Vaccine 36 (2018) 4077–4086

Contents lists available at ScienceDirect

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

Influenza vaccination for HIV-positive people: Systematic review and

network meta-analysis
Wei Zhang a,b,⇑, Huiying Sun b, Mohammad Atiquzzaman c, Julie Sou b, Aslam H. Anis a,b, Curtis Cooper d
a
School of Population and Public Health, University of British Columbia, Canada
b
Centre for Health Evaluation and Outcome Sciences, St. Paul’s Hospital, Canada
c
Faculty of Pharmaceutical Sciences, University of British Columbia, Canada
d
Department of Medicine, University of Ottawa, Canada

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

Article history: Background: People with Human Immunodeficiency Virus (HIV) are highly susceptible to influenza-
Received 22 November 2017 related morbidity and mortality. In order to assess comparative efficacy of influenza vaccine strategies
Received in revised form 30 April 2018 among HIV-positive people, we performed a systematic review and Bayesian network meta-analysis
Accepted 21 May 2018
(NMA).
Available online 30 May 2018
Methods: In this systematic review, we searched MEDLINE, EMBASE, Cochrane Central Register of
Controlled Trials, Cochrane Database of Systematic Reviews, and CINAHL between 1946 and July 2015
Keywords:
for randomized controlled trials (RCTs) on influenza vaccines for HIV-positive adults reporting serocon-
HIV
Influenza vaccine
version or seroprotection outcomes. The NMAs were conducted within a Bayesian framework and logistic
Network meta-analysis models were used for comparing the effect of the vaccine strategies on the two outcomes.
Seroconversion Results: A total of 1957 publications were identified, 143 were selected for full review, and 13 RCTs were
Seroprotection included in our final analysis. Fourteen separate NMAs were conducted by outcomes, vaccine strain, and
different outcome measurement timepoints. For example, compared with the 15 lg single vaccine strat-
egy, the odds ratio was the highest for the adjuvant 7.5 lg booster strategy (2.99 [95% credible interval
1.18–7.66]) when comparing seroconversion for H1N1 at 14–41 days after the last dose of vaccination
and for the 60 lg single strategy (2.33 [1.31–4.18]) when comparing seroconversion for strain B.
Conclusions: The adjuvant 7.5 lg booster and 60 lg single vaccine strategies provided better seroconver-
sion and seroprotection outcomes. These findings have important implications for national and interna-
tional guidelines for influenza vaccination for HIV-positive people and future research.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction positive people [5–8]. The evidence, although limited, suggests

The emergence of antiretroviral therapy (ART) has greatly
reduced human immunodeficiency virus (HIV)-related morbidity
and mortality [1]. Despite this, HIV-positive people remain at
increased risk for susceptibility to infectious diseases including
influenza, given the deficiencies in humoural and cell-mediated
immunity [2]. For example, studies have shown that while ART
has reduced the number of influenza-associated hospitalizations
[3], HIV-positive people still have a higher influenza-attributable
risk of acute cardiopulmonary events and prolonged influenza
symptoms [4,5]. Several systematic reviews and meta-analyses
have been conducted to compare the efficacy, effectiveness, and
safety of influenza vaccination with no vaccination among HIV-
⇑ Corresponding author at: Centre for Health Evaluation and Outcome Sciences,
St. Paul’s Hospital, 588-1081 Burrard Street, Vancouver, BC V6Z 1Y6, Canada.
E-mail address: wzhang@cheos.ubc.ca (W. Zhang).
that influenza vaccination is effective in preventing influenza
infection in HIV-positive adults [5–7]. Therefore, many national
guidelines are recommending influenza vaccination for HIV-
positive people [9–11].
Many different influenza vaccine strategies – characterized by
different dose, dose frequency (single versus booster), mode of
administration (intramuscular versus intradermal), and use of
adjuvant (yes versus no), have been developed and compared
among HIV-positive people in clinical trials. For example, a recent
randomized controlled trial (RCT) compared MF59-adjuvanted A/
H1N1 pandemic influenza vaccine (intramuscular adjuvant 7.5 lg
booster) with an unadjuvanted trivalent influenza vaccine (intra-
muscular 15 lg booster) among HIV-positive people and tested
the efficacy of the first and second dose of the corresponding vac-
cine [12]. Results showed that a greater proportion of HIV-positive
participants achieved seroprotection (93%) and seroconversion

https://doi.org/10.1016/j.vaccine.2018.05.077
0264-410X/Ó 2018 Elsevier Ltd. All rights reserved.
4078 W. Zhang et al. / Vaccine 36 (2018) 4077–4086

(79%) after the first dose of the MF59-adjuvanted vaccine as com-
pared to the unadjuvanted vaccine (87% and 68%, respectively).
After two doses of the adjuvanted vaccine, 100% of HIV-positive
participants were seroprotected while 90% were seroprotected
after two doses of the unadjuvanted vaccine [12]. Previous system-
atic reviews and meta-analyses examining the efficacy of influenza
vaccination among HIV-positive people are out of date and have
identified only up to three RCTs [5–8]. Moreover, they have only
compared the effects of vaccine versus no vaccine without distin-
guishing different vaccine strategies [5–8]. As such, no previous
studies have synthesized evidence on the comparisons between
different vaccination strategies.
Network meta-analysis (NMA) has recently been utilized to
provide clinical evidence even when there is a lack of direct
head-to-head comparisons between treatment options [13,14]. In
a NMA, all treatment options (three or more) are assessed simulta-
neously based on direct evidence between treatment options
within clinical trials and indirect evidence across trials with a com-
mon treatment comparator [14]. This method provides better com-
parative evidence than the conventional, pair-wise meta-analysis
and increases the power of statistical comparisons while allowing
for inferences of comparative effects between treatment options
that have not been compared head-to-head [13,15]. The objective
of this study is to use a NMA to assess the comparative efficacy
of influenza vaccine strategies available at present for HIV-
positive people.
2. Methods
2.1. Search strategy and selection criteria
In this systematic review, a systematic literature search was
performed to identify all studies relevant to influenza vaccination
among HIV-positive people, according to PRISMA guidelines [16].
Databases searched were MEDLINE (1946 to July 22, 2015),
EMBASE (1974 to July 21, 2015), Cochrane Central Register of Con-
trolled Trials (1991 to June 2015), Cochrane Database of Systematic
Reviews (2005 to June 2015), and CINAHL (until July 22, 2015). The
Medical Subject Headings (MeSH) term ‘‘influenza vaccines”; key-
words: ‘‘HIV”, ‘‘human immunodeficiency virus”; and truncated
keywords: ‘‘flu vaccin”, ‘‘influenza vaccin” were used. The specific
search strategy is presented in the appendix. The full text of all rel-
evant review articles were examined to identify potentially over-
looked studies. In the search strategy we limited the studies to
English language only.
Study inclusion was determined by two research assistants
independently. During the first stage, titles and abstracts of all arti-

