Relative and Absolute Effectiveness of High-Dose and Standard-Dose Influenza Vaccine Against Influenza-Related Hospitalization Among Older Adults—United States, 2015–2017

Abstract

Background

Seasonal influenza causes substantial morbidity and mortality in older adults. High-dose inactivated influenza vaccine (HD-IIV), with increased antigen content compared to standard-dose influenza vaccines (SD-IIV), is licensed for use in people aged ≥65 years. We sought to evaluate the effectiveness of HD-IIV and SD-IIV for prevention of influenza-associated hospitalizations.

Methods

Hospitalized patients with acute respiratory illness were enrolled in an observational vaccine effectiveness study at 8 hospitals in the United States Hospitalized Adult Influenza Vaccine Effectiveness Network during the 2015–2016 and 2016–2017 influenza seasons. Enrolled patients were tested for influenza, and receipt of influenza vaccine by type was recorded. Effectiveness of SD-IIV and HD-IIV was estimated using a test-negative design (comparing odds of influenza among vaccinated and unvaccinated patients). Relative effectiveness of SD-IIV and HD-IIV was estimated using logistic regression.

Results

Among 1487 enrolled patients aged ≥65 years, 1107 (74%) were vaccinated; 622 (56%) received HD-IIV, and 485 (44%) received SD-IIV. Overall, 277 (19%) tested positive for influenza, including 98 (16%) who received HD-IIV, 87 (18%) who received SD-IIV, and 92 (24%) who were unvaccinated. After adjusting for confounding variables, effectiveness of SD-IIV was 6% (95% confidence interval [CI] −42%, 38%) and that of HD-IIV was 32% (95% CI −3%, 54%), for a relative effectiveness of HD-IIV versus SD-IIV of 27% (95% CI −1%, 48%).

Conclusions

During 2 US influenza seasons, vaccine effectiveness was low to moderate for prevention of influenza hospitalization among adults aged ≥65 years. High-dose vaccine offered greater effectiveness. None of these findings were statistically significant.

In this observational study during two US influenza seasons, vaccine effectiveness was low to moderate for prevention of influenza hospitalization among adults aged ≥65 years. High-dose vaccine offered greater effectiveness, but this finding was not statistically significant.

Annual influenza epidemics exact a harsh toll on older Americans. Each year, hundreds of thousands of older Americans are hospitalized with influenza [1, 2]. It is estimated that 7−10% of these patients ultimately die from their influenza illnesses [2, 3]. Among those who survive, a substantial proportion is left debilitated and unable to live independently [4–7]. Vaccination is the cornerstone of influenza prevention, and older Americans have long been a priority group for vaccination in the United States [8, 9]. Because older adults mount less robust responses to immunization [10–12], efforts have been made to enhance immunogenicity and improve vaccine response in older adults, including the licensure in the United States of high-dose inactivated influenza vaccine (HD-IIV), with 4 times the antigen content of standard-dose inactivated influenza vaccines (SD-IIV) [13]. In prelicensure immunogenicity studies and a large comparative efficacy trial conducted during the 2011–2013 influenza seasons, HD-IIV provided higher antibody titers and improved protection against clinical influenza illness than SD-IIV [14–16]. Since then, a cluster randomized trial and several observational cohort studies have demonstrated benefit of HD-IIV compared to SD-IIV, with considerable variation in results [17]. Use of HD-IIV has expanded rapidly in the United States [18, 19], providing the opportunity to further evaluate protection provided by this vaccine in adults against hospitalized influenza due to different virus types/subtypes over multiple seasons. In this study, we sought to assess the effectiveness of HD-IIV compared with SD-IIV against laboratory-confirmed influenza-related hospitalizations among older adults during the 2015–2016 and 2016–2017 US influenza seasons.

METHODS

Enrollment

We analyzed data from the US Hospitalized Adult Influenza Vaccine Effectiveness Network (HAIVEN), an observational study of vaccine effectiveness (VE) against laboratory-confirmed, hospitalized influenza conducted at study sites in 4 states. HAIVEN methods have been described previously [20]. Briefly, HAIVEN used a test-negative design with active case finding at 8 hospitals in Michigan (MI), Pennsylvania (PA), Tennessee (TN), and Texas (TX). Study staff reviewed daily admissions to identify patients meeting prespecified eligibility criteria. Patients aged ≥18 years were eligible if they had evidence of acute respiratory infection (ARI) based on review of chief complaints, admitting diagnosis, and summary of initial clinical evaluations. Participating patients had respiratory specimens for influenza testing collected ≤10 days after illness onset and ≤72 hours after admission.

