Bras, M. et al. Towards an integrated approach in surveillance of vector-borne diseases in Europe. Parasit. Vectors. 4, 192, https://
doi.org/10.1186/1756-3305-4-192 (2011).
4. Semenza, J. C. & Zeller, H. Integrated surveillance for prevention and control of emerging vector-borne diseases in Europe. Euro
Surveill. 19, 20757, https://doi.org/10.2807/1560-7917.ES2014.19.13.20757 (2014).
5. Springer, Y. P. et al . Tic-, mosquito-, and rodent-borne parasite sampling designs for the National Ecological ObservatoryNetwor.
Ecosphere 7, e01271, https://doi.org/10.1002/ecs2.1271 (2016).
6. osenblatt, J. E. Laboratory diagnosis of infections due to blood and tissue parasites. Clin. Infect. Dis. 49, 1103–1108, https://doi.
org/10.1086/605574 (2009).
7. Banoo, S. et al. Evaluation of diagnostic tests for infectious diseases: general principles. Nat. ev. Microbiol. 8, S17–S29, https://doi.
org/10.1038/nrmicro1523 (2010).
8. Institute of Medicine. Global Infectious Disease Surveillance and Detection: Assessing the Challenges—Finding Solutions.
Worshop Summary (e National Academies Press, 2007).
9. Drain et al. Diagnostic point-of-care tests in resource-limited settings. Lancet Infect. Dis. 14, 239–249, https://doi.org/10.1016/
S1473-3099(13)70250-0 (2014).
10. Ndao, M. Diagnosis of parasitic diseases: old and new approaches. Interdiscip. Perspect. Infect. Dis. 2009, 278246, https://doi.
org/10.1155/2009/278246 (2009).
11. Yager, P., Domingo, G. J. & Gerdes, J. Point-of-care diagnostics for global health. Annu. ev. Biomed. Eng. 10, 107–144, https://doi.
org/10.1146/annurev.bioeng.10.061807.160524 (2008).
12. Pai, N. P., Vadnais, C., Deninger, C., Engel, N. & Pai, M. Point-of-care testing for infectious diseases: diversity, complexity, and
barriers in low- and middle-income countries. PLoS Med. 9, e1001306, https://doi.org/10.1371/journal.pmed.1001306 (2012).
13. Lent, H. & Wygodzinsy, P. evision of the Triatominae (Hemiptera: eduviidae) and their signicance as vectors of Chagas’
disease. Bull. Am. Mus. Nat. Hist. 163, 123–520 (1979).
14. World Health Organization. Control of Chagas disease. Second report of the WHO Expert Committee. World Health Organ. Tech.
ep. Ser. 905, 1–109 (2002).
15. Abad-Franch, F., Vega, M. C., olón, M. S., Santos, W. S. & ojas de Arias, A. Community participation in Chagas disease vector
surveillance: Systematic review. PLoS Negl. Trop. Dis. 5, e1207, https://doi.org/10.1371/journal.pntd.0001207 (2011).
16. Dias, J. C. et al. 2nd Brazilian Consensus on Chagas Disease, 2015. ev. Soc. Bras. Med. Trop. 49(Suppl. 1), 3–60, https://doi.
org/10.5123/s1679-49742016000500002 (2016).
17. Brenière, S. F. et al. Field application of polymerase chain reaction diagnosis and strain typing of Trypanosoma cruzi in Bolivian
triatomines. Am. J. Trop. Med. Hyg. 53, 179–184, https://doi.org/10.4269/ajtmh.1995.53.179 (1995).
18. Brenière, S. F. et al. Peridomestic colonization of Triatoma longipennis (Hemiptera, eduviidae) and Triatoma barberi (Hemiptera,
eduviidae) in a rural community with active transmission of Trypanosoma cruzi in Jalisco state, Mexico. Act a Trop. 101, 249–257,
https://doi.org/10.1016/j.actatropica.2007.02.007 (2007).