Records identified through database

Identification

searching (n = 1957)
• EMBASE – 1974 to Jul 21, 2015 (n=1387)
Duplicates removed (n = 363)
• MEDLINE – 1946 to Jul 22, 2015 (n=366)
• COCH-CT – Jun 2015 (n=59)
• COCH-SR – 2005 to Jun 2015 (n=18)
• CINAHL Complete – Jul 22, 2015 (n=127)

Studies excluded (n =1451)

• Duplicates (n=23)
Screening

• Irrelevant topic (n=1310)

• Review (n=54)
Titles & abstracts screened (n =1594)
• Editorial (n=12)
• Conference publication (n=5)
• Case study (n=4)
• Letter (n=19)
• Survey (n=1)
• Other studies (n=6)
• Full text unavailable (n=17)
Eligibility

Full-text articles assessed for eligibility Studies excluded (n =125)

(n =143)
• No comparison group among HIV+ people
(n= 115)
• Not our outcomes of interest (n=4)
• Non-injection vaccination for adult (n=1)
• Missing information (n=1)
• Outcomes measured beyond time range
(n=3)
• Not randomized controlled trial (n=1)
Included

• Pediatric studies (n=5)

Studies included for network meta-analysis

(n =13)

Fig. 1. Study inclusion for network meta-analysis.

Table 1
Characteristics of selected studies comparing different influenza vaccine strategies among HIV-positive people (N = 13).

Author Year Country Study Design Vaccine Strains Vaccine Strategies N Male Age Reported Outcomes of
(%) Interest
McKittrick et al. 2013 USA Double blind (DB) randomized Seasonal vaccine: 15 single 95 77 Median (Interquartile Range) Seroconversion (SC);
[22] controlled trial (RCT) A/California/07/2009 (H1N1), A/ Med (IQR): 46 (37–53) seroprotection (SP)
Victoria/210/2009 (H3N2), B/ 60 single 100 64 44 (35–50)
Brisbane/60/2008
Santini-Oliveira 2012 Brazil RCT, open label, phase II Pandemic vaccine: Adjuvant (Adj) 129 69.8 Mean (Standard Deviation) SC; SP
et al. [23] A/California/7/2009 (H1N1) 3.75 booster (SD): 42.2 (8.5)
Adj 7.5 booster 127 62.2 42.0 (9.2)
Cooper et al. [2] 2012 Canada RCT, phase III, multicenter Pandemic vaccine: Adj 3.75 booster 74 82 Mean (SD): 45 (8.1) SC; SP
A/California/7/2009 (H1N1) Adj 3.75 single 76 83
El Sahly et al. [24] 2012 USA RCT, open label Pandemic vaccine: 15 booster 95 80 Mean: 45.7 SC; SP
A/California/7/2009 (H1N1) 30 booster 97 77 45.3
Ansaldi et al. [25] 2012 Italy RCT, open-label study, phase IV Seasonal vaccine: 15 single 26 65.4 Med (IQR): 48.7 (37.0–53.9) SC; SP
A/California/7/2009 (H1N1), A/ Intradermal 9 28 71.5 47.0 (35.7–52.2)
Perth/16/2009 (H3N2), B/ single
Brisbane/60/2008
Cooper et al. [26] 2011 Canada RCT, multicenter Seasonal vaccine: 15 booster 100 88 Mean (SD): 47 (8.5) SC; SP
A/Brisbane/59/2007 (H1N1), A/ 15 single 94 90 47 (8.5)
Uruguay/716/2007 (H3N2), B/ 30 booster 104 92 47 (8.5)

W. Zhang et al. / Vaccine 36 (2018) 4077–4086

Florida/4/2006
Launay et al. [27] 2011 France RCT, phase 2, multicenter, patient- Pandemic vaccine: 15 booster 151 76 Med (IQR): 47.3 (40.2–53.6) SC; SP
blinded A/California/7/2009 (NYMC X- Adj 3.75 booster 155 85 46.5 (38.6–54.0)
179A)
Madhi et al. [28] 2011 South DB RCT; placebo-controlled trial Seasonal vaccine: 15 single 97 16.7 Mean (SD): 36.3 (7.2) SC
Africa A/Solomon Islands/3/2006 Placebo 92 14.3 36.4 (7.2)
(H1N1)-like strain (IVR-145), A/
Brisbane/10/2007 (H3N2)-like
strain (IVR-147), B/Florida/4/
2006-like strain (B/Brisbane/3/
2007)
Gelinck et al. [29] 2009 The Nether- Prospective, randomized study Seasonal vaccine: 15 single 41 71 Mean (range): 46 (22–75) SP
lands A/New York/55/2004 NYMC X- 15 single 39 80 46 (20–60)
157 reassortant, A/New intradermal
Caledonia/20/99 IVR-116
reassortant, B/Jiangsu/10/03

