04 Dengue Kaul 508

CDC ACIP — Vaccine Advisory Committee

Acip

Slides

30

Document text

Summary of two economic
models for dengue vaccine
TAK-003 use in Puerto Rico
Advisory Committee on Immunization Practices 
 
June 22, 2023 
The models summarized are currently undergoing the CDC economic review following the ACIP 
Guidance for Health Economics Studies, so results should be considered as preliminary.RajReni Kaul, PhD
CDC/NCIRD/ISD
The findings and conclusions in this presentation are those of the author(s) and do not 
necessarily represent the views of the Centers for Disease Control and Prevention.
Acknowledgements
•This presentation summarizes work conducted by two modeling 
teams
•Notre Dame team contracted by CDC (ND/CDC Model)
•Guido España , Manar Alkuzweny , Alex Perkins
•Takeda team (Takeda Model)
•J. Shen, R. Hanley, I. Zerda , et al. 
•CDC and ACIP contributors and reviewers
•Dengue ACIP workgroup
•Economists at CDC and colleagues (NCIRD/ISD)
Conflicts of Interest Statements
•RajReni Kaul: None.
•Notre Dame team:
•Dr. Guido España and Dr. Alex Perkins have previously received research 
funding from GlaxoSmithKline to support unrelated research on dengue 
vaccine development.
•Dr. Alex Perkins currently receives research funding and consulting fees
from Emergent Biosciences to support unrelated research on
chikungunya vaccine development.
•Takeda team:
•Takeda is the developer and manufacturer of the TAK -003 vaccine.
•Directly employed by Takeda or consultants employed by Putnam PHMR and 
contracted by Takeda
Terminology
Abbreviation Full term/Meaning
ND/CDC Notre Dame/CDC model
VE Vaccine efficacy
DENV (e.g., DENV -3) Dengue virus (e.g., serotype 3 dengue virus)
PICO Policy question articulated as Population, Intervention, Comparison, Outcomes
Case Medically -attended case
Hosp Hospitalization
Additional hospitalizations Hospitalization induced by vaccine -enhanced disease
NNV Number needed to vaccinate to avert an outcome (e.g., NNV hospitalization)
QALY Quality -adjusted life -years
All values rounded to 3 significant figures. 
Outline
•Recap of PICO questions
•Overview of models
•Comparison of results
•Exploring differences in assumptions
•QALY values used, vaccine efficacy
•Summary of Takeda model results 
•Base case
•Scenario analysis
•Model comparison summary and limitations
•Application to PICO questions
PICO Questions
•Should two doses of TAK -003 be administered routinely to seropositive
persons aged 4 –16 years living in dengue -endemic areas?
•Should two doses of TAK -003 be administered routinely to seronegative
persons aged 4 –16 years living in dengue -endemic areas?
•Should two doses of TAK -003 be administered routinely to seropositive
persons aged 17 –60years living in dengue -endemic areas?
•Should two doses of TAK -003 be administered routinely to seronegative
persons aged 17 –60years living in dengue -endemic areas?
