Articles

Exploring the urban arbovirus landscape in Rio de Janeiro,

Brazil: transmission dynamics and patterns of disease spread
Gabriel Schuab,a,b,p Stephane Tosta,c,p Keldenn Moreno,c,p Vagner Fonseca,d,p Luciane Amorim Santos,e Svetoslav Nanev Slavov,f ,g
Simone Kashima,f Massimo Ciccozzi,h José Lourenço,i,j Eleonora Cella,k Carla de Oliveira,b Andréa Cony Cavalcanti,l Luiz Carlos Junior Alcantara,m
Fernanda de Bruycker-Nogueira,b Ana Maria Bispo de Filippis,b,∗ and Marta Giovanettin,o,∗∗
a
Universidade Federal do Rio de Janeiro, Duque de Caxias, Rio de Janeiro, Brazil
b
Laboratório de Arbovírus e Vírus Hemorrágicos, Instituto Oswaldo Cruz, Rio de Janeiro, Rio de Janeiro, Brazil
c
Programa Interunidades de Pós-Graduação em Bioinformática, Universidade Federal de Minas Gerais, Belo Horizonte,
Minas Gerais, Brazil
d
Department of Exact and Earth Sciences, University of the State of Bahia, Salvador, Brazil
e
Bahiana School of Medicine and Public Health, Salvador, Bahia, Brazil
f
Blood Center of Ribeirão Preto, Ribeirão Preto Medical School, University of São Paulo, Ribeirão Preto, São Paulo, Brazil
g
Butantan Institute, São Paulo, Brazil
h
Unit of Medical Statistics and Molecular Epidemiology, Università Campus Bio-Medico di Roma, Rome, Italy
i
Universidade Católica Portuguesa, Faculdade de Medicina, Biomedical Research Center, Lisboa, Portugal
j
Climate Amplified Diseases and Epidemics (CLIMADE), Portugal, Europe
k
Burnett School of Biomedical Sciences, University of Central Florida, Orlando, FL, 32827, USA
l
Central Laboratory of Health of the Rio de Janeiro State, Brazil
m
Instituto René Rachou, Fundação Oswaldo Cruz, Minas Gerais, Brazil
n
Department of Science and Technology for Humans and the Environment, University of Campus Bio-Medico di Roma, Rome, Italy
o
Oswaldo Cruz Institute, Oswaldo Cruz Foundation, Rio de Janeiro, Brazil

Summary The Lancet Regional

Health - Americas
Background This study focuses on urban arboviruses, specifically dengue (DENV), chikungunya (CHIKV), and Zika
2024;35: 100786
(ZIKV), which pose a significant public health challenge in Rio de Janeiro state, Southeast Brazil. In our research, we
Published Online xxx
highlight critical findings on the transmission dynamics of these arboviruses in Rio de Janeiro, identifying distinct
https://doi.org/10.
patterns of disease spread. 1016/j.lana.2024.
100786
Methods By combining genomic data with case reports from the Brazilian Ministry of Health, we have analysed the
phylogenetics, prevalence and spatial distribution of these endemic viruses within the state.

Findings Our results revealed sustained DENV transmission primarily in the northern part of the state, a significant
ZIKV epidemic in 2016 affecting all mesoregions, and two major CHIKV outbreaks in 2018 and 2019, predominantly
impacting the northern and southern areas. Our analysis suggests an inverse relationship between arboviral case
incidence and urban density, with less populous regions experiencing higher transmission rates, potentially attrib-
uted to a complex interplay of factors such as the efficacy of vector control measures, environmental conditions, local
immunity levels, and human mobility. Furthermore, our investigation unveiled distinct age and gender trends among
affected individuals. Notably, dengue cases were predominantly observed in young adults aged 32, while chikungunya
cases were more prevalent among individuals over 41. In contrast, cases of ZIKV were concentrated around the
33-year age group. Intriguingly, females accounted for nearly 60% of the cases, suggesting a potential gender-based
difference in infection rates.

Interpretation Our findings underscore the complexity of arbovirus transmission and the need for interventions
tailored to different geographical mesoregions. Enhanced surveillance and genomic sequencing will be essential for a
deeper, more nuanced understanding of regional arbovirus dynamics. Identifying potential blind spots within the
state will be pivotal for developing and implementing more effective public health strategies, specifically designed to
address the unique challenges posed by these viruses throughout the state.

*Corresponding author.
**Corresponding author. Department of Science and Technology for Humans and the Environment, University of Campus Bio-Medico di Roma,
Rome, Italy.
E-mail addresses: ana.bispo@ioc.fiocruz.br (A.M. Bispo de Filippis), [email protected] (M. Giovanetti).
p
Denote equal contribution.
Translation: For the Portuguese translation of the abstract see Supplementary Materials section.

www.thelancet.com Vol 35 July, 2024 1

 Articles

Funding This study was supported by the National Institutes of Health USA grant U01 AI151698 for the United World
Arbovirus Research Network (UWARN) and the CRP-ICGEB RESEARCH GRANT 2020 Project CRP/BRA20-03.

Copyright © 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Arboviruses; Dengue; Epidemiology; Rio de Janeiro; Mesoregions

Research in context

Evidence before this study DENV transmission, a significant ZIKV epidemic in 2016, and
We conducted a comprehensive search of the PubMed major CHIKV outbreaks occurred in 2018 and 2019.
database to identify previous evidence on the circulation of Interestingly, our analysis suggests an inverse relationship
arboviruses across the various mesoregions within the state of between arboviral case incidence urban density, with peri-
Rio de Janeiro. Our search utilized the terms “Arboviruses”, urban areas experiencing higher transmission rates,
“DENV”, “CHIKV*”, “ZIKV”, “Rio de Janeiro”, and “Rio de potentially due to factors like vector control measures,
Janeiro’s mesoregions” from inception until November 30th, environmental condition, local immunity levels, and human
2023. Prior to this study, the urban arbovirus landscape in Rio mobility. Demographic analysis revealed specific trends with
de Janeiro, Brazil, was mostly unexplored, focusing only on DENV, CHIKV, and ZIKV affecting distinct age groups—
the description of sporadic epidemics. Despite numerous predominantly young adults around 32, those over 41, and at
studies on the prevalence and control of mosquito-borne 33 years, respectively, with a notable roughly 60% of cases
diseases such as dengue, Zika, and chikungunya across Brazil, being female. These findings underscore the complex
few have delved into the unique urban challenges presented dynamics of arbovirus spread and the need for tailored public
by Rio de Janeiro. This city’s distinctive characteristics—high health strategies.
population density, complex environmental conditions, and
Implications of all the available evidence
varying levels of health infrastructure—create a unique
Our findings highlight the critical need for public health
scenario for arbovirus transmission. Such conditions
strategies that are specifically tailored to address the unique
exacerbate the challenges of disease management and
challenges posed by urbanization, ecological variations, and
control, setting Rio de Janeiro apart from other urban settings
population density. These insights advocate for the
both within Brazil and globally.
development of targeted interventions and policies that
Added value of this study consider the specific ecological and socio-demographic
This study provides a comprehensive analysis of Rio de contexts of urban and peri-urban areas. Moreover, our
Janeiro’s urban arbovirus landscape, uncovering distinct research provides a valuable framework for the global health
patterns in disease transmission influenced by urbanization, community, offering actionable for the effective management
ecological variations, and population density. We identified and control of arboviruses in similarly urbanized settings
characteristic spatial and temporal trends, including sustained worldwide.