Durando et al. 2008 Italy Randomized, comparative study; Seasonal vaccine: 15 single 49 79.6 Mean (SD): 40.1 (6.4) SC
[30] multicenter, open-label A/New Caledonia/20/99 (H1N1), Adj 15 single 46 86.9 41.0 (5.7)
A/California/7/2004 (H3N2), B/
Shanghai/361/2002
Gabutti et al. [31] 2005 Italy RCT, comparative, open study Seasonal vaccine: 15 single 19 73.7 Mean (95% CI): 37.1 (34.5– SC; SP
A/New Caledonia/20/99 (H1N1), 39.7)
A/Moscow/10/99 (H3N2), B/ Adj 15 single 18 77.8 40.2 (35.5–44.9)
Hong Kong/330/2001

Tasker et al. [32] 1999 USA DB RCT Seasonal vaccine: 15 single 55 95 Mean: 33.5 SC; SP
A/Texas/36/91 (H1N1), A/ Placebo 47 96 32.4
Johannesburg/33/94 (H3N2), B/
Harbin/07/94
Diaz et al. [12] 2014 Brazil RCT Pandemic vaccine: 15 booster 32 61 Mean (SD): 42.0 (10.4) SC; SP
A/California/7/2009 (H1N1) Adj 7.5 booster 29 46 42.6 (9.1)