General Model Design
Assumption/Model Characteristic Notre Dame/CDC Takeda
Model type Stochastic individual -based model Deterministic compartmental model
Prevaccination screening Included Not included
Vaccine implementation Age range, varying coverage rate over 
timeSingle age, a catch -up routine in year 
1possible, varying coverage rate over 
first four years then constant 
thereafter
Serotype specific VE point 
estimatePoint estimate estimated using multi -
level Bayesian modelPoint estimate estimated using 
traditional methods from clinical trial*
Serotype specific VE ranges Each simulation used VE inputs 
sampled from the confidence interval 
around the point estimateNo range, only point estimate used
Geographic area San Juan municipality (N=280,000) Puerto Rico (N= 3,256,028 )
DENV caused deaths Not included in QALYs Included in QALYs
Assumption/Model Characteristic Notre Dame/CDC Takeda
Model type Stochastic individual -based model Deterministic compartmental model
Prevaccination screening Included Not included
Vaccine implementation Age range, varying coverage rate over 
timeSingle age, a catch -up routine in year 
1possible, varying coverage rate over 
first four years then constant thereafter
Serotype specific VE point 
estimatePoint estimate estimated using multi -
level Bayesian modelPoint estimate estimated using 
traditional methods from clinical trial*
Serotype specific VE ranges Each simulation used VE inputs 
sampled from the confidence interval 
around the point estimateNo range, only point estimate used
* When the clinical trial estimate was negative or confidence interval included negative values, a VE model input of zero was assumed.Geographic area San Juan municipality (N=280,000) Puerto Rico (N= 3,256,028 )
DENV caused deaths Not included in QALYs Included in QALYsGeneral Model Design
Assumption/Model Characteristic Notre Dame/CDC Takeda
Model type Stochastic individual -based model Deterministic compartmental model
Prevaccination screening Included Not included
Vaccine implementation Age range, varying coverage rate over 
timeSingle age, a catch -up routine in year 
1possible, varying coverage rate over 
first four years then constant thereafter
Serotype specific VE point 
estimatePoint estimate estimated using multi -
level Bayesian modelPoint estimate estimated using 
traditional methods from clinical trial*
Serotype specific VE ranges Each simulation used VE inputs 
sampled from the confidence interval 
around the point estimateNo range, only point estimate used
Geographic area San Juan municipality (N=280,000) Puerto Rico (N= 3,256,028 )General Model Design
* When the clinical trial estimate was negative or confidence interval included negative values, a VE model input of zero was assumed.DENV caused deaths Not included in QALYs Included in QALYs
General Model Design
Assumption/Model Characteristic Notre Dame/CDC Takeda
Model type Stochastic individual -based model Deterministic compartmental model
Prevaccination screening Included Not included
Vaccine implementation Age range, varying coverage rate over 
timeSingle age, a catch -up routine in year 
1possible, varying coverage rate over 
first four years then constant thereafter
Serotype specific VE point 
estimatePoint estimate estimated using multi -
level Bayesian modelPoint estimate estimated using 
traditional methods from clinical trial*
Serotype specific VE ranges Each simulation used VE inputs 
sampled from the confidence interval 
around the point estimateNo range, only point estimate used
Geographic area San Juan municipality (N=280,000) Puerto Rico (N= 3,256,028 )
DENV caused deaths Not included in QALYs Included in QALYs
* When the clinical trial estimate was negative or confidence interval included negative values, a VE model input of zero was assumed.
Economic Model Preliminary Results by PICO
Age 4-16 years
ModelScenario conditionsNumber 
vaccinatedNet number averted with vaccination §
$/QALY 
(ICER)§ Geographic 
Area† AgePrevaccination 
screening Cases Hospitalizations Deaths
ND/CDCSan Juan 
Municipality4-16 Yes 11,700 485 (1.2%) 182 (2.6%) 1 (2.9%) 182,000 *
ND/CDCSan Juan 
Municipality4-16 No 30,300 1,070 (2.5%) 192 (2.8%) 1 (2.9%) 255,000 *
Takeda Puerto Rico8, 
catch -
up 9 -16 No 157,000 46,700 (12%) 9,150 (14 %) 5 (13%) Cost -
saving
†Modeled population size is 280,000 for the San Juan Municipality (ND/CDC model) and 3,256,028 for all of Puerto Rico (Takeda model).
§When compared to no vaccination
*Death is not incorporated into QALYs gained in the ND/CDC model base case assumptions. If QALY gains from averted deaths wer e 
included, then the ICERs would change from $182,000 to $65,000 per QALY for the scenario with prevaccination screening and would 
change from $255,000 to $137,000 per QALY for the scenario without prevaccination screening.
Economic Model Preliminary Results by PICO
Age 4-16 years
ModelScenario conditionsNumber 
vaccinatedNet number averted with vaccination §
$/QALY 
(ICER)§ Geographic 
Area† AgePrevaccination 
screening Cases Hospitalizations Deaths
ND/CDCSan Juan 
Municipality4-16 Yes 11,700 485 (1.2%) 182 (2.6%) 1 (2.9%) 182,000 *
ND/CDCSan Juan 
Municipality4-16 No 30,300 1,070 (2.5%) 192 (2.8%) 1 (2.9%) 255,000 *
Takeda Puerto Rico8, 
catch -
up 9 -16 No 157,000 46,700 (12%) 9,150 (14 %) 5 (13%) Cost -
saving
†Modeled population size is 280,000 for the San Juan Municipality (ND/CDC model) and 3,256,028 for all of Puerto Rico (Takeda model).
