Influenza Vaccine Effectiveness in the United States During the 2016–2017 Season

Abstract

Background

In recent influenza seasons, the effectiveness of inactivated influenza vaccines against circulating A(H3N2) virus has been lower than against A(H1N1)pdm09 and B viruses, even when circulating viruses remained antigenically similar to vaccine components.

Methods

During the 2016–2017 influenza season, vaccine effectiveness (VE) across age groups and vaccine types was examined among outpatients with acute respiratory illness at 5 US sites using a test-negative design that compared the odds of vaccination among reverse transcription polymerase chain reaction–confirmed influenza positives and negatives.

Results

Among 7083 enrollees, 1342 (19%) tested positive for influenza A(H3N2), 648 (9%) were positive for influenza B (including B/Yamagata, n = 577), and 5040 (71%) were influenza negative. Vaccine effectiveness was 40% (95% confidence interval [CI], 32% to 46%) against any influenza virus, 33% (95% CI, 23% to 41%) against influenza A(H3N2) viruses, and 53% (95% CI, 43% to 61%) against influenza B viruses.

Conclusions

The 2016–2017 influenza vaccines provided moderate protection against any influenza among outpatients but were less protective against influenza A(H3N2) viruses than B viruses. Approaches to improving effectiveness against A(H3N2) viruses are needed.

The 2016–2017 influenza vaccines provided moderate protection against any influenza among outpatients but were less protective against influenza A(H3N2) viruses than B viruses.

Influenza vaccine components are frequently updated due to constant changes in circulating influenza viruses. The emergence of antigenically distinct viruses, referred to as antigenic drift, is often associated with decreased vaccine effectiveness (VE). Most recently, this was observed during the 2014–2015 influenza season when predominant A(H3N2) viruses were antigenically different from the recommended A(H3N2) vaccine reference virus [1, 2]. Vaccine immunogenicity and effectiveness are also affected by age, prior vaccination history, virus type/subtype and vaccine type [3–6]. In recent influenza seasons, effectiveness of inactivated vaccines against illness caused by A(H3N2) viruses has generally been lower than VE against the 2009 pandemic A(H1N1) virus (A(H1N1)pdm09) and type B viruses [3, 7, 8], even when vaccine components were antigenically matched to circulating A(H3N2) viruses.

In the United States, annual studies of seasonal influenza VE have been conducted since the 2004–2005 influenza season, and by the US Influenza Vaccine Effectiveness (Flu VE) Network since the 2008–2009 season. The Flu VE Network has also provided interim (within season) VE estimates since the 2012–2013 season [9]. During the 2016–2017 influenza season in the United States, influenza A(H3N2) viruses predominated with co-circulation of influenza B viruses [10]. Interim VE estimates including data from November, 2016, through February, 2017, indicated moderate effectiveness of 48% (95% confidence interval [CI], 37% to 57%) against medically-attended influenza due to any virus type, and 43% (95% CI, 29% to 54%) against influenza A(H3N2)-associated illness [11]. Here, we report updated VE estimates for the complete 2016–2017 season, including VE estimates by virus type/subtype, age group and prior season vaccination.

METHODS

Study Population

Details of the US Flu VE Network have been published previously [2, 6–8, 12]. Briefly, study participants were recruited during the 2016–2017 influenza season (defined by local influenza surveillance) at participating healthcare facilities in 5 research sites in Michigan, Pennsylvania, Texas, Washington, and Wisconsin. Study staff enrolled patients aged ≥6 months seeking outpatient care for acute respiratory illness with a cough of 7 or fewer days’ duration at the time of the medical visit. Patients who had received an antiviral medication in the 7 days before enrollment, who were younger than 6 months of age as of September 1, 2016, or who had enrolled in the study within the previous 14 days were ineligible. Staff interviewed patients for demographic data, current health status, symptoms, and reported receipt of 2015–2016 and 2016–2017 influenza vaccine. Presence of any preexisting chronic health conditions associated with increased risk of severe influenza [6, 13] was defined as an International Classification of Diseases code (Version 10 [ICD-10]) for a high-risk condition assigned to a medical encounter in the year before study enrollment.

The study was approved by institutional review boards at each participating site and the Centers for Disease Control and Prevention (CDC).

Influenza Vaccination History

The virus strains for the 2016–2017 Northern Hemisphere influenza vaccine were an A/California/7/2009 (H1N1)-like virus, an A/Hong Kong/4801/2014 (H3N2)-like virus, and a B/Brisbane/60/2008-like virus (Victoria lineage) in the trivalent formulation, plus a B/Phuket/3073/2013-like virus (Yamagata lineage) in the quadrivalent formulation [14]. For all US Flu VE Network sites, a participant’s vaccination status was based on documented receipt of 2016–2017 influenza vaccine in electronic immunization records, including medical records, state immunization registries, and employee health records. In addition, at 4 sites (excluding Wisconsin), we considered as vaccinated those patients aged ≥9 years who reported timing and place of vaccination without documented receipt [12].