Seasonal vaccine:
A/California/7/2009 (H1N1), A/
Perth/16/2009 (H3N2), B/

4079
Brisbane/60/2008
4080 W. Zhang et al. / Vaccine 36 (2018) 4077–4086
cles identified by the literature search were screened. Articles were
excluded if the study was not about HIV-positive people, influenza
vaccination, or if they were duplicates, reviews, editorials, confer-
ence abstracts, letters, or case studies. Next, the full texts of all arti-
cles deemed eligible from the previous stage were reviewed to
further confirm their eligibility. At this stage, in addition to the cri-
teria considered at the first stage, further exclusion criteria
included no comparison of different flu vaccine strategies among
HIV-positive people, not reporting our outcomes of interest, non-
injectable flu vaccine for adults, and missing information. Any dis-
crepancies between the two reviewers were resolved by consensus
with input from a third reviewer (W.Z.).
After we identified the studies that met the above initial inclu-
sion/exclusion criteria, we further examined the comparability
among these studies in terms of the timing of outcome measure-
ment, study design (RCT or not), and adult or pediatric studies.
Since antibody titres increase after vaccination and then gradually
decrease over time, the outcome measurement timing is very
important to consider. To ensure the comparability of the out-
comes among studies, we chose the measurement timepoints from
14 to 41 days since the first dose (before the booster) and 42–84
days since the first dose (after the booster). After the comparability
examination, we decided to exclude those studies with outcome
measurement beyond 84 days, non-RCT studies and pediatric
studies.
Legend
ID = Intradermal; Adj = Adjuvant
*Numbers in circles indicate the number of studies that compared the
respective vaccine strategies, as indicated by connecting lines
Fig. 2. Network of eligible comparisons for seroconversion of vaccine strain H1N1 between 14 and 41 days.
2.2. Data extraction and outcome definitions
For each study included in the final review, data were extracted
into a table. The information extracted included authors, study
year, country, study design, age, gender, ART use, CD4 cell count,
HIV viral load, study inclusion/exclusion criteria, total number of
participants, vaccine strains, vaccine strategies by arms, follow-
up timepoints, and reported outcomes including seroconversion
(either a pre-vaccination hemagglutination inhibition (HI) anti-
body titre <1:10 and a post-vaccination HI antibody titre >1:40
or a pre-vaccination HI titre >1:10 and a minimum four-fold rise
in post-vaccination HI antibody titre), and seroprotection (HI anti-
body titre >1:40).
The outcomes were the proportion of patients achieving sero-
conversion and the proportion of patients achieving seroprotec-
tion. Consistently, for each study, we adopted the ‘‘intention to
treat” method by assuming any missing data or lost to follow up
as ‘‘failure” to achieve seroconversion and seroprotection. The
two outcomes were extracted by vaccine strains (H1N1, H3N2,
and B) and outcome measurement timepoints. We synthesized
outcomes at three different measurement timepoints: (1) out-
comes at 14–41 days since the first dose for single or booster vac-
cination strategies (prior to booster); (2) outcomes at 42–84 days
since the first dose for single or booster strategies (post booster);
(3) outcomes at 14–41 days after the last dose of vaccination to
 W. Zhang et al. / Vaccine 36 (2018) 4077–4086
compare the best outcomes for different vaccination strategies (i.e.,
outcomes at 14–41 days for single strategies and outcomes at 42–
84 days for booster strategies). We assumed the best outcomes
would be between 14 and 41 days after the last dose of
vaccination.
2.3. Risk of bias assessment
Risk of bias was assessed using the Cochrane Collaboration’s
tool for assessing risk bias in randomized trials [17]. The assess-
ment was conducted from two perspectives: one from a general
perspective for all possible outcomes and the other from our study
perspective for our two outcomes of interest. The assessment was
conducted by two research assistants and any discrepancies were
resolved by consensus with input from a third reviewer (W.Z.).
Six bias domains were considered: random sequence generation,
allocation concealment, blinding of participants and personnel,
incomplete outcome data, and selective reporting. Each study
was assessed as either high, low, or unclear risk for bias for each
domain.
2.4. Data synthesis
We performed Bayesian NMAs using both fixed-effect and
random-effects models. The random-effects model took into
account the correlations between trial-specific vaccine strategies
in multi-arm trials. Three Monte Carlo Markov chains (MCMC)
were used for all models. The convergence was assessed using
Gelman-Rubin diagnostics [18] and inspection of autocorrelation
and history plots. Logistic models were used to compare the
Adj = adjuvant; Ref = reference strategy
*An odds ratio < 1 favours the reference strategy, while an odd ratio > 1 favours the comparators.
Fig. 3. Odds ratios and 95% credible intervals from network meta-analysis of vaccination strategies on seroconversion at 14–41 days by vaccine strains.
4081
treatment effect on the two outcomes. Vague priors were set for
the relative treatment effects and the heterogeneity between stud-
ies. Goodness of fit was measured using the deviance information
criterion (DIC) – the smaller, the better, and difference of  5
points was considered meaningful [19]. DIC was used to compare
fixed-effect models and random-effects models [19]. The posterior
mean of the residual deviance was also consulted to check the
overall fit [20]. The results for the NMA were reported as posterior
median odds ratios with the corresponding 95% credible intervals
(CrIs) and presented by forest plots. Additionally, consistency
between the direct and indirect evidence have been assessed by
the node-splitting method [21]. All analyses were conducted using
WinBUGS version 1.4 (Medical Research Council Biostatistics Unit,
Cambridge) and R 3.4.0 (https://www.r-project.org/).
3. Results
We identified a total of 1957 abstracts (Fig. 1), of these, 143
publications were identified for full text screening and 13 adult-
specific studies were included for final NMA [2,12,22–32]. We
excluded 3 studies reported outcomes at time points ranging from
6 months to 1 year – well beyond the time range of the 13 studies
we finally included and 1 non-RCT study. We also excluded 5 stud-
ies on children and adolescents as they were conducted among dif-
ferent age groups: 6–59 months, 5–18 years, 9–20 years, and 9–26
years. Study characteristics are presented in Table 1 with more
details that can be found in the appendix.
A total of 11 unique vaccination strategies were identified by
different dose, dose frequency (single versus booster), mode of
administration (intramuscular versus intradermal), and use of
4082 W. Zhang et al. / Vaccine 36 (2018) 4077–4086
Adj = adjuvant; Ref = reference strategy
*An odds ratio < 1 favours the reference strategy, while an odd ratio > 1 favours the comparators.
Fig. 4. Odds ratios and 95% credible intervals from network meta-analysis of vaccination strategies on seroprotection at 14–41 days by vaccine strains.
adjuvant (yes versus no): 15 lg single, 15 lg booster, 30 lg boos-
ter, 60 lg single, intradermal 9 lg single, intradermal 15 lg single,
adjuvant 3.75 lg single, adjuvant 3.75 lg booster, adjuvant 7.5 lg
booster, adjuvant 15 lg single, and placebo. All strategies that are
not labeled as intradermal or adjuvanted are intramuscular and
non-adjuvanted. Detailed descriptions can be found in the appen-
dix. This implied that we could compare a maximum of 9 possible
unique vaccine strategies for the 14–41 day measurement time-
point if both seroconversion and seroprotection outcomes were
reported in all 13 included studies when the first dose of booster
vaccine strategies were considered as one single strategy (see
appendix). For example, the first dose of adjuvant 7.5 lg booster
vaccine strategy was considered as an adjuvant 7.5 lg single strat-
egy for outcomes reported at the 14–41 days measurement time-
point. Similarly, we could compare a maximum of 11 strategies
for the 42–84 days outcomes if the outcomes were reported.
The actual number of studies and strategies considered in each
of our separate analyses based on the availability of outcome
reporting is presented in the appendix. We conducted a total of
14 separate analyses by outcome (seroconversion and seroprotec-
tion), vaccine strain (H1N1, H3N2, and B), and outcome measure-
ment timepoint (14–41 days, 42–84 days, and 14–41 days for
single strategies and 42–84 days for booster strategies). For exam-
ple, one set of analyses assessed seroconversion of the H1N1 vac-
cine strain for the 14–41 days timepoint; this analysis consisted
of 11 studies and 8 unique strategies. Fig. 2 plots the corresponding
network of this specific analysis. Other networks are presented in
the appendix.
In NMA, we used a burn-in of 100,000 iterations to ensure the
model convergence was satisfactory for all outcomes and our
results were based on 500,000 further iterations (thin = 5, i.e.,
retaining every 5th parameter value) for each of the three MCMC
chains. The model fit statistics are reported in the appendix. Of
the 14 separate analyses, only DICs for two analyses involving
the seroconversion outcome of strain B from the random-effects
models were meaningfully smaller than those from the fixed-
effect models (86.85 vs. 92.13 and 92.53 vs. 97.75). Therefore, we
reported the results from the fixed-effect models for all analyses
in this report. However, since random-effects models may fit the
data better for the two analyses involving the seroconversion out-
come of strain B, we presented the corresponding random-effects
model results in the appendix.
Fig. 3 plots the pairwise comparison results for vaccine strains
H1N1, H3N2, and B for the seroconversion outcome at the 14–41
days timepoint using the fixed-effect model. Plots showed that
among single dose vaccine strategies, 30 lg, 60 lg, adjuvant 3.75
lg, and adjuvant 7.5 lg were all significantly better than 15 lg,
adjuvant 15 lg, intradermal 9 lg, and placebo for the H1N1 vac-
cine strain but they were not significantly different between each
other. It is important to note that the adjuvant 3.75 lg and adju-
vant 7.5 lg vaccine strategies were adjuvanted monovalent vacci-
nes (H1N1pdm09) while the adjuvant 15 lg was an adjuvanted
trivalent vaccine. Compared with the 15 lg vaccine strategy, the
odds ratio was the highest for the adjuvant 7.5 lg (2.99 [95% CrIs
1.59–5.73]) strategy. The single dose strategies 30 lg and 60 lg
also performed better than 15 lg for strain B, and 60 lg was better
than adjuvant 15 lg. Compared with 15 lg, the odds ratio was the
highest for 60 lg (2.33 [1.31–4.20]). There was no superior vaccine
dose strategy for strain H3N2. In terms of seroprotection (Fig. 4),
we found that adjuvant 3.75 lg was significantly better than 15
 W. Zhang et al. / Vaccine 36 (2018) 4077–4086 4083

