MAJOR ARTICLE
Hemagglu ina ion Inhibi ion An ibody Ti e s
as a Co ela e o P o ec ion Agains Seasonal
A/H3N2 Influenza Disease
Anne Benoi ,1,a Ji i Be an,2Jeanne-Ma ie De as e ,3Me al Esen,4Odile Launay,5Gee Le oux-Roels,6
Jane E. McElhaney,7Lidia Oos ogels,3Ge i A. an Essen,8Manjusha Gaglani,9Lisa A. Jackson,10 Timo Vesika i,11
Ca he ine Leg and,1Fabian Tibaldi,3B uce L. Innis,12 and Wal hè e Dewé3
1
Ins i u de S a is ique, Bios a is ique e Sciences Ac ua ielles, Uni e si é Ca holique de Lou ain, Lou ain-la-Neu e, Belgium;
2
Vaccina ion and T a el
Medicine Cen e, Poliklinika II, H adec K álo é, Czech Republic;
3
GSK Vaccines, Rixensa , Belgium;
4
Ins i u ü T openmedizin, Uni e si ä sklinikum
Tübingen, Ge many;
5
Uni e si é Pa is Desca es, So bonne Pa is Ci é; Assis ance Publique Hôpi aux de Pa is, Hôpi al Cochin, CIC Cochin-Pas eu ; Inse m,
CIC 1417-REIVAC, Pa is, F ance;
6
Cen e o Vaccinology, Ghen Uni e si y and Hospi al, Belgium;
7
Alan M. McGa in Chai in Ge ia ics Resea ch,
Depa men o Medicine, Uni e si y o B i ish Columbia, Vancou e , Canada;
8
Julius Cen e o Heal h Sciences and P ima y Ca e, Uni e si y Medical
Cen e U ech , The Ne he lands;
9
Sec ion o Pedia ic In ec ious Diseases, Baylo Sco and Whi e Heal h, Texas A&M Heal h Science Cen e College o
Medicine, Temple;
10
G oup Heal h Resea ch Ins i u e, Sea le, Washing on;
11
Vaccine Resea ch Cen e , Uni e si y o Tampe e, Finland; and
12
GSK
Vaccines, King o P ussia, Pennsyl ania
Backg ound.To in es iga e he ela ionship be ween hemagglu inin-inhibi ion (HI) an ibody le els o he isk
o influenza disease, we conduc ed a co ela e o p o ec ion analysis using pooled da a om p e iously published
andomized ials.
Me hods.Da a on he occu ence o labo a o y-confi med influenza and HI le els p e- and pos accina ion we e
analyzed om 4 da ase s: 3 da ase s included subjec s aged <65 yea s who ecei ed inac i a ed i alen influenza
accine (TIV) o placebo, and 1 da ase included subjec s aged ≥65 yea s who ecei ed AS03-adju an ed TIV (AS03-
TIV) o TIV. A logis ic model was used o e alua e he ela ionship be ween he pos accina ion i e o A/H3N2 HI
an ibodies and occu ence o A/H3N2 disease. We hen buil a ecei e -ope a ing cha ac e is ic cu e o iden i y a
po en ial cu o i e be ween p o ec ion and no p o ec ion.
Resul s.The baseline odds a io o A/H3N2 disease was highe o subjec s aged ≥65 yea s han <65 yea s and
highe in seasons o s ong epidemic in ensi y han mode a e o low in ensi y. Including age and epidemic in ensi y
as co a ia es, a 4- old inc ease in i e was associa ed wi h a 2- old dec ease in he isk o A/H3N2 disease.
Conclusions.The modeling exe cise confi med a ela ionship be ween A/H3N2 disease and HI esponses, bu
i did no allow an e alua ion o he p edic i e powe o he HI esponse.
Keywo ds.A/H3N2; influenza; modeling; se ologic co ela es; accine.
Inac i a ed and ecombinan p o ein influenza accines
a e licensed based in pa on immunogenici y da a
because egula o y au ho i ies assume pos accina ion
hemagglu ina ion-inhibi ion (HI) an ibody i e s abo e
adefined h eshold a e p edic i e o clinical benefi .
An HI i e h eshold o 1:40 is gene ally ecognized as
co esponding o a 50% educ ion in he isk o influenza,
and his is based on a challenge s udy in adul s conduc ed
by Hobson e al [1] in 1972. Howe e , in he li e a u e,
he e is no consensus on he defini ion o “p o ec ion”,
wi h some s udies defining p o ec ion as a p edefined
isk educ ion (usually 50%) and o he s udies defining
p o ec ion as he i e le els ha p o ide he bes sepa a-
ion be ween influenza cases and noncases [2–5].
To iden i y and alida e any immunological h esh-
old as a co ela e o p o ec ion, i would be desi able
Recei ed 19 No embe 2014; accep ed 11 May 2015.
a
P esen A filia ion: GSK Vaccines, Rixensa , Belgium.