Laboratory Methods

After obtaining informed consent, study staff collected nasal and throat swabs from patients aged ≥2 years or a nasal swab only from children aged <2 years. Specimens were tested by reverse-transcription polymerase chain reaction (RT-PCR) with primers and probes provided by CDC to detect influenza, including additional testing to determine virus subtype or lineage. All network laboratories completed proficiency testing. Patients who tested positive for influenza were designated as case patients, and those who tested negative were designated as non-case patients.

Study sites sent a subset of influenza–positive specimens to CDC for antigenic and/or genetic characterization. We used whole-genome sequencing to obtain full-length hemagglutinin (HA) sequences from original specimen samples and classified viruses into genetic groups based on phylogenetic analyses. We assessed antigenic similarity between viral isolates and vaccine reference strains using the hemagglutination-inhibition assay [15] (for all vaccine components) or a focus reduction assay adapted from Matrosovich et al [16] (for A(H3N2) viruses only) with ferret antisera raised against egg-propagated A(H1N1)pdm09 and B vaccine strains and against egg- and cell-propagated A(H3N2) reference viruses. Circulating viruses were considered antigenically similar to cell culture-derived vaccine reference viruses if hemagglutination-inhibition titers using postinfection ferret reference antisera were within 4-fold of the homologous titer [17].

Statistical Analysis

We excluded from primary analyses patients with inconclusive RT-PCR results, patients with documented or reported vaccination ≤14 days before illness onset, children aged <9 years who were partially vaccinated according to the 2016–2017 Advisory Committee on Immunization Practices (ACIP) recommendations [14], and influenza-negative patients who were enrolled outside the periods of local influenza circulation. For the remaining enrollees, we calculated descriptive statistics separately for influenza case patients and non-case patients, including medians for continuous variables and distributions for categorical variables.

We used a test-negative design to estimate VE, which compared the odds of influenza vaccination among participants with RT-PCR-positive influenza (cases) to the odds of influenza vaccination among participants who were negative for influenza (non-cases) using a logistic regression model [18]. We estimated VE against any influenza, by influenza virus type A or B, or by virus subtype/lineage in separate models and stratified models by participant age. To estimate VE by vaccine type, we compared odds of influenza among patients who received 1 vaccine type to unvaccinated patients, excluding patients who received other vaccine types. To investigate the effects of prior season (2015–2016) vaccination on 2016–2017 VE, we limited analyses to enrollees aged ≥9 years with more than 1 year of electronic records and estimated VE for combinations of current (2016–2017) and prior (2015–2016) season vaccination with an interaction term in the model, as previously described [7, 8]. In addition, we also stratified enrollees by documented 2015–2016 vaccination status and estimated 2016–2017 VE in separate models. We adjusted all logistic regression models, a priori, for network site, calendar time (in 11 bi-weekly intervals), participant age (using a tail-restricted natural cubic spline with 5 percentile knots using month or year of age at enrollment), and high-risk status (any high-risk ICD-10 code in the medical record versus none). Additional covariates were retained if their inclusion changed the odds ratio by at least 5%.

To test the robustness of VE estimates, we included additional analyses using different definitions of vaccination (self-report only, electronic immunization records plus plausible self-report [primary analysis], or electronic immunization records only), restricted VE analyses to patients enrolled <5 days after illness onset, and estimated VE for children aged <9 years including partially vaccinated children.

RESULTS

From November 28, 2016, through April 14, 2017, we enrolled 7410 patients who presented to ambulatory care clinics for acute respiratory illness. We excluded 327 participants (4%) from the primary analysis, including 17 patients with inconclusive RT-PCR results, 85 patients who had been vaccinated less than 14 days before illness onset, 162 partially vaccinated children younger than 9 years of age, and 63 influenza virus-negative patients enrolled outside the periods of influenza circulation. Of the remaining 7083 enrolled patients, 2043 (29%) tested positive for influenza virus (case patients) with the number of cases peaking in epidemiologic week 7, 2017 (Figure 1). The influenza case patients included 1342 (66%) who were infected with A(H3N2), 26 (1%) with A(H1N1)pdm09 and 648 (32%) infected with influenza B viruses (including 577 [89%] infected with B/Yamagata and 63 [10%] infected with B/Victoria-lineage viruses). For VE estimates by influenza virus type/subtype, we excluded 26 (1%) case patients infected with A(H1N1)pdm09, 5 coinfected with more than one influenza virus type/subtype, 31 infected with influenza A virus with undetermined subtype, and 8 infected with influenza B virus of undetermined lineage.