Adj = adjuvant; Ref = reference strategy; SC = seroconversion; SP = seroprotection

*An odds ratio < 1 favours the reference strategy, while an odd ratio > 1 favours the comparators.
Fig. 5. Odds ratios and 95% credible intervals from network meta-analysis of vaccination strategies on seroconversion and seroprotection at 42–84 days for H1N1.

lg, intradermal 15 lg, adjuvant 15 lg, intradermal 9 lg, and 30 lg
for strain H1N1. We calculated the probability of each vaccine
strategy being the best strategy for each separate analysis and pre-
sented the results in the appendix. Overall, the single dose strate-
gies, adjuvant 7.5 lg, adjuvant 3.75 lg, and 60 lg performed best
at the 14–41 days timepoint according to their probability of being
ranked as the best for each outcome and vaccine strain.
Fig. 5 plots the comparison results for H1N1 for both serocon-
version and seroprotection outcomes at 42–84 days. The compar-
isons involved both single and booster strategies. It showed that
in terms of seroconversion, adjuvant 3.75 lg booster and adjuvant
7.5 lg booster vaccine strategies performed significantly better
than 15 lg single, 15 lg booster, and adjuvant 3.75 lg single
strategies. Compared with 15 lg, the odds ratio was the highest
for adjuvant 7.5 lg booster (2.99 [1.19–7.69]). In terms of seropro-
tection, adjuvant 3.75 lg booster was significantly better than 15
lg single and adjuvant 3.75 lg single strategies. Overall, adjuvant
7.5 lg booster, adjuvant 3.75 lg booster, and 30 lg booster were
most effective at 42–84 days.
Figs. 6 and 7 compared seroconversion and seroprotection out-
comes at 14–41 days for single dose vaccine strategies and at 42–
84 days for booster strategies, respectively. In terms of seroconver-
sion for H1N1, 30 lg booster, adjuvant 3.75 lg booster, and adju-
vant 7.5 lg booster strategies were significantly superior to 15
lg single, adjuvant 15 lg single, 15 lg booster, intradermal 9 lg
single, and placebo but similar to 60 lg single and adjuvant 3.75
lg single vaccine strategies. The odds ratio was the highest for
adjuvant 7.5 lg booster (2.99 [1.18–7.66]) when compared with
15 lg. For strain B, the 60 lg single was superior to 15 lg single
(2.33 [1.31–4.18]) and 15 lg booster (2.95 [1.11–7.91]. In terms
of seroprotection, compared to the standard 15 lg single, all other
vaccine strategies were not significantly different except for the
adjuvant 15 lg single, which was significantly worse for both
H1N1 and B strains. Overall, adjuvant 7.5 lg booster and 60 lg sin-
gle strategies were the two best vaccine strategies.
Node-split consistency tests were performed for all the possible
comparisons: among 15 lg single, adjuvant 3.75 lg single, and
adjuvant 7.5 lg single at 14–41 days timepoint, and among 15 lg
booster, adjuvant 3.75 lg booster, and adjuvant 7.5 lg booster
for their best outcome scenario (14–41 days for single strategies
and 42–84 days for booster strategies). There was no evidence of
inconsistency between direct and indirect comparisons. The plot
and p-values have been presented in the appendix.
In terms of risk of bias assessments from a general perspective,
a perspective that considers all reported outcomes, four studies
were considered low risk for all of the six bias domains, seven stud-
ies were considered high risk for at least one of the six domains
(most studies were high risk in the domains of blinding of partici-
pants and personnel and blinding of outcome assessment), and two
studies were considered unclear risk for at least one of the six
domains. However, since our study only considered seroconversion
and seroprotection outcomes, objective measures derived from
blood samples, the impact of the potential risks related to alloca-
tion concealment, blinding of participants and personnel, and
blinding of outcome assessment was minimal. Thus, from our
study perspective, all of the 13 studies were either assessed as
4084 W. Zhang et al. / Vaccine 36 (2018) 4077–4086
Adj = adjuvant; Ref = reference strategy
*An odds ratio < 1 favours the reference strategy, while an odd ratio > 1 favours the comparators.
Fig. 6. Odds ratios and 95% credible intervals from network meta-analysis of vaccination strategies on seroconversion at 14–41 days for single strategies and at 42–84 days
for booster strategies by vaccine strains.
low or unclear risk. All studies were included in our final analyses.
The risk assessment results from the two perspectives and sup-
porting evidence are presented in the appendix.
4. Discussion
This review was the first to assess and compare the efficacy of
different influenza vaccine strategies available at present for HIV-
positive people. A total of 14 separate NMA analyses have been
conducted to compare different strategies by seroconversion and
seroprotection, vaccine strains (H1N1, H3N2, B), and outcome
measurement timepoints (14–41, 42–84, and best outcome scenar-
io). A different number of vaccine strategies were compared in sep-
arate analyses depending on the availability of literature evidence.
Most vaccine strategies were compared for H1N1, where the adju-
vant 7.5 lg booster and 60 lg single strategies consistently
showed better performance in all analyses. Excluding more com-
plex adjuvant and booster strategies due to lack of evidence, the
60 lg single was found to be the most effective strategy for strain
B. We found that no one strategy was significantly better than
other strategies for H3N2.
In this study, we have treated pandemic H1N1 and seasonal
H1N1 the same and synthesized their seroconversion and seropro-