Study Design

We analyzed data from participants aged ≥65 years enrolled in HAIVEN during the 2015–2016 and 2016–2017 influenza seasons (those enrolled during 2015–2016 were included in our previous report of VE for the full HAIVEN 2015–2016 study sample [20]). After providing consent, each participant or his/her surrogate provided information on demographics, symptoms, comorbidities, and influenza vaccination status. Information about clinical course was obtained from electronic medical records. Indicators of comorbidity including the Charlson comorbidity index (CCI) [21] and number of conditions that elevate risk of serious influenza complications (“high-risk conditions”) were derived using diagnosis codes of inpatient and outpatient medical encounters during the year prior to study enrollment. We described patient frailty using a score ranging from 0 (not frail) to 5 (very frail) based on self-reported responses to 5 questions as previously described [20, 22].

Respiratory specimens were tested by polymerase chain reaction (PCR)-based molecular assays. Participating laboratories used standardized Centers for Disease Control and Prevention (CDC) PCR protocols or tested specimens using molecular assays for routine clinical care in laboratories completing CDC proficiency testing. Patients who tested positive for influenza were classified as cases and those testing negative for influenza were controls.

Influenza Vaccination Status

Documentation of vaccination was sought from medical records and immunization registries, as well as from vaccine providers such as pharmacies and grocery stores. At PA and TX sites, occupational health records and insurance billing claims were also used to document vaccination status. Information collected included date of administration, product, and lot number. A participant was defined as vaccinated if date of vaccination was ≥14 days prior to illness onset. We excluded patients whose self-reported vaccination was unverified after record review and patients vaccinated 0–13 days before illness onset. This study differed from our previous study of effectiveness of any vaccination during the 2015–2016 season [20] in that only patients with documented vaccination were included, as documentation was used to determine the type of vaccine received.

Influenza virus strains for 2015–2016 Northern Hemisphere influenza vaccines were A/California/7/2009 (H1N1)pdm09, A/Switzerland/9715293/2013 (H3N2), and B/Phuket/3073/2013 (Yamagata lineage) for trivalent vaccines (SD-IIV3 and HD-IIV3); quadrivalent standard dose vaccines included a B/Brisbane/60/2008-like virus (Victoria lineage). For 2016–2017 influenza vaccines, strains included A/California/7/2009 (H1N1)pdm09, A/Hong Kong/4801/2014 (H3N2), and B/Brisbane/60/2008 (Victoria lineage) for trivalent vaccines; quadrivalent standard dose vaccines included a B/Phuket/3073/2013-like virus (Yamagata lineage). All HD-IIV was trivalent.

Statistical Methods

Differences in patient characteristics by influenza and vaccination status were quantified by standardized mean difference (SMD; difference in means or proportions of a variable divided by its pooled standard deviation) and tested using χ ² test for categorical variables and t-test or Wilcoxon rank sum test for continuous variables. VE was calculated by comparing odds of influenza among patients receiving each vaccine and unvaccinated patients using multivariate logistic regression, with VE = (1–adjusted odds ratio) × 100%. Relative VE (rVE) of HD-IIV compared to SD-IIV (including trivalent and quadrivalent formulations) was calculated by comparing odds of influenza among patients who received HD-IIV versus those who received SD-IIV.

Absolute VE and rVE models were adjusted for age (65–74, ≥75 yr), study site, season, CCI, and prior season influenza vaccination. A priori variables included in the multivariate model included age, study site, and season. Additional variables associated with vaccination status (Table 1) and influenza status (Table 3) were considered as covariates, including race, prior season influenza vaccination, current tobacco smoking, and indicators of health status and comorbidity. Race did not confound the association of interest after controlling for a priori variables; therefore, it was eliminated from the model for parsimony. We controlled for health status using CCI, although results were similar using alternative health status variables. After controlling for CCI, individual comorbidities, smoking, and frailty were not confounders and were deleted from the model. Likewise, calendar time of illness onset (defined by tertiles as prepeak, peak, or postpeak influenza periods to facilitate pooling across seasons) and days from specimen collection to illness onset were not confounders and deleted. Alternative adjustment variables considered during model selection are summarized in Supplemental Table 1. Statistical analyses were performed in SAS v9.4 (SAS Institute, Cary, NC, USA) or R v3.4.1 (R Foundation for Statistical Computing, Vienna, Austria). P values <.05 or odds ratios (ORs) excluding the null value were considered statistically significant.

Table 1.

