Vaccine 39 (2021) 5318–5325

Contents lists available at ScienceDirect

Vaccine
journal homepage: www.elsevier.com/locate/vaccine

Effectiveness of a single-dose mass dengue vaccination in Cebu,

Philippines: A case-control study
Michelle Ylade a,⇑, Kristal An Agrupis a, Jedas Veronica Daag a, Maria Vinna Crisostomo a,
Mark Owen Tabuco a,1, Ava Kristy Sy c, Joshua Nealon d, Denis Macina d, Jesus Sarol b, Jacqueline Deen a,
Anna Lena Lopez a,1
a
Institute of Child Health and Human Development, National Institutes of Health, University of the Philippines – Manila, Philippines, Pedro Gil St., Ermita, Manila, Philippines
b
Interdisciplinary Health Sciences Institute, University of Illinois at Urbana-Champaign, Urbana, IL, USA
c
Department of VirologyResearch Institute for Tropical Medicine, Muntinlupa, Philippines
d
Sanofi Pasteur, Lyon, France

a r t i c l e i n f o a b s t r a c t

Article history: Background: Dengue fever is an important public health problem in the Philippines. In April 2016, the
Received 26 April 2021 Department of Health launched a three-dose school based dengue vaccination program of nine- to
Received in revised form 14 July 2021 fourteen-year-old children in three regions with the highest number of dengue cases using CYD-TDV
Accepted 16 July 2021
(Dengvaxia, Sanofi Pasteur). In July 2017, a community-based dengue vaccination program was imple-
Available online 7 August 2021
mented in Cebu province. The program was discontinued in December 2017 amidst public controversy,
after the first dose had been administered. We assessed the effectiveness of a single dose of CYD-TDV
Keywords:
against hospitalized virologically confirmed dengue (VCD).
Dengue
Dengue fever
Methods: We conducted a case-control study in Cebu province following the dengue mass vaccination.
Dengue vaccine Children who were nine to fourteen years of age during the mass vaccination and subsequently admitted
Effectiveness to any of four participating public hospitals with suspected dengue were enrolled in the study as cases.
Blood for RT-PCR and clinical and socio-demographic information were obtained. To estimate the level of
vaccine protection, vaccination status was compared between children with hospitalized virologically
confirmed dengue and controls of the same six-year age-group as the cases, matched on sex, neighbor-
hood and time of occurrence of cases.
Findings: We enrolled 490 cases and 980 controls. Receipt of one dose of CYD-TDV was associated with
26% (95 % CI,
2 to 47%; p = 0 0675) overall protection against hospitalized virologically confirmed den-
gue and 51% (95 % CI, 23 to 68; p = 0 0016) protection against dengue with warning signs.
Interpretation: A single dose of CYD-TDV given to nine to fourteen-year-old children through a
community-based mass vaccination program conferred protection against dengue with warning signs
and severe dengue but we were unable to conclude on protection against milder illness.
Ó 2021 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/).

1. Introduction 25 months of 66% for those nine years of age or older but an
increased risk of hospitalized and severe dengue in children two
In 2015, the first dengue vaccine, CYD-TDV (DengvaxiaTM, Sanofi to five years.[1] Since only a subset of the trial participants had
Pasteur), was licensed on a three-dose schedule with an indication dengue serostatus assessed at baseline, it could not be confirmed
for nine- to forty-five-year-old individuals in dengue endemic at that time whether the safety signal was related to age or dengue
areas. The multi-country Phase 3 trials showed a three-dose pooled serostatus at time of vaccination. In April 2016, the World Health
vaccine efficacy against virologically-confirmed dengue (VCD) over Organization (WHO) recommended that high dengue transmission
countries consider introducing CYD-TDV vaccine in age groups
with seroprevalence of 70% or greater [2].
⇑ Corresponding author at: Institute of Child Health and Human Development
Starting in 2016, the Philippine Department of Health (DOH)
National Institutes of Health University of the Philippines – Manila, 623 P. Gil St.,
Manila 1000, Philippines.
implemented a three-dose dengue vaccination program among
E-mail address: [email protected] (M. Ylade). nine- to fourteen-year-old children in Central Luzon, Calabarzon
1
Deceased (Dr Anna Lena Lopez died on 12 January 2020, Mark Tabuco died on 21 and Metro Manila. In 2017, the program was extended to Cebu pro-
November 2020).

https://doi.org/10.1016/j.vaccine.2021.07.042
0264-410X/Ó 2021 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/).
M. Ylade, Kristal An Agrupis, Jedas Veronica Daag et al. Vaccine 39 (2021) 5318–5325