§When compared to no vaccination
*Death is not incorporated into QALYs gained in the ND/CDC model base case assumptions. If QALY gains from averted deaths wer e 
included, then the ICERs would change from $182,000 to $65,000 per QALY for the scenario with prevaccination screening, and would 
change from $255,000 to $137,000 per QALY for the scenario without prevaccination screening.
ModelScenario conditionsNumber 
vaccinatedNet number averted with vaccination§
$/QALY 
(ICER)§ Geographic 
Area† AgePrevaccination 
screening Cases Hospitalizations Deaths
ND/CDCSan Juan 
Municipality17-60 Yes 105,000 2,710 (6%) 724 (10%) 5(11%) 397,000 *
ND/CDCSan Juan 
Municipality17-60 No 121,000 3,360 (8%) 928 (13%) 4(14%) 315,000 *
Takeda Puerto Rico17, 
catch -
up 18 -
60No 449,000 67,000 (17%) 13,100 (21%) 8 (21%)Cost -
saving
†Modeled population size is 280,000 for the San Juan Municipality (ND/CDC model) and 3,256,028 for all of Puerto Rico (Takeda model).
§When compared to no vaccination
*Death is not incorporated into QALYs gained in the ND/CDC model base case assumptions. If QALY gains from averted deaths 
were included, then the ICERs would change from $397,000 to $188,000 per QALY for the scenario with prevaccination screening, 
and would change from $315,000 to $153,000 per QALY for the scenario without prevaccination screening.Economic Model Preliminary Results by PICO
Age 17-60 years
ModelScenario conditionsNumber 
vaccinatedNet number averted with vaccination§
$/QALY 
(ICER)§ Geographic 
Area† AgePrevaccination 
screening Cases Hospitalizations Deaths
ND/CDCSan Juan 
Municipality17-60 Yes 105,000 2,710 (6%) 724 (10%) 5(11%) 397,000 *
ND/CDCSan Juan 
Municipality17-60 No 121,000 3,360 (8%) 928 (13%) 4(14%) 315,000 *
Takeda Puerto Rico17, 
catch -
up 18 -
60No 449,000 67,000 (17%) 13,100 (21%) 8 (21%)Cost -
saving
†Modeled population size is 280,000 for the San Juan Municipality (ND/CDC model) and 3,256,028 for all of Puerto Rico (Takeda model).
§When compared to no vaccination
*Death is not incorporated into QALYs gained in the ND/CDC model base case assumptions. If QALY gains from averted deaths 
were included, then the ICERs would change from $397,000 to $188,000 per QALY for the scenario with prevaccination screening, 
and would change from $315,000 to $153,000 per QALY for the scenario without prevaccination screening.Economic Model Preliminary Results by PICO
Age 17-60 years
Exploring Difference
Model Inputs : QALY loss from dengue episodes
Disease OutcomeQuality -adjusted 
days lost
ND/CDC Takeda
Non -hospitalized dengue 11.2§5.5
Hospitalized dengue 12.8§6.8
Persistent dengue after non -hospitalized or 
hospitalized caseNA 4.9*
Dengue caused deaths NAAge-
dependent
§QALY lost for dengue episodes includes persistent dengue in 34% of all cases.
* Model assumes 34% of dengue episodes aged 30 or older develop persistent symptoms.
Disease OutcomeQuality -adjusted 
days lost
ND/CDC Takeda
Non -hospitalized dengue 11.2§6.1†
Hospitalized dengue 12.8§7.2†
Persistent dengue after non -hospitalized or 
hospitalized caseNA -
Dengue caused deaths NAAge-
dependentExploring Difference
Model Inputs : QALY loss from dengue episodes
When persistent 
dengue QALY loss is 
combined with 
acute phase QALY 
loss
§QALY lost for dengue episodes includes persistent dengue in 34% of all cases.