Characteristics of Participants Aged ≥65 Years Hospitalized With Acute Respiratory Infections by Influenza Vaccination Status, Hospitalized Adult Influenza Vaccine Effectiveness Network Study, 2015–2016 to 2016–2017 (n = 1487)

	Unvaccinated	Vaccinated (any vaccine)	P  a	Standardized Mean Differenceb
Number	380	1107
Enrolled in 2016–2017 season	248 (65.3)	739 (66.8)	.64	0.03
Study site			.47	0.10
Texas	91 (23.9)	254 (22.9)
Michigan	79 (20.8)	269 (24.3)
Pennsylvania	127 (33.4)	336 (30.4)
Tennessee	83 (21.8)	248 (22.4)
Age ≥75 y	139 (36.6)	556 (50.2)	<.001	0.28
Male	175 (46.1)	473 (42.7)	.29	0.07
White race	260 (68.4)	914 (82.6)	<.001	0.33
Nursing home residence	29 (7.8)	73 (6.8)	.58	0.04
Immunosuppression	80 (21.1)	364 (32.9)	<.001	0.27
Chronic obstructive pulmonary disease	169 (44.5)	543 (49.1)	.14	0.09
Nonobstructive lung disease	135 (35.5)	483 (43.6)	.007	0.17
Heart disease	265 (69.7)	859 (77.6)	.003	0.18
Congestive heart failure	156 (41.1)	512 (46.3)	.09	0.11
Neurologic disorder	121 (31.8)	431 (38.9)	.02	0.15
Renal disorder	151 (39.7)	565 (51.0)	<.001	0.23
Malignancy	114 (30.0)	423 (38.2)	.005	0.17
Hematologic disorder	33 (8.7)	122 (11.0)	.24	0.08
Metabolic disorder	190 (50.0)	739 (66.8)	<.001	0.35
Diabetes	138 (36.3)	457 (41.3)	.10	0.10
Number of high-risk conditions			<.001	0.34
0	16 (4.2)	13 (1.2)
1–2	48 (12.6)	82 (7.4)
3–4	91 (23.9)	231 (20.9)
5–6	107 (28.2)	296 (26.7)
>6	118 (31.1)	485 (43.8)
Charlson comorbidity index	3.6 (2.7)	4.3 (2.9)	<.001	0.25
Frailty category			.21	0.10
0, least frail	72 (18.9)	168 (15.2)
1–3, mild to moderately frail	225 (59.2)	695 (62.8)
4–5, most frail	83 (21.8)	244 (22.0)
≥1 (all-cause) hospitalization in past year	235 (63.3)	759 (70.3)	.02	0.15
Respiratory hospitalizations in past year			.001	0.23
0	173 (45.5)	389 (35.1)
1	159 (41.8)	518 (46.8)
2+	48 (12.6)	200 (18.1)
Home oxygen use	73 (19.2)	291 (26.3)	.007	0.17
Current tobacco smoking	79 (20.8)	128 (11.6)	<.001	0.25
Prior season influenza vaccination	142 (37.4)	1041 (94.0)	<.001	1.49

^a  P value for test of difference between unvaccinated and vaccinated (with any vaccine type) study participants.

^bStandardized mean difference between unvaccinated and vaccinated (with any vaccine type) study participants.

Table 3.

Characteristics of Participants Aged ≥65 Years Hospitalized With Acute Respiratory Infections by Influenza Status, Hospitalized Adult Influenza Vaccine Effectiveness Network (HAIVEN) Study, 2015–2016 to 2016–2017 (n = 1487)