vince. Subsequently, using a novel diagnostic test that differenti- 2.3. Case-control study
ates between antibodies induced by CYD-TDV and wild-type den-
gue infection,[3] a follow-up analysis of the CYD-TDV Phase 3 The primary research question is: Does receipt of one dose of
trials showed that vaccination conferred protection for at least five CYD-TDV protect against hospitalized VCD? Using a case-control
years among dengue-seropositive participants but resulted in an design, we enrolled cases from the source population and recruited
increased risk for hospitalized and severe dengue among dengue- two matched neighborhood controls per case. Then for both cases
seronegative participants.[4] Consequently the Philippine dengue and controls, we ascertained whether or not they had received a
vaccination program was discontinued, after the first dose had dose of CYD-TDV. Information on vaccination status and other
been administered in Cebu. Herein, we report the results of a exposure variables were obtained by study staff who were una-
case-control study to evaluate the effectiveness of a single-dose ware of how the information on vaccination status was to be used
mass dengue vaccination in Cebu. in the analysis.
All patients with suspected or probable dengue admitted at the
participating hospitals were eligible to be enrolled as a case if they
2. Methods
fulfilled the following criteria: 1) written informed consent and
assent to participate in the study; 2) born between January 2003
This is a matched case-control study.[5] The protocol was
to December 2008; 3) resided in Cebu since June 2017 and eligible
approved by the University of the Philippines Manila-Research
to have received the dengue vaccine; 4) with fever for less than five
Ethics Board. A written informed consent was provided by a parent
days; and 5) submitted a blood sample within five days of fever
or legal guardian and assent was documented with the participant.
onset. For the case to be included in the analysis, the participant
The study was conducted in accordance with the Helsinki Declara-
should meet all the inclusion criteria and have a blood sample pos-
tion and Good Clinical Practice guidelines and is reported accord-
itive for dengue by RT-PCR. Repeat episodes meeting the criteria
ing to the STROBE statement.
were included.
We aimed to recruit controls that are representative of the
2.1. Study site and mass immunization campaign source population from which the cases were selected.[5]
Neighborhood-matching would help ensure similar risk factors
Cebu province, as in the rest of the country, is endemic for den- associated with vector dynamics, including the flight range of adult
gue with increases in dengue cases from July to November, coincid- female Aedes aegypti,[13,14] the risk for clustering of dengue cases
ing with the rainy season.[6] In the Philippine public health attributed to concurrent infection of nearby mosquitoes, and the A.
system, the Rural Health Units (RHU) are the main primary health aegypti behavior of taking a single blood meal from multiple hosts.
care facilities, while hospitals provide secondary care.[7] Prior to [15] A systematic selection procedure was used to recruit two con-
implementation of the mass dengue vaccination in Cebu, the trols per case as soon as the RT-PCR results became available.
DOH carried-out extensive preparation and community engage- Recruitment of controls was started on the third house to the left
ment. RHU staff listed the name, barangay (village), birthdate, of the case’s house (index house), up to a maximum of 500 m or
and sex of healthy children nine to fourteen years of age residing 20 houses, whichever came first. If necessary, the same procedure
in Cebu. A total of 285,242 children were listed and invited to par- was repeated to the right of the index house until two controls per
ticipate in the vaccination campaign. CYD-TDV doses given from case were recruited. Only one control was recruited per household.
June to August 2017 were recorded in the list and maintained as Sex- and neighborhood-matched controls of the same six-year age-
the vaccination registry. Each child received the CYD-TDV (Deng- group as the case (i.e. born between January 2003 to December
vaxia, Sanofi Pasteur) following the manufacturer’s prescribing 2008) were eligible as a control if they had not sought treatment
information. CYD-TDV is a live recombinant vaccine supplied as 5 for a dengue-like illness from June 2017 to the date of onset of
doses/vial. Eligible children were vaccinated using 0.5 ml given the febrile illness of his or her matched case. Eligibility for selection
subcutaneously in the left deltoid area. The vaccine was stored also required the same informed consent, residency, and eligibility
at + 2–8°C and transported using WHO recommended carriers with to receive the dengue vaccine, as applied to the cases. We excluded
ice-packs.[8] A vaccination card was provided to the parents. An those had been previously recruited as a control.
estimated 149,023 (52 2%) children received a single CYD-TDV Demographic, socio-economic, and environmental variables
dose in Cebu before the program was discontinued.[9] were ascertained through questionnaires administered to cases
and controls and their families. The questionnaires did not include
2.2. Post-vaccination surveillance for dengue information on race and ethnicity (all participants were Filipino),
body mass index, smoking status, medical/immunologic status
We invited four government hospitals (of 13)[10] in Cebu to par- and concomitant drug use. Clinical data on the cases were obtained
ticipate in the case-control study. Surveillance for febrile cases at the from source documents in the hospital. Receipt of the dengue vac-
participating hospitals was started on February 15, 2018. Onset of ill- cine during the mass immunization program was ascertained in
ness was defined as the day when fever was reported to have started, face-to-face interviews. The parents of cases and controls were
whether or not documented by a temperature reading of 38 °C or asked to show vaccination cards distributed during the campaign
higher. About five ml of blood was obtained from each patient within and copy was kept with the case report form. For individuals
the first 5 days of illness (acute phase). Blood samples were collected who claimed to have been vaccinated but who were not in posses-
in anticoagulant-free vacutainer tubes, processed, and aliquoted. sion of a card, vaccination status was confirmed by searching the
Sera were stored at 2–8 °C and shipped within seven days to the vaccination registry. Decisions about linkage to the vaccination
Research Institute of Tropical Medicine in Manila for dengue RT- registry were made blinded to case-control status and were based
PCR. Total nucleic acid was extracted from the serum samples using on the subject’s name, barangay, birthdate, and sex.
the QIAmp Viral RNA (QIAGEN, Valencia, CA, USA) kit according to
the manufacturer’s protocol. Dengue detection and serotyping of 2.4. Statistical analysis
samples was done using the Simplexa Dengue assay (Focus Diagnos-
tics, Cypress, CA, USA).[11] Children hospitalized with VCD were fol- The sample size was calculated based on the requirements for a
lowed until discharge and their illness classified according to WHO matched case-control analysis. The study was designed to have
2009 criteria.[12] a statistical power of 80% at a significance level of 0 05 to detect
5319
M. Ylade, Kristal An Agrupis, Jedas Veronica Daag et al. Vaccine 39 (2021) 5318–5325

