3.1. Influenza Outcomes
Three hundred and fifteen subjects (58% females, 42% males) were enrolled and vaccinated in the fall of 2007 and influenza surveillance initiated at two study sites at the start of the influenza season in January 2008. The mean age was 74 years (range 60–92). Most of the participants were healthy adults (82.5%), 14% had heart failure (HF) and 6% had chronic obstructive pulmonary disease (COPD). Of the 315 participants, 90 (25%) reported ILI during the influenza season. The predominant circulating strains in the area of Vancouver, Canada, were influenza A/H3N2, and in the Hartford, CT area were influenza B, and this was reflected in the LCI detected at the two study sites. All LCI cases were documented between 10-weeks and 20-weeks post-vaccination when influenza was circulating in the community. According to the clinical and laboratory criteria, 19 subjects had confirmed influenza (20% of those reporting ILI). Thirteen of the 19 (66%) subjects had influenza A/H3N2 infections, and six (34%) had influenza B infections. An additional two subjects seroconverted to the A/H3N2 vaccine strain, and four subjects to the B strain in the vaccine, but reported no ILI, and thus could not be classified as LCI (i.e., asymptomatic seroconverters). Three LCI subjects had influenza A/H3N2 influenza confirmed by PCR but did not seroconvert to the infection. One influenza A/H3N2 case had a medically attended acute respiratory illness, was exposed to a confirmed LCI case, and showed an almost 4-fold increase in antibody titer but was not available for a nasopharyngeal swab at the time of illness. The results for the ILI and LCI cases are summarized in .
Figure 1 A flow chart of the results of influenza surveillance is shown for the two study sites (University of Connecticut Health Center [UCHC] and Vancouver). Laboratory confirmation for influenza A/H3N2 cases (Flu A) and influenza B (Flu B) are shown indicating (more ...)
3.2. ILI Symptoms Predicting Laboratory Confirmed Influenza
The most commonly reported symptoms by LCI subjects were coryza (94%), sore throat (68%) and cough (56%). Individuals with other respiratory illnesses reported coryza (89%), sore throat (71%) and malaise/fatigue (60%). Fever was the only symptom reported more frequently by the influenza cases (39% vs. 12.5%, p=0.009).
The average time from the onset of symptoms to reporting the illness was 3.7 days for subjects with influenza and 4 days for subjects with ILI without influenza (P=0.83). There was no difference between the numbers of symptoms reported by subjects with influenza A and influenza B (3.9 vs. 3.8). Flu positive subjects who seroconverted were not found to have more systemic complaints when compared to those who did not seroconvert.
As shown in , fever was the best predictor of LCI with the odds ratio of 4.45 and PPV of 44% (P=0.01; 95% CI 1.37–14.4). Fever combined with coryza increased the PPV (54%) for LCI. The presence of HF and COPD was also suggestive of increased risk of influenza but it did not reach statistical significance (OR=3.2, P=0.08) probably due to the limited number of subjects. The symptom complex of fever, coryza and shortness of breath with a history of HF or COPD was strongly predictive of LCI (OR=13.9, P=0.003). None of the other symptoms, signs or subject characteristics could distinguish influenza from other respiratory illnesses.
Clinical symptoms predicting influenza
Subjects with LCI due to influenza A/H3N2 or influenza B were separated in stepwise logistic regression model to identify clinical predictors. For influenza A, both coryza (OR=7.78, P=0.005; 95% CI, 2.0–56.7) and fever (OR 16.3, P=0.0001; 95% CI, 6.8–253) were strong predictors of influenza. For influenza B, only the history of HF or COPD was predictive of influenza illness when compared to the ILI not due to influenza (OR=5.4, P=0.02; 95% CI, 1.4–53). Three of the six LCI due to influenza B had CHF or COPD, while overall only 20% of subjects had CHF or COPD.
3.3. Granzyme B and cytokine responses to influenza infection
To determine the specificity of immune response to influenza, we compared the subset of LCI subjects with influenza A/H3N2 (N=13), to the rest of the ILI subjects including those with influenza B (N=77); as there is no immunologic cross-reactivity between influenza A and influenza B strains, LCI due to influenza B were combined with ILI due to other viral etiologies as the controls for this analysis. A similar analysis for the influenza B LCI group could not be conducted due to the limited number of cases and, thus, the subsequent analysis was restricted to the H3N2 LCI group, and cytokine and GzmB levels in A/H3N2-stimulated PBMC. provides the antibody titers for the H3N2 LCI subjects, showing that three of the 13 subjects did not seroconvert to the A/H3N2 infection as determined by the change in antibody titers from 10-weeks to 20-weeks post-vaccination.
Antibody titers for all LCI subjects*
To determine the predictive value of cytokine and GzmB levels, we analyzed the absolute levels of GzmB, cytokines, and the IFN-γ:IL-10 ratio prior to A/H3N2 infection (10-weeks post-vaccination), and the response to infection (change in level from 10-weeks to 20-weeks post-vaccination) in A/H3N2-stimulated PBMC. GzmB levels in influenza A/H3N2-stimulated PBMC from A/H3N2 LCI cases were lower compared to the other ILI subjects but this difference was not statistically significant (). As shown in , the change in GzmB levels from pre-infection (10-weeks post-vaccination) to post-infection (20-weeks post-vaccination) with influenza A/H3N2 distinguished those older adults with confirmed influenza A/H3N2 infection from those with ILI due to other viruses. GzmB levels were measured using the same validated assay of GzmB activity in both study laboratories [15
]. Mean GzmB activity significantly increased from 581.9 to 878.2 U/mg protein in response to influenza A/H3N2 infection (p=0.02) with a significantly greater increase in GzmB levels in LCI subjects with A/H3N2 compared to the rest of the ILI subjects (p=0.02) in whom no significant change in GzmB levels occurred (P=0.63).