	Influenza-Negative Controls	Influenza-Positive Cases	P  a	Standardized Mean Differenceb
Number	1210	277
Enrolled in 2016–2017 season	785 (64.9)	202 (72.9)	.01	0.18
Study site			.003	0.26
Texas	301 (24.9)	44 (15.9)
Michigan	272 (22.5)	76 (27.4)
Pennsylvania	362 (29.9)	101 (36.5)
Tennessee	275 (22.7)	56 (20.2)
Age ≥75 y	559 (46.2)	136 (49.1)	.42	0.06
Male	510 (42.1)	138 (49.8)	.02	0.15
White race	969 (80.1)	205 (74.0)	.03	0.15
Nursing home residence	85 (7.2)	17 (6.3)	.70	0.04
Immunosuppression	374 (30.9)	70 (25.3)	.08	0.13
Chronic obstructive pulmonary disease	600 (49.6)	112 (40.4)	.007	0.19
Nonobstructive lung disease	535 (44.2)	83 (30.0)	<.001	0.30
Heart disease	929 (76.8)	195 (70.4)	.03	0.15
Congestive heart failure	569 (47.0)	99 (35.7)	.001	0.23
Neurologic disorder	457 (37.8)	95 (34.3)	.31	0.07
Renal disorder	590 (48.8)	126 (45.5)	.36	0.07
Malignancy	448 (37.0)	89 (32.1)	.14	0.10
Hematologic disorder	128 (10.6)	27 (9.7)	.77	0.03
Metabolic disorder	760 (62.8)	169 (61.0)	.63	0.04
Diabetes	492 (40.7)	103 (37.2)	.32	0.07
Number of high-risk conditions			<.001	0.34
0	20 (1.7)	9 (3.2)
1–2	94 (7.8)	36 (13.0)
3–4	260 (21.5)	62 (22.4)
5–6	313 (25.9)	90 (32.5)
>6	523 (43.2)	80 (28.9)
Charlson comorbidity index	4.21 (2.80)	3.79 (2.89)	.03	0.15
Frailty category			.02	0.19
0, least frail	186 (15.4)	54 (19.5)
1	742 (61.3)	178 (64.3)
2, most frail	282 (23.3)	45 (16.2)
≥1 (all-cause) hospitalization in past year	819 (69.4)	175 (64.6)	.14	0.10
Respiratory hospitalizations in past year			.18	0.13
0	450 (37.2)	112 (40.4)
1	548 (45.3)	129 (46.6)
2+	212 (17.5)	36 (13.0)
Home oxygen use	327 (27.0)	37 (13.4)	<.001	0.35
Current tobacco smoking	172 (14.2)	35 (12.6)	.56	0.05
Prior season influenza vaccination	984 (81.3)	199 (71.8)	.001	0.23

^a  P value for test of difference between influenza-positive and influenza-negative study participants.

^bStandardized mean difference between influenza-positive and influenza-negative study participants.

Supplemental Analyses

We conducted 2 supplemental analyses to evaluate potential confounding. First, we examined rVE in a dataset limited to vaccinees (n = 1107) balanced by baseline participant characteristics using a propensity score model (PSM) [23] and inverse probability of treatment weighting [24] implemented with R’s twang package. Differences in SD-IIV and HD-IIV recipients were evaluated by SMD, and characteristics with SMD >0.10 [25] were included in the PSM with the exception of characteristics unassociated with influenza status (as assessed by SMD <≈0.05 across influenza status), which were not confounders. Thus, SD-IIV and HD-IIV vaccinees were balanced by age (65–74 vs ≥75 yr), site, season, location and month of vaccination, number of high-risk comorbidities, presence/history of chronic obstructive pulmonary disease (COPD), home oxygen use, and nursing home residence. Balance of SMD ≤0.10 was achieved on these and all other measured characteristics (Supplemental Table 2). Logistic regression was used with the balanced data set with vaccination type as a predictor of influenza status to obtain an OR. Calendar time of illness onset, days between specimen collection and illness onset, and days between vaccination and illness onset were not confounders of the association of interest and were deleted from the regression model.

Second, we examined the association between vaccine type and risk of respiratory syncytial virus (RSV) as a negative control outcome [26]. Insofar as an association between vaccine type and RSV positivity is biologically implausible, observing such an association would suggest that unmeasured sources of bias may have influenced our results.

RESULTS

Participant Characteristics

The HAIVEN study enrolled 2203 hospitalized patients aged ≥65 years during the 2015–2016 (N = 799) and 2017–2018 (N = 1404) influenza seasons. After exclusions, 1487 patients were included in the analysis (Figure 1), including 56 influenza A(H1N1)pdm09 cases enrolled primarily in 2015–2016, and 150 influenza A(H3N2) cases and 68 influenza B cases enrolled primarily in 2016–2017. Median participant age was 74 years. Comorbidities were common, with 95% of patients having ≥2 chronic medical conditions; median CCI was 4. Seventy-six percent of patients had cardiovascular disease, 48% had COPD, and 30% had an immunosuppressive condition. Median length of hospital stay was 4 days. Ten percent of patients were admitted to intensive care units (ICU), with influenza-positive patients less likely than influenza-negative patients to require intensive care (7% vs 11%; P = .03). Twenty-eight patients (2%) died in hospital, including 4 influenza-positive patients (2 of whom were vaccinated, both with SD-IIV).

Figure 1.

HAIVEN study enrollment and participant inclusion, 2015–2016 and 2016–2017. Abbreviation: HAIVEN, Hospitalized Adult Influenza Vaccine Effectiveness Network.