Fig. 1. Assembly of participants for the case-control study, February 15, 2018 to February 14, 2020.

a 50% vaccine effectiveness assuming a 50% vaccine coverage. Con- a participating hospital and for physicians to hospitalize the child.
sidering a 20% drop-out or subsequent non-participation, the This analysis included VCD cases and their age- and sex-matched
design required at least 144 hospitalized VCD dengue cases and controls whose dates of enrolment were between February 15,
288 controls. Case recruitment is continuing irrespective of num- 2018 to February 14, 2020 with vaccination defined as receipt of
bers achieved to enable sub-analyses. Sample size was calculated one dose, documented either by vaccination card or vaccination
using PASS 14 (Kaysville, UT, USA). registry.
The primary analysis, formulated a priori, addressed the protec- We performed crude analysis to determine the odds ratio (OR)
tion conferred by one dose of vaccine against VCD that was severe of vaccination status by employing conditional logistic regression
enough to have prompted the parents to bring the child for care at on the matched dataset. Selected socio-economic, environmental

5320
M. Ylade, Kristal An Agrupis, Jedas Veronica Daag et al. Vaccine 39 (2021) 5318–5325

Fig. 2. Timeline of 724 enrolled febrile cases and 517 virologically-confirmed dengue cases admitted to study hospitals in Cebu, Philippines, February 15, 2018 to February 14,
2020.

and behavioral variables that were known to be associated with PCR positive. We plotted the monthly distribution of the 724 sus-
dengue were also compared between cases and their matched con- pected dengue episodes and the 517 VCD cases (Fig. 2). Out of the
trols in bivariate analysis using conditional logistic regression to 517 VCD cases, 27 (5 2%) were excluded because appropriate con-
identify potential confounders. Among the variables whose associ- trols could not be recruited. A total of 490 VCD cases and 980 age-
ations with case-control status were statistically significant and and sex-matched controls were included in this analysis.
were known correlated with each other, we selected those that We assessed the characteristics of the 490 cases (Table 1). Aside
had strongest association based on p-values to minimize mutual from fever, the most common presumptive manifestations were
confounding among them. We fitted a logistic regression model nausea or vomiting (349 or 71 2%), anorexia (193 or 39 4%) and
controlling for one selected variable at a time and then performed myalgia (162 or 33 1%). Around two-thirds (329 or 67 1%) of cases
backward stepwise regression in selecting a final model. Variables presented one or more warning signs. The most common warning
were eliminated if the tests for their corresponding coefficients did sign was abdominal pain (225 or 45 9%), followed by lethargy (125
not reach statistical significance, provided their removal did not or 25 5%) and restlessness (83 or 20 9%). Bleeding was noted in 67
markedly change the OR for the vaccination and dengue associa- (13 7%); 49 (71 0%) from the nasal mucosa, 10 (145%) from the
tion. Conditional likelihood estimation was used to derive esti- gums and 10 (14 5%) from the lips and oral mucosa. There were
mates of the coefficients of the model. The estimated coefficient four (0 8%) severe dengue cases and one death (0 2%). All four
for the vaccination variable in this model was exponentiated to DENV serotypes were detected but over half (266 or 54 3%) were
obtain the adjusted odds ratio. The value (1 - adjusted odds DENV 3.
ratio) X 100 percent for the vaccination variable was computed We compared the clinical features of cases by vaccination status
to estimate adjusted levels of vaccine protection. We also consid- (Table 2). There was a significantly lower proportion of bleeding
ered case status based on dengue severity and serotype and like- (p = 0 04) and less severe illness (p = 0 007) among those who
wise fitted conditional logistic regression models for these had received a vaccine dose compared to the unvaccinated cases.
outcomes. All p-values and 95 percent confidence intervals were One case who had been vaccinated died. This was an eleven-
interpreted in a two-tailed fashion. The data analysis was per- year-old girl who was admitted on the fifth day of fever and pre-
formed using SAS software (Ver 9 4., 2016, SAS Institute Inc. Cary, sented with nausea, vomiting, retroorbital pain, malaise and
NC, USA). abdominal pain. She was eventually transferred to a tertiary hospi-
tal where her condition worsened, accompanied by severe bleed-
3. Results ing. Dengue RT-PCR result was positive for DENV serotype 3.
We compared demographic and socio-behavioral variables
From February 15, 2018 to February 14, 2020, there were between cases and controls (Table 3). Cases had slightly higher
33,371 presentations by patients to a study hospital, which were percentage of older children (13 years and above) than controls
assessed for eligibility (Fig. 1). Of these, 32,378 (97 0%) did not fulfil (p = 0 0262). Reporting a household member and neighbor diag-
the study criteria and were excluded. Of the 993 eligible hospital- nosed with dengue during the past seven days was significantly
ized suspected dengue episodes, 269 (27 1%) were excluded for more common among cases than controls (p < 0 0001). Variables
consent refusal or withdrawal or blood collection > 5 days after associated with higher socio-economic status including living in
fever onset. We enrolled 724 (72 9%) suspected dengue episodes a house made of concrete (p < 0 0001), living in a screened house
in 722 children, of which 517 (71 4%) episodes were dengue RT- (p = 0 0002) and ownership of a computer (p < 0 0001), refrigerator