19. irchho, L. V., Votava, J. ., Ochs, D. E. & Moser, D. . Comparison of PC and microscopic methods for detecting Trypanosoma
cruzi. J. Clin. Microbiol. 34, 1171–1175 (1996).
20. Shianai-Yasuda, M. A., Ochs, D. E., Tolezan, J. E. & irchho, L. V. Use of the polymerase chain reaction for detecting Trypanosoma
cruzi in triatomine vectors. Trans. . Soc. Trop. Med. Hyg. 90, 649–651, https://doi.org/10.1016/s0035-9203(96)90419-8 (1996).
21. Dorn, P. L. et al. Utility of the polymerase chain reaction in detection of Trypanosoma cruzi in Guatemalan Chagas’ disease vectors.
Am. J. Trop. Med. Hyg. 60, 740–745, https://doi.org/10.4269/ajtmh.1999.60.740 (1999).
22. Pizarro, J. C., Lucero, D. E. & Stevens, L. PC reveals signicantly higher rates of Trypanosoma cruzi infection than microscopy in
the Chagas vector, Triatoma infestans: high rates found in Chuquisaca, Bolivia. BMC Infect. Dis. 7, 66, https://doi.org/10.1186/1471-
2334-7-66 (2007).
23. Lardeux, F., Aliaga, C. & Depicère, S. Bias due to methods of parasite detection when estimating prevalence of infection of Triatoma
infestans by Trypanosoma cruzi. J. Vector Ecol. 41, 285–291, https://doi.org/10.1111/jvec.12224 (2016).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
9
SCIEntIFIC REpORTS | (2018) 8:151 | DOI:10.1038/s41598-017-18532-2
24. Miller, D. A. et al. Improving occupancy estimation when two types of observational error occur: non-detection and species
misidentication. Ecology 92, 1422–1428, https://doi.org/10.1890/10-1396.1 (2011).
25. Burnham, .P. & Anderson, D.. Model Selection and Multimodel Inference: A Practical Information-theoretic Approach.
(Springer, 2002).
26. Chen, G., éry, M., Zhang, J. & Ma, . Factors aecting detection probability in plant distribution studies. J. Ecol. 97, 1383–1389,
https://doi.org/10.1111/j.1365-2745.2009.01560.x (2009).
27. Abad-Franch, F. et al. Modeling disease vector occurrence when detection is imperfect: infestation of Amazonian palm trees by
triatomine bugs at three spatial scales. PLoS Negl. Trop. Dis. 4, e620, https://doi.org/10.1371/journal.pntd.0000620 (2010).
28. Abad-Franch, F., Valença-Barbosa, C., Sarquis, O. & Lima, M. M. All that glisters is not gold: sampling-process uncertainty in
disease-vector surveys with false-negative and false-positive detections. PLoS Negl. Trop. Dis. 8, e3187, https://doi.org/10.1371/
journal.pntd.0003187 (2014).
29. éry, M., Gardner, B. & Monnerat, C. Predicting species distributions from checlist data using site-occupancy models. J. Biogeogr.
37, 1851–1862, https://doi.org/10.1111/j.1365-2699.2010.02345.x (2010).
30. Poley, L. G. et al. Occupancy patterns of large mammals in the Far North of Ontario under imperfect detection and spatial
autocorrelation. J. Biogeogr. 41, 122–132, https://doi.org/10.1111/jbi.12200 (2014).
31. McClintoc, B. T. et al. Seeing a second opinion: uncertainty in disease ecology. Ecol. Lett. 13, 659–674, https://doi.
org/10.1111/j.1461-0248.2010.01472.x (2010).
32. Schmidt, B. ., éry, M., Ursenbacher, S., Hyman, O. J. & Collins, J. P. Site occupancy models in the analysis of environmental DNA
presence/absence surveys: a case study of an emerging amphibian pathogen. Methods Ecol. Evol. 4, 646–653, https://doi.
org/10.1111/2041-210X.12052 (2012).