Of 1487 patients, 1107 (74%) patients were vaccinated. Enrollees who were aged ≥75 years, white, and had greater comorbidity were more likely to be vaccinated (Table 1; P < .05 for each), as were patients with prior season influenza vaccination (P < .001). Current tobacco smokers were less likely to be vaccinated (P < .001). Among vaccinees, 485 (44%) received SD-IIV (21% SD-IIV3, 79% SD-IIV4), and 622 (56%) received HD-IIV. Vaccinees who received HD-IIV were more likely to be aged ≥75 years, enrolled in 2016–2017 or at the PA and TN sites, to have COPD or metabolic disorders, and to have received vaccine at pharmacies, grocery stores, or workplaces (Table 2; P < .05 for each). Frailty was unassociated with vaccine type.

Table 2.

Characteristics of Vaccinated Participants Aged ≥65 Years Hospitalized With Acute Respiratory Infections by Type of Influenza Vaccine Received, HAIVEN Study, 2015–2016 to 2016–2017 (n = 1107)

	SD-IIV Vaccinees	HD-IIV Vaccinees	P  a	Standardized Mean Differenceb
Number	485	622
Enrolled in 2016–2017 season	288 (59.4)	451 (72.5)	<.001	0.28
Study site			<.001	0.53
Texas	162 (33.4)	92 (14.8)
Michigan	132 (27.2)	137 (22.0)
Pennsylvania	112 (23.1)	224 (36.0)
Tennessee	79 (16.3)	169 (27.2)
Age ≥75 y	226 (46.6)	330 (53.1)	.04	0.13
Male	215 (44.3)	258 (41.5)	.37	0.06
White race	406 (83.7)	508 (81.7)	.42	0.05
Nursing home residence	42 (8.9)	31 (5.1)	.02	0.15
Immunosuppression	166 (34.2)	198 (31.8)	.44	0.05
Chronic obstructive lung disease	211 (43.5)	332 (53.4)	.001	0.20
Nonobstructive lung disease	210 (43.3)	273 (43.9)	.89	0.01
Heart disease	369 (76.1)	490 (78.8)	.32	0.07
Congestive heart failure	225 (46.4)	287 (46.1)	.98	0.01
Neurologic disorder	190 (39.2)	241 (38.7)	.93	0.01
Renal disorder	249 (51.3)	316 (50.8)	.91	0.01
Malignancy	185 (38.1)	238 (38.3)	1.00	0.00
Hematologic disorder	58 (12.0)	64 (10.3)	.43	0.05
Metabolic disorder	296 (61.0)	443 (71.2)	<.001	0.22
Diabetes	206 (42.5)	251 (40.4)	.52	0.04
Number of high-risk conditions			.03	0.20
0	5 (1.0)	8 (1.3)
1–2	49 (10.1)	33 (5.3)
3–4	103 (21.2)	128 (20.6)
5–6	132 (27.2)	164 (26.4)
>6	196 (40.4)	289 (46.5)
Charlson comorbidity index	4.4 (2.9)	4.2 (2.8)	.30	0.06
Frailty category			.23	0.10
0, least frail	67 (13.8)	101 (16.2)
1–3, mild to moderately frail	318 (65.6)	377 (60.6)
4–5, most frail	100 (20.6)	144 (23.2)
≥1 (all-cause) hospitalization in past year	328 (69.2)	431 (71.1)	.54	0.04
Respiratory hospitalizations in past year			.11	0.13
0	171 (35.3)	218 (35.0)
1	239 (49.3)	279 (44.9)
2+	75 (15.5)	125 (20.1)
Home oxygen use	116 (23.9)	175 (28.1)	.13	0.10
Current tobacco smoking	61 (12.6)	67 (10.8)	.40	0.06
Prior season influenza vaccination	452 (93.2)	589 (94.7)	.36	0.06
Days between vaccination and illness onset	124 (43)	127 (44)	.17	0.08
Vaccination month			.15	0.18
Aug	7 (1.4)	19 (3.1)
Sep	133 (27.4)	204 (32.8)
Oct	233 (48.0)	270 (43.4)
Nov	71 (14.6)	85 (13.7)
Dec	28 (5.8)	27 (4.3)
Jan	13 (2.7)	17 (2.7)
Vaccination location			<.001	0.44
Unknown	46 (9.5)	63 (10.1)
Outpatient clinic/Hospital	374 (77.3)	424 (68.3)
Pharmacy/Grocery/Occupational	36 (7.4)	125 (20.1)
Nursing home/Assisted living/VA facilities	28 (5.8)	9 (1.4)

Abbreviations: HD-IIV, high-dose inactivated influenza vaccine; SD-IIV, standard-dose inactivated influenza vaccine; VA, Veteran’s Administration.

