Comparative Effectiveness of High-Dose Versus Standard-Dose Influenza Vaccines Among US Medicare Beneficiaries in Preventing Postinfluenza Deaths During 2012–2013 and 2013–2014

Abstract

Background.

Recipients of high-dose vs standard-dose influenza vaccines have fewer influenza illnesses. We evaluated the comparative effectiveness of high-dose vaccine in preventing postinfluenza deaths during 2012–2013 and 2013–2014, when influenza viruses and vaccines were similar.

Methods.

We identified Medicare beneficiaries aged ≥65 years who received high-dose or standard-dose vaccines in community-located pharmacies offering both vaccines. The primary outcome was death in the 30 days following an inpatient or emergency department encounter listing an influenza International of Classification of Diseases, Ninth Revision, Clinical Modification code. Effectiveness was estimated by using multivariate Poisson regression models; effectiveness was allowed to vary by season.

Results.

We studied 1039645 recipients of high-dose and 1683264 recipients of standard-dose vaccines during 2012–2013, and 1508176 high-dose and 1877327 standard-dose recipients during 2013–2014. Vaccinees were well-balanced for medical conditions and indicators of frail health. Rates of postinfluenza death were 0.028 and 0.038/10000 person-weeks in high-dose and standard-dose recipients, respectively. Comparative effectiveness was 24.0% (95% confidence interval [CI], .6%–42%); there was evidence of variation by season (P = .12). In 2012–2013, high-dose was 36.4% (95% CI, 9.0%–56%) more effective in reducing mortality; in 2013–2014, it was 2.5% (95% CI, –47% to 35%).

Conclusions.

High-dose vaccine was significantly more effective in preventing postinfluenza deaths in 2012–2013, when A(H3N2) circulation was common, but not in 2013–2014.

influenza vaccines, influenza, human, comparative effectiveness, death.

(See the editorial commentary by Monto on pages 500–2.)

Much of the impetus to expand influenza immunization programs has been prompted by a desire to reduce serious complications of influenza infections, including death. It has been recognized at least since the 1957 A(H2N2) pandemic that older persons and those with some chronic health conditions, including pulmonary and cardiac compromise, are at the greatest risk of severe influenza outcomes [1, 2]. Recent randomized trials of influenza vaccines conducted in older populations have used more common outcomes as endpoints, typically laboratory-confirmed infections [3, 4]. Trials assessing influenza-associated mortality are unfeasible, given the low likelihood of death following influenza infection during seasonal epidemics, even among the older persons at greatest risk. In fact, the best evidence available from randomized placebo-controlled studies of inactivated influenza vaccines among older persons demonstrated vaccine efficacy of 58% (95% confidence interval [CI], 26%–77%) for the prevention of symptomatic clinical illness associated with serologic evidence of influenza illness [5].

Several approaches to improving the clinical effectiveness of influenza vaccines for older adults are being investigated. In December 2009, the US Food and Drug Administration (FDA) licensed an injectable inactivated trivalent influenza vaccine containing more antigen (60 µg vs 15 µg per strain) for use among adults aged ≥65 years, hereafter referred to as “high-dose vaccine” [6]. High-dose vaccine was approved under FDA accelerated approval regulations, for which licensure was based on studies showing that high-dose vaccine elicited higher hemagglutination inhibition (HAI) titers in adults aged ≥65 years than standard-dose vaccine for influenza A(H1N1) and A(H3N2) viruses, and noninferior titers for B viruses [7–9]. A manufacturer-sponsored postlicensure randomized trial of high-dose vs standard-dose vaccine among approximately 30000 subjects aged ≥65 years was conducted during 2011–2012 and 2012–2013 [3]. It demonstrated that high-dose vaccine was more efficacious in preventing laboratory-confirmed influenza infections (relative efficacy, 24.2% [95% CI, 9.7%–36.5%]). Despite its large size, this trial was not powered to estimate the comparative efficacy of these vaccines against more serious influenza illnesses as primary outcomes. In an observational study among Medicare beneficiaries aged ≥65 years who were vaccinated in 2012–2013 with high-dose vs standard-dose vaccines in pharmacies [10], we found that high-dose vaccine was 22% (95% CI, 15%–29%) more effective than standard-dose vaccines in preventing influenza-related office visits. This finding was consistent with the relative efficacy of high-dose vaccine in the postlicensure trial, which used culture- or polymerase chain reaction (PCR)–confirmed influenza infection as the primary outcome [3]. We also found a 22% (95% CI, 16%–27%) reduction in influenza hospitalizations in the high-dose group [10].