Introduction country, and subsequently, in 1986, serotype 1 was also

Urban arboviruses such as dengue (DENV), chikungu- reported in Rio de Janeiro state.4 Since then, dengue
nya (CHIKV), and Zika (ZIKV), continue to pose has become endemic in Brazil, with the country
significant public health concerns in Brazil, expanding accounting for approximately 60% of reported cases in
in conditions favoured by a confluence of human and the Americas.5 The state of Rio de Janeiro, with a pop-
ecological factors.1 The latter include adequate levels of ulation of over 16 million, has emerged as a focal point
temperature, humidity, and rainfall, all exacerbated by for urban arbovirus transmission, necessitating a deeper
the uneven distribution of water treatment services1,2 understanding of the epidemiological history and situ-
creating ideal environments for the reproduction and ation in the region.5,6 The introduction of DENV-1
dispersal of Aedes spp. mosquitoes, which are the genotype V in the late 1980s marked the onset of sus-
primary vectors for transmission of these arboviruses.1 tained dengue activity in the state.5 The emergence of
Notably, the first reported epidemic compatible with a DENV-2 in the early 1990s,4 identified by its Asian/
dengue viral infection occurred in 1845,3 marking the American genotype (genotype III) within the metropol-
beginning of its long history in Brazil. The reinfestation itan area, led to significant epidemics and introduced
of Brazil by the Aedes aegypti mosquito in 1977 signifi- the first severe forms of the disease.6 The early 2000s
cantly contributed to the reintroduction of DENV in the saw the introduction of DENV-3 genotype III, causing to
1980s.4 Initially, serotypes 1 and 4 were identified in date the largest epidemic in the summer of 2001–2002,6
the state of Roraima located in the Northern part of the which has also recently experienced a significant

2 www.thelancet.com Vol 35 July, 2024

 Articles

resurgence.7,8 DENV-4’s identification in 2011 in Niterói spatial distribution of these viruses in the Rio de Janeiro
and its subsequent spread highlights the ongoing state, seeking to enhance our understanding of their
evolution of dengue’s epidemiological landscape in Rio historical and contemporary transmission patterns. This
de Janeiro.6,9 This period was particularly marked by the comprehensive approach is vital for informing public
co-circulation of two distinct genotypes of DENV-4, health responses and mitigating the impact of these
genotypes I and II, with genotype II becoming the viruses on affected populations.
predominant strain in the state.9 This emergence
further complicated the epidemiological scenario, lead-
ing to a notable increase in dengue cases, which showed Methods
a significant rise in the total number of reported cases Sequence data collection
compared to the year preceding the arrival of DENV-4.10 We retrieved all available viral genome sequences of
Furthermore, between 2011 and 2013, DENV-4 was DENV 1–4, CHIKV, and ZIKV from the state of Rio de
identified in 44.2% of dengue-positive cases, under- Janeiro from public repositories such as NCBI21 and
scoring its substantial impact on the epidemic in Rio de GISAID (https://www.gisaid.org/) up to November
Janeiro.6,9 30th, 2023. Prior to any phylogenetic inference, we
CHIKV and ZIKV, first identified in Rio de Janeiro undertook a rigorous quality assessment of these
in 2015,10–12 have introduced new challenges at both local sequences. This process involved excluding incomplete
and national levels. CHIKV is notably linked to chronic or problematic sequences, as well as those lacking
conditions that significantly affect the quality of life, comprehensive metadata (for example collection date;
especially due to the persistent joint pain experienced by country of origin). Sequences underwent genotype
patients post-recovery.13 While CHIKV’s incidence in assignment using the Genome Detective Tool.22 Addi-
2023 has seen a decline, the rising cases in neighbour- tionally, we removed sequences containing more than 10
ing states signal a potential threat of re-emergence.14 ambiguous bases. To assess the spread of the virus
ZIKV, circulating since 2015,10 drew widespread within the state, we considered other existing and
concern during Brazil’s 2016 epidemic,12 primarily due representative sequences collected globally. Acknowl-
to its association with Guillain-Barré syndrome15 and edging the impact of sampling bias on ancestral state
congenital Zika syndromes.12,16,17 In recent years, the reconstruction methods, we aimed to reflect the global
overlapping clinical symptoms among these arbovi- diversity of genomes in our sampled datasets. In Brazil,
ruses, the potential for cross-immunity between ZIKV given the high number of arbovirus cases, we adopted a
and DENV,18 and changes in surveillance during downsampling approach to ensure balanced representa-
the COVID-19 pandemic have contributed to a drop in tion by selecting reference sequences based on the
case reports. However, as the COVID-19 pandemic reported incidence of cases across Brazilian regions and
progressed and subsequent decline, a resurgence in the availability of genome sequences, thereby mitigating
arbovirus cases was observed.19,20 potential bias from uneven sequence data distribution.
The evolving landscape of arboviral diseases in both For countries in the Americas other than Brazil, we did
the state and the country has been significantly high- not employ the same downsampling strategy due to the
lighted by the introduction, re-introduction and circu- limited number of available genome sequences. For
lation of different dengue virus serotypes and other countries outside the Americas not directly linked to the
arboviruses. The introduction of DENV-3 and DENV-4 epidemics in Rio de Janeiro and Brazil, our down-
alongside the already circulating DENV-1 and DENV-2, sampling criteria were based on metadata availability and
has notably complicated the epidemiological landscape. sequence quality, as previously described (Figure S1).
Since late 2014, the emergence of other arboviruses, The Americas was categorised into four regions for
such as CHIKV, ZIKV, has further diversified the simplification: i) North America, consisting solely of the
arboviral challenges faced by public health systems. This United States; ii) the Caribbean (n = 19), including
scenario presents a complex interplay of multiple countries such as Anguilla, Antigua and Barbuda,
viruses, underscoring the critical need for integrated Aruba; iii) Central America (n = 8), covering Belize,
surveillance and control strategies to effectively manage Costa Rica, El Salvador, Guatemala; and iv) South
the spread of these diseases. America (n = 10), encompassing Argentina, Bolivia,
Given Rio de Janeiro’s role as a significant hub for Chile, Colombia. The final datasets for each viral path-
business and tourism in Brazil’s southeastern region, ogen contained the following numbers of sequences:
attracting an influx of 1.6 million airline passengers DENV-1 accounted for n = 2580 sequences (including
annually, understanding the urban arboviral epidemi- 91 genomes from Rio de Janeiro); DENV-2 accounted
ology and spatial dynamics within this context is essential. for n = 2176 sequences (including 144 genomes from
The lack of detailed data on the genomic epidemiology of Rio de Janeiro); DENV-3 accounted for n = 2220
these co-circulating viruses highlights an urgent need for sequences (including 13 genomes from Rio de Janeiro);
focused public health strategies and preparedness plans. DENV-4 accounted for n = 1579 sequences (including
Our study aims to explore the phylogenetic history and 14 genomes from Rio de Janeiro); CHIKV accounted for

