Interpreting reports of Covid-19 vaccine adverse reactions
Updated: Jan 12, 2021
There have been a few incidents reported in the media of adverse reactions associated with COVID-19 mRNA vaccines. From cosmetic filler reactions, to death by low platelets. Some of these events have been tragic and close to home, adding to the fear and uncertainty we have all experienced the past year. Some of you have contacted me for my opinion. Let me try to help interpret these events for you:
Lets start thinking slow
The most widespread model of cognitive science posits that our brains process information using two different systems. System 1 and system 2.
System 1 is our immediate response - fast thinking. It's the “gut feeling” we get. The process of evolution has fine-tuned quick responses to gives us fast solutions to survive in the world around us. Our emotions play a large part in directing system 1 thinking. Hear a lion roar, feel fear, and run. A situation where time is critical. Very little thinking is required. This worked out well for early humans living on the savanna plains of Africa. It solved simple problems, and we are here to tell the tale.
Unfortunately, system 1 makes mistakes and it is not suited to figure out more complex problems. Problems that involve abstract thinking, large amounts of data, or multiple variables. For example, the causal dynamics of rare events linked to vaccination in millions of people.
System 2, on the other hand, is our logic and reasoning - slow thinking. It is located in our frontal lobes. As the name implies, it takes time, but yields much more accurate results. System 2 thinking acknowledges the potential errors of system 1 and tries to correct for them.
(If you want to read more on systems thinking read, “Thinking, Fast and Slow” by Daniel Kahneman.)
When we hear the story, “Someone was recently vaccinated, and shortly after, something bad happened to them”, our first response (system 1), is to experience fear, and run from the lion. This is normal. We all experience this (including myself). The next step is to slow down and engage system 2. Check if system 1 made any mistakes. So let’s do that...
Humans seem to find anecdotal evidence very compelling. Learning stories from others helped us survive on the savanna, "there is clean drinking water over there". Consequently, we tend to accept anecdotes at face value without critical evaluation. Knowing this, we must engage system 2, and learn to examine anecdotes properly. We must attempt to find out all the details of the anecdote, and whether the details are accurate before we start making any inferences. We need to ask the following questions:
What is the source of the anecdote?
Is the source reliable?
Does the source have expertise in medicine, epidemiology, and statistics to determine causation?
Has the anecdote been examined by an independent agent?
With respect to the sudden and tragic death of the Miami Beach doctor, the autopsy has not yet been finalized or published. Before making an inference that the vaccine caused his untimely demise, should we not await the details of formal evaluation? Maybe pathological evidence will surface that demonstrates another cause of his low platelets? Or demonstrate that his condition was pre-existing? Has someone checked his bloodwork prior to the event? Do we have enough accurate details to make a valid inference? At this point, NO.
If we, or someone close to us, experiences an event, we perceive that event to be more likely to happen than it actually is. This is known as the availability heuristic - a classic system 1 error. As an example, suppose a friend of yours died from a brain tumor in the past, and you began to develop a headache. You are more likely to think you have a brain tumor than if you didn’t have a friend that died from a brain tumor. Yet, the likelihood of you having a brain tumor is a matter of statistics, independent of the experiences of the people around you. As it happens, the likelihood of a malignant brain tumor in the general population is low (less than 1 in 200 lifetime risk), and there are exceedingly more common things that cause a headache (migraines, tension headaches, sinus congestion, TMJ, etc.). To make it even less likely that you have a brain tumor, most brain tumors present with neurologic symptoms - not headaches.
The death of the Miami Beach doctor certainly was close to home. This may incite the availability heuristic in some of us. Consequently we begin to believe that the likelihood of death from COVID-19 vaccine is high. Yet, this is an isolated event, given that 20,000,000 vaccines have been administered worldwide. We should not infer that this is a common occurrence just because it happened so close to home.
Correlation does not imply causation
Someone takes the vaccine, and then something bad happens afterwards. It is easy to see how system 1 thinking can infer causality. We are pattern seeking creatures. It’s the quick and easy answer. If event A happens after event B, then B caused A. However; this is a logical fallacy. Probably one of the most common logical fallacies.
Why is this a fallacy? The sun comes up every day after the rooster crows. However, one cannot infer that the rooster caused the sun to rise. It is possible that A caused B (reverse causality), or another element C caused both A and B (confounding), or that A and B have nothing to do with each other (coincidence). We must entertain all of these possibilities before we infer causality.
This fallacy explains why many people so strongly believe that: vaccines cause autism, Hydroxychloroquine cures COVID-19, prayer cures cancer, acupuncture helps infertility, and reiki works. They inferred causality because they, or someone they know, got better or worse after a treatment.
When it comes to the COVID-19 vaccine, these recent isolated cases of adverse reactions are most likely coincidences. Why do I say this?
The law of very large numbers
There have been 20,000,000 adults vaccinated thus far. This is a very large number. It’s not that rare events and coincidences CAN happen with such a large number, it’s guaranteed that they WILL happen. This is known as the law of very large numbers.
Let’s conduct a thought experiment. We will choose 20,000,000 adults randomly and follow their health over the next 2 months - starting now. At the end of the two months how many sudden unexplained deaths, episodes of Bell’s Palsy, episodes of unexplained low platelets, episodes of cosmetic fillers causing swelling, will occur? The answer is: many. For instance, there will be about 150 cases of Bell’s Palsy and 200 cases of unexplained low platelets. Now lets reverse the clock back to today, and give those 20,000,000 adults the COVID-19 vaccine. You can see how these same events that occurred out of the blue would be linked to the administration of the vaccine.
In fact, in the Pfizer trial of 40,000 patients (See published trial here) there were two fatalities reported in the vaccine group: both died of heart attacks after the vaccine. But these events were not caused by the vaccine. Why? Because four patients died in the placebo group - two from unknown causes, one from a stroke, and one from a heart attack. Point here is, stuff happens. Even to the placebo group given sugar water. Frankly, I am surprised that after the vaccination of 20,000,000 people there are only a handful of adverse events associated with the vaccine. There should be more, just by pure coincidence alone.
Unfortunately, if you are one the few that experiences these coincidences, system 1 thinking will engage. I don’t think anyone or anything could convince you otherwise.
Control groups are important
Because there is a baseline event rate of disease in people - bad things happen to good people every day - we need to compare the vaccine group to a control group. This is exactly what was done in the two large trials with 40,000 people each. During these trials there were no reported events of Bell’s palsy, or death from low platelets. This does not prove that the vaccine doesn’t cause these adverse reactions. However, it suggests that if the vaccine did indeed cause these adverse events, they are rare. Less than 1 in 10,000.
What can we learn from the above:
Anecdotes often contain inaccurate and incomplete information.
Despite this, anecdotes trigger strong emotion, they are very compelling, and hard to dispel.
Even more so with anecdotes closer to home.
Always wait for a formal examination of an anecdote.
Always look for alternative explanations for an anecdote.
Realize that coincidences are more common than you would expect.
Realize that coincidences with very large numbers are inevitable.
Controlled trials are the most reliable form of evidence to determine causality.
Is the vaccine causing these reported events? The answer is, we don’t know. But if it is, the events are rare. Wait for each report to be investigated properly. At this point, the likelihood of coincidence is high.
(The above exercise was a critical thinking approach to medicine. For a more in-depth explanation of critical thinking in medicine see here.)