Figure 1.

Number of enrolled patients with acute respiratory illness testing positive for influenza A(H3N2) (red bars) and influenza B (gray bars), and percent of enrollees testing positive for influenza (black line) by epidemiologic week during the 2016–2017 influenza season.

The proportion of patients with influenza differed by study site, age group, race/ethnicity, presence of any high-risk condition, and interval from illness onset to enrollment (Table 1). Case patients were less likely than non-case patients to have asthma or another pulmonary high-risk condition, to have any high-risk condition, or to have received 2016–2017 influenza vaccine. Vaccination status differed by study site, sex, age group, race/ethnicity, presence of chronic underlying medical conditions, and exposure to cigarette smoke (Table 2). The proportion of ARI patients vaccinated with 2016–2017 seasonal influenza vaccine was 42% among influenza-positive case patients and 52% among non-case patients (Table 3). Among 3190 patients aged ≥6 months with documented receipt of vaccine of known type, 65% received a quadrivalent inactivated influenza vaccine (IIV4), whereas 35% received a trivalent vaccine (IIV3). Among 809 vaccinated patients aged ≥65 years, vaccine type was recorded for 754 (93%) patients: 53% received high-dose inactivated influenza vaccine (HD-IIV3), 25% received standard-dose IIV4, 21% received standard-dose IIV3, and <1% received adjuvanted IIV3.

Table 1.

Demographic and Clinical Characteristics of Enrolled Patients, by Influenza Virus Infection—2016–2017 Influenza Season

Characteristic	Test Result Status				P-valuea
	Influenza Positive		Influenza Negative
	n = 2043	Col %	n = 5040	Col %
Study site					<.001
Michigan	361	18	790	16
Pennsylvania	383	19	810	16
Texas	255	12	1072	21
Washington	461	23	1097	22
Wisconsin	583	29	1271	25
Sex					<.001
Female	1106	54	2984	59
Male	937	46	2056	41
Age group, years					<.001
6 mo–8	338	17	1181	23
9–17	403	20	608	12
18–49	528	26	1637	32
50–64	447	22	915	18
≥65	327	16	699	14
Race/ethnicityb					.002
White, non-Hispanic	1580	77	3757	75
Black, non-Hispanic	165	8	385	8
Other, non-Hispanic	142	7	429	9
Hispanic	147	7	460	9
Unknown	9	0	9	0
High-risk conditionc
Any	964	47	2572	51	.003
Asthma/pulmonary	367	18	1127	22	<.001
Cardiovascular	198	10	563	11	.07
Diabetes	159	8	409	8	.43
Morbid obesityd	118	6	349	7	.10
Other	631	31	1626	32	.32
Interval from onset to enrollment					<.001
<3 days	781	38	1499	30
3–4 days	834	41	1956	39
5–7 days	428	21	1585	31
Self-rated general health statuse					.004
Excellent, very good	1434	70	3340	66
Good	474	23	1287	26
Fair, poor	134	7	411	8
Missing	1	0	2	0
Self/household exposure to tobacco smokef					<.001
Yes	362	18	1075	21
No	1677	82	3958	79
Missing	4	0	7	0
Receipt of 2016–2017 influenza vaccineg	866	42	2629	52	<.001
Reported current health assessment score, median (IQR)h	50	(40–70)	60	(45–75)	<.001

Data are no. (%) of subjects, unless otherwise indicated.

Abbreviations: ACIP, Advisory Committee on Immunization Practices; BMI, body mass index; Col %, column percent; ICD, International Classification of Diseases; IQR, interquartile range.

^a P-value comes from χ² comparing frequency of participants testing influenza-positive and frequency of participants testing influenza-negative by characteristic.

^bData were missing for 18 enrollees.

^cPresence of a high-risk health condition is defined as the presence of ≥1 medical record-documented ICD-10 high risk code from October 1, 2015, to enrollment, as defined by the ACIP guidance for conditions that increase risk for complications from influenza.

^d Defined as a body mass index (BMI) ≥ 40 calculated as kg/(m²) from height and weight recorded in the electronic medical record. Calculated for adults aged ≥18 years only.

^eData were missing for 3 enrollees.

^fData on exposure to tobacco smoke were missing for 11 enrollees.

^gDocumented vaccination or plausible self-report of location of vaccination ≥14 days prior to illness onset.

^hPossible values range from 1 (the worst) to 100 (the best). Data were missing for 6 enrollees.

Table 2.