Here we assess the risk of death after a diagnosed influenza infection among Medicare beneficiaries who received influenza vaccines in 2012–2013 or 2013–2014. Based on results from previous randomized and observational studies [3, 10, 11], we hypothesized that high-dose vaccine would prevent 20%–30% more influenza-associated deaths than standard-dose vaccine.

METHODS

Data Sources

Data were collected from Medicare administrative files (see https://www.medicare.gov/sign-up-change-plans/decide-how-to-get-medicare/whats-medicare/what-is-medicare.html for a summary). Medicare provides US government–sponsored health insurance to >40 million US residents aged ≥65 years and to approximately 9 million persons aged <65 years who are disabled or have end-stage renal disease. We linked enrollment data for beneficiaries receiving fee-for-service care with claims from inpatient (Medicare Part A) and outpatient care (Part B) settings, as well as claims for prescription drugs dispensed in outpatient settings (Part D). Pharmacy data were drawn from the Provider Enrollment, Chain, and Ownership System, the National Plan and Provider Enumeration System, and the National Council for Prescription Drug Programs databases. The proportion of respiratory samples testing positive for influenza viruses among samples submitted to laboratories collaborating with the National Respiratory and Enteric Virus Surveillance System (NREVSS) was used to monitor the intensity of influenza activity during the study period [12].

Participants

The base population was drawn from Medicare beneficiaries aged ≥65 years enrolled in fee-for-service care through Medicare parts A and B. Enrollment on the date of vaccination and for at least 6 months prior to vaccination was required, so that data were available to detect comorbid medical conditions. Beneficiaries with Part C coverage (managed care plan enrollment) were excluded because regulations regarding coverage for medications and vaccines are more homogenous among those receiving fee-for-service care. Beneficiaries who first enrolled in Medicare for any reason other than reaching 65, specifically being disabled or having end-stage renal disease, were excluded. We selected beneficiaries who received inactivated influenza vaccines from 5 August 2012 through 31 January 2013 for the 2012–2013 season, and from 4 August 2013 through 31 January 2014 for the 2013–2014 season. Dates were formatted to reflect the reporting of NREVSS influenza virus data—weekly, from Sunday through Saturday. Beneficiaries recorded as receiving both high-dose and standard-dose vaccines between 1 August and 31 May of the following year were excluded. From this group of beneficiaries, we obtained the study population. They were beneficiaries who received an influenza vaccination at a community-located pharmacy. This restriction was applied to identify beneficiaries who met a minimum standard of physical and mental health, demonstrated by an ability to visit a pharmacy and request an influenza vaccination [10]. To help adjust for temporal and geographical factors that influenced the availability of or access to high-dose vaccine, the study population was further restricted to those beneficiaries who received a standard- or high-dose vaccine at a community-located pharmacy that vaccinated at least 1 Medicare beneficiary with the alternative influenza vaccine in the 14 days preceding or following each vaccination date. Figure 1 summarizes the cohort selection process.

Figure 1.

Cohort selection process. Abbreviations: HD, high-dose influenza vaccine; SD, standard-dose influenza vaccines.