www.thelancet.com Vol 35 July, 2024 3

 Articles

n = 1627 sequences (including 130 genomes from Rio utilized the curated dataset from 2015 onwards to ensure
de Janeiro); ZIKV accounted for n = 481 sequences robustness and accuracy. According to the Ministry’s
(including 30 genomes from Rio de Janeiro). definition, a case is considered “confirmed” either by
laboratory criteria (molecular and serological) or clinical-
Phylogenetic reconstruction epidemiological criteria, considering the clinical symp-
Sequences were aligned using MAFFT employing toms and epidemiological history of that patient. It is also
default parameters.23 The alignment was manually noted by the Ministry that during epidemic seasons, most
curated to remove artefacts at the terminal regions and cases are confirmed based solely on clinical criteria. The
within the alignment using Aliview.24 Phylogenetic dataset included metadata such as the date of sample
analysis of these sequences was performed using the collection, the identified virus, patient gender, age, and
maximum likelihood method implemented in IQ-TREE the municipality and state of residence. We processed
v.2.25 The tree was inferred with the general time this data to categorise each municipality into one of the
reversible (GTR) model of nucleotide substitution and a six mesoregions of the state. The northern part of the
proportion of invariable sites (+I) as selected by the state encompassed Noroeste Fluminense (NWF) and Norte
ModelFinder application. Branch support was assessed Fluminense (NOF). The central region included Baixadas
by the approximate likelihood-ratio test based on the (CL) and Centro Fluminense (FC), while the southern
bootstrap and the Shimodaira–Hasegawa-like procedure region comprised Metropolitana do Rio de Janeiro (MR)
(SH-aLRT) with 1000 replicates. The resulting tree and Sul Fluminense (SF). Monthly climate data for Rio de
topologies were inspected for temporal molecular clock Janeiro was obtained from Copernicus.eu satellite climate
signals using the clock functionality of TreeTime.26 data (https://www.copernicus.eu/en). We derived the
ML-trees were then transformed into time scaled phy- temperature data summary by computing the yearly
logenies in TreeTime. Outlier sequences that deviated lowest, mean, and maximum values. In addition, to
from the strict molecular clock assumption as flagged by evaluate cross-correlations of the spatial distributions of
TreeTime were removed with the Ape package in R27 the three different viruses we utilised the spatialEco
until a good time scaled phylogeny was obtained. The package V2.0-0 in R (https://github.com/jeffreyevans/
percentage of outliers ranged from 0.8% for DENV-1 to spatialEco), which follows the recent advancements of
1.5% for DENV-2, to 1.8% for DENV-3, to 1.3% for Chen YG in spatial cross-correlation analyses.29 For each
DENV-4, to 1.2% for CHIKV, and 1.7% for ZIKV. For pair of viruses, after aggregating incidence across the
each viral population, we produced time scaled tree observation period, we used the cross-Correlation func-
topologies and performed discrete ancestral state tion to quantify Local Indicators of Spatial Association
reconstruction (of locations) to infer the global dissem- (LISA) clusters (high–high, high-low, low-high, low–low)
ination of each virus using the mugration package and the spatial autocorrelation global Moran’s I.
extension of TreeTime under a GTR model. The
mugration package systematically explores the entire Role of funding source
tree, including internal nodes and tips, to evaluate the The funding source supported the design and execution
potential and putative events. Finally, using a custom of this study, including data collection, analysis, and
python script, we counted the number of state changes interpretation. It had no role in the writing of the
from the resulting annotated tree topologies by iterating manuscript or the decision to submit it for publication.
over each phylogeny from the root to the external tips.
We count state changes when an internal node transi-
tions from one location to a different location in the Results
resulting inferred internal node or tip(s). The timing of The analysis of weekly notified Dengue cases from 2015
putative transition events was recorded, which served as to 2023 highlights distinct epidemic peaks, with signif-
the estimated import or export event. All results were icant outbreaks occurring in early 2015, from late 2015
visualized using ggplot libraries.28 to early 2016, in early 2019, and again in early 2023
(Fig. 1a). This pattern aligns with trends observed both
Epidemiological data within the country and internationally,30,31 where a
Data regarding weekly reported cases of DENV, ZIKV, notable decrease in Dengue notifications was seen
and CHIKV in the state of Rio de Janeiro were sourced following 2016. This reduction coincides with the
from the Sistema de Informação de Agravos de Notifi- emergence of the ZIKV (Fig. 1a). Concurrently, the state
cação (SINAN) (https://portalsinan.saude.gov.br/dados- witnessed the introduction of CHIKV in mid-2015.
epidemiologicos-sinan), as provided by the Brazilian However, CHIKV’s most significant outbreak did not
Ministry of Health.19 While SINAN has been collecting occur until the 2018–2019 period, during which the
epidemiological data on arboviral infections since its weekly case count surpassed 5000 in 2019.
inception, it is important to note that the reliability and Analysis of genomic data over the years (Fig. 1b),
comprehensiveness of this data have significantly identified that Dengue serotypes presented temporal
improved post-2021. Accordingly, our analysis specifically dynamics with an oscillatory behaviour characterised by