Characteristics of Enrolled Patients, by Vaccination Status, 2016–2017

Characteristic	Vaccination Status				P-valuea
	Vaccinated		Unvaccinated
	No.	(%)	No.	(%)
Overall	3495	49	3595	51
Study site					<.001
Michigan	597	52	554	48
Pennsylvania	599	50	601	50
Texas	578	44	749	56
Washington	860	55	698	45
Wisconsin	861	46	993	54
Sex					<.001
Female	2134	52	1959	48
Male	1361	45	1636	55
Age group, years					<.001
6 mo–8	674	44	850	56
9–17	364	36	647	64
18–49	909	42	1257	58
50–64	739	54	624	46
≥65	809	79	217	21
Race/ethnicityb					<.001
White, non-Hispanic	2714	51	2628	49
Black, non-Hispanic	229	42	323	58
Other, non-Hispanic	282	49	289	51
Hispanic	265	44	342	56
Unknown	5	28	13	72
High-risk conditionc					<.001
Any	2118	60	1401	40
Asthma/pulmonary	887	59	596	41	<.001
Cardiovascular	544	71	212	29	<.001
Diabetes	402	71	164	29	<.001
Morbid obesityd	292	63	173	37	<.001
Other	1455	64	789	36	<.001
Self/household exposure to tobacco smokee					<.001
Yes	529	37	911	63
No	2962	53	2677	47
Missing	4	36	7	64

Data are no. (%) of subjects, unless otherwise indicated.

Abbreviations: ACIP, Advisory Committee on Immunization Practices; BMI, body mass index; ICD, International Classification of Diseases.

^a P-value comes from χ² comparing frequency of vaccinated participants and frequency of unvaccinated participants by characteristic.

^bData were missing for 18 enrollees.

^cPresence of a high-risk health condition is defined as the presence of ≥1 medical record-documented ICD-10 high risk code from October 1, 2015, to enrollment, as defined by the ACIP guidance for conditions that increase risk for complications from influenza.

^dDefined as a body mass index (BMI) ≥ 40 calculated as kg/(m²) from height and weight recorded in the electronic medical record. Calculated for adults aged ≥18 years only.

^eData on exposure to smoke were missing for 11 enrollees.

Table 3.

Unadjusted and Adjusted Estimates of Influenza Vaccine Effectiveness, Overall and Stratified by Age and Virus Type, Subtype, or Lineage—United States, 2016–2017 Influenza Season

Influenza (Sub)Type/Age Group	Influenza Positive		Influenza Negative		Unadjusted		Adjusteda
	No. Vaccinated/Total	%	No. Vaccinated/Total	%	VE %	(95% CI)	VE %	(95% CI)
Influenza A and B
All ages ≥6 months	866/2043	42	2629/5040	52	33	(25 to 39)	40	(32 to 46)
6 months–8 years	93/338	28	581/1181	49	61	(49 to 70)	57	(43 to 68)
9–17 years	121/403	30	243/608	40	36	(16 to 51)	36	(15 to 52)
18–49 years	198/528	38	711/1637	43	22	(4 to 36)	19	(0 to 34)
50–64 years	206/447	46	533/915	58	39	(23 to 51)	40	(24 to 53)
≥65 years	248/327	76	561/699	80	23	(−5 to 44)	20	(−11 to 43)
Influenza A(H3N2)
All ages ≥6 months	604/1342	45	2629/5040	52	25	(15 to 33)	33	(23 to 41)
6 months–8 years	60/190	32	581/1181	49	52	(34 to 66)	49	(28 to 64)
9–17 years	76/259	29	243/608	40	38	(15 to 54)	33	(7 to 52)
18–49 years	138/351	39	711/1637	43	16	(−7 to 33)	13	(−11 to 32)
50–64 years	148/304	49	533/915	58	32	(12 to 48)	31	(9 to 47)
≥65 years	182/238	77	561/699	80	20	(−14 to 44)	21	(−15 to 45)
Influenza B
All ages ≥6 months	236/648	36	2629/5040	52	47	(38 to 56)	53	(43 to 61)
Influenza B/Yamagata
All ages ≥6 months	213/577	37	2629/5040	52	46	(36 to 55)	52	(42 to 61)
6 months–8 years	22/113	20	581/1181	49	75	(60 to 85)	73	(54 to 84)
9–17 years	39/122	32	243/608	40	29	(−7 to 53)	37	(1 to 60)
18–49 years	46/137	34	711/1637	43	34	(5 to 54)	30	(−3 to 53)
50–64 years	53/129	41	533/915	58	50	(27 to 66)	63	(42 to 76)
≥65 years	53/76	70	561/699	80	43	(4 to 66)	34	(−24 to 65)
Influenza B/Victoria
All ages ≥6 months	21/63	33	2629/5040	52	54	(22 to 73)	56	(23 to 75)

Vaccine effectiveness was estimated by comparing the odds of vaccination influenza case patients and non-case patients and calculated as 100 × (1 – odds ratio) in logistic regression models. Vaccination includes any 2016–2017 inactivated influenza vaccine.

