This week’s hot Internet story was Julia Ioffe writing in The New Republic about her whooping cough. [Mention on DK FP; Volokh Conspiracy; Discover] As she points out, whooping cough (also known as pertussis) had almost disappeared in the United States, but cases are on the rise. One reason is that a growing number of Americans are refusing to vaccinate their children.
As expected by anyone who has ever seen vaccination discussed on the Internet, Ioffe’s comments section rapidly filled with anti-vaccine cranks. Some approach from the right, with claims of personal autonomy that trump even the libertarian night-watchman state's power of quarantine. Others come from a narcissistic pseudo-Green belief that their Superman immune systems, fortified with expensive supplements, can take on all those pesky germs. How dare we ordinary people complain that in reality they are free-riding on our assuming the non-zero (albeit small) risks of vaccination! [Note: autism is not one of these risks.]
Either way the denialists have a major statistical problem: the near-eradication of several diseases in the course of vaccination campaigns. Flailing attempts to solve this problem include: Big Pharma has persuaded doctors to reclassify paralytic polio as something else; sanitation and hygiene improved; the disease rates were going down before vaccines; and people who are vaccinated still get sick (no one, of course, denied this).
No pseudo-scientific campaign is complete without pseudo-statistical nonsense, and in following the antivax links I encountered some real doozies. A few of them are generalizable to bad statistical reasoning in general, so for the details and maybe a good laugh, follow below the cronut.
1. Easy Warmup — What’s wrong here?
Chart of polio cases over time
Sorry about the poor jpeg, but if I made this mistake, I'd use a low-quality format too. This chart is supposed to show that polio incidence went up after vaccination.
This is strong evidence showing that the Polio vaccine is not responsible for the decline and should negate the belief and statement that the Polio vaccine demonstrates vaccine success. Below, see research that reflects what it stated above.
Now get out your magnifying glass. The chart shows the introduction of the vaccine against
diphtheria and
pertussis but the Y-axis measures the number of cases of
polio! You would think someone would have noticed this before spreading the chart to three different slickly-produced antivax web sites, with many links from Facebook and Twitter. In case you were wondering, the
polio vaccine was introduced at just the time this chart ends, 1954 and 1955. And they end their chart because polio cases sink like a rock: from 36,000 in 1953 to under 200 in 1961.
2. Missing Data Are Data They Don't Want You To See
The anti-vaxers pick the narrative back up in 1961. We shall see why the 1950s got disappeared.
In 1977, Dr Jonas Salk, who developed the first polio vaccine, testified along with other scientists that mass inoculation against polio was the cause of most polio cases throughout the USA since 1961 [Emphasis added].
Quite true! The Salk vaccine could cause polio; indeed one of the first batches was screwed up by the lab and caused ten deaths and much illness before being identified and withdrawn. But we’re talking about “most” of 100 cases, 10 cases, and in recent years no cases whatsoever. The author is trying to leave the false impression that it’s most of the tens of thousands of cases from the early 1950s. Here is the real picture.
Polio cases 1951–2003
3. A Little Bit of Data is a Dangerous Thing
Obviously, looking at that last graph, something happened around 1954. An alternative (in every sense) explanation is that doctors substituted diagnoses of aseptic meningitis for polio.
Here’s their evidence (using the word loosely).
Date |
Viral or Aseptic Meningitis |
Polio |
July 1955 |
50 |
273 |
July 1961 |
161 |
65 |
July 1963 |
151 |
31 |
September 1966 |
256 |
5 |
October 1966 |
312 |
3 |
Oh, and these numbers are only Los Angeles County!
Why do you suppose five months for just one county were cherry-picked out of years of data? Why does the Nigerian guy want your bank data? It’s a swindle: all you have to do is wait for a local outbreak of aseptic meningitis after 1955 from some cause other than polio, and presto, a data point. Unfortunately for this theory, recent data for viral meningitis in Los Angeles are
online, and between 2008 and 2011 the largest number of cases in
any month was 102, and the mean (average) less than 50, notwithstanding that the population increased by half. So what are they diagnosing polio as now? Chopped liver? Like most conspiracy theories, this one falls apart fast under scrutiny.
