Post #741: Flu vaccine, other viral vaccines, first two factoids

Source:  US Centers for Disease Control.

I keep trying to write a science-based post on flu vaccine and vaccines for other viral illnesses.  So far, the more I learn, the less I know.  More-or-less everything I thought I knew about flu, and flu vaccine, and vaccine effectiveness, was wrong.  And you can take it from there.

So I’m going to start just by putting out a few factoids, in separate posts.  Just stuff that I stumbled across, in the course of this.

To recap, a lot of people seem to look at flu vaccine as the model for a COVID-19 vaccine.  So I am too.  But I’ll tell you up front, I don’t think there’s any particularly sound basis for thinking that flu vaccine is a good model.  A model, sure, but no guarantee it’s a good model.

First, the economics are completely different.  We’ve seen a lot of commentary about the long time it will take to produce a COVID-19 vaccine.  All of that was based on the production of other vaccines.  And, in my opinion, all of that was dead wrong.  Production of a COVID-19 vaccine is in no way analogous to prior situations, as I outlined in Post #698  .  The UK is still on track to have AstraZeneca deliver the first doses of the Oxford vaccine in September 2020.  As in, two months from now.  And the US has bought into a big share of the first billion doses of that vaccine.

Second, the vaccine itself is completely different.  The seasonal flu vaccine consists of an injection of the actual flu virus itself, either dead or weakened.  By contrast, the leading COVID-19 vaccine candidate (the Oxford/Astrazeneca vaccine), is a different live virus, genetically modified to express one of the key surface proteins of the COVID-19 virus.  (So, presumably, antibodies to that protein will then flag COVID-19, if you are infected.)

Different production techniques, different approach, but presumably some similarity in method of action.

But in any case, I learned so many new facts that I thought I would share.  So here are the highlights of two of them:

Asymptomatic flu infections are common.

Flu vaccine is only modestly effective even when the CDC correctly predicts which flu strains will be prevalent in flu season.

Asymptomatic flu cases are common.

The first time I’d come across the term “asymptomatic case” was in the COVID-19 context.  “Asymptomatic”, recall, means people who never, ever have symptoms (Post #711).  The Chinese epidemological studies said that was rare; US analysis seems to suggest that it’s common.

I thought this issue was somehow a unique feature of COVID-19.  It’s not.  It’s not even rare.  On average, 16% of flu infections are asymptomatic cases. 

I could belabor the details.  How do you know they had the flu if they never had symptoms?  (Testing for antibodies, mostly.)  How to you find them if they never had symptoms?  (Usually as a byproduct of some other type of screening.)

But, really, the details don’t matter.  What matters is that asymptomatic viral infections are a) not unique to COVID-19, b) common, and c) well-know, well-documented, and well-studied.

The only mystery is why seemingly important public health figures downplayed the potential for asymptomatic infection with COVID-19.  Did they truly not believe it, the clear example of flu nothwithstanding?  Did they not want to frighten the US population?  No clue.  Probably never will know.

The question of whether such cases are common is separate from the question of whether or not they are infectious (Again, recall, asymptomatic means never, ever having any symptoms, and does not include pre-symptomatic individuals in the period between infection and symptom onset.  Pre-symptomatic individuals are clearly contagious.)  It’s not even clear whether the US CDC has an official opinion on whether asymptomatic individuals can spread infection.  It’s also not clear that the question of transmission of flu, by asymptomatic individuals, has ever been adequately answered.

Flu vaccine is only modestly effective, at best, even when the CDC correctly predicts which flu strains will be prevalent in the flu season.

Some viral vaccines are nearly 100% effective, such as the vaccine for human papillomavirus.  Even Ebola — a deadly disease of long standing, for which most attempts at finding a vaccine failed — now has a vaccine estimated to be at least 95% effective.  But for many other economically important viral diseases, such as HIV, no effective vaccine has yet been produced.

And I thought flu vaccine was like that.  I thought that the mediocre average effectiveness of flu vaccine came from those years when the CDC guessed wrong about what flu strains would be circulating during flu season. So I thought that the (e.g.) 50% average effectiveness was the result of some years being 100%, and some years being more-or-less 0%.

Dead wrong.  At best, 60% flu vaccines effectiveness is what you get in a good year, when the CDC has guessed right, and everything works as intended.  See graph and citation as to source, above.

In years when the CDC guesses wrong, the effectiveness is far below that.  And sometimes, even when the CDC guesses right, the effectiveness is far below that.

So, when everything works perfectly, the best you can hope for is 60%.  And even that varies by patient characteristics, with flu vaccine being less effective in the elderly than in a younger population.

To be continued …

And here I start running into things that I still don’t quite understand.  So all I’ll say is, to be continued.  If I can figure them out.


Do I even know what “effectiveness” means?  No, I sure don’t.  And, so it’s a pretty good guess that most of the people writing for the popular press don’t know what it means, either.  Like “asymptomatic” (Post #711), “effectiveness” is a term that everybody thinks they understand.  But nobody bothers to define clearly, and most people really have no clue what it actually means.

I mean, you probably thought vaccine effectiveness was something like:

Effectiveness rate = People vaccinated who didn’t get flu / People vaccinated?

So, like, if 60% of those vaccinated didn’t get flu, then the vaccine was 60% effective?  Nope, not even close. Not even in the ballpark.  It’s not what the CDC means when they say the seasonal flu vaccine was 60% effective.

And, wait yet again.  Just how, exactly, does the CDC measure flu vaccine effectiveness?  Is it a controlled clinical trial?  No.  Not even close.  It’s based on observational data, and in a very weird way.  Does the CDC even measure the fraction of vaccinated individuals who didn’t get sick with flu?  Again, no, not even in the ballpark.

At present, the CDC’s preferred method is the “Test Negative Design”.  They only test individuals who show up at some health care provide with some acute respiratory illness.  (Here, acute merely means “not chronic”.  It doesn’t necessarily mean severe.)  They test those sick people for flu, and ask whether they got the flu shot.  And with that self-selected and ill cohort, they measure vaccine effectiveness as:

Flu Vaccine Effectiveness = 1 – (% of flu positives that were vaccinated/percent of flu negatives that were vaccinated).

E.g., 100 people show up at the clinic with acute (non-chronic) respiratory illness.  Half of them test positive for flu.  Half don’t.  Among the with-flu cohort, 40% got the flu shot.  Among the without-flu cohort, 60% got the flu shot.  From that, I think the CDC concludes that the flu vaccine was 33% effective (1 – (.4/.6)).

If that all made sense to you, you’re ahead of me on this one.  I can do the math.  I’m just not convinced that the math is telling me what I need to know.  Despite the fact that I’m a Ph.D. health economist, with lots of exposure to methods, I still can’t quite grasp why the CDC thinks that vaccination rate, conditional on illness, is is a good approach to measuring the effectiveness of flu vaccine.  As opposed to, say, something that’s convenient and cheap for them to do.

More to follow, if I can ever figure this out.