Co espondence: Wal hè e Dewé, MSc, GSK Vaccines, Rue de l′Ins i u 89,
B-1330 Rixensa , Belgium (wal he e.p.de[email p o ec ed]).
Open Fo um In ec ious Diseases
© The Au ho 2015. Published by Ox o d Uni e si y P ess on behal o he In ec ious
DiseasesSocie yo Ame ica.ThisisanOpenAccessa icledis ibu edunde he e ms
o he C ea i e Commons A ibu ion-NonComme cial-NoDe i s licence (h p://
c ea i ecommons.o g/licenses/by-nc-nd/4.0/), which pe mi s non-comme cial
ep oduc ion and dis ibu ion o he wo k, in any medium, p o ided he o iginal wo k
is no al e ed o ans o med in any way, and ha he wo k is p ope ly ci ed. Fo
comme cial e-use, please con ac [email p o ec ed].
DOI: 10.1093/ofid/o 067
Influenza Vaccine and Se ologic Co ela es o P o ec ion •OFID •1
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://o id.ox o djou nals.o g/Downloaded om
o influenza e ficacy accine ials o be adequa ely and consis-
en ly designed o allow co ela e o p o ec ion (COP) analyses,
ei he wi hin a single ial o a e pooling o da a ac oss di -
e en ials. The findings o such analyses would inc ease he
clinical ele ance o subsequen s udies based on se ologic
endpoin s gene a ed by he same labo a o y and assay p o ocol.
The e o e, we pooled da a om se e al ials o assess he
ele ance o HI an ibody le els o p o ec ion.
In his s udy, we desc ibe a COP analysis o pooled da a om
andomized ials o seasonal influenza accines including 7730
subjec s [6–9]. In 1 ial, A/H3N2 was he mos common influ-
enza i us de ec ed o e all; he e o e, his COP analysis ocused
on A/H3N2 [9]. This mul i ial app oach suppo s he sea ch
o an immunological measu emen ha is p edic i e o accine
e ficacy ac oss a ious se ings.
METHODS
Ma e ials and Me hods
The analysis was based on 2 ials in subjec s aged 18–64 yea s, 1
ial in subjec s aged 18–49 yea s, and 1 s udy in subjec s aged
≥65 yea s. In each o hese e ficacy s udies, immunogenici y anal-
yses we e pe o med on pe -p o ocol immunogenici y subco-
ho s (including subjec s who me eligibili y c i e ia, complied
wi h he p o ocol, ecei ed any dose o ei he accine o placebo,
and o whom da a we e a ailable o a gi en endpoin ). An o e -
iew o he ials is as ollows. (1) Be an e al [6]pe o med a an-
domized, double-blind, placebo-con olled s udy o he e ficacy
o i alen influenza accine (TIV) agains cul u e-confi med in-
fluenza in heal hy adul s aged 18–64 yea s. A o al o 5103 and
2549 subjec s ecei ed TIV o placebo, espec i ely, du ing he
2006–2007 season in Czech Republic and Finland. The immuno-
genici y subcoho in ou analysis included 291 and 148 subjec s
in he TIV and placebo g oups, espec i ely (ClinicalT ials.go
NCT00363870). (2) Be an e al [7] pe o med a andomized,
double-blind, placebo-con olled s udy o he e ficacy o TIV
agains cul u e-confi med influenza in heal hy adul s aged
18–64 yea s. A o al o 4137 and 2066 subjec s ecei ed TIV o
placebo, espec i ely, du ing he 2005–2006 season in he Czech
Republic. The immunogenici y subcoho in ou analysis includ-
ed 632 and 315 subjec s in he TIV and placebo g oups, espec-
i ely (ClinicalT ials.go NCT00197223). (3) Jackson e al [8]
pe o med a andomized, double-blind, placebo-con olled e fi-
cacy s udy o TIV agains cul u e-confi med influenza in heal hy
adul s aged 18–49 yea s. In his s udy, a o al o 3783 and 3828
subjec s ecei ed TIV o placebo, espec i ely, du ing he 2005–
2006 and 2006–2007 seasons in he Uni ed S a es. The immuno-
genici y subcoho in ou analysis included 1298 and 216 subjec s
in he TIV and placebo g oups, espec i ely (ClinicalT ials.go
NCT00216242). (4) The Influence65 ial was a andomized,
obse e -blinded s udy o he ela i e e ficacy o AS03-TIV s TIV
agains polyme ase chain eac ion (PCR)-confi med influenza
in heal hy adul s aged ≥65 yea s. The s udy included 43 695
subjec s om 15 coun ies who ecei ed AS03-TIV o TIV du -
ing he 2008–2009 and 2009–2010 seasons. The immunogenic-
i y subse included 2422 and 2408 subjec s in he AS03-TIV and
TIV g oups, espec i ely, and his analysis included immunoge-
nici y da a om he 2008–2009 season (ClinicalT ials.go
NCT00753272) [9].