Influenza Vaccine Exposure

Participants were defined as exposed to high-dose or standard-dose vaccines by using Healthcare Common Procedure Coding System (HCPCS) and Current Procedural Terminology (CPT) codes for high-dose vaccine (CPT 90662 code) or standard-dose vaccine (CPT codes 90655–90661 or 90724; HCPCS codes G0008, Q2035–Q2038). The 2012–2013 influenza vaccines dispensed in the northern hemisphere contained antigens representing the following viruses: an A/California/7/2009 (H1N1)pdm09-like virus, an A/Victoria/361/2011(H3N2)-like virus, and a B/Wisconsin/1/2010-like virus [13]. For the 2013–2014 season, trivalent vaccines contained the following antigens: an A/California/7/2009(H1N1)pdm09-like virus, an A(H3N2) virus antigenically like the cell-propagated prototype virus A/Victoria/361/2011(A/Texas/50/2012), and a B/Massachusetts/2/2012-like virus [14]. B/Wisconsin/1/2010 and B/Massachusetts/2/2012 are both B/Yamagata lineage viruses, and thus serologic cross-reactivity was expected with these 2 antigens. The vaccine components used in the 2012–2013 and 2013–2014 influenza vaccines were similar, suggesting that pooled estimates of vaccine effects might be possible.

Outcomes

The primary outcome was postinfluenza death, defined as a death occurring in the 30 days following a Medicare claim for an inpatient hospitalization or an emergency department visit with a diagnosis of influenza, identified using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes 487.xx and 488.xx. To permit comparison of our findings with those from previous randomized and observational studies [3, 10, 11], we included 2 secondary outcomes: (1) hospitalizations and emergency department visits listing an ICD-9-CM code for influenza; and (2) likely influenza-related office visits, defined by a visit with claims for a rapid influenza diagnostic test (CPT code 87804) and for dispensing of a treatment regimen of oseltamivir (75 mg twice daily for 5 days) within 2 days of the test. The latter outcome was examined in a cohort restricted to participants who were enrolled in Medicare Part D on the day of influenza vaccination, so that we could identify prescriptions for a therapeutic course of oseltamivir.

Person-time Under Observation

Follow-up time began 14 days after the date of the vaccination, to permit the development of vaccine-specific immunity [15], and continued until the first of the following events: disenrollment from Medicare Part A or B; the outcome of interest; or death. For the influenza-related illness outcome, beneficiaries were also censored if they disenrolled from Medicare Part D. Using numbers of postinfluenza deaths in the entire cohort, we estimated we had adequate power to detect a 30% reduction in postinfluenza deaths in the high-dose vaccine group (Supplementary Table 1).

Variables of Interest

Demographic characteristics (age, sex, race, and ethnicity), and region of residence as characterized by the Department of Health and Human Services (DHHS) were collected for each participant. During the 6-month period prior to receipt of an influenza vaccine, we categorized preexisting medical conditions with the Hierarchical Condition Categories used for Medicare risk adjustment [16]. Participants were identified in the following categories: asthma, blood disorders, chronic lung disease, diabetes, heart disease, kidney disorders, liver disorders, neurological conditions, and immune system dysfunction. Several indicators of frail physical health were also collected for each participant (see Supplementary Materials).

Data Analyses

To determine whether the 2 cohorts were similar, differences in baseline covariates were evaluated using standardized mean differences, calculated as the difference in means or proportions of a variable divided by its pooled standard deviation. Standardized mean differences are not influenced by population size, and may provide a more meaningful measure of differences in large cohorts than standard hypothesis tests. A standardized mean difference of ≥0.1 between groups was considered meaningful [17].

Each calendar week during the study period in each DHHS region was classified into low, medium, or high influenza periods, based on detections of influenza viruses during that week and in that region. High, medium, and low periods of influenza activity were defined as weeks when the proportion of respiratory samples that tested positive for influenza was the ≥75th, 55th–75th, or <55th percentile for each region/season, respectively [10].