Figure 2 PBMC stimulated 20 hours with live influenza virus and granzyme B (GzmB) activity measured in PBMC lysates stimulated with the A/H3N2 strain. Geometric mean GzmB levels represent combined data for the two study sites are shown for the pre- (0 wks) and (more ...)
Changes in Granzyme B among ILI subjects with or without H3N2 LCI
In contrast to the GzmB results, the change in the IFN-γ:IL-10 ratio from pre- to post-infection, reflecting the T helper type 1 (IFN-γ) relative to the T helper type 2/3 (IL-10) response, was not significantly different between the two groups. These results may be explained by the different methods for the cytokine assays used at the two sites. Cytokine levels were measured by a validated multiplex bead array assay method in the Vancouver laboratory, and ELISA-based assays in the Connecticut laboratory. We have previously reported that the limited dynamic range of the ELISA-based assays often necessitates dilution of the sample, which also dilutes any inhibitors in the sample and may contribute to an inflation of the calculated cytokine levels [16
]. The increased dynamic range of multiplex bead arrays minimizes the issues of sample dilution and affords the evaluation of multiple cytokines in a relatively small sample volume. Multiplex vs. ELISA-based assay techniques may account for the lack of difference in the IFN-:IL-10 ratio between H3N2 LCI and the rest of the ILI subjects when the results for the two laboratories were combined.
3.4. Fever correlates with low levels of GrzB and IFN-γ:IL-10 ratios
The next analyses were performed to establish linkages between the different immunologic measures and the clinical profile of LCI due to A/H3N2 infection. Due to the issues with the ELISA-based assays from the Connecticut site, we limited our analysis to the nine A/H3N2 cases from the Vancouver site. We postulated that fever was a sign of more severe influenza illness and analyzed the association between pre-infection levels of GzmB and cytokines (measured at 10-weeks post-vaccination, and the immune response to the infection (the change in levels from 10-weeks to 20 weeks post-vaccination). As shown in , H3N2 LCI subjects who seroconverted to the A/H3N2 infection (Sero), compared to H3N2 LCI subjects who were PCR+ but did not seroconvert to the infection (PCR+), showed a significant increase from pre- to post-infection in both IFN-γ and IL-10 levels in A/H3N2-stimulated PBMC (p=0.056 and p=0.04, respectively). H3N2 LCI subjects who were PCR+ had a lower IFN-γ:IL-10 ratio at the pre- and 4-week post-vaccination but the ratio was not statistically different from the rest of the subjects. However, A/H3N2 PCR+ subjects appeared to be a unique subset of the H3N2 LCI cases in that there was a statistically significant correlation between low levels of GzmB in A/H3N2-stimulated PBMC prior to infection (10 weeks post-vaccination) and the report of fever (vs. no fever) during influenza A/H3N2 infection, only in the A/H3N2 PCR+ group (r=1.000, p=0.01). In addition, these low GzmB levels were highly correlated with the IFN-γ:IL-10 ratio (r=0.999, p=0.03). Thus, illness severity (presence of fever) in the A/H3N2 PCR+ group compared to the A/H3N2 Sero group, may correspond to the low levels of GzmB and the IFN-γ:IL-10 ratio and the inability to mount an antibody response to infection.
Figure 3 Serum antibody titers to the three vaccine strains and IFN-γ and IL-10 levels in PBMC stimulated with A/H3N2 virus pre- (0 wks) and post-vaccination (4-, 10- and 20-wks) are shown for the Vancouver site. In subjects who developed influenza illness (more ...)
3.5 Other cytokines do not show a response to influenza infection
The analysis of other cytokine levels in influenza-stimulated PBMC supernatants by multiplex bead arrays was explored to determine which cytokine levels changed in response to influenza vaccination and infection in older adults. The IFN-γ and IL-10 response from pre- to 4-weeks post-vaccination was statistically significant (p<0.001 for both) but the IFN-γ:IL-10 ratio did not significantly change over this time period (). As shown in , a significant response from pre- to 4-weeks vaccination could be demonstrated for IL-1β, IL-2, IL-5, IL-6, IL-12, IL-17, and TNF-γ (p≤ 0.01) but a clear increase from pre-infection (10-weeks post-vaccination) to post-infection (20-weeks post-vaccination) with A/H3N2 could not be demonstrated with the limited number of A/H3N2 cases available for this analysis.
Figure 4 Cytokine levels in PBMC stimulated with A/H3N2 virus pre- (0 wks) and post-vaccination (4-, 10- and 20-wks) are shown for the Vancouver site. In subjects who developed influenza illness, subjects who were PCR+ only (PCR+, n=3) and those who seroconverted (more ...)