In he ials including adul s aged 18–64 yea s and 18–49 yea s,
eligible subjec s we e heal hy. In he Influence65 ial including
adul s aged ≥65 yea s, subjec s wi h como bidi ies we e eligible
o inclusion i hey we e ambula o y, hei heal h was s able,
and hey we e wi hou acu e illness a he ime o accina ion. In-
clusion and exclusion c i e ia and e hics s a emen s o he ials
in his analysis ha e been p e iously published [6–9].
All accines we e manu ac u ed by GSK Vaccines. In he
placebo-con olled s udies, subjec s we e andomized o ecei e
(1) 1 0.5 mL dose o TIV con aining 15 μg each o he 3 hemag-
glu inin an igens (HAs) ecommended by he Wo ld Heal h
O ganiza ion (WHO) o he gi en influenza season o (2) saline
placebo. In he Influence65 ial, subjec s we e andomized o e-
cei e one 0.7 mL dose o AS03-TIV o 0.5 mL TIV. De ails o he
accines used in each s udy ha e been p e iously desc ibed [6–9].
Case Defini ions and Labo a o y Me hods
The e we e di e ences among he ials, bu hey all ollowed
he same gene al p o ocol. Du ing he s udy pe iods, subjec s
we e moni o ed o influenza-like illness (ILI) by ac i e su eil-
lance ( elephone con ac /s udy cen e isi /home isi s by s udy
pe sonnel) and by passi e su eillance (whe eby subjec s no i-
fied he s udy cen e i hey expe ienced ILI symp oms). Case
defini ions in each s udy ha e been p e iously desc ibed [6–9].
Nasal and h oa swabs we e ob ained om subjec s epo ing
ILI o he labo a o y iden ifica ion o influenza i uses as p e i-
ously desc ibed [6–8].
In all s udies, se um samples we e aken be o e accina ion and
21 days a e accina ion o he assessmen o se um an ibodies
o each accine-homologous HA. All es ing was pe o med a a
GSK’s labo a o y (La al, Canada and D esden, Ge many) acco d-
ing o an es ablished me hod [10]. The high-sensi i i y HI assay
used 25 µL se um and 25 µL an igen om a solu ion concen a -
ed a 4 HA uni s/50 µL (ie, 2 HA uni s). The e y h ocy e (chicken)
concen a ion was 0.45%. Hemagglu ina ion-inhibi ion assay-
based an ibody esponses we e desc ibed as he an ilog o he
a i hme ic mean o he log
10
- ans o med i e s (geome ic
mean i e s [GMTs]). A i e o 1:5 was assigned o samples wi h
a alue below a cu o i e o 1:10.
S a is ical Analysis
The ollowing a iables we e conside ed in ou analyses: gende ,
age, seasonal influenza accina ion his o y wi hin 2 p e ious
yea s, A/H3N2 in ec ion s a us by he end o he s udy
season, p e- (Day 0) and pos accina ion (Day 21) HI i e s
2•OFID •Benoi e al
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://o id.ox o djou nals.o g/Downloaded om
agains A/H3N2, p e accina ion immuni y s a e (ie, A/H3N2
HI i e ≥1:40), A/H3N2 epidemic in ensi y (“s ong in ensi y”
o “low/mode a e in ensi y”), and A/H3N2 disease occu ence.
In Be an e al [6,7] and Jackson e al [8], he epidemic in ensi y
was based on he WHO influenza su eillance (FluNe ) da a-
base and by e alua ing he magni ude o he epidemics in he
co esponding coun ies a he ime he s udies we e conduc ed.
In he Influence65 ial, epidemic in ensi y was based on na-
ional su eillance da a and a ack a es in he s udy, as assessed
by he Adjudica ion S ee ing Commi ee o he influenza peak
season, including expe s in he field o influenza and influenza
accina ion who we e independen o he s udy in es iga o s
and he s udy sponso .
In he desc ip i e analysis, he dis ibu ion o a iables was
cha ac e ized. Fo con inuous a iables, he numbe o obse a-
ions, mean, s anda d de ia ion, and minimum and maximum
alues we e compu ed. Fo HI i e s, GMTs and hei coe ficien
o a ia ion we e also calcula ed a e a log
10
ans o ma ion.
F equency s a is ics, including coun s and p opo ions we e
ob ained o he ca ego ical a iables. A p elimina y g aphical
analysis was pe o med o assess each co a ia e e sus A/H3N2
a ack a es. The p opo ion o subjec s wi h labo a o y-
confi med A/H3N2 influenza was calcula ed o each dilu ion
ac o o he pos accina ion HI esponse agains A/H3N2.
A logis ic eg ession model was used o assess he e ec o
co a ia es on A/H3N2 disease occu ence. The ollowing co a -
ia es we e conside ed: p e accina ion immuni y s a e ( i e
≥1:40 defined as p o ec ed), Day 21 pos accina ion A/H3N2
log
10
i e s, gende , his o y o accina ion ( accina ion 1 and
2 yea s be o e s udy s a ), accine ecei ed, and age (Influ-
ence65 ial ≥65 yea s s o he ials <65 yea s). The epidemic
in ensi y was also included in he model as a po en ial e ec
modifie because he p o ec ion le el associa ed o a pa icula
HI i e could depend on he le el o exposu e [11]. A manual
s epwise a iable selec ion was pe o med based on he Baye-
sian in o ma ion c i e ion o selec he bes combina ion o co-
a ia es o desc ibe he disease occu ence. One o he objec i es
o ou analysis was o y o iden i y he HI i e ha bes sep-
a a es he subjec s who we e p o ec ed om hose who we e no .