^a  P value for test of difference between SD-IIV and HD-IIV recipients.

^bstandardized mean difference between SD-IIV and HD-IIV recipients.

The percentage of patients testing positive for influenza was 24% among unvaccinated enrollees, 18% among SD-IIV recipients, and 16% among HD-IIV recipients (P = .003). Influenza positivity was higher among males, nonwhite patients, those enrolled at the PA site, and among patients with lesser frailty (Table 3, P < .05 for each). Individuals with more comorbidities overall (measured by CCI or number of high-risk conditions) and with cardiorespiratory conditions specifically were less likely to test influenza-positive (P < .05 for each), as were patients with prior season influenza vaccination (P = .001).

Vaccine Effectiveness

After adjusting for site, season, age group, CCI, and prior season influenza vaccination, VE for any vaccination was 21% (95% confidence interval [CI] −15, 46), 6% (95% CI −42, 38) for SD-IIV and 32% (95% CI −3, 54) for HD-IIV against all influenza types and subtypes; relative VE of HD-IIV versus SD-IIV was 27% (95% CI −1, 48) (Table 4). VE against influenza A(H1N1)pdm09 was 30% (95% CI −54, 68) for any vaccine, 23% (95% CI −84, 68) for SD-IIV and 36% (95% CI −54, 74) for HD-IIV, with rVE of 17% (95% CI −76, 61). Against influenza A(H3N2), VE was 8% (95% CI −54, 45) for any vaccine, −6% (95% CI −86, 40) for SD-IIV and 18% (95% CI −42, 53) for HD-IIV, with rVE of 23% (95% CI −19, 50). VE against influenza B/Yamagata was 50% (95% CI −1, 75) for any vaccine, 31% (95% CI −50, 68) for SD-IIV, and 62% (95% CI 16, 83) for HD-IIV, with rVE of 44% (95% CI −13, 73). RVE against any influenza was similar across seasons (24% and 27% in 2015–2016 and 2016–2017, respectively) and age groups (25% and 27% among those 65–74 yr and ≥75 yr, respectively). Sample sizes were insufficient to estimate rVE for influenza B/Victoria or by subtype within seasons.

Table 4.

Absolute and Relative Influenza Vaccine Effectiveness for Prevention of Laboratory-Confirmed Influenza Hospitalization Among Adults Aged ≥65 Years, HAIVEN Study, 2015–2016 to 2016–2017

	N	Influenza-Positive Cases/Total (%)			Adjusted VE (95% CI)a			Relative VEb
		Unvaccinated	SD-IIV	HD-IIV	Any Vaccine	SD-IIV	HD-IIV	(95% CI)
Influenza type/Subtype
All influenza A/B	1487	92/380 (24)	87/485 (18)	98/622 (16)	21 (−15, 46)	6 (−42, 38)	32 (−3, 54)	27 (−1, 48)
Influenza A(H1N1)pdm09	1266	23/311 (7)	18/416 (4)	15/539 (3)	30 (−54, 68)	23 (−84, 68)	36 (−54, 74)	17 (−76, 61)
Influenza A(H3N2)	1360	46/334 (14)	45/443 (10)	59/583 (10)	8 (−54, 45)	−6 (−86, 40)	18 (−42, 53)	23 (−19, 50)
Influenza B/Yamagata	1264	20/308 (6)	17/415 (4)	17/541 (3)	50 (−1, 75)	31 (−50, 68)	62 (16, 83)	44 (−13, 73)
Season
2015–2016	500	28/132 (21)	28/197 (14)	19/171 (11)	35 (−31, 68)	26 (−58, 66)	44 (−23, 75)	24 (−46, 61)
2016–2017	987	64/248 (26)	59/288 (20)	79/451 (18)	19 (−27, 48)	3 (−58, 41)	29 (−15, 56)	27 (−8, 50)
Age group
65–74 y	792	56/241 (23)	42/259 (16)	43/292 (15)	14 (−41, 48)	0 (−73, 43)	25 (−30, 57)	25 (−22, 54)
≥75 y	695	36/139 (26)	45/226 (20)	55/330 (17)	32 (−21, 62)	18 (−54, 56)	40 (−10, 67)	27 (−15, 54)

Abbreviations: CI, confidence interval; HAIVEN, Hospitalized Adult Influenza Vaccine Effectiveness Network; HD-IIV, high-dose inactivated influenza vaccine; SD-IIV, standard-dose inactivated influenza vaccine; VE, vaccine effectiveness.