5321
M. Ylade, Kristal An Agrupis, Jedas Veronica Daag et al. Vaccine 39 (2021) 5318–5325

Table 1 Table 2
Characteristics of the virologically confirmed dengue cases. Clinical features of the virologically confirmed dengue cases, by vaccination status.

Case (N = 490) Vaccinated Not vaccinated Χ2 test

(n = 94) (n = 396) (p-
No Percent
value)
Hospital where case was admitted:
Presumptive signs and symptoms:
Cebu Provincial Hospital – Balamban 91 18 6
Nausea/vomiting 63 (6 7 0) 286 (7 2 2) 10
Cebu Provincial Hospital – Bogo 192 39 2
(0 3 2)
Cebu Provincial Hospital – Danao 122 24 9
Rash 28 (30) 118 (2 9 8) 00
Eversley Child’s Sanitarium and General Hospital 85 17 4
(0 9 9)
Number of days of fever on admission:
Headache 7 (7 5) 52 (1 3 1) 23
1 1 02
(0 1 3)
2 68 13 9
Retroorbital pain 25 (2 6 6) 103 (26) 00
3 127 25 9
(0 9 1)
4 222 45 3
Myalgia 26 (2 7 7) 137 (3 4 6) 27
5 72 14 7
(0 2 5)
Presumptive signs and symptoms:
Anorexia 39 (4 1 5) 155 (3 9 1) 04
Nausea/vomiting 349 71 2
(0 8 2)
Rash 145 29 6
Arthralgia 19 (2 0 2) 104 (2 6 3) 26
Headache 59 12 0
(0 2 7)
Retroorbital pain 126 25 7
Malaise 53 (5 6 4) 243 (6 1 4) 08
Myalgia 162 33 1
(0 3 7)
Anorexia 193 39 4
Watery stools 28 (2 9 8) 127 (3 2 1) 02
Arthralgia 122 24 9
(0 9 0)
Malaise 296 60 4
Flushed skin 19 (2 0 2) 86 (2 1 7) 01
Watery stools 155 31 6
(0 7 5)
Flushed skin 104 21 2
Warning signs:
Warning signs: 329 67 1
Abdominal pain 35 (3 7 2) 191 (4 8 2) 40
Abdominal pain 225 45 9
(0 1 3)
Persistent vomiting 69 14 1
Persistent vomiting 15 (1 6 0) 56 (1 4 1) 02
Lethargy 125 25 5
(0 6 5)
Restlessness 83 20 9
Lethargy 21 (2 2 3) 103 (2 6 0) 05
Bleeding 67 13 7
(0 4 6)
Enlarged liver 1 02
Restlessness 18 (1 9 1) 84 (2 1 2) 02
Signs of severe dengue: 4 08
(0 6 6)
Shock 2 04
Bleeding 7 (7 4) 61 (1 5 4) 40
Respiratory signs of fluid accumulation 2 04
(0 0 4)
Plasma leakage 2 04
Enlarged liver 1 (1 1) 0 (0 0) 13 0
Severe bleeding 2 02
(0 0 1)
Died 1 02
Signs of severe dengue:
Serotype:
Shock 0 (0 0) 2 (0 5 1) 17
DENV 1 36 73
(0 4 3)
DENV 2 158 32 2
Respiratory signs of fluid 1 (1 1) 1 (0 2 5) 17
DENV 3 266 54 3
accumulation (0 4 3)
DENV 4 22 45
Plasma leakage 1 (1 1) 1 (0 2 5) 17
Indeterminate (2 serotypes/RT-PCR test) 8 16
(0 4 3)
Severe bleeding 1 (1 1) 1 (0 2 5) 06
(0 7 3)
(0 0041) and car (0 0008) were significantly more common among Severity of illness:
Dengue without warning 42 (4 4 8) 119 (3 0 1) 74
cases than controls. The percentage reporting stagnant water in the
signs (0 0 1)*
surroundings was higher for cases than in controls (p = 0 0004). Dengue with warning sign 50 (5 3 2) 275(6 9 4)
Variables associated with dengue were treated as potential con- (s)
founders and were included in a logistic regression model individ- Severe dengue 2 (2 1) 2 (0 5)
ually and simultaneously as control variables. Using backward Outcome
Recovered 93 (9 8 9) 396 (1 0 0) 42
stepwise elimination, presence of dengue in the household and (0 0 4)
neighborhood, house with screen, stagnant water, main housing Died 1 (1 1) 0
material and possession of computer were found to be indepen-
* Dengue with warning sign(s) combined with severe dengue for validity of chi-
dently associated with dengue (p h0 0 5). The results of logistic square test.
regression analyses that controlled for these variables individually
and simultaneously suggested these variables did not confound the
observed effect of vaccination on dengue incidence.
We compared the odds ratios for single-dose CYD-TDV vaccina- cination between cases of dengue with warning signs and their
tion between cases and their matched controls, overall, by severity matched controls was 0 52 (95% CI: 0 35 to 0 78) and the adjusted
of dengue illness, and by DENV serotype (Table 4). The overall OR was 0 49 (95% CI: 0 32 to 0 77). When combined with severe
crude OR was 0 73 (95% CI: 054 to 0 99) and adjusted OR was 0 dengue, the crude and adjusted ORs were 0 54 (95% CI: 0 36 to 0
74 (95% CI: 0 53 to 1 02). This was equivalent to a vaccine effective- 80) and 0 52 (95% CI: 0 34 to 0 80), respectively. The vaccine con-
ness of 26% (95% CI:
2% to 47%). The crude and adjusted ORs dur- ferred 51% (95% CI: 23% to 68%) protection against dengue with
ing the first year of surveillance were 0 54 (0 35 to 0 82) and 0 51 (0 warning signs and 48% (95% CI: 20% to 66%) protection against den-
32 to 0 82), respectively. The crude and adjusted ORs during the gue with warning signs combined with severe dengue. The crude
second year of surveillance were 1 03 (0 67 to 1 60) and 1 14 (0 and adjusted ORs for CYD-TDV vaccination between cases and con-
70 to 1 83), respectively. This was equivalent to vaccine effective- trols and the vaccine effectiveness were calculated by DENV sero-
ness of 49% (19 to 68%) during the first year and
14% (-83 to 30%) type but no conclusions could be reached due to insufficient
during the second year. The crude OR for single-dose CYD-TDV vac- sample size.