4 www.thelancet.com Vol 35 July, 2024

 Articles

a b DENV-1 DENV-2 DENV-3 DENV-4 CHIKV ZIKV

c Female Male

DENV
5000 1.00
CHIKV
ZIKV 34% 66%
ZIKV

Viruses proportion
4000
0.75
Total cases

Viral type
3000
DENV 45% 55%
0.50

2000

0.25
1000 CHIKV 38% 62%

0.00
0
2016 2018 2020 2022 2023 1986 1995 2001 2008 2014 2023 0% 25% 50% 75% 100%
Sampling date Collection year Gender rate (%)

d >90
e f
>90 Gender >90 Gender
Gender
Female Female
80−90 Female 80−90
80−90 Male
Male
Male
70−80 70−80 70−80

60−70 60−70 60−70

Age−group

Age−group
50−60 50−60 50−60
Age−group

40−50 40−50 40−50

30−40 30−40 30−40

20−30 20−30 20−30

10−20 10−20 10−20

0−10 0−10 0−10 ZIKV

DENV CHIKV
20 10 0 10 20 20 10 0 10 20 20 10 0 10 20
Percentage (of cases in each age−group) Percentage (of cases in each age−group) Percentage (of cases in each age−group)

Fig. 1: Dynamics of circulating urban arboviruses in Rio de Janeiro, Southeast Brazil. a) A time series from 2015 to 2023 showing weekly
reported cases in the state, categorized by virus type (DENV in grey, CHIKV in dark green, ZIKV in purple). b) Proportion of available viral
genomes from samples collected in the state of Rio de Janeiro, highlighted by year and virus type, with colors representing each virus (DENV in
scale grey, CHIKV in dark green, ZIKV in purple); c) Gender rate of the reported cases for DENV, CHIKV and ZIKV; d–f) Breakdown of reported
cases by gender and age for DENV, CHIKV and ZIKV respectively.

recurrent peak genome prevalence. DENV-2 was the incidence was observed in the North mesoregions
most sequenced viral strain in the state, followed by (NWF, NOF) and the South (SF). Intriguingly, the
DENV-1. DENV-3 was only present in 2001 and Central mesoregions (CL and FC) also experienced an
2004–2005, while DENV-4 was only detected in increase in case numbers during this period (Fig. 2a).
2011–2013. Notably, CHIKV and ZIKV genomes were However, this increase did not align with human den-
documented in years following their emergence, but sity, as evidenced by a Pearson correlation coefficient
seemingly disappearing from public databases thereafter. of −0.16 and a P value of 1.15 × 10ˆ−11. This discrepancy
When examining the age profile of infected in- prompted further investigation into the role of climatic
dividuals, we found that the median age varied by virus: conditions, particularly temperature. Interestingly, the
32 years for DENV, 41 years for CHIKV, and 33 years correlation between dengue incidence and temperature
for ZIKV. Notably, there was a slight female predomi- was found to be low. Despite this, our results highlight
nance, with females accounting for more than 55% of that peri-urban regions, among others, have reported
the cases, reaching up to 66% (Fig. 1c). Cases were most the highest mean temperatures over the past nine years
frequently reported in two age groups for both females (Figure S2).
and males 30–40 years across all viral infections and From 2020 to 2021, there was a notable decrease in
20–30 years (all but CHIKV), with those aged 40–60 case counts across all mesoregions. However, an
years being particularly affected by CHIKV (Fig. 1d–f). upsurge was detected in 2023, with the North (NWF)
However, the 0–10-year age group was the least affected, region reporting the highest case numbers as of
and the distribution of cases was slightly higher among September 2023, as shown in Fig. 2a. It is important to
females compared to males (Fig. 1d–f). note that during the COVID-19 pandemic, arbovirus
notifications decreased, likely influenced by the imple-
Dengue virus transmission dynamics in the state of mented sanitary measures and the presence of over-
Rio de Janeiro lapping symptoms between SARS-CoV-2 and arboviral
The analysis of spatial distribution patterns of dengue infections.
cases across the mesoregions of Rio de Janeiro, depicted Fig. 2b presents time-stamped maximum likelihood
in Figs. 1 and 2a, revealed distinct epidemic waves in phylogenetic trees for all DENV serotypes/genotypes
2015, 2016, and again in 2023. In 2015, the highest detected in the region since 1986, including DENV1

www.thelancet.com Vol 35 July, 2024 5

 Articles

Fig. 2: Dengue transmission dynamics in Rio de Janeiro. a) The first map indicates the different mesoregion of Rio de Janeiro state: Baixadas
(CL), Centro Fluminense (FC), Metropolitana do Rio de Janeiro (MR), Noroeste Fluminense (NWF), Norte Fluminense (NOF) e Sul Fluminense
(SF). Weekly notified dengue cases normalized per 100 K individuals by mesoregion in 2015–2023. b) Time-stamped Maximum Likelihood trees
for dengue serotypes (1–4) progress from left (serotype 1) to right (serotype 4). Tips on the trees are annotated with circles and colored based
on their locations. Colors denote distinct sampling locations in accordance with the legend located to the left of each tree. Only significant
sampling locations involved in DENV transmission with Rio de Janeiro have been highlighted, while others are grouped as ‘Others.’ Rio de
Janeiro state genomes are highlighted in teal blue. c) Heatmaps display the number of inferred virus transitions, across the entire study
timeframe, ranging from lower numbers in light blue to higher numbers in dark blue. Inside each square there is the number of the inferred
transitions. The y-axis axis represents sinks/destinations, and the x-axis represents sources/origins, with one heatmap per serotype.