A ecei e -ope a ing cha ac e is ic (ROC) cu e, p esen ing he
sensi i i y agains one minus he specifici y a a ious A/H3N2
HI i e cu o alues, was de i ed. Sensi i i y was defined as he
p opo ion o subjec s wi h a pos accina ion i e below he
cu o alue among hose wi h confi med A/H3N2 influenza;
specifici y was defined as he p opo ion o subjec s wi h a pos -
accina ion i e equal o o g ea e han he cu o alue among
hose wi hou confi med A/H3N2 influenza. The Youden index,
iden i ying he lowes i e a which he sum o he specifici y
and sensi i i y was maximum, was used o assess such a h esh-
old [12]. The ROC a ea unde he cu e quan ifies he o e all
abili y o he es o disc imina e be ween hose indi iduals
wi h A/H3N2 disease and hose wi hou disease; i ep esen s
he p obabili y ha a andomly selec ed influenza case will
ha e a lowe esul (es ima ion om he model aking in o
accoun all selec ed co a ia es) han a andomly selec ed subjec
wi hou disease. Because he noncases we e a mix u e o p o-
ec ed subjec s and no su ficien ly exposed o nonp o ec ed
subjec s, we also de i ed a cu o pos accina ion i e alue gi -
ing mo e weigh o he cases de ec ed, which we defined as he
HI i e cu o alues o he de ec ion o A/H3N2 influenza
wi h 90% sensi i i y.
RESULTS
Desc ip i e Analysis
An o e iew o subjec s included in he analysis is shown in
Figu e 1. The demog aphic cha ac e is ics and GMTs o subjec s
included in he analysis by ial a e shown in Table 1.P e-and
Figu e 1. O e iew o analysis popula ion (desc ip i e analysis). Abb e ia ion: TIV, inac i a ed i alen influenza accine.
Influenza Vaccine and Se ologic Co ela es o P o ec ion •OFID •3
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://o id.ox o djou nals.o g/Downloaded om
pos accina ion GMTs in subjec s who ecei ed TIV o AS03-
TIV we e 14.03–17.4 and 172.3–285.6, espec i ely, and in sub-
jec s who ecei ed placebo we e 14.12 and 14.14, espec i ely.
P e accina ion, 5405 (71.1%) subjec s had an an ibody i e
agains A/H3N2 ha was <1:40 (Supplemen a y Da a 1)and
2309 (29.9%) had a i e ha was ≥1:40. Six een subjec s aged
≥65 yea s did no ha e p e accina ion i e da a a ailable.
The A/H3N2 disease a es by age and epidemic in ensi y a e
shown in Table 2. The equency o A/H3N2 cases and pos ac-
cina ion HI i e s agains A/H3N2 a e shown in Figu e 2,
among which 1098 and 6632 subjec s, espec i ely, had pos ac-
cina ion i e s o <1:40 and ≥1:40. O 1098 o 7730 (14.2%) sub-
jec s wi h pos accina ion HI i e s o <1:40, 24 o 1098 (2.2%)
subjec s had confi med A/H3N2 disease; among 6632 o 7730
(85.8%) subjec s wi h pos accina ion i e s o ≥1:40, 50 o 6632
(0.75%) had confi med A/H3N2 disease.
Logis ic Reg ession Analysis
Table 3shows he pa ame e es ima es used in he selec ed mod-
els, which includes pos accina ion HI i e s, epidemic in ensi y,
and age as co a ia es. The odds a io o A/H3N2 disease in high
e sus mode a e o low epidemic in ensi y was 3.4 (95% confi-
dence in e al [CI], 2.1–5.6). The odds a io o A/H3N2 disease
in subjec s aged ≥65 yea s s <65 yea s was 3.5 (95% CI, 1.9–6.3).
Including pos accina ion HI i e and age as co a ia es, he odds
a io o A/H3N2 disease in subjec s aged ≥65 yea s s <65 yea s
was 4.2 (95% CI, 2.3–7.8).
In he ull model, including epidemic in ensi y and age ca ego-
y, a 4- old inc ease in pos accina ion HI i e was associa ed
wi h a 49.0% dec ease in he isk o in ec ion, and including
only age ca ego y as a co a ia e, a 4- old inc ease in i e was
also associa ed wi h a 49.4% dec ease in he isk o in ec ion. Con-
sis ency o he HI esponse ac oss he HI ange was an assump ion
o he s a is ical model, which appea ed accep able based on he
cases and HI i e s obse ed (Figu e 3).