5322
M. Ylade, Kristal An Agrupis, Jedas Veronica Daag et al. Vaccine 39 (2021) 5318–5325

Table 3
Comparison of the virologically confirmed dengue cases and their controls.

Cases Controls Wald Χ2 test (p-value)

(N = 490) (N = 980)
No. Percent No. Percent
Sex:
Female 238 48 6 476 48 6 Not done (matching variable
Male 252 51 4 504 51 4
Age in years:
9–10 77 15 7 190 19 4
11–12 182 37 1 379 38 7 4 9 (0 0 3)
13–14 166 33 9 275 28 0
15–17 64 13 1 136 13 9
17 1 02 0 00
Reported history of past dengue:
No 468 95 5 929 94 8 0 4 (0 5 4)
Yes 22 45 51 52
Hospital admission during previous dengue episode:
Yes 22 45 41 42 0 01 (0 9 9)
No 0 00 10 10
Household member diagnosed with dengue during the past 7 days 9 94 20 20 33 8 (<0 0001)
Neighbor diagnosed with dengue during the past 7 days 170 34 7 171 17 5 62 1 (<0 0001)
Household head with > 6 years of schooling 458 93 5 891 90 9 3 3 (0 0 7)
Main housing material:
Concrete 289 59 0 440 44 9 27 1 (<0 0001)
Wood 198 40 4 528 53 9
Thatch (Nipa leaves) 2 04 6 06
Bamboo 1 02 5 0 51
Galvanized iron sheet 0 00 1 01
Environmental conditions:
With screens 61 11 8 59 60 13 5 (0 0 1)
Presence of stagnant 188 38 4 289 29 5 12 5 (0 0 1)
water
Ownership of specific household appliances:
Radio 266 54 3 506 51 6 1 0 (0 3 2)
Television 423 86 3 814 83 1 2 7 (0 1 0)
Refrigerator 226 46 1 380 38 8 8 3 (0 0 1)
Bicycle 188 38 4 319 32 6 5 1 (0 0 2)
Motorcycle 258 52 7 486 49 6 1 4 (0 2 5)
Mobile phone 474 96 7 934 95 3 1 7 (0 1 9)
Desktop/handheld 85 17 4 90 92 22 2 (<0 0001)
computer
Electricity 477 97 4 948 96 7 0 5 (0 5 0)
Car 37 76 35 36 11 2 (0 0 1)
Migrated to the current residence during the past 2 years 28 57 36 37 3 8 (0 0 5)

Household members use of topical insect repellant:

No 407 83 1 754 76 9 8 1 (0 0 4)
<3 days/week 40 82 118 12 0
3–5 days/week 19 39 41 42
Everyday 24 49 67 68
Burn mosquito coil* during the day:
No 214 43 7 432 44 1 3 3 (0 3 5)
<3 days/week 88 18 0 167 17 0
3–5 days/week 66 13 5 107 10 9
Everyday 122 24 9 274 28 0
Use insecticide spray at home:
No 399 81 4 819 83 6 4 1 (0 2 5)
<3 days/week 62 12 7 114 11 6
3–5 days/week 11 22 26 27
Everyday 18 37 20 20
Fogging in the neighborhood during the past month: 21 43 42 43 0 0 (1 0 0)

*Spiral made from a dried paste of pyrethrum powder, which when lit burns slowly to produce a mosquito-repellent smoke.

