Experts are puzzled by a new study in which influenza vaccination seemed to provide little or no protection against flu in the 2010-11 season—and in which the only participants who seemed to benefit from the vaccine were those who hadn´t been vaccinated the season before.

The investigators recruited 328 households in Michigan before the flu season started and followed them through the season. Overall, they found that the infection risk was nearly the same in vaccinated and unvaccinated participants, indicating no significant vaccine-induced protection, according to their report in Clinical Infectious Diseases. That contrasted sharply with several other observational studies that found the vaccine to yield about 60% protection during the same season.

In trying to figure out why the effectiveness was so low, the researchers sifted their data in different ways, said Arnold S. Monto, MD, of the University of Michigan, senior author of the study. "We discovered that if you separated out those that had not been vaccinated the previous year, you got percentages close to what were seen in the major vaccine effectiveness studies," he told CIDRAP News.

"We were playing with this for a long time, and there was clear interaction of sequential vaccination and vaccine effectiveness, looking at it in a strictly statistical way," he added. "We felt it had to be separated out."

The vaccine was found to be 62% effective in those who hadn´t been vaccinated the previous year. That was similar to findings in the other observational studies and also to the results of a recent, rigorous meta-analysis of randomized controlled trials. In contrast, those who had been vaccinated 2 years in a row (before both the 2009-10 and 2010-11 seasons) got no significant protection.

An additional finding was that the vaccine did not seem to protect participants who were exposed to flu in their own household, though the numbers in that arm of the study were small.

Researchers from the US Centers for Disease Control and Prevention and the University of Hong Kong collaborated with University of Michigan researchers on the study, with Suzanne E. Ohmit, DrPH, of Michigan as the lead author.

The findings come amid a growing number of studies that raise questions about flu vaccine effectiveness (VE). They include, among others, last week´s CDC report that this year´s vaccine has worked poorly in elderly people and three recent European studies showing that vaccine-induced immunity in the 2011-12 season waned after 3 to 4 months. Other studies have cast doubt on the long-standing belief that a close match between the vaccine virus strains and circulating strains improves VE.

In an editorial commentary accompanying the Michigan study, John Treanor, MD, and Peter Szilagyi, MD, both of the University of Rochester Medical Center, wrote, "As we are currently struggling through one of the most vigorous influenza seasons in recent memory, the apparent failure of influenza vaccine under optimal conditions seen in this study is indeed troubling."

And Edward Belongia, MD, a Wisconsin clinician-researcher and member of the CDC´s Influenza Vaccine Effectiveness Network, said he was perplexed by the low overall VE in the study, given the approximate 60% protection levels found in studies by the network the same season. "I don´t know what to make of it," he told CIDRAP News.

Other researchers have said that additional studies suggesting a negative effect of prior-year vaccination on flu VE will be emerging in coming months, but they declined to give any details.

Aiming to detect all cases
The researchers used a prospective cohort design in an effort to detect all flu cases in the study group, regardless of whether or not participants were sick enough to seek medical attention.

The team sought to recruit households that had at least four members with at least two children and that received medical care through the University of Michigan Health System, based in Ann Arbor. Out of a target group of 4,511 households, the authors recruited 328, with 1,441 members.

Participants were instructed to report any acute respiratory illnesses throughout the flu season. Individuals with symptoms went to a study site for collection of a throat swab for flu testing. The researchers followed the illnesses to collect data on disease course, including whether the volunteers sought medical attention. Specimens were tested using polymerase chain reaction (PCR).

Among the 1,441 participants, 866 (60%) had documentation of receiving a flu shot for the 2010-11 season, with coverage lower among younger adults and higher in those with high-risk health conditions. Of those vaccinated, 88% received an inactivated vaccine and 12% the live-attenuated vaccine.

During the season, 624 individuals reported 1,028 acute respiratory illnesses, leading to the collection of 983 specimens. Of those, 130 specimens from 125 participants (13%) were positive for flu. By subtype, 45% were influenza A/H3N2, 34% were type B, and 20% were 2009 H1N1. Thirty-two percent of the cases led to medical attention.