The a ea unde he cu e o he ROC including age and sea-
son s eng h as co a ia es was es ima ed a 0.7735 (Supplemen-
a y Da a 2). The Youden index HI i e cu o s and he 90%
sensi i i y cu o alues a e shown in Table 4.
DISCUSSION
In ag eemen wi h p e ious epo s, we ound a co ela ion be-
ween HI i e s and he isk o A/H3N2 disease [1,13]. Includ-
ing age and epidemic in ensi y as co a ia es, we showed ha a
4- old inc ease in i e was associa ed wi h a 2- old dec ease in
he isk o A/H3N2 disease, wi h a simila di e ence in isk ob-
se ed when including only age as a co a ia e. The Youden
index cu o alues in a season o high epidemic in ensi y
we e 1:40 and 1:640 in subjec s aged <65 yea s and ≥65 yea s,
Table 1. Demog aphic Cha ac e is ics and A/H3N2 HI An ibody Ti e s P e accina ion (Day 0) and Pos accina ion (Day 21) (Desc ip i e
Analysis)
Adul s aged 18–64 yea s [6–8] Adul s aged ≥65 yea s [9]
TIV N = 2221 Placebo N = 679 AS03-TIV N = 2422 TIV N = 2408
Mean age (SD) ange, yea s 35.78 (12.2) 18–64 34.53 (11.3) 18–64 73.2 (6.0) 65–95 73.4 (6.3) 65–100
Vaccina ion his o y, n (%)
1 y 243 (10.9%) 81 (11.9%) 1647/2197 (75.0%) 1650/2199 (75.0%)
2 y 187 (8.4%) 32 (4.7%) 1569/2119 (74.0%) 1578/2127 (74.2%)
GMT, ( ange)
Day 0 14.03 (5–1810) 14.12 (5–640) 17.4 (5–1810) 17.4 (5–1280)
Day 21 178.61 (5–7240) 14.14 (5–905) 285.6 (5–20480) 172.3 (5–20480)
Abb e ia ions: AS03, ocophe ol-based oil-in-wa e Adju an Sys em; GMT, geome ic mean i e ; HI, hemagglu ina ion inhibi ion; n, numbe o subjec s ul illing
de ini ion; N, numbe o subjec s in g oup; SD, s anda d de ia ion; TIV, inac i a ed i alen in luenza accine.
Table 2. A/H3N2 In ec ion Ra es by Age and Epidemic In ensi y in Subjec s Pooled F om 4 Vaccine E ficacy T ials (Pe -p o ocol Immu-
nogenici y Subcoho s) (Desc ip i e Analysis)
Age Epidemic In ensi y Subjec s A/H3N2 Cases In ec ion Ra e (%)
≥65 y McElhaney e al [9] Low o mode a e 2939 20 0.68
High 1891 40 2.12
18–64 y Be an e al [6,7] Jackson e al [8] Low o mode a e 1873 4 0.21
High 1027 10 0.97
4•OFID •Benoi e al
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://o id.ox o djou nals.o g/Downloaded om
espec i ely. Al hough we ound ha age did no appea o
a ec he se ological esponse o accina ion based on GMTs,
olde subjec sseemed oha eag ea e isko in ec iona
simila i e scompa edwi hyounge subjec s.
Al hough age defined he popula ion s udied in each ial
included in he analysis, case de ec ion me hods and ci cula -
ing i uses may a ec he suscep ibili y o a popula ion o in-
fluenza disease. In addi ion, because hese ac o s di e ed
be ween he ials, he e ec o age on HI i e s as a COP
mus be in e p e ed wi h cau ion. Indeed, he e a e se e al
ac o s ela ed o he di e ences among he ials ha a e
con ounded in ou analysis: he na u e o he compa ison
(placebo o ac i e ea men , adju an ed o no ), influenza
case defini ions, labo a o y me hods o i al de ec ion (cul-
u e o PCR), HI measu emen s, and he age o pa icipan s
(18–64 yea s, 18–49 yea s, ≥65 yea s). In addi ion, cell-
media ed immuni y con ibu es o p o ec ion agains influen-
za, bu he eliabili y o he HI i e as an index o bo h
humo al and cellula immuni y is unknown and likely o di e
wi h ad ancing age. We belie e ha he mos impo an di e -
ences among he ials we e he di e ences in he popula ion
s udied eflec ed by age, which is a su oga e o unmeasu ed
Figu e 2. Numbe o subjec s in each i e ca ego y and numbe o A/H3N2 cases (A) and p opo ion o subjec s in each i e ca ego y wi h labo a o y-
confi med A/H3N2 in ec ion (B) (desc ip i e analysis).
Table 3. Es ima es o he Logis ic Reg ession Model Pa ame e s
Pa ame e Es ima e PValue Odds Ra io 95% Con idence In e al on he Odds Ra io
Baseline isk o he e e ence ca ego y
a
−3.922 <.0001
Pos accina ion log i e −1.1199 <.0001 0.3263 .2311, .4607
Epidemic in ensi y 1.2366 <.0001 3.4439 2.0995, 5.6491
Age as co a ia es
b
1.2388 .0001 3.4515 1.8863, 6.3154
a
Null pos accina ion i e , low mode a e epidemic in ensi y, and <65 yea s.
b
T ial (In luence65 ial [≥65 yea s] s o he ials [<65 yea s])/age (≥65 s <65).