tection outcomes. The same strategy was utilized in our evaluation
of H3N2 and B immunogenicity. This was a pragmatic decision.
However, we recognize that different strains of influenza may dif-
fer in immunogenicity and that past history of natural influenza
infection and/or influenza vaccine exposure influences immune
response to subsequent influenza vaccinations [33–35]. We
believe, therefore, that it is important to include this among mul-
tiple variables that may influence vaccine immunogenicity.
The outcome seroprotection does not adjust for the baseline
titres, which may be less comparable among different studies than
seroconversion. It is well recognized that at baseline, a proportion
of individuals will already possess strain-specific titres [36]. In a
proportion of vaccine recipients, this reflects carry-over titres from
previous natural infection and/or immunization [37]. In the 2009
pandemic year it was recognized that some individuals possessed
titres to pandemic H1N1. Cross reactivity with prior H1N1 strains
and long-term immunity obtained in individuals exposed to natu-
ral H1N1 infection prior to 1957 are thought to explain this obser-
vation [38,39]. We elected not to account for this phenomenon in
our analyses as available data lacked this level of detail.
We did not consider other outcomes such as influenza-like ill-
ness (ILI) and laboratory confirmed influenza, hospitalization, mor-
tality, pneumonia, or safety in our NMA. According to previous
reviews [5–8] and our review, we did not find that there were
enough studies to enable NMAs for these outcomes. Seroconver-
sion and seroprotection are related to fewer occurrences of
influenza-like illness later on and are well accepted as the standard
in evaluation of influenza vaccine immunogenicity [40]. Individual
publications and past reviews have suggested no evidence of safety
concerns with the vaccines, doses, dosing frequencies, or adjuvants
evaluated in these studies [2,5,8]. Instead, we reviewed these other
outcomes for the top vaccine strategies in five of our included stud-
ies (i.e., adjuvant 7.5 lg single and booster, 60 lg single, and adju-
vant 3.75 lg single and booster) [2,12,22,23,27]. Three studies
measured ILI and laboratory confirmed influenza and found ILI
cases were balanced between different vaccine strategies and very
 W. Zhang et al. / Vaccine 36 (2018) 4077–4086
Adj = adjuvant; Ref = reference strategy
*An odds ratio < 1 favours the reference strategy, while an odd ratio > 1 favours the comparators.
Fig. 7. Odds ratios and 95% credible intervals from network meta-analysis of vaccination strategies on seroprotection at 14–41 days for single strategies and at 42–84 days for
booster strategies by vaccine strains.
few cases were laboratory confirmed [2,23,27]. In terms of adverse
events, it showed that higher proportion of patients experienced
local or systemic adverse events in the groups with higher dose
(i.e., 60 lg single vs. 15 lg single [22], adjuvant 7.5 lg booster
vs. adjuvant 3.75 lg booster [23]) or the adjuvanted vaccine
groups (i.e., adjuvant 3.75 lg booster vs. 15 lg booster [27], adju-
vant 7.5 lg booster vs. 15 lg booster [12]). There was a trend
towards less reactogenicity after the second vaccine dose com-
pared with after the first dose [2,12,23,27]. However, the majority
of the adverse events were mild or moderate [2,12,22,23,27]. There
were very few severe adverse events and they were balanced
between different vaccine strategy groups. No deaths occurred in
any of these five studies. McKittrick et al. reported 3 hospitaliza-
tions (unrelated to the vaccine intervention) occurring in the stan-
dard 15 lg single vaccine group when compared with 60 lg single
[22] and Launay et al. reported 1 hospitalization for ILI after the
second unadjuvanted 15 lg booster when compared with adjuvant
3.75 lg booster [27]. Overall, all of these vaccination strategies
were well tolerated [2,12,22,23,27] and did not affect HIV viral
loads and CD4 cell counts [2,12,23,27].
In conclusion, seasonal influenza represents an ongoing medical
risk to HIV-positive people. Given suboptimal immunogenicity
with standard, licensed vaccines, different vaccination strategies
have been developed to help prevent influenza infections. Our
study provides evidence that the adjuvant 7.5 lg booster and 60
lg single vaccine strategies provide better seroconversion and
seroprotection outcomes. These findings have important implica-
tions for national and international guidelines for influenza vacci-
nation in HIV-positive people and may inform future research
evaluating novel vaccination strategies.
4085
Conflict of interest
CC received research funding from Merck.
Acknowledgements
We would like to thank Judy Chiu for conducting the first round lit-
erature search for this systematic review. We would also like to
thank Nazrul Islam and Canice Ma for their contributions to article
screening.
Author Contributions
WZ, HS, AA, and CC designed the study. Under WZ’s guidance,
MA and JS performed the literature search, screened and reviewed
publications, extracted the data from all included studies, and
assessed the risk of bias for all included studies. HS analyzed the
data. WZ, AA, and CC guided the data analyses and interpreted
the results. WZ drafted the manuscript, and HS, CC, AA, JS, and
MA provided significant contributions to the manuscript.
Funding
This work was not supported by a funding source.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at https://doi.org/10.1016/j.vaccine.2018.05.
077.
4086 W. Zhang et al. / Vaccine 36 (2018) 4077–4086