Rates of each study outcome were calculated as the number of events during influenza periods divided by person-time measured in weeks. To improve specificity, our analyses focused on events occurring during high influenza periods. Three Poisson regression models were used. In the first, comparative vaccine effectiveness (VE) was estimated during high influenza periods with a covariate for season (2012–2013 or 2013–2014). A second model included all covariates listed in Table 1, to adjust for potential residual confounding after restricting participation to beneficiaries who were vaccinated in pharmacies. In the first 2 models, we pooled outcomes from 2012–2013 and 2013–2014. A third added a vaccine-season interaction term to the second, to evaluate whether effectiveness varied substantially by season. A log-likelihood ratio test was used to assess whether the interaction term statistically improved model fit; a P value of <.1 was considered significant. VE was estimated as [(1 – rate high-dose recipients/rate standard-dose recipients) × 100] and 95% CIs were estimated. Analyses were conducted using R 3.2.0 (R Foundation for Statistical Computing, Vienna, Austria) and SAS version 9.4 (SAS Institute Inc, Cary, North Carolina). All analyses were performed with de-identified data collected for administrative purposes by the Centers for Medicare and Medicaid Services, and thus informed consent was not required by the US DHHS.

Table 1.

Characteristics of Study Participants for Postinfluenza Mortality and Influenza Hospitalization Outcomes, by Influenza Vaccine Type and Season

Characteristic	2012–2013 Season			2013–2014 Season
	Standard Dose	High Dose	Standardized Mean Difference	Standard Dose	High Dose	Standardized Mean Difference
Population	1683264	1039645		1877327	1508176
Age group, y
65–74	52.1%	49.8%	0.05	51.2%	49.4%	0.04
75–84	34.7%	36.6%	0.04	34.9%	36.7%	0.04
≥85	13.1%	13.7%	0.02	13.9%	14.0%	0.00
Sex
Male	40.7%	42.1%	0.03	40.3%	41.7%	0.03
Female	59.3%	57.9%	0.03	59.7%	58.3%	0.03
Race/ethnicity
White	93.6%	93.1%	0.02	92.9%	92.7%	0.01
Black	2.6%	2.9%	0.02	2.8%	3.0%	0.01
Asian/Pacific Islander	1.3%	1.4%	0.01	1.4%	1.5%	0.01
Hispanic	0.6%	0.7%	0.01	0.8%	0.8%	0.00
Other, non–North American Native	1.3%	1.4%	0.01	1.3%	1.4%	0.01
North American Native	0.1%	0.2%	0.01	0.1%	0.1%	0.00
Unknown	0.4%	0.4%	0.01	0.7%	0.6%	0.01
DHHS region of residence
Region 1: CT, ME, MA, NH, RI, VT	6.2%	4.0%	0.10	6.7%	4.6%	0.09
Region 2: NJ, NY, PR, VI	7.8%	5.6%	0.09	8.3%	6.6%	0.07
Region 3: DE, DC, MD, PA, VA, WV	8.1%	10.0%	0.07	8.4%	10.4%	0.07
Region 4: AL, FL, GA, KY, MS, NC, SC, TN	23.4%	27.2%	0.09	21.8%	26.8%	0.12
Region 5: IL, IN, MI, MN, OH, WI	19.6%	12.9%	0.18	19.6%	13.7%	0.16
Region 6: AR, LA, NM, OK, TX	11.2%	10.9%	0.01	12.2%	10.7%	0.05
Region 7: IA, KS, MO, NE	4.9%	3.3%	0.08	5.0%	3.2%	0.09
Region 8: CO, MT, ND, SD, UT, WY	2.9%	4.7%	0.10	2.6%	4.3%	0.09
Region 9: AZ, CA, HI, NV, AS, FS, GU, PU	11.9%	14.9%	0.09	11.5%	13.8%	0.07
Region 10: AK, ID, OR, WA	3.9%	6.6%	0.12	3.7%	5.9%	0.10
Proportion of beneficiaries enrolled in Medicare Part D	56.5%	55.9%	0.01	67.7%	66.5%	0.02
Low-income subsidy statusa
Proportion with 15% copay	0.8%	0.8%	0.00	0.7%	0.6%	0.00
Proportion with high copay	4.8%	4.9%	0.01	4.4%	4.3%	0.01
Proportion with low copay	4.9%	5.2%	0.02	4.5%	4.5%	0.00
Proportion with zero copay	1.1%	1.1%	0.01	1.0%	1.0%	0.01
Chronic medical conditionsb
At least 1 high-risk disorder	61.9%	62.9%	0.02	63.0%	64.6%	0.03
Asthma	3.8%	3.9%	0.01	4.0%	4.3%	0.02
Blood disorders	15.1%	15.6%	0.01	15.5%	16.3%	0.02
Chronic lung disease	13.8%	14.4%	0.02	13.9%	14.7%	0.02
Diabetes	20.1%	20.3%	0.01	21.1%	21.4%	0.01
Heart disease	33.1%	34.1%	0.02	33.7%	35.1%	0.03
Kidney disorders	6.9%	7.3%	0.01	7.7%	8.2%	0.02
Liver disorders	2.2%	2.3%	0.00	2.3%	2.4%	0.01
Neurological or neurodevelopmental conditions	11.1%	11.3%	0.01	11.3%	11.6%	0.01
Weakened immune system	12.1%	12.4%	0.01	12.1%	12.8%	0.02
Indicators of frail health statusc
Home oxygen use	2.4%	2.5%	0.01	2.4%	2.5%	0.00
Wheelchair use	0.1%	0.1%	0.00	0.1%	0.1%	0.00
Walker use	1.1%	1.1%	0.00	1.1%	1.0%	0.01
Dementia	3.2%	3.1%	0.00	3.4%	3.2%	0.02
Urinary catheter use	0.1%	0.1%	0.00	0.1%	0.1%	0.00
Falls	3.1%	3.1%	0.00	3.3%	3.2%	0.00
Fractures	1.1%	1.1%	0.00	1.1%	1.1%	0.00