Abbreviations: CI, confidence interval; REF, referent group (1.0); VE, vaccine effectiveness

^aModels were adjusted for study site, patient age in months, presence of any high-risk health condition and calendar time (two week intervals).

Antigenic and Genetic Characterization of Viruses

In all, 1038 influenza viruses (939 A(H3N2), 10 A(H1N1)pdm09 and 89 B viruses, including 44 B/Yamagata lineage) were genetically characterized from US Flu VE case patients. Among 939 A(H3N2) viruses analyzed, the HA gene of 886 (94%) belonged to the 3C.2a genetic group (the same genetic group as the A/Hong Kong/4801/2014 vaccine reference virus), including 616 (66%) belonging to the 3C.2a1 subgroup, whereras 53 (6%) belonged to the 3C.3a genetic group (represented by the 2015–2016 vaccine reference strain A/Switzerland/9715293/2013(H3N2)). A subset of 63 A(H3N2) viruses were also antigenically characterized; 61 (97%) were antigenically similar to the cell-propagated A(H3N2) vaccine reference strain by HI or FRA. However, only 24 (38%) of the 63 A(H3N2) viruses tested were antigenically similar to the egg-propagated A/Hong Kong/4801/2014(H3N2) vaccine reference strain by HI or FRA. All antigenically characterized influenza A(H1N1)pdm09 viruses (n = 9), B/Victoria-lineage viruses (n = 12), and B/Yamagata-lineage viruses (n = 29) were similar to vaccine reference viruses.

Overall Vaccine Effectiveness

After adjustment for potential confounders, the estimated VE against any influenza virus was 40% (95% CI, 32 to 46) (Table 3). Overall protection was similar for standard-dose IIV3 vaccines (VE, 37%; 95% CI, 22 to 49) and IIV4 vaccines (VE, 39%; 95% CI, 30 to 47). VE against any influenza virus ranged from 36% (95% CI, 15 to 52) among 9–17 year olds, to 57% (95% CI, 43 to 68) among fully vaccinated children aged 6 months to 8 years, and was not statistically significant among patients aged 18–49 years (VE, 19%; 95% CI, 0 to 34) or ≥65 years (VE, 20%; 95% CI, −11 to 43). Among patients aged ≥65 years, point estimates for VE against any influenza were higher for standard dose IIV3 (VE, 38%; 95% CI, −10 to 65) compared to high dose IIV3 (VE, 15%; 95% CI, −32 to 45) or standard-dose IIV4 (VE, 18%; 95% CI, −32 to 49), although none indicated statistically significant protection.

VE against A(H3N2) viruses was 33% (95% CI, 23 to 41), ranging from 13% (95% CI, −11 to 32) among 18–49 year olds to 49% (95% CI, 28 to 64) among those aged 6 months to 8 years; VE against A(H3N2) viruses was 21% (95% CI, −15 to 45) among patients aged ≥65 years. Against any influenza B viruses, VE was 53% (95% CI, 43 to 61). VE against influenza B/Yamagata viruses (not included in trivalent vaccines) was nonsignificantly higher for IIV4 (VE, 59%; 95% CI, 48 to 68) than for standard-dose IIV3 (VE, 34%; 95% CI, 9 to 51). Against B/Victoria-lineage viruses, VE was 56% (95% CI, 23 to 75).

In analyses using definitions of vaccination status other than documented doses or plausible self-reported vaccination, estimates of VE against any influenza virus among patients aged ≥6 months were similar based on documented doses only (VE, 39%; 95% CI, 32 to 46), electronic immunization records excluding patients with plausible self-report of vaccination (VE, 40%; 95% CI 33 to 47) and vaccination by self-report only excluding the WI network site (VE, 40%; 95% CI 32 to 48). Restricting analysis to patients aged ≥6 months enrolled <5 days after illness onset also resulted in similar overall VE (VE, 38%; 95% CI 30 to 46), and inclusion of partially vaccinated children resulted in similar VE among patients aged 6 month to 8 years (58%; 95% CI, 45 to 68).