References [23] Santini-Oliveira M, Camacho LAB, Souza TML, Luz PM, Vasconcellos MTL,
Giacoia-Gripp CBW, et al. H1N1pdm09 adjuvanted vaccination in HIV-infected
adults: a randomized trial of two single versus two double doses. PLoS One
[1] Granich R, Gupta S, Hersh B, Williams B, Montaner J, Young B, et al. Trends in
2012;7:e39310. https://doi.org/10.1371/journal.pone.0039310.
AIDS deaths, new infections and ART coverage in the top 30 countries with the
[24] El Sahly HM, Davis C, Kotloff K, Meier J, Winokur PL, Wald A, et al. Higher
highest AIDS Mortality Burden; 1990–2013. PLoS One 2015;10():e0131353.
antigen content improves the immune response to 2009 H1N1 influenza
https://doi.org/10.1371/journal.pone.0131353.
vaccine in HIV-infected adults: a randomized clinical trial. J Infect Dis
[2] Cooper C, Klein M, Walmsley S, Haase D, MacKinnon-Cameron D, Marty K, et al.
2012;205:703–12. https://doi.org/10.1093/infdis/jir837.
High-level immunogenicity is achieved vaccine with adjuvanted pandemic
[25] Ansaldi F, Valle L, de Florentiis D, Parodi V, Murdaca G, Bruzzone B, et al. Phase
H1N1(2009) and improved with booster dosing in a randomized trial of HIV-
4 randomized trial of intradermal low-antigen-content inactivated influenza
infected adults. HIV Clin Trials 2012;13:23–32. https://doi.org/10.1310/
vaccine versus standard-dose intramuscular vaccine in HIV-1-infected adults.
hct1301-23.
Hum Vaccin Immunother 2012;8:1048–52. https://doi.org/10.4161/hv.20347.
[3] Neuzil KM, Coffey CS, Mitchel EF, Griffin MR. Cardiopulmonary hospitalizations
[26] Cooper C, Thorne A, Klein M, Conway B, Boivin G, Haase D, et al.
during influenza season in adults and adolescents with advanced HIV
Immunogenicity is not improved by increased antigen dose or booster
infection. J Acquir Immune Defic Syndr 2003;34:304–7.
dosing of seasonal influenza vaccine in a randomized trial of HIV infected
[4] Sheth AN, Althoff KN, Brooks JT. Influenza susceptibility, severity, and
adults. PLoS One 2011;6:e17758. https://doi.org/10.1371/journal.
shedding in HIV-infected adults: a review of the literature. Clin Infect Dis
pone.0017758.
2011;52:219–27. https://doi.org/10.1093/cid/ciq110.
[27] Launay O, Desaint C, Durier C, Loulergue P, Duval X, Jacomet C, et al. Safety and
[5] Remschmidt C, Wichmann O, Harder T. Influenza vaccination in HIV-infected
immunogenicity of a monovalent 2009 influenza A/H1N1v vaccine adjuvanted
individuals: systematic review and assessment of quality of evidence related
with AS03A or unadjuvanted in HIV-infected adults: a randomized. Controlled
to vaccine efficacy, effectiveness and safety. Vaccine 2014;32:5585–92.
Trial J Infect Dis 2011;204:124–34. https://doi.org/10.1093/infdis/jir211.
https://doi.org/10.1016/j.vaccine.2014.07.101.
[28] Madhi SA, Maskew M, Koen A, Kuwanda L, Besselaar TG, Naidoo D, et al.
[6] Atashili J, Kalilani L, Adimora AA. Efficacy and clinical effectiveness of influenza
Trivalent inactivated influenza vaccine in African adults infected with human
vaccines in HIV-infected individuals: a meta-analysis. BMC Infect Dis
immunodeficient virus: double blind, randomized clinical trial of efficacy,
2006;6:138. https://doi.org/10.1186/1471-2334-6-138.
immunogenicity, and safety. Clin Infect Dis 2011;52:128–37. https://doi.org/
[7] Beck CR, McKenzie BC, Hashim AB, Harris RC, University of Nottingham
10.1093/cid/ciq004.
Influenza and the ImmunoCompromised (UNIIC) Study Group, Nguyen-Van-
[29] Gelinck LBS, van den Bemt BJF, Marijt WAF, van der Bijl AE, Visser LG, Cats HA,
Tam JS. Influenza vaccination for immunocompromised patients: systematic
et al. Intradermal influenza vaccination in immunocompromized patients is
review and meta-analysis by etiology. J Infect Dis 2012;206:1250–9. https://
immunogenic and feasible. Vaccine 2009;27:2469–74. https://doi.org/
doi.org/10.1093/infdis/jis487.
10.1016/j.vaccine.2009.02.053.
[8] Anema A, Mills E, Montaner J, Brownstein JS, Cooper C. Efficacy of influenza
[30] Durando P, Fenoglio D, Boschini A, Ansaldi F, Icardi G, Sticchi L, et al. Safety and
vaccination in HIV-positive patients: a systematic review and meta-analysis.
immunogenicity of two influenza virus subunit vaccines, with or without
HIV Med 2008;9:57–61. https://doi.org/10.1111/j.1468-1293.2008.00515.x.
MF59 adjuvant, administered to human immunodeficiency virus type 1-
[9] BC CDC. Influenza (Flu) Vaccine n.d. <http://www.bccdc.ca/health-info/
seropositive and -seronegative adults. Clin Vaccine Immunol 2008;15:253–9.
immunization-vaccines/vaccines-in-bc/influenza-flu-vaccine> (accessed
https://doi.org/10.1128/CVI.00316-07.
December 19, 2016).
[31] Gabutti G, Guido M, Durando P, De Donno A, Quattrocchi M, Bacilieri S, et al.
[10] CDC. HIV/AIDS and the Flu n.d. <https://www.cdc.gov/flu/protect/hiv-flu.htm>
Safety and immunogenicity of conventional subunit and MF59-adjuvanted
(accessed December 19, 2016).
influenza vaccines in human immunodeficiency virus-1-seropositive patients.
