Does flu vaccination work for children?

Flu season is upon us and with it the perennial question, should I get a flu shot. In 2006 the Advisory Committee on Immunization Practices (ACIP) extended its 2004 recommendation that children 6 months to 23 months be vaccinated (2 doses) to now include children 6 months to 59 months (<5 yyears old). I'm in a targeted group to get vaccinated (over 65 years) and my three grandchildren are in the targeted group for children (one will slide into the window during flu season). What do we know about the usefulness of getting vaccinated? It's been studied, but unfortunately the results are confusing. The latest confusing study to appear is from the New Vaccine Surveillance Network that CDC established in 1999 to look at exactly these kinds of questions. The Conclusions section in a just published paper says they were unable to demonstrate vaccine effectiveness in the 6 month to 59 month age group. That's not very good news, so we took a hard look at the paper and found much less there than appears in news summaries or even the paper's own abstract. After our own summary of the paper (it is available online so by all means make your own judgments if you have a vital interst in this) we'll let tell you what we are going to do and what what advice I have about our grandchildren, and what the advice is based on. First, the new study.

This is a hybrid design called a case - cohort study. It was designed to estimate the effectiveness of inactivated flu vaccine to prevent either inpatient (hospitalized) or outpatient (Emergency Room and doctor's office visits) flu-related events. It was carried out in three counties (around Rochester, NY; Memphis, Tennessee; and Cincinnati, Ohio) and compared how often laboratory confirmed cases of influenza were vaccinated with vaccination rates in children of the same age and at the same time who didn't have a flu-related event. If the vaccine works, we would expect the rate of immunization to be lower higher in the children that didn't have flu-related events compared to those that did. Since there were a variety of important differences within and between the groups, statistical methods were used to make it most likely the researchers were comparing like with like. In almost all the comparisons (between counties, age groups, types of insurance, etc.) the researches did in fact see that immunization rates were lower in the kids who got sick compared to the ones who didn't. This sounds very much like a demonstration that the vaccine was effective, something that has been shown in some other studies of the pediatric population as well. Yet the paper concludes:

Each year, US children aged 6 to 59 months experience high rates of hospitalizations, ED visits, and outpatient visits due to influenza. Despite this, we were unable across 3 large communities to demonstrate that influenza vaccination was effective in preventing influenza-related inpatient/ED visits or outpatient visits during 2 consecutive seasons (2003-2004 and 2004-2005) among 6- to 23-month-olds, 24- to 59-month-olds, or the entire age span. (Szilagyi et al., Arch Pediatr Adolesc Med. 2008;162(10):943-951)

In our view this is misleading. Examination of the paper itself reveals many reasons, all acknowledged by the authors, why the efficacy of vaccination might not have been as high in this population. They include the use of inactivated vaccine, thought to be less antigenic than live virus vaccines in children; known poor vaccine match in the years studies; uncertain but large potential biases in the inefficient case - cohort design, especially for accommodating bias from propensity of care seeking behavior; and relatively weak statistical power. The first three are inherent biases (systematic errors) but the last pertains to evaluation of random error. It simply is not true, examining the tables in this paper, to say that the data do not show vaccine efficacy. They in fact show better results for vaccinated children. But because the results are not "statistically significant," what remains in doubt is whether the apparent benefit of the vaccine is due to some random fluctuation one would get in choosing any sample from two groups who are otherwise the same (e.g., getting 6 heads and 4 tails in a coin flip ten times) or whether it is due to one of the biases just mentioned (although most of those effects would probably tend to mask effectiveness, not produce it) or whether it is due to the vaccine really being effective in preventing flu-related medical visits, or all three, perhaps pushing in different directions. This study is essentially non-informative of the main question. They tried hard. It is well done. They got it published. But it still didn't provide much information.

So what am I going to do, and why? I will, this year as in other years, get my flu vaccination. The data on whether it helps or not is a bit murky but I am betting it does. We know the vaccine raises neutralizing antibodies against influenza A virus and what we know about immunology suggests that affords protection. The cost of getting the flu is high. The risks of being vaccinated are quite low. For the same reasons I am going to advise the parents of my grandchildren to get them vaccinated. At least one was vaccinated last year, when he was 7 months old. When the one who is 3 months old gets to 6 months, in mid winter, I'll advise my daughter to get him vaccinated, too. Same for the almost 5 year old.

That's if the parents ask. I've discovered that while the world at large considers me an expert in these matters, my reputation doesn't extend to my nuclear family. At least they consider me harmless.