† Calculated as if 34% of cases ages 30 or older develop persistent symptoms. 
Exploring Difference
Model Inputs : Vaccine Efficacy 
68.1%
(27.2, 86.2)
Takeda 75.8% 98.5% 0%# 0%§ 75.8% 98.5% 71.4% 100%-4
Non -
hospitalized 
Symptomatic 
caseND/CDC*29.7% 
(6.7, 48.5)96.3% 
(87.1, 100)30.0%
(7.8, 49.4)-1.1%
(-73.5,64.4)63.1% 
(52.7, 82.2)85.2% 
(75.6, 94.4)42.5% 
(26.4, 55.8)60.2%Vaccine efficacy inputs by serostatus and by dengue serotype
Seronegative at vaccination Seropositive at vaccination
DENV -1 DENV -2 DENV -3 DENV -4 DENV -1 DENV -2 DENV -3 DENV
(19.8, 89.6)
Takeda† 30.8 % 69.8 % 0%#0%#46.7 % 69.8 % 42.6 % 61.2 %
Hospitalized 
caseND/CDC*84.6% 
(54.5, 98.7)99.0% Model
(95.6, 100)-30.3% 
(-91.6, 25.6)39.2% 
(-20.4, 81.0)54.1% 
(13.3, 82.2)99.5% 
(97.8, 100)70.2% 
(45.6, 90.4)Disease 
Outcome
29.7% 
(6.7, 48.5)96.3% 
(87.1, 100)30.0%
(7.8, 49.4)-1.1%
(-73.5,64.4)63.1% 
(52.7, 82.2)85.2% 
(75.6, 94.4)42.5% 
(26.4, 55.8)60.2%
(19.8, 89.6)
Takeda† 30.8 % 69.8 % 0%#0%#46.7 % 69.8 % 42.6 % 61.2 %
Hospitalized 
caseND/CDC*84.6% 
(54.5, 98.7)99.0% 
(95.6, 100)-30.3% 
(-91.6, 25.6)39.2% 
(-20.4, 81.0)54.1% 
(13.3, 82.2)99.5% 
(97.8, 100)70.2% 
(45.6, 90.4)68.1%
(27.2, 86.2)
Takeda 75.8% 98.5% 0%# 0%§ 75.8% 98.5% 71.4% 100%Exploring Difference
Model Inputs : Vaccine Efficacy
* Point estimate with 95% confidence interval
† Non -hospitalized case
§Due to lack of data* Point estimate with 95% confidence interval
† Average VE over 5 years post -vaccination. Includes vaccine waning.
§Due to lack of data
#Value used because pivotal clinical trial estimate or confidence interval included negative valuesDisease 
OutcomeModelVaccine efficacy inputs by serostatus and by dengue serotype
Seronegative at vaccination Seropositive at vaccination
DENV -1 DENV -2 DENV -3 DENV -4 DENV -1 DENV -2 DENV -3 DENV -4
Non -
hospitalized 
Symptomatic 
caseND/CDC*
Disease 
OutcomeModelVaccine efficacy inputs by serostatus and by dengue serotype
Seronegative at vaccination Seropositive at vaccination
DENV -1 DENV -2 DENV -3 DENV -4 DENV -1 DENV -2 DENV -3 DENV -4
Non -
hospitalized 
Symptomatic 
caseND/CDC*29.7% 
(6.7, 48.5)96.3% 
(87.1, 100)30.0%
(7.8, 49.4)-1.1%
(-73.5,64.4)63.1% 
(52.7, 82.2)85.2% 
(75.6, 94.4)42.5% 
(26.4, 55.8)60.2%
(19.8, 89.6)
Takeda† 30.8 % 69.8 % 0%#0%#46.7 % 69.8 % 42.6 % 61.2 %
Hospitalized 
caseND/CDC*84.6% 
(54.5, 98.7)99.0% 
(95.6, 100)-30.3% 
(-91.6, 25.6)39.2% 
(-20.4, 81.0)54.1% 
(13.3, 82.2)99.5% 
(97.8, 100)70.2% 
(45.6, 90.4)68.1%
(27.2, 86.2)
Takeda 75.8% 98.5% 0%# 0%§ 75.8% 98.5% 71.4% 100%
* Point estimate with 95% confidence interval
† Average VE over 5 years post -vaccination. Includes vaccine waning. 