^aAdjusted for site, season, age, prior season influenza vaccination, Charlson comorbidity index. Models stratified by season do not adjust for season. Models stratified by age group do not adjust for age.

^bRelative VE of HD-IIV compared to SD-IIV was calculated by comparing odds of influenza among patients who received HD-IIV versus those who received SD-IIV in a model using the whole study sample (including unvaccinated participants).

Supplemental Analyses

In a data set in which HD-IIV and SD-IIV recipients were balanced with respect to baseline characteristics, we found rVE of HD-IIV of 22% (95% CI −10, 44). Using RSV as a negative control outcome, we found no effect of HD-IIV compared to SD-IIV for prevention of laboratory-confirmed RSV-associated hospitalization, with estimated rVE of 2% (95% CI −76, 46).

DISCUSSION

During the 2015–2017 US influenza seasons, we found low to moderate effectiveness of influenza vaccination (of any type) for preventing influenza-associated hospitalizations among adults ≥65 years of age, with absolute VE ranging from 8% against influenza A(H3N2) to 50% against influenza B/Yamagata, all with wide CIs. VE was 21% against influenza of any type/subtype. With the exception of 62% effectiveness of HD-IIV against influenza B/Yamagata, we found no statistically significant effectiveness of any vaccine against any influenza type/subtype. Estimates of VE for HD-IIV exceeded those for SD-IIV in all subgroups examined, and, overall, we found that HD-IIV was 27% (95% CI −1, 48) more effective than SD-IIV for preventing influenza-associated hospitalization among adults aged ≥65 years in this study. However, estimates of rVE were not statistically significant. Although our study was conducted during 2 dissimilar seasons, the first being a mild season predominated by an influenza A(H1N1)pdm09 strain antigenically similar to the vaccine strain and the second a more severe season predominated by influenza A(H3N2), we saw similar rVE of HD-IIV in these 2 seasons (24% and 27%); neither was statistically significant.

Our estimate of rVE for HD-IIV of 27% is of similar magnitude to estimates of 24% and 18% for prevention of laboratory-confirmed influenza illness and influenza-associated cardiorespiratory hospitalizations observed in a large comparative efficacy trial of HD-IIV versus SD-IIV [16, 27]. Our findings are also similar to a retrospective cohort study that found HD-IIV 31% more effective than SD-IIV in preventing laboratory-confirmed influenza hospitalizations during the 2016–2017 season in Oregon [28]. Relative VE of HD-IIV ranged from 6 to 11% for prevention of (nonlaboratory confirmed) influenza hospitalizations among US Medicare beneficiaries during the 2015–2016 to 2017–2018 influenza seasons [18, 19]. Given the wide CIs for our rVE point estimates, we cannot conclude that our results differ from those published previously.

Our results suggested that benefit conferred by HD-IIV may have been greater for protection against influenza B/Yamagata and influenza A(H3N2) illness than for influenza A(H1N1)pdm09 illness, a pattern consistent with that seen in the comparative efficacy trial [16] and observational studies reporting larger relative benefit for HD-IIV in A(H3N2)-predominant seasons [18, 19], but an explanation for this finding is unclear. Immunogenicity trials have suggested that HD-IIV may elicit better antibody responses against influenza A(H3N2) than influenza B viruses [14, 29, 30]. However, evaluating relative effectiveness of HD-IIV against specific influenza types/subtypes has been hindered by lack of laboratory confirmation of influenza infection and virus subtyping in some postlicensure studies and lack of statistical power in ours. Although a strength of our study is availability of influenza subtype, we lacked serologic evidence with which to examine subtype-specific antibody response. Serologic investigation of vaccine failures in observational studies, as has been proposed recently [31], may also help compare vaccine-induced protection against specific influenza viruses.

