Incidence of Influenza A(H1N1)pdm09 Infection, United Kingdom, 2009–2011

We conducted a longitudinal community cohort study of healthy adults in the UK. We found significantly higher incidence of influenza A(H1N1)pdm09 infection in 2010–11 than in 2009–10, a substantial proportion of subclinical infection, and higher risk for infection during 2010–11 among persons with lower preinfection antibody titers.

The incidence of infection over the third pandemic wave was significantly higher (p = 0.02) than over the second wave (Figure 1). Among participants with prewave titers <8, the incidence of infection was significantly higher over the third wave than over the second wave (p<0.001); incidence did not differ for participants with prewave titers >8 (  Table 2).
During an illness episode, 20% of infected participants reported fever or ILI, 17% visited their general practitioner, and none visited a hospital ( Figure 2). Because predictions Light gray bars indicate influenza A of all subtypes; dark gray bars indicate the number of A(H1N1)pdm09 cases detected by virologic national surveillance. Healthy adults were recruited after the first pandemic wave (April-August 2009) had ended in the United Kingdom and were followed over 2 influenza seasons, with serum samples collected before and at the end of each influenza season. The median time between visits is shown. The second wave was defined as baseline (September-November 2009) to first follow-up (February-April of a small third pandemic wave were disproved (4), the reasons for this large wave remained unclear. Multivariate logistic regression was undertaken with infection as the dependent variable and age, sex, and prewave titers as independent variables. Each doubling increase in prewave HI titers, after adjustment for age and sex, was associated with significantly lower risk for infection (odds ratio 0.92, 95% CI 0.9-1.0, p = 0.04) during the third, but not the second, wave ( Table 2, Appendix).

Conclusions
Incidence of A(H1N1)pdm09 infection was significantly higher among healthy adults during the third pandemic wave (2010-11) than during the second wave (2009-10). This study complements and corroborates clinical surveillance data and population-sampling seroepidemiology from the United Kingdom (4,6,7), United States (8) and elsewhere (9).
The reasons for this unexpectedly larger third wave in the postpandemic season remain unclear. We show an increased risk for A(H1N1)pdm09 infection associated with lower antibody levels at the start of the season, irrespective of age, during the third, but not the second, wave. Because no substantial viral genetic change occurred between the waves (7), our finding suggests that the third wave was driven by infection among susceptible persons remaining antibody-naive at the end of the second wave. This thesis is supported by serosurveillance data showing lower infection rates over the third wave among age groups with the highest infection rates over previous pandemic waves (7,8). Our interpretation is further strengthened by a meta-analysis of serologic data from 19 countries that showed 20%-27% incidence of infection during the first pandemic year, suggestive of a large population susceptible to infection in subsequent seasons (10).
Incidence in our cohort was lower than that estimated for England by cross-sectional serosurveys (7,11). This finding may reflect our accounting for individual-level vaccination status and baseline antibody titers; data usually unobtainable with cross-sectional population-sample serosurveys. However, our study did not include children or elderly persons, which limits the generalizability of our findings. A major advantage of longitudinal cohort studies recording clinical data is identification of subclinical and asymptomatic infections. More than 80% of participants did not seek primary care or have surveillance-defined ILI indicating a high proportion of subclinical infection among healthy adults undetectable by routine case-based surveillance. We also describe persons shedding virus without antibody seroconversion, a phenomenon recently reported in Vietnam and the United Kingdom (4,12). Although these nonseroconverters might have antibodies detectable by microneutralization assay, such nonseroconverters, undetectable by serosurveys using the standard HI assay, further highlight the possibility of underestimating community infection rates when cross-sectional serosurveys alone are used.
Despite our intensive symptom ascertainment, 4 participants with influenza reported no symptoms. Cross-reactive cellular immune responses that are highly prevalent in the population (13) have recently been shown to be associated with protection against symptomatic illness (14).
Our analysis of pandemic influenza in a community cohort over successive seasons offers insight into contributors of the unexpectedly larger third pandemic wave. Our analysis also highlights the necessity of using cohorts to complement routine case-based surveillance to estimate influenza burden. , and entire study period, United Kingdom. Proportion of persons with reported symptoms over the study period is combined from the second and third waves. Symptoms were recorded by a Web-based symptom questionnaire emailed to participants every 3 weeks. Symptoms associated with illness episode were acute respiratory infection (ARI; illness episode with any symptoms), influenza-like illness (ILI; episode with fever plus cough or sore throat), fever (recorded temperature >38°C) alone, or visit to a general practitioner (GP). The graph depicts the average with 95% CIs calculated by using binomial distribution.