Abbreviations: AK, Alaska; AL, Alabama; AR, Arkansas; AS, American Samoa; AZ, Arizona; CA, California; CO, Colorado; CT, Connecticut; DC, Washington, D.C.; DE, Delaware; DHHS, Department of Health and Human Services; FL, Florida; FS, Federated States of Micronesia; GA, Georgia; GU, Guam; HI, Hawaii; IA, Iowa; ID, Idaho; IL, Illinois; IN, Indiana; KS, Kansas; KY, Kentucky; LA, Louisiana; MA, Massachusetts; MD, Maryland; ME, Maine; MI, Michigan; MN, Minnesota; MO, Missouri; MS, Mississippi; MT, Montana; NC, North Carolina; ND, North Dakota; NE, Nebraska; NH, New Hampshire; NJ, New Jersey; NM, New Mexico; NV, Nevada; NY, New York; OH, Ohio; OK, Oklahoma; OR, Oregon; PA, Pennsylvania; PR, Puerto Rico; PU, Republic of Pulau; RI, Rhode Island; SC, South Carolina; SD, South Dakota; TN, Tennessee; TX, Texas; UT, Utah; VA, Virginia; VI, U.S. Virgin Islands; VT, Vermont; WA, Washington; WI, Wisconsin; WV, West Virginia; WY, Wyoming.

^aLow-income subsidy status among beneficiaries enrolled in Medicare Part D. This subsidy provides assistance with the premium, deductible, and co-payments of the program. See Supplementary Materials for details.

^bMedical conditions defined by the appearance of specific International Classification of Diseases, Ninth Revision, Clinical Modification codes in the 6 months prior to vaccination; see Supplementary Materials for details.

^cThese indicators were defined by the appearance of specific Healthcare Common Procedural Codes in the 6 months prior to vaccination; see Supplementary Materials for details.

RESULTS

During the study period, >33 million Medicare beneficiaries aged ≥65 years received influenza vaccinations (Figure 1). In 2012–2013, 17% of vaccinees received the high-dose vaccine (2.8 million beneficiaries), while 24% received the high-dose vaccine in 2013–2014 (3.9 million beneficiaries). We studied 5797090 persons who received high-dose or standard-dose vaccines at 29669 community-located pharmacies during the 2 seasons. During 2012–2013, 1039645 of 2722909 (38.2%) subjects meeting entry criteria received high-dose vaccine; during 2013–2014, 1508176 of 3385503 (44.5%) received high-dose vaccine. For the influenza-related illness outcome, requiring outpatient prescription-drug coverage reduced the number of subjects to 1530966 and 2273211 in the 2012–2013 and 2013–2014 seasons, respectively.

