Samples for this evaluation were collected from an observational cohort study of influenza vaccination in HCP described earlier [7]. In brief, HCP at 2 United States medical centers with voluntary vaccination programs were prospectively enrolled and offered the 2010–2011 IIV3 containing A/California/7/2009 p(H1N1), A/Perth/16/2009 (H3N2), and B/Brisbane/60/2008 strains. Preseason (S1) sera were collected on all participants, and postvaccination (S2) sera were collected approximately 30 days postvaccination on participants who chose to be vaccinated. End-of-season (S3) sera were also collected on all participants, regardless of vaccination status. Participants were monitored weekly for the development of symptoms of acute respiratory illness (ARI) defined as fever or feverishness and cough, and participants reporting ARI were tested for influenza infection within 7 days of illness onset via real-time reverse-transcriptase polymerase chain reaction (PCR) on nasal, nasopharyngeal, and oropharyngeal swabs. Participants were interviewed approximately 10–14 days after sample collection and asked the date when their illness resolved. Illness duration was defined as the days from self-reported illness onset to illness resolution. For the few HCP who remained ill at the follow-up interview, illness resolution was the date of the interview plus 1 day. We report here the results of NAI-antibody testing of the available S1 and S2 sera and on S3 sera collected from participants with PCR-confirmed influenza illness. We also report HAI-antibody results on the same set of S1 and S2 sera.

Because HA-specific antibodies can sterically interfere with NA activity [8], we used pseudotyped viruses that contained an irrelevant HA as the antigens for NAI testing. To generate the pseudotyped viruses, a Madin-Darby Canine Kidney (MDCK) cell line that constitutively expressed the HA (MDCK-HA) of A/Vietnam/1203/04 (H5N1) influenza virus was used [9]. We used single-cycle infectious influenza A viruses (sciIAV) in the background of A/WSN/33 (H1N1), in which the viral HA gene was replaced by that of green fluorescent protein as previously described [10], and the viral NA gene was replaced by either that of the A/California/04/2009 (pH1N1) or the A/Wyoming/03/2003 (H3N2) viruses [11]. Infection of the A/Vietnam/1203/04 H5 MDCK-HA cells with these sciIAV results in the production of H5N1- and H5N2-pseudotyped viruses. Before using these sciIAVs in NAI assays, we confirmed that samples from the study did not have any HAI activity against the H5-pseudotyped sciIAVs.

The NA protein sequences of A/Wyoming/03/2003 used here and A/Perth/16/2009 found in 2010–2011 IIV3 are 97.6% identical, varying in 12 of the 496 amino acids. Murine models of NA DNA vaccines using 2 different but 89% identical N2 subtypes demonstrated cross-protective antibody production between H3N2 influenza strains when challenged [12]. Furthermore, studies using ferret antisera against drifted variants of N2 demonstrated that sera generated using contemporary NA antigens retain NAI activity against older strains [13]. Thus, the NAI specificity demonstrated in our approach likely recapitulates the activity against matched N2 strains.

To measure the anti-NA antibody, we used a modified lectin-based assay described previously [4, 5]. Ninety-six well, flat-bottom immuno-polystyrene plates with Maxisorp surface (Nunc-Immuno) were coated with a fetuin plate-coating solution (50 µg/mL; Sigma-Aldrich). Sera were tested in serial 2-fold dilutions from 1:16 to 1:32 768. Serial dilutions of heat-inactivated sera were combined with each pseudotyped virus and incubated at room temperature for 2 hours. The antibody-antigen volume was then transferred to fetuin-coated plates and incubated for 16–18 hours at 37°C in dry heat, followed by peroxidase-labeled peanut lectin (Sigma-Aldrich) for 2 hours, and bound lectin detected with 3,3′,5,5′ tetramethylbenzidine (Sigma-Aldrich). All sera were tested in duplicate and with each NA antigen. The reciprocal value of the dilution producing a 50% reduction in the optical density of pseudotyped virus alone was defined as the serum's NAI antibody titer. When comparing pre- and postvaccination titers, a 2-fold increase in antibody titer was considered significant, in alignment with previously published lectin-based assays for NA immunogenicity [5, 14].