genotype V (DENV1-V), DENV2 genotype III (DENV2- and Northeast Brazil (8 introductions), as shown in
III), DENV3 genotype III (DENV3-III), and DENV4 ge- Fig. 2c.
notype II (DENV4-II). Phylogenetic reconstruction also pointed to the likely
Analysis indicated six probable introduction events introduction of DENV2-III into Rio de Janeiro from
of DENV1-V into Rio de Janeiro, with the initial dispersal events in the Caribbean and South America
introduction suggested to have a South American during the 1990s, as depicted in Fig. 2b. The Rio de
origin in the late 1980s. The Rio de Janeiro isolates Janeiro isolates are grouped into three distinct clades.
exhibited significant clustering with those from The earliest identified DENV2-III genome in Rio de
different Brazilian regions and other South American Janeiro is linked to its introduction from the American
countries, indicative of both internal and cross-border region, followed by its spread to Northeast Brazil and
transmission dynamics. Notably, a substantial number subsequent reintroductions in the state and other South
of transition events were traced back to Rio de American countries. Similar to DENV1-V, the most
Janeiro, predominantly originating from Southeast frequent introduction events for DENV2-III originated
Brazil (13 events), with Northeast Brazil and other from Southeast Brazil, totalling 13 events. The dissem-
South American countries each accounting for four ination of DENV2-III from Rio de Janeiro notably
events. Fewer transitions were observed from North affected other Brazilian regions, particularly Southeast
(2 events) and Midwest Brazil (1 event). In contrast, Brazil (14 events) and Northeast Brazil (9 events), and
transitions originating from Rio de Janeiro were also extended to other South American regions
primarily directed towards Southeast Brazil (17 events) (9 events), as shown in Fig. 2c.

6 www.thelancet.com Vol 35 July, 2024

 Articles

Concerning DENV3-III, our findings showed a pre- Chikungunya virus transmission dynamics in Brazil,
dominant association and clustering with regions in 2015–2023
Southeast Brazil (two events) and the Caribbean (one Fig. 3a illustrates the spatial distribution of CHIKV
event) (Fig. 2b). Following its first introduction, our data cases across the mesoregions of Rio de Janeiro. From
indicates that DENV3-III likely spread throughout the 2016 to the end of 2017, the prevalence of CHIKV
Northeast and Southeast mesoregions of Brazil. Finally, remained relatively stable but with a significant surge in
for DENV4-II, our analysis revealed that the isolates cases in the North (NWF and NOF) mesoregions in
from Rio de Janeiro were predominantly associated with 2018, followed by notable increases in the Central (FC,
those from Southeast and North Brazil. This observa- CL), and South (MR) mesoregions in 2019. In 2020,
tion, despite being based on a limited dataset, provides reported CHIKV cases were relatively low, likely influ-
evidence that DENV4-II was introduced into Rio de enced by the COVID-19 pandemic. Additionally, it is
Janeiro from these regions. Specifically, we identified interesting to note that, similar to previous observations,
two independent introduction events from Southeast the incidence of CHIKV does not appear to be related to
Brazil and one from North Brazil, as illustrated in human density, as indicated in Fig. 3a (Pearson coeffi-
Fig. 2c. cient correlation = −0.15, P value = 1.70 × 10ˆ −7).

a 2
NWF 2015 2016 2017 2018
Population Density per km
0 300 600 900 1200
CHIKV
NOF
incidence
FC
SF 0-10

CL 10-20
MR 20-100
2019 2020 2021 2022 2023
100-200

200-600

>600

b c

10.0
Southeast 11 1
Number of inferred transition

Brazil

7.5

5.0
Sink

Rio de 3 0
Janeiro

2.5

Northeast 8 1 0.0
Northeast Brazil
Southeast
Rio de Janeiro
Northeast Rio de Janeiro Southeast
Others
Brazil Brazil
2014 2016 2018 2020 2022 Source

Fig. 3: Chikungunya transmission dynamics in Rio de Janeiro. a) The first map indicates the different mesoregion of Rio de Janeiro state:
Baixadas (CL), Centro Fluminense (FC), Metropolitana do Rio de Janeiro (MR), Noroeste Fluminense (NWF), Norte Fluminense (NOF) e Sul
Fluminense (SF). Weekly notified chikungunya cases normalized per 100 K individuals by mesoregion in 2015–2023. b) Time-stamped Maximum
Likelihood tree for chikungunya. Tips on the tree are annotated with circles and colored based on their locations. Colors denote distinct
sampling locations in accordance with the legend located to the left of each tree. Only significant sampling locations involved in CHIKV
transmission with Rio de Janeiro transmission history are listed, while others are grouped as ‘Others.’ Rio de Janeiro state genomes are
highlighted in teal blue. c) Heatmaps display the number of inferred virus transitions, across the entire study timeframe, ranging from lower
numbers in light blue to higher numbers in dark blue. Inside each square there is the number of the inferred transitions. The y-axis axis
represents sinks/destinations, and the x-axis represents sources/origins.

www.thelancet.com Vol 35 July, 2024 7

 Articles

Phylogenetic analysis suggests that CHIKV was Fig. 3c), with a bidirectional exchange pattern evident, as
initially introduced to Rio de Janeiro through a dispersal three introduction events from the Northeast back into
event from Brazil’s Northeast mesoregion in early 2015 Rio de Janeiro were also identified.
(Fig. 3b). This corresponds to the same region where it
was first detected in mid-2014.32 Subsequent dispersion Zika virus transmission dynamics in Brazil,
from Rio de Janeiro to other regions of Brazil occurred 2015–2023
in 2018 and 2019 (Fig. 3b), highlighting the state’s Fig. 4a presents the geographical distribution of ZIKV
pivotal role in the continued expansion and resurgence cases across the mesoregions of Rio de Janeiro. In
of CHIKV, especially in the Northeast and Southeast contrast to the patterns observed with DENV and
mesoregions of Brazil. However, the paucity of genome CHIKV, ZIKV exhibited a single peak in cases in 2016
sequences from Rio de Janeiro in 2019 may have led to that affected nearly all macroregions across the state.32
an underestimation of the state’s impact on the broader Following that peak, from 2017 to 2020, there were
Brazilian CHIKV context. The majority of inferred minor increases in the number of ZIKV cases, primarily
transmission events, involving Rio de Janeiro, were observed in the North (NOF), Central (MR), and South
from Rio de Janeiro to Northeast Brazil (eight events, (SF) regions. Interestingly, as with DENV and CHIKV,

a NWF
Population Density per km
2 2015 2016 2017 2018
0 300 600 900 1200
NOF
FC ZIKV
SF incidence