Subjects receiving high-dose or standard-dose vaccine were similar with respect to baseline covariates, including health conditions, low-income status, and indicators of frail health. Low and virtually identical proportions of high-dose and standard-dose recipients used home oxygen, wheelchairs or walkers, or had claims for falls, fractures, or dementia (Table 1), suggesting that substantial confounding of vaccine exposure by health status was unlikely. There were differences by region of residence among high-dose and standard-dose vaccinees (Table 1).When we restricted the population further to those who received prescription drug coverage, the findings were very similar (Supplementary Table 2). The characteristics of beneficiaries who received standard- or high-dose vaccine in the first and the second study seasons did not differ (Supplementary Table 3).

Rates of postinfluenza death, hospitalized influenza, and influenza-related visits were plotted by vaccine type and by low, medium, and high influenza period in Figure 2. Rates for each outcome were higher during the first season, when A(H3N2) viruses predominated in the United States. The likelihood of postinfluenza death was 57% lower during 2013–2014, when A(H1N1)pdm09 viruses predominated.

Figure 2.

Outcome rates (per 10000 person-weeks) for each of 3 influenza outcomes, by influenza season and during periods of high, medium, and low influenza activity.

During both seasons, 83 postinfluenza deaths occurred in the high-dose group during 30079255 person-weeks of observation (rate, 0.028/10000 person-weeks), and 162 deaths in the standard-dose group during 42696182 person-weeks (rate, 0.038/10000 person-weeks), representing a risk difference of –0.01/10000 person-weeks (95% CI, –.019 to –.002), and an incidence rate ratio of 0.73 (95% CI, .59–.95). When examined by season of occurrence, differences in the rate of each outcome by vaccine type were greater in 2012–2013 (Figure 2).

Estimates of comparative effectiveness for influenza-related illness, hospital-diagnosed influenza, and postinfluenza death were similar in models that adjusted only for season and those that included covariates for potential confounders (Table 2). However, comparative effectiveness varied significantly by season for influenza-related illness (P value for season-vaccine interaction = .006) and for hospital-diagnosed influenza (P = .041); for postinfluenza death, the P value was .12. Given the consistently higher estimates of comparative effectiveness during the first season of the study, we presented season-specific estimates for each outcome (Table 2). During 2012–2013, comparative VE for the primary outcome of postinfluenza death was 36.4% (95% CI, 9%–56%), while during the following season it was minimal and not significant (2.5% [95% CI, –47% to 35%]).

Table 2.

Comparative Effectiveness of High-Dose Versus Standard-Dose Influenza Vaccine in Preventing Each of 3 Influenza Outcomes

Outcome	Simple Poisson Modela		Covariate-Adjusted Poisson Modelb		Covariate-Adjusted Poisson Model With Season-Vaccine Interactionc
					2012–2013 Season		2013–2014 Season		Term for Season- Vaccine Interaction
	CE    (95% CI)		CE    (95% CI)		CE    (95% CI)		CE     (95% CI)		P Value
Postinfluenza death	23.5	(.3–41.3)	24.0	(.6–41.8)	36.4	(9–55.6)	2.5	(–46.8 to 35.3)	.12
Hospitalized influenza	18.8	(14.4–23)	18.6	(14.1–22.9)	22.1	(16.6–27.3)	12.7	(4.9–19.9)	.041
Influenza-related illness	14.2	(8.5–19.4)	15.3	(9.7–20.6)	22.0	(14.8–28.6)	6.8	(–2.3 to 15.1)	.006

Abbreviations: CE, comparative effectiveness; CI, confidence interval.

^aModel included vaccine status and a covariate for season.

^bAdded to simple model were covariates for each of the factors listed in Table 1.

^cAn interaction term for vaccine status and study season was added to the covariate-adjusted model.