4. Discussion conferred by three doses in this age group over a 25-month period
(80%) against hospitalized dengue.[1]
We found that CYD-TDV conferred protection against more seri- This effectiveness study conducted under the conditions of real-
ous dengue illness over two dengue seasons, with protection life public health program incorporated several features to help
higher during the first compared to the second year of surveillance, ensure the validity of the results. Patients underwent systematic
although the effectiveness estimate for the second year did not evaluation and confirmation of dengue by RT-PCR. Vaccination his-
reach statistical significance. Our estimate of single dose effective- tory was prospectively documented and verified by interview, vac-
ness against dengue with warning signs of 51% within 30 months cination card and registry. Controls were selected in a matched
after vaccination is, as expected, lower than the vaccine efficacy fashion, and extensive information about potentially confounding

5323
M. Ylade, Kristal An Agrupis, Jedas Veronica Daag et al. Vaccine 39 (2021) 5318–5325

Table 4
Single dose CYD-TDV vaccine effectiveness in Cebu, Philippines.

Number (%) vaccinated Crude odds ratio Adjusted* odds ratio Vaccine effectiveness p-value
among the: (95% CI) (95% CI) (95% CI)
Cases Matched controls
Overall 94/490(1 9 2) 228/980(2 3 3) 0 73(0 54 to 0 99) 0 74(0 53 to 1 02) 26% (-2 to 47%) 0 07
February 15, 2018 to February 14, 2019 47/254(1 8 5) 136/508(2 6 8) 0 54(0 35 to 0 82) 0 51(0 32 to 0 82) 49% (19 to 68%) 0 01
February 15, 2019 to February 14, 2020 47/236(1 9 9) 92/472(1 9 5) 1 03(0 67 to 1 60) 1 14(0 70 to 1 83)
14% (-83 to 30%) 0 60
According to severity of illness:
Dengue without warning sign 42/161(2 6 1) 75/322(2 3 3) 1 21(0 74 to 1 96) 1 32(0 76 to 2 28) –32% (-128 to 24) 0 32
Dengue with warning sign 50/326(1 5 3) 153/652(2 3 5) 0 52(0 35 to 0 78) 0 49(0 32 to 0 77) 51% (23 to 68) 0 01
Dengue with warning sign / Severe dengue 52/329(1 5 8) 153/658(2 3 3) 0 54(0 36 to 0 80) 0 52(0 34 to 0 80) 48% (20 to 66) 0 01
According to DENV serotype:
DEN 1 4/36(1 1 1) 8/72(1 1 1) 0 48(0 13 to 1 76) 0 42(0 10 to 1 80) 58% (-80 to 90%) 0 24
DEN 2 30/158(1 9 0) 75/316(2 3 7) 0 79(0 47 to 1 30) 0 79(0 46 to 1 35) 21% (-35 to 54%) 0 39
DEN 3 55/266(2 0 7) 129/532(2 4 2) 0 73(0 48 to 1 10) 0 76(0 48 to 1 20) 24% (-20 to 52%) 0 24
DEN 4 4/22(1 8 2) 13/44(2 9 5) 0 86(0 23 to 3 25) 0 89(0 15 to 5 29) 11% (-43 to 85%) 0 89

*Adjusted for dengue in the household, dengue in the neighborhood, house with screen, stagnant water, main housing material, possession of computer after backward
selection.