Effects of Previous Influenza Vaccination

Among patients aged ≥9 years vaccinated in 2016–2017, 72% had received a 2015–2016 vaccine. Compared to patients not vaccinated in either 2015–2016 (prior season) or 2016–2017 (current season), VE against A(H3N2)-related illness was 35% (95% CI, 18 to 48) for those vaccinated only in 2016–2017 and 26% (95% CI, 12 to 38) for those vaccinated in both current and prior season (Table 4); this difference was not statistically significant (P-value, .10). Vaccination in 2015–2016 only (without current vaccination) offered no significant protection (VE, 14%; 95% CI, −10 to 33). Against any influenza B virus, VE estimates were nonsignificantly higher for patients vaccinated in the current season only (VE, 54%; 95% CI, 35 to 68) than for patients vaccinated in both seasons (VE, 42%; 95% CI, 26 to 55) compared to those unvaccinated in both seasons (P-value, .07), whereas prior season vaccination only offered no significant protection (Table 4). We observed the same nonsignificant effects for influenza B/Yamagata viruses (data not shown). In analyses stratified by prior season (2015–2016) vaccination status, point estimates for 2016–2017 VE against A(H3N2) and B viruses were lower among patients vaccinated the prior season, but differences were not statistically significant (Supplemental Table).

Table 4.

Vaccine Effectiveness Against Influenza A(H3N2) and B Virus Related Illness According to Combinations of Prior Season (2015–2016) and Current Season (2016–2017) Influenza Vaccination Status Among Patients Aged ≥9 Years

	Influenza-positive Cases		Influenza-negative Controls		Unadjusted		Adjusteda
	No. Cases/Row Total	(%)	No. Controls/Row Total	(%)	VE %	(95% CI)	VE %	(95% CI)
Influenza A(H3N2)b
Vaccinated current 2016–2017 only	130/708	18.4	578/708	81.6	35%	(19 to 48)	35%	(18 to 48)
Vaccinated current 2016–2017 and prior 2015–2016	399/1804	22.1	1405/1804	77.9	18%	(4 to 29)	26%	(12 to 38)
Vaccinated prior 2015–2016 only	106/444	23.9	338/444	76.1	9%	(−16 to 29)	14%	(−10 to 33)
Not vaccinated either 2015–2016 or 2016–2017	467/1820	25.7	1353/1820	74.3	REF		REF
Influenza B c
Vaccinated current 2016–2017 only	49/627	7.8	578/627	92.2	53%	(35 to 66)	54%	(35 to 68)
Vaccinated current 2016–2017 and prior 2015–2016	151/1556	9.7	1405/1556	90.3	40%	(26 to 52)	42%	(26 to 55)
Vaccinated prior 2015–2016 only	47/385	12.2	338/385	87.8	23%	(−8 to 45)	22%	(−12 to 46)
Not vaccinated either 2015–2016 or 2016–2017	243/1596	15.2	1353/1596	84.8	REF		REF

Vaccine effectiveness was estimated by comparing the odds of influenza for each category of current or prior vaccination versus the odds of influenza among enrollees not vaccinated in either current (2016–2017) or prior season (2015–2016), and calculated as 100 × (1 – odds ratio) in logistic regression models. Vaccination includes any inactivated influenza vaccine for each season.

Abbreviations: CI, confidence interval; REF, referent group (1.0); VE, vaccine effectiveness.

^aModels were adjusted for study site, patient age in months, presence of any high-risk health condition and calendar time (two week intervals).

^bThe P-value for interaction of prior (2015–2016) and current (2016–2017) season vaccination was 0.10.

^cThe P-value for interaction of prior (2015–2016) and current (2016–2017) season vaccination was 0.07.

DISCUSSION

Overall, estimates from the US Flu VE Network indicated that influenza vaccination reduced influenza-associated illness among outpatients by 40% during the 2016–2017 season. Based on this level of protection, CDC estimated that vaccination prevented an estimated 84700 influenza-related hospitalizations and 2.6 million outpatient illnesses during the 2016–2017 season [19], comparable to estimates of 39000 to 86000 influenza-associated hospitalizations prevented in several recent influenza seasons [19, 20]. However, protection observed against A(H3N2) viruses (33%) was lower than against influenza B viruses (53%). In addition, VE varied by age group; estimates of VE against A(H3N2) viruses were lower among adults (especially those aged 18–49 years) than young children. Also, VE estimates were nonsignificantly lower among persons vaccinated the previous season. Annual studies of VE continue to be important to evaluate protection conferred by different vaccine components and highlight the need for improved influenza vaccines.

