In 2004–2005, the first year of our ongoing study, subject enrollment fell short of projected targets; however, the influenza season was a typical one, with various indicators above epidemic threshold levels [3
]. A drifted type A (H3N2) virus, A/California/07/2004, and type B viruses of both lineages circulated. In our study in that year, the attack rate in the placebo group as measured by virus isolation and/or real-time PCR identification was 7.8%, and thus we had sufficient statistical power to assess the significant efficacy of either vaccine against placebo. The vaccine efficacy for this combined end point was 75% (95% CI, 42%–90%) for the inactivated vaccine and 48% (95% CI, −7% to 74%) for the live attenuated vaccine [7
In 2005–2006, more participants than projected were enrolled; however, the influenza season was protracted and of low intensity. Attack rates in the placebo group, as determined by virus isolation and/or real-time PCR identification, were only 1.8%—far lower than expected in most years [8
]. With such low attack rates, statistically significant reductions produced by vaccine were difficult to demonstrate, and in fact, neither vaccine showed significant benefit over placebo for the “virus isolation and/or real-time PCR identification” end points. The primary end point (which was determined when the Investigational New Drug agreement under which the study was carried out was approved) —virus isolation combined with serologic identification of infection—produced an attack rate in the placebo group of 4.7%, and higher efficacy of the vaccines was demonstrated. The absolute efficacy of the inactivated vaccine was 54% and significantly greater than that of the placebo, whereas the absolute efficacy of the live attenuated vaccine was modestly less. Interpretation of efficacy estimates using serologic end points must be approached with caution; it is known that, for several reasons, they will favor the inactivated vaccine over the live attenuated vaccine [12
In 2004–2005, the live attenuated vaccine performed less well than the inactivated vaccine when each was compared with the placebo group, but in the second year of the study there was little evidence of a difference. How can this be explained? First, in contrast to the first year, the low attack rates in 2005–2006 made statistically significant reductions produced by either vaccine difficult to demonstrate. Second, in the first year, the differences between the 2 vaccines appeared to be related, in part, to differences in efficacy against type B influenza, with the live attenuated vaccine performing less well; the 2005–2006 season did not offer an opportunity to test efficacy against type B influenza. Likewise, the type A/H3 virus circulating in the first year of the study was drifted from the vaccine strain, whereas in the second year of the study, the circulating virus was similar to the vaccine strain. Finally, it is possible that there was variable susceptibility to the live attenuated vaccine virus at the time of vaccination in year 1 of the study, compared with year 2; for example, the A/H3 vaccine virus in 2004–2005 was similar to the Fujian strain, a virus that had caused major outbreaks during the previous year [14
With the low attack rate in 2005–2006, the efficacy estimates were unstable and varied with the particular laboratory criteria that were used. Still, we did demonstrate continuing efficacy at least for the inactivated vaccine. Because vaccine efficacy can change from year to year and because, given the experience in the previous year, efficacy may not be consistently related to drift, further evaluation of vaccines in different situations is warranted. In particular, it is important to determine additional correlates of protection, of which serologic antibody titer is only one. All of this will help to develop improved and perhaps more predictably efficacious vaccines.