Among the 125 people who tested positive for flu, 59% had been vaccinated at least 14 days before their illness onset, long enough for an immune response. The infection risk in the vaccinated people was 8.5% (74 of 866), versus 8.9% (51 of 575) in the unvaccinated individuals.

Community vs household transmission
The researchers estimated VE separately for community and household exposures. Ninety-seven flu cases were classified as community-acquired and included in the analysis. After adjustments for age and high-risk medical conditions, the all-ages VE was estimated at a nonsignificant 31%, (95% confidence interval [CI], –7% to 55%). VE estimates by age-group were similar and likewise nonsignificant.

The result was very different when the team stratified the participants according to whether they´d had a flu immunization the previous season. As noted above, estimated VE in those with no prior-year immunization was 62% overall (95% CI, 17% to 82%), whereas VE in those who did get vaccinated the year before was low in all age-groups and came out to –45% overall (95% CI, –226% to 35%).

The team defined a household-acquired case as one that occurred within a week after another case of the same subtype in the same household. On this basis they determined that 30 flu cases were household-acquired. The estimated VE for this group was –51% overall (95% CI, –254% to 36%), and the age-group estimates were all low.

"Adults were at particular risk of infection despite vaccination," the report says. "In fact, 9 of 11 (82%) adults with household acquired influenza were vaccinated, compared with 11 of 19 (58%) children." In this group the team found no major differences related to prior-season vaccination.

The authors found that the flu risks were similar for adults who were vaccinated in both years and those who weren´t vaccinated in either year. The pattern was slightly different in children under 9 years old, in that those with no vaccination either year had the highest risk of infection.

Summing up, the report notes that VE estimates against community-acquired flu of all severities were all less than 40% and "not statistically different than zero" (because of confidence intervals that overlapped zero). "This unexpected finding was seen in a season with circulation of influenza strains that were considered matched to vaccine strains, and where evaluation of vaccine effectiveness using case-control designs indicated significant reductions of 52 to 60% in medically attended influenza outcomes in vaccinated patients of all ages."

Monto said possible explanations for the low VE within households include that the vaccine may be "overwhelmed" by continual exposure to an infected family member, particularly since children shed more virus than adults.

He said his team is working on further studies of flu VE in the community and households and is collecting blood samples to examine immune responses to vaccination and infection, a step that was not possible in the current study. That may help shed some light on the unexpected findings, he said. For now, "We can only speculate about what´s really going on from an immunology standpoint."

Monto commented that the study raises tough questions. "We recommend vaccination every year because we know the duration of protection is relatively short. What are we to do if we know that being vaccinated every year is perhaps not the best way to get good vaccine effectiveness?" he said.

Intriguing and troubling
In the accompanying commentary, Treanor and Szilagyi call the findings "intriguing" as well as troubling. They suggest some factors that may help explain the differences between the current findings and other VE studies, but they make it clear there are no easy answers.

Treanor and Szilagyi contrast the approach used in the Michigan study with the test-negative case-control design, which several large research networks have been using to assess flu VE. In the latter design, patients seeking care for an acute respiratory illness are tested for flu and their vaccination status is determined. The case-control approach has important pluses, but it is "somewhat incomplete" because it is limited to medically attended cases.

In comparison, Ohmit and colleagues were able to assess VE against both medically attended and unattended illnesses, as is true in randomized controlled trials (RCTs), Treanor and Szilagyi write. But the findings in this case were "strikingly different" from those in some recent RCTs and case-control studies of flu VE.

Various undetected biases might help explain the low VE found in the study, Treanor and Szilagyi say. For example, people who choose to be vaccinated may be more health conscious and more likely to report illnesses, compared with those who don´t get vaccinated. Also, the households that enrolled in the study—only 7% of the target group—may differ from the general population in some way.