Testing for HAI titers to influenza A/California/7/2009 (pH1N1) and A/Perth/16/2009 (H3N2) was performed by Battelle Laboratory (Aberdeen, Maryland) using standard methods (World Health Organization 2011) [15]. An analysis of the pH1N1-specific HAI response in the study population has been reported previously [7]. However, the results shown in the current report are restricted to those samples that were also tested for NAI antibody.

Because the predominant vaccine at both sites was inactivated, we excluded live-attenuated influenza vaccine (LAIV) recipients. Antibody endpoints are presented as geometric mean titers (GMTs). Due to GMT sample distributions being highly left-skewed, base-2 log-transformed GMT data were used to calculate means and confidence intervals (CIs), and these estimates were then back-transformed to the original GMT scale.

Geometric mean ratios ([GMRs] geometric mean fold ratio of postvaccination S2 titer to pre-vaccination S1 titer) were calculated using multivariate linear mixed-effect models to account for the correlation in repeated measures (S1 and S2). Compound symmetric covariance error structure was assumed. Multivariate analyses adjusted for sex, race, ethnicity, and study site. All multivariate analyses were conducted using base-2 log-transformed GMTs.

Percentage of participants associated with positive seroresponse and respective binomial CIs were calculated by age group and prior vaccination status. In assessing the correlation between duration of influenza illness and NAI or HAI GMTs, a nonparametric measure of correlation (Spearman's rank correlation coefficient) was used because sample sizes were small. All data analyses were conducted using SAS version 9.3 (SAS Institute Inc., Cary, North Carolina).

Serum samples were available for testing on a total of 1417 participants who were enrolled in the observational study during the 2010–2011 influenza season; 68 who received LAIV were excluded. Of the remaining 1349 tested, 916 (68%) elected vaccination and provided preseason (S1) and postvaccination (S2) sera for NAI antibody testing, and 433 (32%) were not vaccinated and submitted an S1 serum alone. The characteristics of the study population are shown in Table 1. A majority of participants in both groups were white females, with those vaccinated more often age 50 years and older, married, having a high-risk condition, working in outpatient setting, and involved in patient care for >10 years. A greater number of vaccinated than unvaccinated HCP had received the prior seasonal 2009–2010 IIV-3 or monovalent pH1N1 vaccine. The unvaccinated HCP were more often employed in the emergency department. The number of children in the household, occupation, educational level, self-rated health status, or smoking status did not appear to differ substantially between groups. Table 1.

Descriptive Statistics for Demographic, Occupational, Medical Care Setting, Health Status, and Prior Influenza Vaccination Variable for Healthcare Personnel Population (N = 1349)

The serum NAI antibody responses to vaccination assessed by ELLA assay, and the serum HAI antibody responses, are shown in Table 2. After vaccination, we observed a 2-fold increase in the serum N1 and N2 NAI antibody titers in 63.7% and 47.3%, and a 4-fold response in 28.4% and 16.7% of vaccine recipients, respectively. After adjusting for sex, race, ethnicity, and study site, we examined the change in NAI-adjusted GMT ratios (aGMRs) across 4 age groups (Table 2) and compared them with the HA aGMRs of the pH1N1 and A(H3N2)Wisconsin by HAI testing. Although baseline N1 GMTs were higher in the HCP aged 50 years and older, in comparison to the 3 younger age groups, after vaccination this older age group showed a smaller aGMR to N1. Baseline GMTs for the pH1N1 HA were relatively similar across age groups, and after vaccination the greatest aGMRs occurred in the youngest age group and declined in a step-wise fashion with each increasingly older age, with overlapping CIs. In contrast to this result, the baseline N2 antibody level and the aGMR were highest in the youngest age group, but they were similar regardless of age. Table 2.