0-1
CL
1-10
MR
2019 2020 2021 2022 2023 10-30

30-80

80-200

>200

b c
Southeast
0 1 11 0
Brazil
10.0

Number of inferred transition

Rio de 4 0 7 1
Janeiro
7.5

Northeast 1 0 0 0
Sink

Brazil 5.0

Midwest 0 10 1 0
Brazil 2.5

0 5 0 2 0.0
Caribbean
Northeast
Midwest
Caribbean Midwest Northeast Rio de Southeast
Southeast Caribbean
Rio de Janeiro Others Brazil Brazil Janeiro Brazil

2014 2016 2018 Source

Fig. 4: Zika transmission dynamics in Rio de Janeiro. a) The first map indicates the different mesoregion of Rio de Janeiro state: Baixadas (CL),
Centro Fluminense (FC), Metropolitana do Rio de Janeiro (MR), Noroeste Fluminense (NWF), Norte Fluminense (NOF) e Sul Fluminense (SF).
Weekly notified Zika cases normalized per 100 K individuals by mesoregion in 2015–2023. b) Time-stamped Maximum Likelihood tree of Zika
isolates. Tips on the tree are annotated with circles and colored based on their locations. Colors denote distinct sampling locations in accordance
with the legend located to the left of each tree. Only significant sampling locations involved in ZIKV transmission with Rio de Janeiro
transmission history are listed, while others are grouped as ‘Others.’ Rio de Janeiro state genomes are highlighted in teal blue. c) Heatmaps
display the number of inferred virus transitions, across the entire study timeframe, ranging from lower numbers in light blue to higher numbers
in dark blue. Inside each square there is the number of the inferred transitions. The y-axis axis represents sinks/destinations, and the x-axis
represents sources/origins.

8 www.thelancet.com Vol 35 July, 2024

 Articles

we also noticed that the incidence of ZIKV does not high–high) where identified in the northeastern region,
appear to be related to human density (Pearson coeffi- while clusters of mutual low incidence where generally
cient correlation = −0.15, P value = 8.16 × 10ˆ−5) identified in the south and southwestern regions
(Fig. 4a). This consistent observation across different (Fig. 5a–c). This result was in accordance with the
viruses suggests that factors other than human popula- observation that arboviral incidence was typically asso-
tion density might play a significant role in the spread of ciated with lower human density (Fig. 5d). In fact, all of
these diseases in the region. Starting in 2020 and the municipalities with high human density were
continuing through 2023, the reported incidence of Zika identified as belonging to the LISA clusters corre-
cases continuously decreased. Our phylogenetic tree of sponding to mutual low incidence independently of the
ZIKV (Fig. 4b) revealed multiple clades, underscoring pair of viruses analysed (Fig. 5a–c).
the occurrence of several introduction events in the
country over time.
There were three probable introduction events into Discussion
the state of Rio de Janeiro. ZIKV isolates originating In this study, we delved into the complex landscape of
from Rio de Janeiro are clustered with sequences from urban arboviruses in the state of Rio de Janeiro,
the Midwest, Northeast, and Southeast regions of Brazil, Southeast Brazil, focusing on the public health impli-
as well as with those from Caribbean regions. This cations of DENV, CHIKV, and ZIKV. Our analysis le-
complex pattern of virus dissemination highlights the verages genomic and incidence data sourced from the
intricate dynamics within Brazil and across the conti- Brazilian Ministry of Health to reveal spatial–temporal
nent. The majority of inferred transmission events patterns of these viruses within the region.
linked to Rio de Janeiro involved the influx of in- Our findings indicate that DENV has transitioned to
dividuals moving into Rio de Janeiro, primarily from an endemic status in Southeast Brazil, with significant
Northeast Brazil (seven events, as depicted in Fig. 4c), outbreaks occurring in early 2015, late 2015–early 2016,
and from the Caribbean (four events). early 2019, and again in 2023. Contrary to expectations,
Finally, we assessed the spatial cross-correlation of our analysis revealed no direct correlation between
incidence between the different viruses. Fig. 5 shows dengue incidence and population density. Further
LISA clusters for each pair of viruses. Across all pairs of investigation into climatic conditions, particularly tem-
viruses, clusters of mutual high incidence (LISA clusters perature, also revealed a surprisingly low correlation.

a Population
2
b
NWF Density per km LISA cluster
CHIKV-ZIKV (2015-2023)
FC 1200 High-High
NOF
SF
High-Low
900
Low-High
600
Low-Low
CL
300 NA
MR 0

c d
LISA cluster DENV-ZIKV (2015-2023)
LISA cluster
DENV-CHIKV (2015-2023)
High-High High-High
High-Low High-Low
Low-High Low-High
Low-Low Low-Low
NA NA

Fig. 5: Spatial cross-correlation between virus incidence and human density for DENV, CHIKV, and ZIKV (2015–2023). a) The first map
displays the population density per km2 calculated for the different mesoregions of Rio de Janeiro state: Baixadas (CL), Centro Fluminense (FC),
Metropolitana do Rio de Janeiro (MR), Noroeste Fluminense (NWF), Norte Fluminense (NOF), and Sul Fluminense (SF). b–d) Panels b–d illustrate
the spatial cross-correlation calculated for each pair of viruses (CHIKV-ZIKV; DENV-CHIKV; DENV-ZIKV) after aggregating incidence across the
observation period (2015–2023), utilizing the cross-correlation function to quantify Local Indicators of Spatial Association (LISA) clusters
(high–high, high-low, low-high, low–low), and the spatial autocorrelation global Moran’s I.