variables was collected and controlled for in the analyses. The ana- both a systematic difference in controls compared to the source
lytic plan was formulated a priori. population and a shift in their characteristics over the first two
Despite these measures, the study is not a double-blind, con- years of surveillance. The cause of these differences is not evident
trolled, randomized trial and subject to the limitations of an obser- from our available data but raises the possibility they could have
vational study. Participants (and their parents) were aware of their impacted the precision of our estimates.
vaccination status. The suspension of the mass dengue vaccination As shown in the CYD-TDV clinical trials, breakthrough illness
program occurred amidst intense media coverage and public out- after vaccination will occur in areas such as our study sites with
rage.[9] Children could be more likely to be brought and admitted high dengue transmission. [20,21] Since the vaccination program
to hospital if they had been vaccinated (than non-vaccinated) in the Philippines was conducted without prior knowledge of den-
because they were perceived to be at an increased risk. In addition, gue serostatus, our study measured the protective effect of the vac-
starting in March 2018, the DOH established ‘‘dengue express cine on a population of mixed serostatus at time of vaccination. In a
lanes” for CYD-TDV recipients who present to public and private separate longitudinal cohort study in which we enrolled 2,996 chil-
hospitals with any illness, with expenses paid by the Philippine dren 9 to 14 years of age in Bogo and Balamban, Cebu prior to the
Health Insurance Corporation. This bias would result in an under- mass dengue vaccination, we found that nearly 90% were seropos-
estimation of the protective effect of the vaccine. While not itive for dengue.[22] Thus, it is not unexpected to find that CYD-
entirely absent, this bias was likely less important in case detection TDV demonstrated a sizeable protective effect against more serious
among children with more serious illness (dengue with warning forms of dengue.
signs and severe dengue) whose condition more clearly required This study is continuing for five years after vaccination so to
hospital presentation and admission, regardless of vaccination sta- provide longer-term effectiveness data. In the meanwhile, public
tus. As a result, the validity of the estimated vaccine protective health programs in areas highly-endemic for dengue will need to
effect against dengue with warning signs and severe dengue is weigh the costs and benefits of deploying CYD-TDV. A reduction
probably less affected by health-seeking behavior, perhaps in hospitalizations by 26% (even after only one dose) for a disease
explaining the higher protection observed with these outcomes. of such high incidence has to be balanced against the risks to those
We compared the demographic characteristics between cases of who are seronegative at vaccination. Ideally, vaccinating only those
dengue without warning signs and cases of dengue with warning with prior dengue infection would be the preferred strategy but
signs and severe dengue (Supplementary Table 1) and found that depends on the development of sensitive and specific point-of-
the former were more likely to have a household head with more care tests [20]. Since 2018 the WHO now recommends that coun-
than six years of schooling (156/161 or 96 9% versus 302/329 or tries considering CYD-TDV vaccination as part of their dengue con-
91 8%; p = 0 03). This further supports the possibility of bias trol program should include pre-vaccination screening, so that only
between the groups; level of education may be associated with dengue-seropositive persons are vaccinated but that the limita-
likelihood of being vaccinated,[16] risk of infection[17,18] and like- tions of such screening should be clearly communicated to those
lihood of presenting/being admitted to hospital, particularly in offered vaccination.[23]
milder forms of dengue. As we are unable to determine if and to In summary, we found that one dose of the CYD-TDV vaccine
what extent the health seeking behavior of the milder dengue conferred protection against dengue with warning signs and severe
cases was biased by vaccination status and/or by socio-economic dengue. While our analysis did not find any statistically significant
level, the estimated vaccine effectiveness overall and against den- protective effect overall and against dengue of lesser severity,
gue without warning signs can only be considered with caution, potential biases preclude drawing any conclusion about this out-
while we can be more confident in the validity of our estimates come. We plan to continue the case-control study for five years
of effectiveness against dengue with warning signs and severe den- after the mass dengue vaccination campaign.
gue. The DOH continues to monitor the recipients of CYD-TDV,[19]
and the results of this study will contribute to documenting the
long-term safety profile of CYD-TDV. We also noted that controls 5. Role of the funding source
in our study were less frequently vaccinated than the overall
source population, and that controls enrolled in the second year The study sponsor is the University of the Philippines – Manila
of surveillance were less frequently vaccinated than controls and its staff developed the protocol, implemented the study, man-
enrolled in the first year (Table S1). These differences may signal aged and analyzed the data and wrote the manuscript. The Philip-
pine government paid for the vaccine and the DOH implemented
5324
M. Ylade, Kristal An Agrupis, Jedas Veronica Daag et al. Vaccine 39 (2021) 5318–5325