Several limitations of our study warrant mention. As with all observational studies of the relative benefit of HD-IIV, allocation of vaccine type was not randomized, and results could be influenced by confounding arising from differences in characteristics of individuals who received SD-IIV versus HD-IIV. Similar to previous reports, we found that individuals vaccinated with HD-IIV differed from SD-IIV recipients, including being more likely to have been vaccinated in retail settings such as grocery stores and pharmacies, whereas individuals vaccinated with SD-IIV were more likely to have been vaccinated in long-term care facilities [18, 32]. Although this finding may reflect differences in vaccine cost or purchasing decisions by healthcare institutions, it may also indicate subtle differences in the populations of individuals receiving high-dose and standard-dose vaccines. We controlled for these differences by adjusting for age, underlying health status, and other participant characteristics. Substantial differences between vaccinated and unvaccinated participants in regard to certain characteristics and the size of our sample precluded use of statistical balancing methods for estimating absolute VE, likely because our sample provided too few individuals to balance multiple characteristics simultaneously. In a supplemental analysis in which we successfully balanced observed characteristics of SD-IIV and HD-IIV recipients (excluding unvaccinated participants), we found a similar rVE of 22% (95% CI −10, 44). Our analysis of RSV-related hospitalization as a negative control outcome was null, which provided reassurance that our findings were not substantially attributable to uncontrolled confounding.

We relied on vaccine documentation to determine type of vaccine received and excluded patients who reported receiving vaccine but for whom no documentation of vaccination was found, which differed from methods used in our previous study of effectiveness of any vaccination during the 2015–2016 season and may partially explain why absolute VE results in this analysis are lower than those reported previously [20]. However, if SD-IIV and HD-IIV recipients were excluded similarly for this reason, relative VE estimates should be unaffected. We saw no differences in age, race, or comorbidities between patients included in the analysis compared with those excluded due to missing vaccine documentation. However, excluded patients differed by sex and enrollment site and had a shorter average time between vaccination and illness onset; whether these differences were more (or less) pronounced among SD-IIV compared to HD-IIV recipients is unknown. Additionally, we lacked detailed information on participants’ vaccination history. More years of data with sufficiently detailed vaccination history will be needed to assess if relative benefit conferred by HD-IIV is influenced by repeated vaccination. Most notably, limitations imposed by sample size in our study prevented us from precisely estimating VE and emphasize the need for multiple years of data, especially for estimating relative VE, which, given its smaller effect size, will require greater numbers of patients to detect statistically significant differences. We estimate that detecting an average rVE of HD-IIV of 18%, the pooled estimate for prevention of influenza hospitalization reported in a recent meta-analysis [17], would likely require over 4500 study participants, including 1000 influenza cases, using a similar test-negative study design [33]. Finally, we were unable to examine if vaccine type was associated with attenuation of the most severe manifestations of influenza illness such as ICU admission and death.

Although a universal influenza vaccine that provides highly effective and sustained protection remains the ultimate goal of influenza vaccination research [34], incremental advancements in VE can offer sizeable reductions in the burden of influenza hospitalizations [35, 36]. Although more evaluation of HD-IIV is needed to understand differences in relative benefit of HD-IIV across studies, seasons, and influenza subtypes, growing evidence suggests that HD-IIV can provide improvement in protection against serious influenza illness and its consequences among the large and vulnerable population of older adults.

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

Acknowledgments. The authors thank Drs David Shay and Jerry Tokars of the Centers for Disease Control and Prevention (CDC) Influenza Division and the Hospitalized Adult Influenza Vaccine Effectiveness Network (HAIVEN) study investigators, staff, and participants.

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.

Financial Support. This study was funded by the Centers for Disease Control and Prevention (cooperative agreement IP15-002). This study was also funded in part at Pittsburgh, supported by the National Institutes of Health (NIH) Clinical and Translational Science Award (CTSA) program, grant UL1 TR001857.

Potential conflicts of interest. M. G. reports grants from Centers for Disease Control and Prevention during the conduct of the study; grants from Astra Zeneca—MedImmune outside the submitted work. E. T. M. reports grants from Centers for Disease Control and Prevention during the conduct of the study; grants from Merck & Co., grants from Pfizer, outside the submitted work. D. B. M. reports grants from Centers for Disease Control and Prevention during the conduct of the study; grants from Merck & Co., Sanofi Pasteur, and Pfizer, outside the submitted work. A. M. reports personal fees from Sanofi Pasteur and from Seqirus outside the submitted work. F. P. S. reports grants from Centers for Disease Control and Prevention during the conduct of the study; grants from Merck & Co., Sanofi Pasteur, and Pfizer outside the submitted work. H. K. T. reports grants from Centers for Disease Control and Prevention during the conduct of the study; grants from Sanofi Pasteur, outside the submitted work. R. K. Z. reports grants from Centers for Disease Control and Prevention during the conduct of the study; grants from Merck & Co., Sanofi Pasteur and Pfizer outside the submitted work. J. M. F. reports nonfinancial support from Foundation for Influenza Epidemiology (funded by unrestricted grants from Sanofi Pasteur) outside the submitted work. All other authors report no potential conflicts. 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.