www.thelancet.com Vol 35 July, 2024 9

 Articles

Despite this, our data highlight that peri-urban regions, to DENV, suggesting overlapping transmission dynamics.
among others, have reported the highest mean tem- Interestingly, females represented 60% of the cases across
peratures over the last nine years. Given the low corre- all viruses, indicating a difference in infection rates based
lation, our findings suggest that a complex interplay of on gender.
factors influences the dynamics of dengue transmission.
These factors likely include vector control initiatives, Limitations
local immunity levels, and human mobility patterns. The limited number of complete genome sequences
The interaction between these factors and climatic available for these viruses, despite their high incidence,
conditions may further complicate the transmission points to a significant gap in our understanding of the
dynamics of such arboviruses at the national and arbovirus landscape in Rio de Janeiro. This gap is further
regional levels. This revelation has critical implications underscored by the recent approval of the live-attenuated
for public health strategies and resource allocation. It QDENGUA vaccine by Takeda, highlighting the impor-
indicates that interventions cannot solely focus on high- tance of prevention and effective treatment strategies
density urban areas but must also address the complex tailored to the mesoregion level.41 While our study pro-
interplay of factors contributing to DENV transmission vides insight into the spatial–temporal dynamics of
in less densely populated areas. Tailoring control mea- DENV, CHIKV, and ZIKV in Rio de Janeiro, it is con-
sures to account for these nuanced dynamics, including strained by the limited availability of whole genome se-
the enhancement of vector surveillance and control quences and incomplete metadata, which may hamper a
programs, public health education, and the development full understanding of transmission dynamics and restrict
of healthcare infrastructure to manage outbreaks, be- accurate estimates related to divergence time and ances-
comes pivotal. tral location reconstructions. Future sequencing efforts
Conversely, ZIKV presented a different trend, char- are essential for addressing these limitations.
acterised by a single, widespread surge of infection in
2016 that affected all mesoregions in the state, leading Conclusions
to a substantial outbreak.33,34 The national decrease in Considering Rio de Janeiro’s significance as a hub for
ZIKV cases thereafter is attributed to the onset of herd human mobility, our findings underscore the intricate
immunity, aligning with modelling studies’ forecasts dynamics of arbovirus transmission in the region and
that herd immunity would significantly curb trans- the urgent need for targeted public health interventions
mission within the virus’s first decade of emergence.33,35 to mitigate the substantial challenges posed by these
The pattern of inverse correlation between arbovirus diseases.
incidence and urban density, initially noted for DENV,
extends to CHIKV. This suggests that transmission is Contributors
Conception and design: A.M.B.d.F., and M.G.
more pronounced in northern areas, despite the south-
Development of methodology: G.S., S.T., K.M., V.F., E.C., J.L., and
ern mesoregions’ larger population. Factors such as M.G.; Acquisition of data: G.S., S.T., K.M., V.F., L.A.S., S.N.S., M.C.,
socioeconomic conditions, and human mobility, partic- J.L., E.C., L.C.J.A., F.d.B.N., A.M.B.d.F., J.L., and M.G.
ularly proximity to state borders, appear to facilitate the Analysis and Interpretation: G.S., S.T., K.M., V.F., E.C., J.L., and
sustained transmission of these viruses.36 M.G.
Verification of the primary data and results: V.F., E.C., J.L., and
Our study further documents significant outbreaks of M.G.
CHIKV in 2018 and 2019, followed by a notable decline in Writing, review, and/or revision of the manuscript: All authors.
cases. The 2019 epidemic, particularly severe in the North Study supervision: A.M.B.d.F., and M.G.
(NWF and NOF) mesoregions, represents the most sig-
nificant CHIKV outbreak in Rio de Janeiro to date.37,38 The Data sharing statement
Input files (e.g. alignments and metadata), used in the study are shared
concurrent COVID-19 pandemic is likely to have led to a
publicly on GitHub (https://github.com/genomicsurveillance/Arbovirus-
decrease in arbovirus reporting, a result of both a shift in genomic-surveillance/tree/main/Arbo-RJ).
surveillance priorities and the symptom overlap between
SARS-CoV-2 and arboviral infections.39,40 This situation Editor note
underscores the importance of maintaining robust sur- The Lancet Group takes a neutral position with respect to territorial
claims in published maps and institutional affiliations.
veillance systems, which are crucial for developing effec-
tive regional intervention strategies. Declaration of interests
Our demographic descriptive analysis revealed distinct We declare no competing interests.
patterns in the age and gender of individuals affected by
these arboviruses in Rio de Janeiro. DENV predominantly Acknowledgements
affected young adults around 32 years old, consistent with M. Giovanetti’s funding is provided by PON “Ricerca e Innovazione”
the age group’s susceptibility in urban settings. CHIKV 2014–2020. The authors would also like to acknowledge the Global
Consortium to Identify and Control Epidemics—CLIMADE, (T.O.,
cases were more common among individuals over 41, L.C.J.A., E.C.H., J.L., M.G.) (https://climade.health/). Authors addi-
likely due to the more severe symptoms experienced by tionally would like to acknowledge the CNPq and the Brazilian Ministry
older populations. ZIKV showed a similar age distribution of Health.