The 2016–2017 VE estimates against A(H3N2) viruses in the United States were similar to end-of-season estimates from the United Kingdom [21] and interim estimates against A(H3N2) viruses from the US Flu VE Network [11], several European countries [22], and Canada [23] for the 2016–2017 Northern Hemisphere season. The 2017 Southern Hemisphere vaccines contained the same A(H3N2) vaccine component. The interim estimate of VE against A(H3N2) viruses in Australia was only 10% and not statistically significant [24]. In all of these studies, the predominant A(H3N2) viruses belonged to genetic group 3C.2a, including subgroup 3C.2a1. As in the United States, antigenic characterization of A(H3N2) viruses showed that the majority of circulating viruses were well-inhibited by ferret antisera to cell-grown A(H3N2) 3C.2a reference viruses, including A/Hong Kong/4801/2014, suggesting limited antigenic change among circulating A(H3N2) viruses. Several influenza laboratories also reported that antisera raised against egg-adapted A/Hong Kong/4801/2014 used for the majority of influenza vaccine in the United States resulted in lower antibody titers against circulating A(H3N2) viruses [24, 25]. These results suggest that immune responses to egg-based vaccines may have been less effective against circulating A(H3N2) viruses [26, 27]. However, egg-adaptive changes do not explain higher VE against A(H3N2) viruses among young children compared to adult age groups. It is unknown whether licensed vaccines that do not require egg-adapted candidate vaccine viruses, such as those produced using recombinant technology or cell-grown viruses, would provide improved protection. One study suggested that a recombinant HA protein vaccine was 30% more effective relative to a standard egg-grown vaccine during the 2014–2015 influenza season; however, this vaccine also contains a higher antigen dose than standard vaccines [28]. Thus, the increase in VE may not be attributed solely to vaccine manufacture.

In addition to egg-adaptive changes, there are several additional factors that could be contributing to lower VE against A(H3N2) viruses. Since the emergence of A(H3N2) viruses in humans in 1968, the HA surface protein has become increasing glycosylated, which facilitates viral escape from the human immune response [29]. Acquisition of a glycosylation motif in a rare site on the HA globular head of group 3C.2a viruses aided the virus’ evasion of the immune response against previous A(H3N2) viruses by shielding an immunodominant HA antigenic site B [27]. Interestingly, loss of this glycosylation site (at HA position 158–160) in the egg-adapted 3C.2a vaccine virus exposes an immunodominant antigenic site present in earlier A(H3N2) viruses, possibly misdirecting immune responses towards previously recognized epitopes. Genetic changes in other HA antigenic sites may have contributed to 3C.2a viruses’ ability to evade immune responses. In addition, prior infections with earlier A(H3N2) viruses and vaccination likely affect immune responses generated against vaccine components [27] and may contribute to observed differences in VE by age group. During 2016–2017, VE against A(H3N2) was not significantly lower among patients vaccinated in two consecutive seasons compared to those vaccinated in 2016–2017 only. Studies in the United Kingdom [21] and Australia [24] also observed a trend toward lower VE against A(H3N2) viruses among adult patients vaccinated in 2015–2016 and 2016–2017 compared to patients vaccinated only in 2016–2017, as we observed in this study. Additional studies are needed to both identify optimal vaccine candidates that elicit protective responses against dominant genetic variants and minimize egg adaptive changes, and to better understand the immune response to initial and subsequent infections to A(H3N2) viruses and vaccination in all age groups.

The estimation of VE is subject to several limitations. Misclassification of vaccination status likely occurred, as vaccination was not documented for all adults in the main analysis. Consistency of sensitivity analyses and vaccine-type specific estimates using only documented vaccinations support the findings of the main analyses. Analyses by vaccine type may be confounded by factors associated with the choice of vaccine type, including factors associated with the individual patient or healthcare system. Single season analyses have limited power to detect differences in VE by vaccine type and depend on vaccine uptake. Finally, patient characteristics, vaccine use and influenza virus circulation can affect VE estimates and differ among study sites from the US population.

Ongoing VE studies are important to evaluate how well each annual seasonal vaccine and new vaccine component protect against influenza caused by circulating viruses. Data from VE studies were among several factors that contributed to an updated A(H3N2) vaccine component (A/Singapore/INFIMH-16–0019/2016) during the September 2017 WHO Southern Hemisphere vaccine consultation meeting [30]. With increased use of more immunogenic vaccines for the elderly and non-egg based vaccines in the general population, VE studies are needed to determine protection against circulating influenza viruses and assess differences in VE by vaccine type or formulation. In addition, VE studies have highlighted the need for better influenza vaccines (especially against A(H3N2) viruses) for all ages, as well as gaps in our understanding of immunologic and virologic factors that contribute to vaccine protection.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