4. You Say Morbidity, I Say Mortality, Let’s Call the While Thing Off
Another frequently-cited antivax web site explains “there is no credible scientific evidence that the vaccine caused polio to disappear.” This claim’s original source is a parenting book by eccentric pediatrician Dr. Robert Mendelsohn, who, despite the importance he gave to diet over medicine, died aged 62 from a heart attack complicated by diabetes. Pro-vaccine Dr. Spock lived to 94. For some reason, Dr. Mendelsohn did not accompany his statement that polio declined in European countries that did not vaccinate as fast those that did with any data or even a source citation. The web site, however, continues with this remarkable chart.
Polio death rates
This is a chart of polio
mortality, i.e., death rates, normalized to 1920=100%. This is a bait and switch: polio vaccination reduces
morbidity, i.e., the number of cases. Both our chart and, ironically, the chart somehow claiming that the diphtheria-pertussis vaccine caused more cases of polio agree that polio
morbidity was climbing. One reason for the steep decline after 1920 is the invention of the
Iron Lung, that saved the lives of polio victims who would otherwise have died on the spot. Patients also had better access to doctors and hospitals and, for the latter half of this chart, antibiotic treatment of secondary infections. Why this is a point against polio vaccination which can drive mortality all the way down to zero is a complete mystery—unless, of course, the idea is to fool you into thinking that in the pre-vaccine era polio morbidity, with associated permanent paralysis, was going to zero at the same time, even when their own adjacent chart shows the opposite. Death was not the only life-changing outcome from polio: there were 20,000 cases of permanent paralysis from polio in the United States,
just in 1952.
5. At the Last Gasp, Botch the Math
Given the evidence that the number of cases of these diseases declined drastically after the introduction of vaccines, antivaxers tend to change the subject. Their reasoning skills, however, don’t improve.
Table 1 at the end of Dr Witt’s document shows the percent of cases in the vaccinated, and it as follows: 86% age 2-7, 86% age 8-12, 62% age 13-18, 81% age 2-18. So now you know who gets more pertussis. It is not the unvaccinated.
The implication is that the pertussis vaccine doesn’t work, but the
correct conclusion is that the pertussis vaccine does not
always work. (No one ever claimed otherwise.) The numbers here are about what we expect when the vaccine is only 80% effective. Let's do the math.
Take 1000 kids. 95% vax rate. Make the vaccine 80% effective. That says the susceptible population is 50 unvaccinated kids and 190 unsuccessfully vaccinated kids. That means that, on average, the cohort of sick kids will be 190/240=79% vaccinated! In other words, the very fact that the number of unvaxed kids is minuscule makes it more likely that those kids who do get sick come from unsuccessful vaccinations.
For vaccines that are more effective, even this seeming paradox does not occur. In the recent measles outbreak in the New York Ultra-Orthodox Jewish community, every victim was unvaccinated, either too young or by refusal!
The percentage to look at is not the number of vaccinated children in the sick group, but the comparative likelihood that an unvaccinated child gets sick. Dr. Witt’s complete study is available free online. The very same Table 1 shows that unvaccinated or undervaccinated children were, with 99% significance, twice as likely to get sick. Indeed, that was the purpose of Table 1; see the screenshot for yourself. Vaccines work.
Pertussis cases in Marin County outbreak by vaccination status
The author of this particular antivax travesty is a genuine M.D. turned quack. You can not become a doctor in the United States without a statistics class; in fact, she needed to understand the table well enough to derive the fact the majority of sick children were correctly vaccinated. So, why would she do this? A trip through antivax country shows that not only do they not
believe vaccines work, they
don’t want vaccines to work. The reasoning behind this selfish, fortunately false, belief is a good essay for another day.