§Due to lack of data
#Value used because pivotal clinical trial estimate or confidence interval included negative valuesExploring Difference
Model Inputs : Vaccine Efficacy
Disease 
OutcomeModelVaccine efficacy inputs by serostatus and by dengue serotype
Seronegative at vaccination Seropositive at vaccination
DENV -1 DENV -2 DENV -3 DENV -4 DENV -1 DENV -2 DENV -3 DENV -4
Non -
hospitalized 
symptomatic 
caseND/CDC*29.7% 
(6.7, 48.5)96.3% 
(87.1, 100)30.0%
(7.8, 49.4)-1.1%
(-73.5,64.4)63.1% 
(52.7, 82.2)85.2% 
(75.6, 94.4)42.5% 
(26.4, 55.8)60.2%
(19.8, 89.6)
Takeda† 30.8 % 69.8 % 0%#0%#46.7 % 69.8 % 42.6 % 61.2 %
Hospitalized 
caseND/CDC*84.6% 
(54.5, 98.7)99.0% 
(95.6, 100)-30.3% 
(-91.6, 25.6)39.2% 
(-20.4, 81.0)54.1% 
(13.3, 82.2)99.5% 
(97.8, 100)70.2% 
(45.6, 90.4)68.1%
(27.2, 86.2)
Takeda 75.8% 98.5% 0%# 0%§ 75.8% 98.5% 71.4% 100%Exploring Difference
Model Inputs : Vaccine Efficacy 
* Point estimate with 95% confidence interval
† Average VE over 5 years post -vaccination. Includes vaccine waning. 
§Due to lack of data
#Value used because pivotal clinical trial estimate or confidence interval included negative values
Disease 
OutcomeModelVaccine efficacy inputs by serostatus and by dengue serotype
Seronegative at vaccination Seropositive at vaccination
DENV -1 DENV -2 DENV -3 DENV -4 DENV -1 DENV -2 DENV -3 DENV -4
Non -
hospitalized 
symptomatic 
caseND/CDC*29.7% 
(6.7, 48.5)96.3% 
(87.1, 100)30.0%
(7.8, 49.4)-1.1%
(-73.5,64.4)63.1% 
(52.7, 82.2)85.2% 
(75.6, 94.4)42.5% 
(26.4, 55.8)60.2%
(19.8, 89.6)
Takeda† 30.8 % 69.8 % 0%#0%#46.7 % 69.8 % 42.6 % 61.2 %
Hospitalized 
caseND/CDC*84.6% 
(54.5, 98.7)99.0% 
(95.6, 100)-30.3% 
(-91.6, 25.6)39.2% 
(-20.4, 81.0)54.1% 
(13.3, 82.2)99.5% 
(97.8, 100)70.2% 
(45.6, 90.4)68.1%
(27.2, 86.2)
Takeda 75.8% 98.5% 0%# 0%§ 75.8% 98.5% 71.4% 100%Exploring Difference
Model Inputs : Vaccine Efficacy 
* Point estimate with 95% confidence interval
† Average VE over 5 years post -vaccination. Includes vaccine waning. 
§Due to lack of data
#Value used because pivotal clinical trial estimate or confidence interval included negative values
Exploring Difference
Impact of vaccine efficacy inputs on outcome
•What if the ND/CDC model uses the Takeda VE inputs in the scenario 
where 4 –16-year -olds are vaccinated without screening?
•Takeda’s VE inputs did not substantially impact the results of the ND/CDC 
model. 
VE 
inputsNumber 
vaccinatedNet number averted with vaccination Vaccination enhanced 
disease NNV 
hosp.QALYs 
gained*$/QALY 
(ICER)*
Cases Hosp. DeathsAdditional 
casesAdditional 
hosp.
ND/CDC† 30,300 1,070
(651 -1,470) 192 
(112 -300)1 
(0.5 -1.5)45
(35-52)32 
(29-37)158 
(101 -270)35 255,000 
Takeda† 30,300 1,030 
(699 -1,300) 287 
(197 -425)1
(0.9 -2.1)NA NA 105 
(71-153)34 262,000 
*QALY loss due to dengue caused death is not incorporated into ICER calculation.