33. Elmore, S. A. et al. Toxoplasma gondii exposure in arctic-nesting geese: a multi-state occupancy framewor and comparison of
serological assays. Int. J. Parasitol. Parasites Wildl. 3, 147–153, https://doi.org/10.1016/j.ijppaw.2014.05.005 (2014).
34. Ochola, L. B., Vounatsou, P., Smith, T., Mabaso, M. L. & Newton, C. . e reliability of diagnostic techniques in the diagnosis and
management of malaria in the absence of a gold standard. Lancet Infect. Dis. 6, 582–588, https://doi.org/10.1016/S1473-
3099(06)70579-5 (2006).
35. Lachish, S., Gopalaswamy, A. M., nowles, S. C. L. & Sheldon, B. C. Site-occupancy modelling as a novel framewor for assessing
test sensitivity and estimating wildlife disease prevalence from imperfect diagnostic tests. Methods Ecol. Evol. 3, 339–348, https://doi.
org/10.1111/j.2041-210X.2011.00156.x (2012).
36. Miller, D. A. W., Talley, B. L., Lips, . . & Campbell Grant, E. H. Estimating patterns and drivers of infection prevalence and intensity when
detection is imperfect and sampling error occurs. Methods Ecol. Evol. 3, 850–859, https://doi.org/10.1111/j.2041-210X.2012.00216.x (2012).
37. Cohen, J. e Earth is round (p <0 .05). Am. Psychologist 49, 997–1003, https://doi.org/10.1037/0003-066X.49.12.997 (1994).
38. World Health Organization. Communicable Disease Surveillance and esponse Systems. Guide to Monitoring and Evaluating (WHO/
CDS/EP/LYO/2006.2, World Health Organization, 2006).
39. Abad-Franch, F., Diotaiuti, L., Gurgel-Gonçalves, . & Gürtler, . E. On bugs and bias: improving Chagas disease control
assessment. Mem. Inst. Oswaldo Cruz 109, 125–130, https://doi.org/10.1590/0074-0276140338 (2014).
40. Vace, P. M. The effect of conditional dependence on the evaluation of diagnostic tests. Biometrics 41, 959–968, https://doi.
org/10.2307/2530967 (1985).
41. To, N., Jørgensen, E. & Højsgaard, S. Diagnosing diagnostic tests: evaluating the assumptions underlying the estimation of sensitivity
and specicity in the absence of a gold standard. Prev. Vet. Med. 68, 19–33, https://doi.org/10.1016/j.prevetmed.2005.01.006 (2005).
42. amírez, J. C. et al. Analytical validation of quantitative real-time PC methods for quantication of Trypanosoma cruzi DNA in
blood samples from Chagas disease patients. J. Mol. Diagn. 17, 605–615, https://doi.org/10.1016/j.jmoldx.2015.04.010 (2015).
43. Minuzzi-Souza, T. T. C. Detecção Parasitológica e Molecular de Tripanossomatídeos em Triatomíneos Sinantrópicos e Primatas
Neotropicais no Brasil Central. (PhD esis, University of Brasília, 2016).
44. Coura, J. . (ed) Manual de Capacitação na Detecção de Trypanosoma cruzi para Microscopistas de Malária e Laboratoristas da
ede Pública (PAHO-WHO/Fiocruz, 2011).
45. Perlowagora-Szumlewicz, A. & Moreira, C. J. C. In vivo dierentiation of Trypanosoma cruzi – 1. Experimental evidence of the
influence of vector species on metacyclogenesis. Mem. Inst. Oswaldo Cruz 89, 603–618, https://doi.org/10.1590/S0074-
02761994000400018 (1994).