US Flu VE investigators. University of Michigan School of Public Health: Joshua G. Petrie, Ryan E. Malosh, E. J. McSpadden, Hannah E. Segaloff, Caroline K. Cheng, Rachel Truscon, and Emileigh Johnson; Henry Ford Health System: Lois E. Lamerato; Marshfield Clinic Research Institute: Lynn C. Ivacic, Jennifer P. King, Jennifer K. Meece, Madalyn M. Palmquist and Sherri A. Guzinski; Baylor Scott and White Health: Anne Robertson, Ashley Kossie, Michael Smith, Vanessa Hoelscher, Lydia Clipper, Kimberley Walker, Marcus Volz, Arundhati Rao, Robert Fader, Yolanda Munoz-Maldonado and Michael Reis; University of Pittsburgh: John V. Williams, Goundappa K. Balasubramani, Evelyn C. Reis, Heather Eng, Samantha Ford, Todd M. Bear, Edmund M. Ricci, Robert W. Hickey, Krissy K. Moehling, Jonathan M. Raviotta, Theresa M. Sax, Michael Susick, Monika Johnson, Rose Azrak; Centers for Disease Control and Prevention: LaShondra Berman, Angie Foust, Wendy Sessions, Juliana DaSilva, Thomas Stark, John Barnes.

Acknowledgments. University of Michigan School of Public Health: Anne Kaniclides, Joey Lundgren, Erika Chick, Lindsey Benisatto, Tosca Le, and Dexter Hobdy; Henry Ford Health System: Heather R. Lipkovich, Nishat Islam, Michelle Groesbeck, Shirley Zhang, Andrea Lee, Kristyn Brundidge, Christina Rincon, Stephanie Haralson, Jennifer Hessen, and Ahn Trinh; Marshfield Clinic Research Institute: Elizabeth Armagost, Deanna Cole, Terry Foss, Dyan Friemoth, Katherine Graebel-Khandakani, Linda Heeren, Tami Johnson, Tara Johnson, Nicole Kaiser, Diane Kohnhorst, Sarah Kopitzke, Ariel Marcoe, Karen McGreevey, Vicki Moon, Suellyn Murray, Rebecca Pilsner, DeeAnn Polacek, Emily Redmond, Miriah Rotar, Carla Rottscheit, Jacklyn Salzwedel, Samantha Smith, Sandra Strey, Jane Wesely, Jennifer Anderson, Klevi Hoxha, Tammy Koepel, Nan Pan, Annie Steinmetz, Gregg Greenwald; Baylor Scott and White Health: Crystal Hodges, Teresa Ponder, Ineshia Jackson, Deborah Furze, Martha Zayed, Melissa Zdroik, Kevin Dunlap, Mary Kylberg, Lea Mallett, Hania Wehbe-Janek, Madhava Beeram, Jennifer Thomas, Jaime Walkowiak, Jeremy Ray, Renee Day, Deborah Price, Jennifer Fox and Robert Probe; University of Pittsburgh: Donald S. Burke, MD, Edward Garafolo, MD, Philip Iozzi, MD, Donald B. Middleton, MD, Joe Suyama MD, Leonard Urbanski, MD, Stephen Wisniewski, PhD, Bret Rosenblum MD; Kaiser Permanente Washington Health Research Institute: C. Hallie Phillips, Stacie Wellwood, Lawrence Madziwa, Matt Nguyen, Erika Kiniry, Suzie Park, Julia Anderson; Centers for Disease Control and Prevention: Erin Burns, Jackie Katz, Daniel Jernigan, Dave Wentworth, Mark Thompson, Jerome Tokars.

Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Funding. This work was supported by the Centers for Disease Control and Prevention through cooperative agreements with the University of Michigan (1U01 IP001034), Kaiser Permanente Washington Research Institute (1U01 IP001037), Marshfield Clinic Research Institute (1U01 IP001038), University of Pittsburgh (1U01 IP001035), and Baylor Scott and White Healthcare (1U01 IP001039). At the University of Pittsburgh, the project was also supported by the National Institutes of Health through grant UL1TR001857.

Potential conflicts of interest. M. G. reports grants from CDC, during the conduct of the study; grants from CDC, grants from MedImmune/AstraZeneca, grants from Pfizer, outside the submitted work. M. P. N. reports grants from CDC, during the conduct of the study; grants from Pfizer, Inc., grants from Merck & Co., Inc., outside the submitted work. L. A. J. reports grants from US CDC, during the conduct of the study; grants from Sanofi Pasteur, outside the submitted work. R. K. Z. reports grants from CDC, during the conduct of the study; grants from Merck & Co, grants from Pfizer Inc, grants from Sanofi Pasteur, outside the submitted work. A. S. M. reports personal fees from Sanofi Pasteur, personal fees from Seqirus, outside the submitted work. K. M. reports grants from CDC, during the conduct of the study; other from MedImmune/AstraZeneca, outside the submitted work. All other authors have no conflicts of interest to report. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.