†Range of model results due to stochasticity
Outline
•Recap of PICO questions
•Overview of models
•Comparison of results
•Exploring differences in assumptions
•QALY values used, vaccine efficacy
•Summary of Takeda model results 
•Base case
•Scenario analysis
•Model comparison summary and limitations
•Application to PICO questions
Takeda Base Case Preliminary Results
Routine vaccination at age 8
Base case conditions
•Entire population of Puerto Rico
•20-year time horizon
•Routine vaccination at age 8
•Vaccination rate increases from 
0% at year 0 to 60% at year 4.
•Using slightly different QALY loss 
than previously presented*Base Case
Number vaccinated 449,000
Averted cases 123,000 (15%)
Averted hospitalizations 25,000 (20%)
Averted deaths 15 (20%)
NNV hospitalization 17.6
QALY sgained 2,070
$/QALY Cost -saving
$/hospitalization Cost -saving
*Values reported in supplemental slides
Scenario Analysis
Takeda Model: Routine vaccination at age 8
*Used a fixed vaccine coverage parameter of 21.5% to approximate vaccination 
coverage that increases from 0% to 43% over 10 years.Scenario
Number 
vaccinatedNumber averted with vaccination
NNV 
hospQALYs 
gainedICER 
($/QALY)Time 
(years)TAK-003 
routine 
ageTAK-003 
catchup 
ageTAK-003 
coverage*Cases Hosp Deaths
10 8 None 22% 87,000 24,000 (6%) 4,510 (7%) 3 (8%) 19.3 402Cost -
saving
10 8 9-16 22% 157,000 47,000 (12%) 9,150 (14%) 5 (13%) 17.1 838Cost -
saving
Overall Cost -effectiveness
Preliminary results
Incremental Cost (in thousands)
QALYs gained
Overall Cost-effectiveness
Preliminary results
                               
                
      
      
Incremental Cost (in thousands)
QALYs gained
4-16 
w/prevaccination
screening17-60 w/o 
prevaccination
screeningR8 
Catch-up 9- 16R17 
Catch-up 18- 60
Limitations
•Parameter Uncertainty
•Large confidence interval (>100%) for DENV -3 and DENV -4 VE estimates for 
those seronegative at vaccination.
•Given vaccine’s range in serotype specific efficacy, actual outcome will 
be heavily influenced by the dominant circulating serotype.
•In the ND/CDC model, deaths were not part of the QALY calculation.
•Using Takeda’s model, we cannot assess benefits and risks of pre -
vaccination screening strategies vs strategies without pre -vaccination 
screening.The models summarized are currently undergoing the CDC economic review following the ACIP 
Guidance for Health Economics Studies, so results should be considered as preliminary.
Only seropositives 
(prescreening)All individuals 
(no prescreening)
4–16 years
Children/AdolescentsAnswered by ND/CDC ModelAnswered by ND/CDC Model
& Takeda Model
17–60 years
AdultsAnswered by ND/CDC ModelAnswered by ND/CDC Model
& Takeda ModelModels answering the PICO Questions
Acknowledgements
•This presentation summarized work conducted by two modeling 
teams
•Notre Dame team contracted by CDC (ND/CDC Model)
•Guido España , Manar Alkuzweny , Alex Perkins
•Takeda team (Takeda Model)
•J. Shen, R. Hanley, I. Zerda *, Z.  Janusz *, E. Kharitonova*, A. Rosas, A. Garcia, M. Curtis, 
M. Lu, C. Ruff, V. Tricou , M. Sharma, R. Kastner, E. Lloyd, G. Perez, S. Biswal, T. Tsai, I. 
Willits*
•Name with (*) are employed by Putnam PHMR and contracted by Takeda. All others are 
directly employed by Takeda. 
•CDC and ACIP contributors and reviewers
•Dengue ACIP workgroup
•Economists at CDC and colleagues (NCIRD/ISD)
The findings and conclusions in this presentation are those of the author(s) and do not 
necessarily represent the views of the Centers for Disease Control and Prevention.