the dengue mass vaccination program. The University of the Philip- [8] Guidelines in the Community-Based Immunization with Tetravalent Dengue
Vaccine in Cebu Province including Cebu City, Mandaue City, and Lapu-Lapu
pines - Manila received funding from Sanofi Pasteur to undertake
City in Region VII. Philippines: Department of Health; 2017 May 16, 2017.
the assessment of vaccine effectiveness. Sanofi Pasteur was [9] Larson HJ. Politics and public trust shape vaccine risk perceptions. Nature
involved in discussions of the trial design and contributed to the Human Behaviour. 2018;2:318.
manuscript. [10] Region 7 Hospitals Manila, Philippines: Department of Health; [01 December
2020]. Available from: https://www.doh.gov.ph/sites/default/files/basic-page/
Region%207%20Hospitals.pdf.
Declaration of Competing Interest [11] Sasmono RT, Aryati A, Wardhani P, Yohan B, Trimarsanto H, Fahri S, et al.
Performance of Simplexa dengue molecular assay compared to conventional
and SYBR green RT-PCR for detection of dengue infection in Indonesia. PLoS
The authors declare that they have no known competing finan- One. 2014;9(8):e103815.
cial interests or personal relationships that could have appeared [12] WHO/TDR. Dengue: Guidelines for Diagnosis, Treatment, Prevention and
to influence the work reported in this paper. Control: New Edition. Dengue: Guidelines for Diagnosis, Treatment,
Prevention and Control: New Edition. WHO Guidelines Approved by the
Guidelines Review Committee. Geneva2009.
Acknowledgements [13] Honório Nildimar Alves, Silva Wellington da Costa, Leite Paulo José, Gonçalves
Jaylei Monteiro, Lounibos Leon Philip, Lourenço-de-Oliveira Ricardo. Dispersal
of Aedes aegypti and Aedes albopictus (Diptera: Culicidae) in an urban
We thank the children and their parents, the study field staff endemic dengue area in the State of Rio de Janeiro. Brazil. Mem Inst Oswaldo
and the health workers. We acknowledge Dr Leon Ochiai (Sanofi Cruz. 2003;98(2):191–8.
Pasteur) for his contribution to the study design. We are grateful [14] Muir LE, Kay BH. Aedes aegypti survival and dispersal estimated by mark-
release-recapture in northern Australia. American Journal of Tropical Medicine
to the members of the Research Advisory Board, Dr In-Kyu Yoon
and Hygiene 1998;58(3):277–82.
(Coalition for Epidemic Preparedness Innovations, Washington, [15] Aldstadt J, Yoon IK, Tannitisupawong D, Jarman RG, Thomas SJ, Gibbons RV,
USA) and Dr Lorenz von Seidlein (Mahidol-Oxford Tropical Medi- et al. Space-time analysis of hospitalised dengue patients in rural Thailand
cine Research Unit, Bangkok, Thailand), for their valuable input reveals important temporal intervals in the pattern of dengue virus
transmission. Trop Med Int Health. 2012;17(9):1076-85.
and guidance. [16] Hajizadeh M. Socioeconomic inequalities in child vaccination in low/middle-
income countries: what accounts for the differences? J Epidemiol Community
Health. 2018;72(8):719-25.
Appendix A. Supplementary data [17] Kikuti M, Cunha GM, Paploski IA, Kasper AM, Silva MM, Tavares AS, et al.
Spatial Distribution of Dengue in a Brazilian Urban Slum Setting: Role of
Supplementary data to this article can be found online at Socioeconomic Gradient in Disease Risk. PLoS Negl Trop Dis. 2015;9(7):
e0003937.
https://doi.org/10.1016/j.vaccine.2021.07.042.
[18] Wijayanti SP, Porphyre T, Chase-Topping M, Rainey SM, McFarlane M,
Schnettler E, et al. The Importance of Socio-Economic Versus Environmental
References Risk Factors for Reported Dengue Cases in Java, Indonesia. PLoS Negl Trop Dis.
2016;10(9):e0004964.
[1] Hadinegoro SR, Arredondo-García JL, Capeding MR, Deseda C, [19] Wijayanti SP, Porphyre T, Chase-Topping M, Rainey SM, McFarlane M,
Chotpitayasunondh T, Dietze R, et al. Efficacy and Long-Term Safety of a Schnettler E, et al. The Importance of Socio-Economic Versus Environmental
Dengue Vaccine in Regions of Endemic Disease. New England Journal of Risk Factors for Reported Dengue Cases in Java, Indonesia. (1935-2735
Medicine 2015;373(13):1195–206. (Electronic)).
[2] Meeting of the Strategic Advisory Group of Experts on immunization, April [20] Wilder-Smith Annelies, Flasche Stefan, Smith Peter G. Vaccine-attributable
2016 - conclusions and recommendations. Releve epidemiologique severe dengue in the Philippines. Lancet (London, England). 2019;394
hebdomadaire. 2016;91(21):266-84. (10215):2151–2.
[3] Nascimento EJM, George JK, Velasco M, Bonaparte MI, Zheng L, DiazGranados [21] Flasche Stefan, Wilder-Smith Annelies, Hombach Joachim, Smith Peter G.
CA, et al. Development of an anti-dengue NS1 IgG ELISA to evaluate exposure Estimating the proportion of vaccine-induced hospitalized dengue cases
to dengue virus. Journal of Virological Methods 2018;257:48–57. among Dengvaxia vaccinees in the Philippines. Wellcome Open Res.
[4] Sridhar S, Luedtke A, Langevin E, Zhu M, Bonaparte M, Machabert T, et al. Effect 2019;4:165. https://doi.org/10.12688/wellcomeopenres10.12688/
of Dengue Serostatus on Dengue Vaccine Safety and Efficacy. The New England wellcomeopenres.15507.1.
journal of medicine. 2018;379(4):327–40. [22] Lopez Anna Lena, Adams Cameron, Ylade Michelle, Jadi Ramesh, Daag Jedas
[5] Verani JR, Baqui AH, Broome CV, Cherian T, Cohen C, Farrar JL, et al. Case- Veronica, Molloy Caitlyn T, et al. Determining dengue virus serostatus by
control vaccine effectiveness studies: Preparation, design, and enrollment of indirect IgG ELISA compared with focus reduction neutralisation test in
cases and controls. Vaccine. 2017;35(25):3295–302. children in Cebu, Philippines: a prospective population-based study. The
[6] Bravo L, Roque VG, Brett J, Dizon R, L’Azou M. Epidemiology of dengue disease Lancet Global Health. 2021;9(1):e44–51.
in the Philippines (2000-2011): a systematic literature review. PLoS Negl Trop [23] WHO. Meeting of the Strategic Advisory Group of Expertys on Immunization,
Dis. 2014;8(11):e3027. April 2018 - conclusions and recommendations. Wkly Epidemiol Rec. 2018
[7] Dayrit MM LL, Picazo OF, Pons MC, Villaverde MC. The Philippines Health (23):329-44.
System Review. . New Delhi: World Health Organization, Regional Office for
South-East Asia. 2018;8 No. 2.

5325