Influenza Vaccine and Se ologic Co ela es o P o ec ion •OFID •5
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://o id.ox o djou nals.o g/Downloaded om

di e ences in immuni y based on pas exposu e, and di e -
ences in he defini ion o influenza disease.
Influenza occu ence depends upon he immuni y o a gi en
popula ion as well as hei exposu e o ci cula ing i uses.
Because di e en coun ies ha e di e en accina ion policies
ha influence he ansmission and exposu e o i uses (eg, ac-
cina ion o child en), we included epidemic in ensi y as a co a -
ia e based on su eillance in each coun y as an indica o o
exposu e. Howe e , exposu e may also change he le el o
an ibody needed o p e en illness o any se e i y [11]. This is
an impo an concep because in adul s, mos illnesses a e el-
a i ely mild, bu he occu ence o se e e illness esul ing in
hospi aliza ion and ad e se ou comes inc eases wi h ad ancing
age. In his pooled analysis, we did no ha e a sys ema ic p o-
spec i e classifica ion o mode a e o se e e illness, and because
he le el o an ibody co ela ing wi h p o ec ion agains mode -
a e o se e e illness may be less han ha needed o p o ec
agains mild illness, ou analysis may be con ounded. In addi-
ion, egional and seasonal a iabili y o ci cula ing i uses and
a ia ions in he se e i y o influenza illness a e di ficul o ac-
coun o in a COP analysis.
An objec i e o ou analysis was o y o iden i y he HI i e
ha bes sepa a es he 2 subjec g oups—p o ec ed and no p o-
ec ed. By conside ing a ROC app oach, selec ing a cu o poin
in ol es a ade-o be ween sensi i i y and specifici y. The You-
den index gi es he same weigh o bo h sensi i i y and specifici y
because i defines he cu o poin as he i e alue ha maximiz-
es he sum o he sensi i i y and specifici y [12]. The Youden
index me hod depends upon he sepa abili y o he p o ec ed
and nonp o ec ed popula ions. Howe e , he a e o in ec ion
among subjec s wi h low i e s may be s ongly associa ed wi h
he chance o exposu e and disease p e alence, which a y
among seasons and loca ions as well as social beha io . The e-
o e, he HI i e densi y cu es o subjec s who we e no in ec ed
a e a mix u e o subjec s who we e p o ec ed and hose who we e
possibly unp o ec ed bu also unexposed. The me hodology we
used elies on he belie ha alse nega i es a e likely o occu ,
and hus sensi i i y ( ue cases) should de e mine he cu o
alue. We epo ed he cu o o 90% sensi i i y, which was
1:453 in subjec s <65 yea s and 1:5120 in subjec s aged ≥65
yea s in a season o high epidemic in ensi y. This means ha
subjec wi h a highe p e accina ion isk (ie, olde subjec s)
will need highe an ibody i e s o ha e he same le el o p o ec-
ion as subjec s wi h a lowe p e accina ion isk such ha ou
model p o ides a ying cu o poin s o p o ec ion.
Fu he me hods used o assess influenza accine p o ec ion
include he scaled logis ic eg ession model sugges ed by
Dunning [14] in which he p obabili y o he subjec de eloping
influenza is he p obabili y ha he subjec is suscep ible
o influenza mul iplied by he p obabili y ha suscep ible
Figu e 3. A/H3N2 HI an ibody i e and es ima ed isk o A/H3N2 influ-
enza disease in subjec s aged ≥65 yea s in an epidemic o low o mode -
a e in ensi y (A) o high in ensi y (B), and in subjec s aged <65 yea s in an
epidemic o low o mode a e in ensi y (C) o high in ensi y (D). Do s
ep esen he obse ed p opo ions o cases and shading shows 95% con-
fidence in e al (logis ic eg ession). Abb e ia ion: HI, hemagglu ina ion
inhibi ion
Table 4. Youden Index Cu o and 90% Sensi i i y Cu o Values
o A/H3N2 Hemagglu ina ion Inhibi ion An ibody Ti e s (Logis ic
Reg ession)
Age
Season
S eng h
Youden Index
Cu o Ti e
90% Sensi i i y
Cu o Ti e
<65 y Low/mode a e 1:5 1:28
<65 y High 1:40 1:453
≥65 y Low/mode a e 1:40 1:453
≥65 y High 1:640 1:5120
6•OFID •Benoi e al
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://o id.ox o djou nals.o g/Downloaded om
indi iduals de elop disease. In addi ion, Li e al [15] de eloped
a dicho omiza ion me hod based on he maximiza ion o he
co ela ion be ween he 2 popula ions and he dicho omous
a iable. In he noncases popula ion, he me hods included a
pa ame e defining he p obabili y ha he obse a ion a ises
om he case popula ion (unp o ec ed bu no exposed).