Response to Seasonal 2010–2011 Trivalent-Inactivated Influenza Vaccine by age Group and HAI and NAI Antibody Titer Types (95% Confidence Intervals)

To evaluate the impact of influenza vaccine exposure during the previous season (2009–2010), we compared the HAI and NAI responses of vaccinated participants who had received both the monovalent pH1N1 vaccine and seasonal IIV3 (n = 411) in the previous year with those of participants who had not received either vaccine previously (n = 82). We observed higher baseline N1 antibody levels for those previously vaccinated, and we noted a similar but less substantial trend for N2 (Table 3). However, there was a significantly larger aGMR (P < .01) to both N1 and N2 antigens after the 2010–2011 IIV3 among those who did not receive either IIV3 or pH1N1 vaccine in 2009–2010. Higher baseline HAI titers and lower aGMR among those vaccinated the prior year were also observed for the HAI response. Table 3.

Hemagglutination-Inhibition and NAI Antibody Response to Seasonal 2010–2011 Trivalent-Inactivated Influenza Vaccine by Prior Season Vaccination Status

To compare the response to inactivated vaccine with the response to infection with circulating influenza virus, we assessed pre-exposure antibody (ie, S1 for unvaccinated participants, and S2 for vaccinated participants) and postinfection antibody (ie, S3) levels in participants with PCR-documented influenza. Postseason sera were available on 6 vaccinated and 3 unvaccinated participants who had experienced documented H1N1 illness and on 22 vaccinated and 6 unvaccinated participants who had experienced H3N2 illness. For both viruses, infection stimulated a strong NAI antibody response to the relevant NA antigen. For unvaccinated participants, the GMR comparing S1 to S3 was 10.08 (95% CI = .07, 1452) for N1 in pH1N1 infected and 8.0 (95% CI = 2.2, 29.4) for N2 in participants infected with H3N2. For vaccinated participants, these GMRs were 8.98 (95% CI = 1.9, 42.5) and 3.01 (95% CI = 1.9, 4.82), respectively. Compared with the vaccine responses shown in Table 3, these GMRs are slightly higher than those seen in response to vaccine (no prior vaccine GMR 3.24 for N1 and 2.14 for N2; history of prior vaccine GMR 1.9 for N1 and 1.5 for N2), but the differences were not statistically significant.

Assessment of the potential role of NAI antibody in protection against influenza was complicated by the relatively small number of laboratory-documented cases that occurred in the study. For this analysis, participants who reported respiratory illness but who tested negative for influenza are removed to avoid possible confusion from false-negative PCR tests. Figure 1 shows the distributions of serum HAI and NAI antibody against H1 and N1, or against H3 and N2 antigens, in the pre-exposure (S1 in unvaccinated and S2 in vaccinated) sera. Although a protective titer in the ELLA assay has not been defined, participants with laboratory-documented influenza had relatively lower pre-exposure serum HAI and NAI antibody; these differences were not statistically significant. Figure 1.

Number of participants in each strata of pre-exposure serum antibody titer. A and B show titers against H1N1 (left, HAI titer; right, NAI titer), whereas C and D show titers against H3N2 (right, HAI titer; left, NAI titer). In each panel, the white bars show the S1 titers in unvaccinated participants, and the gray bars show S2 titers in vaccinated participants. The red bar indicates those with polymerase chain reaction (PCR)-documented influenza H1N1 (A and B) or H3N2 (C and D) in each strata.

Among the unvaccinated, having higher S1 N1 titers was associated with a shorter duration of influenza illness among all influenza A positives, but this was not seen for N2 (Table 4). However, among the vaccinated participants, although a higher N1 antibody may correlate with a shorter duration of illness for pH1N1 infection, the N2 antibody correlated with a shorter period of illness among all influenza A positives and also for pH1N1 infection, for unclear reasons. It is interesting to note that these correlations were not seen with either H1 or H3 HAI titers. Table 4.

Spearman Correlations Between N1 and N2 NAI, and A(H1N1)pdm09 HAI, and A(H3N2) HAI Antibody Titers and Duration of Subtypes of Influenza A Virus Illness Among Vaccinated and Unvaccinated Healthcare Personnel^a