10 www.thelancet.com Vol 35 July, 2024

 Articles

Appendix A. Supplementary data 18 Pinotti F, Giovanetti M, de Lima MM, et al. Shifting patterns of
Supplementary data related to this article can be found at https://doi. dengue three years after Zika virus emergence in Brazil. medRxiv.
org/10.1016/j.lana.2024.100786. 2023. https://doi.org/10.1101/2023.05.29.23290597.
19 SINANWEB - Página inicial. https://portalsinan.saude.gov.br/.
20 Estado do Rio tem aumento do número de casos de dengue. Agência
Brasil; 2023. https://agenciabrasil.ebc.com.br/saude/noticia/2023-03/
References estado-do-rio-tem-aumento-do-numero-de-casos-de-dengue.
1 Lee SA, Economou T, de Castro Catão R, Barcellos C, Lowe R. The 21 Home - Nucleotide - NCBI. https://www.ncbi.nlm.nih.gov/nuccore.
impact of climate suitability, urbanisation, and connectivity on the 22 Vilsker M, Moosa Y, Nooij S, et al. Genome detective: an automated
expansion of dengue in 21st century Brazil. PLoS Negl Trop Dis. system for virus identification from high-throughput sequencing
2021;15:e0009773. data. Bioinformatics. 2019;35:871–873.
2 Falta de saneamento básico é considerado como um dos vilões da 23 Katoh K. MAFFT: a novel method for rapid multiple sequence
proliferação de Aedes aegypti. FUNDACENTRO. https://www.gov. alignment based on fast Fourier transform. Nucleic Acids Res.
br/fundacentro/pt-br/comunicacao/noticias/noticias/2016/4/falta-de- 2002;30:3059–3066.
saneamento-basico-e-considerado-como-um-dos-viloes-da-proliferacao- 24 Larsson A. AliView: a fast and lightweight alignment viewer and
de-aedes-aegypti editor for large datasets. Bioinformatics. 2014;30:3276–3278.
3 Fares RCG, Souza KPR, Añez G, Rios M. Epidemiological scenario 25 Nguyen L-T, Schmidt HA, von Haeseler A, Minh BQ. IQ-TREE: a
of dengue in Brazil. BioMed Res Int. 2015;2015. fast and effective stochastic algorithm for estimating maximum-
4 Nogueira RMR, de Araújo JMG, Schatzmayr HG. Dengue likelihood phylogenies. Mol Biol Evol. 2015;32:268–274.
viruses in Brazil, 1986-2006. Rev Panam Salud Publica. 2007;22: 26 Sagulenko P, Puller V, Neher RA. TreeTime: maximum-likelihood
358–363. phylodynamic analysis. Virus Evolution. 2018;4.
5 de Bruycker-Nogueira F, Souza TMA, Chouin-Carneiro T, et al. 27 Paradis E, Claude J, Strimmer K. APE: analyses of phylogenetics
DENV-1 Genotype V in Brazil: spatiotemporal dispersion pattern and evolution in R language. Bioinformatics. 2004;20:289–290.
reveals continuous co-circulation of distinct lineages until 2016. Sci 28 Wickham H. ggplot2. Springer International Publishing; 2016.
Rep. 2018;8(1):17160. https://doi.org/10.1038/s41598-018-35622-x. https://doi.org/10.1007/978-3-319-24277-4.
6 Torres MC, de Bruycker Nogueira F, Fernandes CA, et al. Re-intro- 29 Chen YG. A new methodology of spatial cross-correlation analysis.
duction of dengue virus serotype 2 in the state of Rio de Janeiro after PLoS One. 2015;10(5):e0126158. https://doi.org/10.1371/journal.
almost a decade of epidemiological silence. PLoS One. 2019;14(12): pone.0126158.
e0225879. https://doi.org/10.1371/journal.pone.0225879. 30 Adelino TÉR, Giovanetti M, Fonseca V, et al. Field and classroom
7 Nogueira RMR, Miagostovich MP, de Filippis AMB, Pereira MAS, initiatives for portable sequence-based monitoring of dengue virus
Schatzmayr HG. Dengue virus type 3 in Rio de Janeiro, Brazil. in Brazil. Nat Commun. 2021;12:2296. https://doi.org/10.1038/
Mem Inst Oswaldo Cruz. 2001;96:925–926. s41467-021-22607-0.
8 Naveca FG, Santiago GA, Maito RM, et al. Reemergence of dengue 31 Brito AF, Machado LC, Oidtman RJ, et al. Lying in wait: the resur-
virus serotype 3, Brazil, 2023. Emerg Infect Dis. 2023;29. gence of dengue virus after the Zika epidemic in Brazil. Nat Commun.
9 Heringer M, Souza TMA, Lima MDRQ, et al. Dengue type 4 in Rio 2021;12:2619. https://doi.org/10.1038/s41467-021-22921-7.
de Janeiro, Brazil: case characterization following its introduction 32 Nunes MR, Faria NR, de Vasconcelos JM, et al. Emergence and po-
in an endemic region. BMC Infect Dis. 2017;17:410. http://medrxiv. tential for spread of Chikungunya virus in Brazil. BMC Med.
org/lookup/doi/10.1101/2023.11.11.23298412, 2023. 2015;13:102.
10 Lowe R, Barcellos C, Brasil P, et al. The Zika virus epidemic in 33 Lourenço J, Maia de Lima M, Faria NR, et al. Epidemiological and
Brazil: from discovery to future implications. Int J Environ Res ecological determinants of Zika virus transmission in an urban
Public Health. 2018;15:96. setting. Elife. 2017;9:62.
11 Simião AR, Barreto FK, Oliveira RD, et al. A major chikungunya 34 Giovanetti M, Faria NR, Lourenço J, et al. Genomic and epidemi-
epidemic with high mortality in northeastern Brazil. Rev Soc Bras ological surveillance of Zika virus in the amazon region. Cell RepM.
Med Trop. 2019;52:e20190266. 2020;18:30.
12 Giovanetti M, Goes de Jesus J, Lima de Maia M, et al. Genetic 35 Michlmayr D, Kim EY, Rahman AH, et al. Comprehensive
evidence of Zika virus in mother’s breast milk and body fluids of a immunoprofiling of pediatric Zika reveals key role for monocytes
newborn with severe congenital defects. Clin Microbiol Infect. in the acute phase and No effect of prior dengue virus infection.
2018;24(10):1111–1112. https://doi.org/10.1016/j.cmi.2018.06.008. Cell Rep. 2020;28:31.
13 Xavier J, Giovanetti M, Fonseca V, et al. Circulation of chikungunya 36 Baker RE, Mahmud AS, Miller IF, et al. Infectious disease in an era
virus East/Central/South African lineage in Rio de Janeiro, Brazil. of global change. Nat Rev Microbiol. 2022;193:205.
PLoS One. 2019;14:e0217871. 37 Rio com saúde. http://www.riocomsaude.rj.gov.br/site/conteudo/.
14 Peixoto VGMNP, Azevedo JP, Luz KG, Almondes KM. Cognitive 38 World Health Organization (WHO). Dengue monitoring; 2024.
dysfunction of chikungunya virus infection in older adults. Front https://www.who.int/emergencies/disease-outbreak-news/item/2023-
Psychiatr. 2022;13:823218. DON475.
15 Zika and guillain-barre syndrome. CDC; 2014. https://www.cdc.gov/ 39 Borre F, Borri JI, Cohen YZ, Gasparoto M, Gurung TB. Impact of
zika/healtheffects/gbs-qa.html. the COVID-19 pandemic on infectious diseases in Brazil: a case
16 Aromolaran A, Araujo K, Ladines-Lim JB, et al. Unequal burden of Zika- study on dengue infections. Epidemiologia. 2022;3:97–115.
associated microcephaly among populations with public and private 40 Silva SJRD, Magalhães JJFD, Pena L. Simultaneous circulation of
healthcare in Salvador, Brazil. Int J Infect Dis. 2022;120:201–204. DENV, CHIKV, ZIKV and SARS-CoV-2 in Brazil: an inconvenient
17 de Magalhães-Barbosa MC, Prata-Barbosa A, Robaina JR, truth. One Health. 2021;12:100205.
Raymundo CE, Lima-Setta F, da Cunha AJ. Trends of the micro- 41 CDC. Dengue vaccination: what everyone should know | CDC. Centers
cephaly and Zika virus outbreak in Brazil, january–july 2016. Trav for Disease Control and Prevention; 2022. https://www.cdc.gov/
Med Infect Dis. 2016;14:458–463. vaccines/vpd/dengue/public/index.html.

www.thelancet.com Vol 35 July, 2024 11