In ou logis ic eg ession model, we ound ha he isk o dis-
ease was highe o olde han younge subjec s (p e accina ion
isk) and highe in a season o high epidemic in ensi y han
mode a e o low epidemic in ensi y. Subjec s wi h a g ea e
isk (ie, olde subjec and/o in a s ong epidemic in ensi y)
will need highe an ibody i e s o ha e he same le el o p o ec-
ion as subjec s wi h a lowe isk (p e accina ion o linked o
he season), meaning ha ou model p o ides a ying cu o
poin s o p o ec ion. We did no find a significan in e ac ion
be ween pos accina ion i e s and subjec - ela ed co a ia es,
al hough he s udy lacks powe . Howe e , we ound a simila
ela ionship o ha epo ed in he li e a u e: a ou - old in-
c ease in pos accina ion i e s was associa ed wi h a 2- old de-
c ease in he isk o in ec ion.
CONCLUSIONS
The s a is ical modeling exe cise confi med ha he e is a ela-
ionship be ween he occu ence o A/H3N2 disease and HI an-
ibody esponses bu did no allow us o e alua e he p edic i e
powe o he HI esponse. An al e na i e o pe o ming pooled
analyses is o conduc influenza accine e ficacy ials ha a e de-
signed o p o ide da a o COP analyses. In addi ion o ac-
coun ing o age and epidemic in ensi y, op imizing he
COP model would in ol e collec ing se um samples om all
pa icipan s, a he han om a subcoho , iden i ying influ-
enza disease based on a consis en case defini ion, and using
a consis en labo a o y me hod o i al sub yping. The anal-
ysis should also ake in o accoun influenza disease se e i y
and he le el o an igenic misma ch be ween he in ec ing
i us and he accine s ains.
Supplemen a y Ma e ial
Supplemen a y ma e ial is a ailable online a Open Fo um In ec ious Diseas-
es (h p://OpenFo umIn ec iousDiseases.ox o djou nals.o g/).
Acknowledgmen s
We a e indeb ed o he pa icipa ing s udy olun ee s and hei pa en s,
clinicians, nu ses, labo a o y echnicians a he s udy si es, and au ho s om
he p ima y publica ions, especially P. Van Belle, A. T o a, A. Ka onen,
E. Kalisko a, N. Lindblad, and M. Pee e s. We a e also g a e ul o all
Influence65 s udy g oup membe s and he sponso 's p ojec s a o
suppo and con ibu ions h oughou he s udy, especially M. Albanese,
N. Della-Vecchia, M. Dupelle, S. Fannoy, N. Lega e, L. Pesche, M. Ribo ,
S. Papagiannis, J. Roge , M. Lanoue, and V. Wansa d o pa icipa ion in clin-
ical es ing. We hank V. Dodeu , L. Hollinge , K. Pee e s ( eelance, Spain, on
behal o GSK Vaccines), and W. Talbo o in ol emen in he s udy
coo dina ion and s udy managemen . Finally, we hank A. Moon (Moon
Medical Communica ions L d, UK) o p o iding medical w i ing se ices
and B. Dumon (Business and Decision Li e Sciences, on behal o GSK Vac-
cines, Rixensa , Belgium) o edi o ial assis ance and manusc ip coo dina ion.
Au ho con ibu ions. All au ho s had ull access o he da a. All au-
ho s pa icipa ed in he implemen a ion o he s udy including subs an ial
con ibu ions o concep ion and design, he ga he ing o he da a, o analysis
and in e p e a ion o he da a. All au ho s we e in ol ed in c i ically e ising
he manusc ip o impo an in ellec ual con en and app o ed he submi -
ed manusc ip .
Financial suppo . This wo k was suppo ed by GlaxoSmi hKline Bio-
logicals SA. GlaxoSmi hKline Biologicals SA was in ol ed in all s ages o he
s udy conduc and analysis and unded he de elopmen and he publishing
o he p esen manusc ip .
Po en ial conflic s o in e es . J.-M. D., L. O., F. T., B. L. I., and W. D. a e
employees o he GSK g oup o companies and epo owne ship o s ock
op ions and/o es ic ed sha es. A he ime o he s udy, A. B. epo s
g an s om GSK g oup o companies o hei ins i u ions o he conduc
o he s udy, g an s om he IAP Resea ch ne wo k P7/06 o he Belgian
S a e (Belgian Science Policy), and epo s g an s om he “P oje d’Ac ions
de Reche che Conce ées”11/16–039 o he “Communau ée F ançaise de
Belgique”, allowed by he Académie Uni e si ai e Lou ain. A. B. also epo s
she is now an employee o he GSK g oup o companies. G. L.-R., L. J., M. G.,
and O. L. epo g an s om GSK g oup o companies o hei ins i u ions
o he conduc o he s udy. J. E. M. epo s g an s om GSK g oup o com-
panies o hei ins i u ions o he conduc o he s udy, was eimbu sed o
a el and accommoda ion ela ed o ac i i ies o he publica ion s ee ing
commi ee, and epo s ha ing ecei ed pe sonal ees om Sanofi o
Da a Moni o ing Boa d and Ad iso y Boa d, ou side he submi ed
wo k. M. E. epo s g an s om GSK g oup o companies o he ins i u ions
o he conduc o he s udy and a el g an o mee ing o publica ion
s ee ing commi ee and g an s om Bax e and BMBF, ou side he submi -
ed wo k. T. V. epo s g an s om GSK g oup o companies o hei ins i-
u ions o he conduc o he s udy, pe sonal ees o membe ship on
Ad iso y Boa ds, and a el suppo o p esen esul s. C. L. epo s a
g an o suppo a PhD s uden (A. B.) unde he supe ision on his a icle.