[11] NHS. HIV and AIDS - Living with 2015. <http://www.nhs.uk/Conditions/HIV/
J Int Med Res 2005;33:406–16. https://doi.org/10.1177/
Pages/living-with.aspx> (accessed December 19, 2016).
147323000503300406.
[12] Diaz RS, Tenore SB, da Silva MMG, da Cunha CA. A Phase III, randomized study
[32] Tasker SA, Treanor JJ, Paxton WB, Wallace MR. Efficacy of influenza vaccination
to evaluate the immunogenicity and safety of an MF59Ò-adjuvanted A/H1N1
in HIV-infected persons a randomized, double-blind, placebo-controlled trial.
pandemic influenza vaccine in HIV-positive adults. Trials Vaccinol
Ann Internal Med 1999;131:430–3.
2014;3:114–20. https://doi.org/10.1016/j.trivac.2014.07.001.
[33] Leroux-Roels I, Roman F, Forgus S, Maes C, De Boever F, Dramé M, et al.
[13] Mills EJ, Thorlund K, Ioannidis JPA. Demystifying trial networks and network
Priming with AS03 A-adjuvanted H5N1 influenza vaccine improves the
meta-analysis. BMJ 2013;346:f2914. https://doi.org/10.1136/bmj.f2914.
kinetics, magnitude and durability of the immune response after a
[14] Li T, Puhan MA, Vedula SS, Singh S, Dickersin K. Network meta-analysis-highly
heterologous booster vaccination: an open non-randomised extension of a
attractive but more methodological research is needed. BMC Med 2011;9:79.
double-blind randomised primary study. Vaccine 2010;28:849–57. https://
https://doi.org/10.1186/1741-7015-9-79.
doi.org/10.1016/j.vaccine.2009.10.017.
[15] Mills EJ, Ioannidis JPA, Thorlund K, Schünemann HJ, Puhan MA, Guyatt GH.
[34] Plennevaux E, Sheldon E, Blatter M, Reeves-Hoché M-K, Denis M. Immune
How to use an article reporting a multiple treatment comparison meta-
response after a single vaccination against 2009 influenza A H1N1 in USA: a
analysis. JAMA 2012;308:1246–53. https://doi.org/10.1001/2012.jama.11228.
preliminary report of two randomised controlled phase 2 trials. Lancet
[16] Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gotzsche PC, Ioannidis JPA, et al.
2010;375:41–8. https://doi.org/10.1016/S0140-6736(09)62026-2.
The PRISMA statement for reporting systematic reviews and meta-analyses of
[35] Popova L, Smith K, West AH, Wilson PC, James JA, Thompson LF, et al.
studies that evaluate healthcare interventions: explanation and elaboration
Immunodominance of antigenic site B over Site A of hemagglutinin of recent
b2700–b2700. BMJ 2009;339. https://doi.org/10.1136/bmj.b2700.
H3N2 influenza viruses. PLoS One 2012;7:e41895. https://doi.org/10.1371/
[17] Higgins JPT, Altman DG, Gøtzsche PC, Jüni P, Moher D, Oxman AD, et al. The
journal.pone.0041895.
Cochrane collaboration’s tool for assessing risk of bias in randomised trials.
[36] Cooper CL. Pandemic H1N12009 influenza and HIV: a review of natural
BMJ 2011;343:d5928.
history, management and vaccine immunogenicity. Curr Opin Infect Dis
[18] Brooks SP, Gelman A. General methods for monitoring convergence of iterative
2012;25:26–35. https://doi.org/10.1097/QCO.0b013e32834ef56c.
simulations. J Comput Graph Statist 1998;7:434–55.
[37] Kajaste-Rudnitski A, Galli L, Nozza S, Tambussi G, Di Pietro A, Pellicciotta G,
[19] Dias S, Sutton AJ, Ades AE, Welton NJ. Evidence synthesis for decision making
et al. Induction of protective antibody response by MF59-adjuvanted 2009
2: a generalized linear modeling framework for pairwise and network meta-
pandemic A/H1N1v influenza vaccine in HIV-1-infected individuals. AIDS
analysis of randomized controlled trials. Med Decis Making 2013;33:607–17.
2011;25:177–83. https://doi.org/10.1097/QAD.0b013e328341afa8.
https://doi.org/10.1177/0272989X12458724.
[38] Labrosse B, Tourdjman M, Porcher R, LeGoff J, de Lamballerie X, Simon F, et al.
[20] Spiegelhalter DJ, Best NG, Carlin BP, van der Linde A. Bayesian Measures of
Detection of extensive cross-neutralization between pandemic and seasonal
Model Complexity and Fit. J Roy Statist Soc Ser B (Statist Methodol)
A/H1N1 influenza viruses using a pseudotype neutralization assay. PLoS One
2002;64:583–639.
2010;5:e11036. https://doi.org/10.1371/journal.pone.0011036.
[21] Dias S, Welton NJ, Caldwell DM, Ades AE. Checking consistency in mixed
[39] Van Kerkhove MD, Hirve S, Koukounari A, Mounts AW. H1N1pdm serology
treatment comparison meta-analysis. Statist Med 2010;29:932–44. https://
working group. Estimating age-specific cumulative incidence for the, influenza
doi.org/10.1002/sim.3767.
pandemic: a meta-analysis of A(H1N1)pdm09 serological studies from 19
[22] McKittrick N, Frank I, Jacobson JM, White CJ, Kim D, Kappes R, et al. Improved
countries. Influenza Other Respir Viruses 2009;2013(7):872–86. https://doi.
immunogenicity with high-dose seasonal influenza vaccine in HIV-infected
org/10.1111/irv.12074.
persons: a single-center, parallel, randomized trial. Ann Intern Med
[40] Committee for Proprietary Medicinal Products (CPMP). Note for Guidance on
2013;158:19–26. https://doi.org/10.7326/0003-4819-158-1-201301010-
Harmonisation of Requirements for Influenza Vaccines; 1996.
00005.