DISCUSSION

In our study population, we found a significant reduction of approximately 35% in postinfluenza deaths associated with receipt of high-dose vs standard-dose vaccines in 2012–2013, but not in 2013–2014. The 2012–2013 reduction was accompanied by significant decreases of approximately 20% in influenza-related visits and hospital-based influenza diagnoses. The latter 2 findings are consistent with the results from our previous study (which used slightly different eligibility criteria and only 2012–2013 data) [10], and those from analyses of a manufacturer-sponsored trial [3, 11]. The internal and external consistency of our findings suggests that the use of high-dose vaccine in Medicare beneficiaries during 2012–2013 prevented additional influenza-related deaths. Despite 31 million person-weeks of follow-up time during 2012–2013, the magnitude of this reduction in mortality was not precisely estimated (95% CI, 9%–56%). As the outcomes of greatest interest when considering alternative vaccination strategies are often rare (eg, influenza-related mortality or serious vaccine adverse events), large postlicensure observational studies are vital in understanding the comparative effects of new vaccine formulations—randomized trials are unable to answer all relevant policy questions. Multiple seasons of observational data analyzed with appropriate longitudinal methods will be critical to providing valid information to those considering alternative influenza vaccine recommendations.

Persons aged 65 years and older experience the highest rates of many serious complications of influenza—including prolonged hospitalization, the need for intensive care, and death—during most interpandemic influenza seasons [18–21]. During seasonal epidemics in which influenza A(H3N2) viruses predominate, rates of serious complications are higher, often by a factor of 2 to 3 [21, 22]. Influenza A(H3N2) virus circulation was common during the 2012–2013 influenza season [14], leading to higher rates of influenza-associated outcomes; thus, we had improved power during the first season to detect a difference in our rarest outcome, postinfluenza mortality. During the 2013–2014 influenza season, receipt of high-dose vaccine was associated with a significant reduction only in the rate of hospitalized influenza. This reduction was smaller in magnitude than that found in the 2012–2013 season. There are several plausible reasons that the comparative effectiveness of high-dose vaccine was reduced or absent during 2013–2014, including differences in the individuals who received high-dose vaccine by season, differences in the circulation patterns of specific influenza viruses by season, and differences in the antigenic relatedness of vaccine and wild-type viruses by season.

Although more high-dose vaccine was used in 2013–2014, both in terms of total vaccinations administered and in the proportion of participants who received it, we found no evidence that the characteristics of high-dose recipients differed by season. On the other hand, A(H3N2) viruses predominated in 2012–2013 [14], and 90% of influenza A detections in 2013–2014 were A(H1N1)pdm09 viruses [23]. Data from the randomized trial suggested that relative efficacy of high-dose vaccine might be greater for A(H3N2) than for A(H1N1) viruses, but this difference was not statistically significant (23.3% [95% CI, 6.0%–37.5%] and 11.1% [95% CI, –160% to 70.2%], respectively) [3]. Therefore, it is uncertain if our findings might reflect a difference in comparative effectiveness by A subtype. Influenza B viruses circulated in both seasons, representing 20%–30% of influenza detections [14, 23], so differences in B virus activity are unlikely to explain our findings. There was no evidence of antigenic drift in A(H1N1)pdm09 viruses during the 2 study seasons: 99% of A(H1N1)pdm09 viruses characterized were antigenically similar to A/CA/7/2009-like viruses [14, 23]. The great majority of A(H3N2) viruses characterized in 2012–2013 were antigenically similar to cell culture–propagated A/Victoria/361/2011 viruses [24]. A change in the A(H3N2) vaccine strain from A/Victoria/361 in 2012–2013 to A/Texas/50 in 2013–2014 was recommended because ferret antisera made with egg-propagated A/Victoria/361 did not recognize more recent cell culture–propagated A(H3N2) viruses well. Ferret antisera made with egg-propagated A/Texas/50/2012 (which was genetically closely related to A/Victoria/361) better recognized the viruses that were tested [24]. Low effectiveness of the 2012–2013 vaccine against reverse transcription PCR–confirmed A(H3N2) virus infections was found [25]. Low effectiveness was based mainly on mutations associated with adaptation to growth in eggs, and not on mutations associated with antigenic drift [25]. Without samples of postvaccination sera and of the infecting influenza viruses from participants, we cannot conclude that 2012–2013 high-dose vaccine led to a broader or improved immune response than did standard-dose vaccine, although we can expect that its administration did lead to higher HAI titers to that season’s wild-type H3N2 viruses on a population level, based on published immunogenicity data [3, 26, 27].