All au ho s ha e submi ed he ICMJE Fo m o Disclosu e o Po en ial
Conflic s o In e es . Conflic s ha he edi o s conside ele an o he con-
en o he manusc ip ha e been disclosed.
Re e ences
1. Hobson D, Cu y RL, Bea e AS, Wa d-Ga dne A. The ole o se um
haemagglu ina ion-inhibi ing an ibody in p o ec ion agains challenge
in ec ion wi h influenza A2 and B i uses. J Hyg (Lond) 1972;70:767–77.
2. Ba e PN, Be ezuk G, F i sch S, e al. E ficacy, sa e y, and immunoge-
nici y o a Ve o-cell-cul u e-de i ed i alen influenza accine: a mul i-
cen e, double-blind, andomised, placebo-con olled ial. Lance 2011;
377:751–9.
3. Black S, Nicolay U, Vesika i T, e al. Hemagglu ina ion inhibi ion an i-
body i e s as a co ela e o p o ec ion o inac i a ed influenza accines
in child en. Pedia In ec Dis J 2011; 30:1081–5.
4. Sibe GR. Me hods o es ima ing se ological co ela es o p o ec ion.
De Biol S and 1997; 89:283–96.
5. Skow onski DM, Mose FS, Janjua NZ, e al. H3N2 and o he influenza
epidemic isk based on age-specific es ima es o se o-p o ec ion and
con ac ne wo k in e ac ions. PLoS One 2013; 8:e54015.
6. Be an J, Vesika i T, We zo a V, e al. E ficacy o inac i a ed spli - i us
influenza accine agains cul u e-confi med influenza in heal hy adul s:
a p ospec i e, andomized, placebo-con olled ial. J In ec Dis 2009;
200:1861–9.
7. Be an J, We zo a V, Honeg K, e al. Challenge o conduc ing a place-
bo-con olled andomized e ficacy s udy o influenza accine in a sea-
son wi h low a ack a e and a misma ched accine B s ain: a conc e e
example. BMC In ec Dis 2009; 9:2.
8. Jackson LA, Gaglani MJ, Keyse ling HL, e al. Sa e y, e ficacy, and im-
munogenici y o an inac i a ed influenza accine in heal hy adul s: a
Influenza Vaccine and Se ologic Co ela es o P o ec ion •OFID •7
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://o id.ox o djou nals.o g/Downloaded om
andomized, placebo-con olled ial o e wo influenza seasons. BMC
In ec Dis 2010; 10:71.
9. McElhaney JE, Be an J, De as e JM, e al. AS03-adju an ed e sus non-
adju an ed inac i a ed i alen influenza accine agains seasonal
influenza in elde ly people: a phase 3 andomised ial. Lance In ec
Dis 2013; 13:485–96.
10. Hehme N, Künzel W, Pe schke F, e al. Ten yea s o expe ience wi h he
i alen spli -influenza accine, Flua ix™.ClinD ugIn es 2002;
22:751–69.
11. Tsang TK, Cauchemez S, Pe e a RA, e al. Associa ion be ween an ibody
i e s and p o ec ion agains influenza i us in ec ion wi hin house-
holds. J In ec Dis 2014; 210:684–92.
12. Kelly MJ, Duns an FD, Lloyd K, Fone DL. E alua ing cu poin s o he
MHI-5 and MCS using he GHQ-12: a compa ison o fi e di e en
me hods. BMC Psychia y 2008; 8:10.
13. Coude ille L, Bailleux F, Riche B, e al. Rela ionship be ween haemag-
glu ina ion-inhibi ing an ibody i es and clinical p o ec ion agains
influenza: de elopmen and applica ion o a bayesian andom-e ec s
model. BMC Med Res Me hodol 2010; 10:18.
14. Dunning AJ. A model o immunological co ela es o p o ec ion. S a
Med 2006; 25:1485–97.
15. Li S, Pa nes M, Chan IS. De e mining he cu o based on a con inuous
a iable o define wo popula ions wi h applica ion o accines. J Biopha m
S a 2013;23:662–80.
8•OFID •Benoi e al
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://o id.ox o djou nals.o g/Downloaded om