Treanor and Szilagyi say concerns about the possible effect of previous vaccination on VE have been raised before, particularly in a 1979 study of students in British boarding schools. But later randomized trials did not show a consistent effect.

Given the many persistent questions about flu VE, it may be time to rethink the view that randomized trials are unethical, the two commentators suggest. "Given that the effectiveness of the vaccine is unclear, [that] the subjects in such studies are typically at extremely low risk of serious disease, and that effective antiviral therapy is available, perhaps [the ethics] should be reconsidered," they write.

More research needed
Angus Nicoll, MB, director of the influenza program at the European Centre for Disease Prevention and Control in Stockholm, praised the study and said the question of prior-year vaccination clearly needs more investigation.

"Our bottom line is that immunization is the most effective single thing you can do to protect yourself [from flu], and this isn´t going to change what we say," Nicoll said. But he added, "It´s an important finding, and this does now need to be looked at in the longer term and a larger cohort." He commented that the question calls for study in a stable community where the turnover of residents is not too high.

The study also won praise from Belongia, who has studied flu VE extensively at the Marshfield Clinic Research Foundation in Wisconsin. "I think they did a fine job with the study," he said. "I applaud them for trying to do a community-based study, which is hard to do these days."

He agreed that the finding of an effect of prior-year vaccination is important. "It needs to be looked at in other populations and seasons," he said. "The numbers are relatively small in this study. As the authors note, the majority of people who get the vaccine get it year after year, so there may be important differences between those who get vaccinated repeatedly and those who just recently chose to do it."

As noted above, Belongia was particularly puzzled that the overall adjusted VE in the Ohmit study, at 31%, was only about half what was found in case-control studies the same season. "I think a key message is that we need more community-based studies, with PCR-confirmed outcomes," he said.

Another flu vaccine researcher, Heath Kelly, of the Victoria Infectious Diseases Reference Laboratory in Melbourne, Australia, said the suggestion that prior-year vaccination affects flu VE is not new, pointing to a study of British children in 1979. He noted that another research group developed a model suggesting that this effect is related to the antigenic distance between the current and previous vaccines and the circulating viruses.

Kelly said he found it "intriguing" that the Michigan study failed to find a significant protective effect of vaccination, "given that many observational studies in Europe, Canada, and the US found moderate protection against medically attended, PCR-confirmed influenza in the 2010-11 season."

He remarked that the 62% effectiveness seen in those who were not vaccinated the previous year is similar to other published estimates, mainly from sentinel surveillance programs. "Although it seems unlikely, could it be that the sentinel schemes include a majority of people who were not previously vaccinated?" he asked.

Another flu expert, Michael T. Osteholm, PhD, MPH, said the findings further complicate the already difficult challenge of framing flu vaccination recommendations. Osterholm, director of the University of Minnesota´s Center for Infectious Disease Research and Policy, which publishes CIDRAP News, was the lead author of a lengthy 2012 report on the flu vaccine landscape and the need for better vaccines.

"We´re at a major crossroads in integrating our current influenza vaccine science with our current flu vaccine recommendations," he said. "The issues of vaccine efficacy by age and by vaccine [formulation] as well as the concept of waning immunity in a given season, the lack of correlation between vaccine virus match with circulating viruses and protection, and the potential for repeated annual vaccination to lower one´s protection, versus not being repeatedly vaccinated, are all immense challenges for us today.

"If we don´t go back and revisit our current vaccine recommendations, I think we stand to lose a great deal of credibility with both the medical community and even the general public as to the trustworthiness of what public health concludes and promotes," he said. "This is exactly why we need game-changing influenza vaccines."

Ohmit SE, Petrie JG, Malosh RE, et al. Influenza vaccine effectiveness in the community and the household. Clin Infect Dis 2013 Feb 14 (early online publication) [Abstract]

Treanor JJ, Szilagyi P. Influenza vaccine—glass half full or half empty? (Editorial Commentary) Clin Infect Dis 2013 Feb 14 (early online publication)