During 2012–2013, estimates of influenza VE were 39% (95% CI, 29%–47%) for medically attended A(H3N2) infections among US subjects of all ages; estimates were lower and nonsignificant among those aged ≥65 years (11% [95% CI, –41% to 43%]) [28]. Given the greater relative severity of A(H3N2) seasons in older adults, this low effectiveness estimate (which should be interpreted in light of its broad CI) highlights the need to improve VE for recent A(H3N2) strains, especially among those aged ≥65. We lacked data on infecting viruses, and cannot make subtype-specific VE estimates. However, how comparative VE varies during seasons dominated by different influenza types/subtypes deserves further investigation as new vaccines are introduced.

This study has limitations. The lack of laboratory results in Medicare data means that none of our outcomes represented laboratory-confirmed influenza infections. As our comparative effectiveness results for influenza-related visits (defined by a claim for a rapid influenza test followed by receipt of oseltamivir) were concordant with those from a randomized trial that used laboratory-confirmed influenza illness as a primary outcome [3], we may have identified a valid proxy for a medically attended influenza infection in the Medicare population. Our 2 more serious outcomes were defined only by influenza diagnoses made in a hospital setting. Some data suggest that use of influenza diagnostic tests by clinicians has increased since the 2009 pandemic; thus, the diagnosis of serious influenza infections may have improved recently [29, 30]. As in all observational research, it is possible that our findings were biased with residual confounding. In vaccine studies focusing on older adults, residual confounding associated with chronic conditions which are not well characterized by a simple dichotomous indicator—for example, a clinical history of chronic heart failure—can be especially problematic [31, 32]. The consistency of our findings with those of the randomized trial of high-dose vaccine suggests that we have successfully addressed this issue through restriction of the study population to those vaccinated in community pharmacies. It is possible that knowledge by a care provider of the influenza vaccine type a participant received affected decisions to test for or to diagnose influenza. However, the lack of published clinical results for high-dose vaccine during the study period and the lack of recommendations to favor high-dose over standard-dose influenza vaccines make such a bias unlikely. A final limitation is that restricting participation to those Medicare beneficiaries vaccinated in pharmacies may decrease the generalizability of our findings. On the other hand, it is essential that comparative effectiveness studies address the possibility of confounding by indication, and the use of restriction to minimize this bias in large observational studies has been advocated [33, 34].

Our findings suggest that high-dose influenza vaccines and perhaps other vaccines designed to elicit higher HAI immune responses among older adults may yield the most benefits during seasons when influenza A(H3N2) viruses are widespread. A recent meta-analysis estimated that the effectiveness of standard-dose inactivated vaccines among adults aged >60 years for laboratory-confirmed A(H3N2) infection was only 24% [35]. The authors concluded that influenza vaccines offering better protection against A(H3N2) infection are critically needed. The availability of A(H3N2) vaccines offering substantially better protection for older adults and their widespread use in this population could lead to meaningful reductions in influenza-associated morbidity and mortality.

Notes

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 (CDC).

Financial support.  This work was supported by the CDC and the FDA. This study was performed as part of the SafeRx Project, a joint initiative of the Centers for Medicare and Medicaid Services, the FDA, and the CDC.

Potential conflicts of interest.  J. F. reports receiving limited travel support to meetings unrelated to the submitted work from Sanofi Pasteur. All other authors report no potential conflicts of interest. The 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.

Author notes

Presented in part: Options for the Control of Influenza IX, Chicago, Illinois, August 2016. Abstract P-112.

Correspondence: D. K. Shay, Centers for Disease Control and Prevention, 1600 Clifton Rd NE, MS A-20, Atlanta, GA 30333 ([email protected]).