46. Lyman, D. F. et al. Mitochondrial DNA sequence variation among triatomine vectors of Chagas’ disease. Am. J. Trop. Med. Hyg. 60,
377–386, https://doi.org/10.4269/ajtmh.1999.60.377 (1999).
47. Schijman, A. G. et al. Dierential detection of Blastocrithidia triatomae and Trypanosoma cruzi by amplication of 24Sα ribosomal
NA genes in faeces of sylvatic triatomine species from rural northwestern Argentina. Acta Trop. 99, 50–54, https://doi.
org/10.1016/j.actatropica.2006.06.010 (2006).
48. Ndao, M. et al. Trypanosoma cruzi infection of squirrel moneys: comparison of blood smear examination, commercial enzyme-
lined immunosorbent assay, and polymerase chain reaction analysis as screening tests for evaluation of money-related injuries.
Comp. Med. 50, 658–665 (2000).
49. Wittwer, C. T., Herrmann, M. G., Moss, A. A. & asmussen, . P. Continuous fluorescence monitoring of rapid cycle DNA
amplication. BioTechniques 22, 130–138 (1997).
50. utledge, . G. & Côté, C. Mathematics of quantitative inetic PC and the application of standard curves. Nucleic Acids es. 31,
e93, https://doi.org/10.1093/nar/gng093 (2003).
51. Hines, J. E. PESENCE – Soware to estimate patch occupancy and related parameters, v. 11.8. USGS – PWC. www.mbr-pwrc.
usgs.gov/soware/presence.html. (2006)
52. Cuba Cuba, C. A. evisión de los aspectos biológicos y diagnósticos del Trypanosoma (Herpetosoma) rangeli. ev. Soc. Bras. Med.
Trop. 31, 207–220, https://doi.org/10.1590/S0037-86821998000200007 (1998).
53. de Sousa, M. A. Morphobiological characterization of Try panosoma cruzi Chagas, 1909 and its distinction from other trypanosomes.
Mem. Inst. Oswaldo Cruz 94, 205–210, https://doi.org/10.1590/S0074-02761999000700031 (1999).
54. Schaub, G. A. Interactions of tr ypanosomatids and triatomines in Advances in Insect Physiology, vol. 37 (ed. Simpson, J.S., Casas, J.)
177–242 (Academic Press, 2009).
Acknowledgements
We thank the vector and parasite surveillance sta of Goiás State Health Department and the Federal District
Environmental Surveillance Agency, Brazil. We also thank F. das Chagas, D.A. Rocha, and V.J. de Mendonça for
assistance. M.R.F. de Oliveira made useful comments on an earlier dra of the manuscript, and R.N. is work
was funded by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES, grant 1276/2011) and
the Fundação de Amparo à Pesquisa do Distrito Federal (FAP-DF, grant 6098/2013), Brazil. Additional support
came from the Instituto René Rachou and the Vice-Presidência de Pesquisa e Laboratórios de Referência (both
at Fiocruz, Brazil).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
10
SCIEntIFIC REpORTS | (2018) 8:151 | DOI:10.1038/s41598-017-18532-2
Author Contributions
F.A.-F. and R.G.-G. conceived the project, R.G.-G., C.A.C.C. and N.N. designed the experiments, T.T.C.M.-S.,
C.S., L.H., M.K., M.S., M.R., N.N. and T.E.V. performed the experiments, and F.A.-F. analysed the data. R.G.-G.
and N.N. provided resources, and F.A.-F., R.G.-G. and T.T.C.M.-S. draed the manuscript and prepared the
gures. All authors reviewed and contributed to the nal version of the manuscript.
Additional Information
Supplementary information accompanies this paper at https://doi.org/10.1038/s41598-017-18532-2.
Competing Interests: e authors declare that they have no competing interests.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre-
ative Commons license, and indicate if changes were made. e images or other third party material in this
article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons license and your intended use is not per-
mitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
© e Author(s) 2017
Content courtesy of Springer Nature, terms of use apply. Rights reserved

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com