Wednesday, March 12, 2014

What doctors and anti-vaxxers have in common: Part 2

When I openly pose the post-vaccine fever scenario, most people consider it more likely than not that the vaccine caused the fever. That is, most people think the chance that the vaccine caused the fever is > 50%. Almost every medical student in a group of 20 that I was teaching a few weeks ago thought that. After all, fever is a well-recognized side effect of vaccine injections and your child was perfectly fine until a few days after the injection. Who wouldn’t reasonably conclude that the vaccine likely caused the fever?

Notice that by focusing responses on 1-49%, I set the question up to give you a better chance at being correct than they were but still, nobody got the question right. Don't worry. You are in good company. For now, just remember the feeling you had when you made your choice.

Back in the 80’s, investigators were concerned that combining mumps, measles, and rubella vaccines into one (MMR) might increase the risk of side effects, so some doctors in Finland did an ingenious study to determine what was not just temporally associated with vaccine injection, but caused by vaccine injection.

They studied 581 pairs of twins and randomly gave the MMR injection to one of the twins and a placebo injection to the other. Then, so as to not deprive the children that got a placebo injection of the vaccine's proven benefits, three weeks later, they gave each twin the opposite of what they had earlier received. Parents, who were blinded (they didn’t know when the kids were getting the placebo or the MMR injection), were instructed to vigilantly check for fever and a variety of other potential side effects after each injection.

In the first 6 days after placebo injection, 17% of children had a fever. That’s the background frequency of fever in vaccine-age children, and it's pretty high, huh? On the other hand, in the first 6 days after MMR injection, 17.2% of children got a fever. The difference - 0.2% - is what can be causally attributed to the active ingredients in the vaccine.

Conclusion: >99% of fevers that occur in the first week after MMR injection have nothing to do with the vaccine at all

However, we are very likely to erroneously attribute the fever to the vaccine. We do this because we have a powerful intuition that leads us to identify a causal relationship when 2 events that could possibly be causally related follow each other in time. We are mistaking what is merely possible for what is probable. Unfortunately, children develop mysterious illnesses like MS, autism, epilepsy, hepatitis, arthritis, etc. with regular frequency. Sometimes, those illnesses will appear fairly soon after getting a vaccination, and that’s when that intuition of ours can do some serious damage. That same intuition does damage when mysterious conditions like these disappear -as they often spontaneously do- after interventions that have no effect on the disease like prayer, acupuncture, chiropractic neck manipulations, homeopathic remedies, etc. Only randomized controlled trials can sort this stuff out.

Imagine once again your son who got a fever early after the MMR injection, only now, he is a healthy 18-year-old man who was spared the ravages of mumps, measles, and rubella. At any point in his upbringing, would you have ever had cause to stop and be thankful for the vaccine? How could you? You’d have had no way of knowing whether your child would have contracted one of these illness had he not been vaccinated!

See what’s happening here? The way that we experience life offers us no way to identify the situation where a vaccine prevents a horrible illness or death. On the other hand, we have a tendency to erroneously attribute adverse events to vaccines when they follow each other in time. Remember how you felt about your child's fever after MMR vaccine? Parents considering whether to vaccinate their children can anticipate the regret that they will experience if their child does develop an illness like autism afterwards, but they cannot anticipate the relief that they cannot experience when their child is spared a vaccine-preventable illness. Anti-vaxxers are much more likely to anticipate the former type of regret than the latter because they tend to overestimate the risks of vaccines and underestimate their benefits. Our intuitions erroneously set vaccines up to be unattractive. 

And so it is with physicians and oral anticoagulants (OAC’s). Bleeding events are fairly common. Whenever somebody experiences a bleeding event on an OAC, they and the doctors tasked with treating the bleeding blame the OAC. The reality is that most bleeding events that happen while on an OAC would have happened otherwise: just as only 0.2% of fevers early after the MMR injection were caused by it, only 0.2-0.3% of bleeds /yr are caused by the OAC. But physicians can anticipate the regret that they will experience when they prescribe an OAC to a person with atrial fibrillation (AF) who then develops a bleeding event. The finger of blame will point to the OAC and the physician who prescribed it. Unfortunately, nobody ever returns to the prescribing physician to pat them on the back and thank them for the stroke that they and the OAC prevented because, just like the case of mumps that your son avoided, there is no way to recognize a stroke that would otherwise have happened. Doctors' intuitions erroneously set OACs up to be unattractive.

Our ancestors evolved on an African Savannah with no pressures to select for intuitions that address the types of complicated primary prevention questions posed by vaccination and OAC use in AF. It seems that we did evolve intuitions that lead us to readily identify patterns and infer causal relations where there often isn't one. On the whole, this rudimentary heuristic has done well to protect us from certain kinds of dangers, but, as Sam Harris has written, “we have flown the perch built for us by evolution”. If we are to make advances with complex questions, we simply have to recognize when our intuitions lead us astray, and they regularly do, whether we are specialized physicians or lay people. None of us are spared the consequences of irrationality. It is a struggle we all must recognize and participate in.

If you are faced with complex decisions like vaccinating or taking an OAC, I hope that you are most heavily weighing the evidence from RCTs. If you aren't, then I hope that you are seeking the involvement of people who are. Otherwise, you're just being irrational.


  1. But wait. If that study was done over again, does anybody really think that the frequency of fever after placebo and MMR would again be exactly 17% and 17.2% in each group? There is going to be some variation around those numbers because random circumstances are going to be different the next time around. Maybe it would be 18% and 15% (3% fewer fevers in the MMR group) or 14% and 16% (2% more fevers in the MMR group). A statistical test called a Chi-squared-test is used to tell us if the observed difference is likely to be due to random chance or to a true difference, and in this case, it would tell us that the 17% and 17.2% frequency of fevers likely represents no detectable true difference (P=0.87). The 0.2% difference is very easily explained by random variation rather than a miniscule causal effect of the vaccine on fever. In fact, while fever attributable to the vaccine occurred 0.2% of the time, we can be 95% confident that if we repeated the experiment many times, the results would fall between the vaccine *preventing* fever 2.8% of the time up to the vaccine causing fever 2.8% of the time : a wash. That is, the available data doesn't prevent us from excluding a protective effect whereby the vaccine prevents 2.8% of children from developing a fever in the first 6 days after injection, nor a harmful effect where the vaccine causes 3.3% of children to develop a fever. (Thanks to my friend, Steve Wilton, for calculating the P value and the 95% confidence intervals. These stats were not performed back in the 80's when the original paper was published in The Lancet.)

  2. By the way, MMR vaccine does cause about 5% of kids to get a fever, but it happens between days 7 and 12 (not 1-6). Interestingly, 10-15% fewer children experience cough and runny nose between days 7 and 12 after MMR vaccine than after placebo injection, implying that the MMR vaccine may actually be providing a protective effect against these symptoms.

  3. The RCTs of novel anticoagulants show a risk of *brain* bleeds of less than 1% per year, but the risk of major bleeds that will land you in the hospital with blood gushing out of someplace you'd rather it wasn't is much higher. Also, please note that patients in clinical trials are cherry-picked to make the drug look as good as possible, but the drug is then prescribed to people who average much older and have more comorbidities. Those with reduced kidney function may be more at risk from Pradaxa, for example. Other countries approved two regular doses of dabigatran, but our FDA saw fit to decree that all users, whatever their age and state of health, should have the same values.

  4. Meta-analysis of the 4 RCTs of NOACs vs warfarin indicates a reduced rate of major bleeding associated with NOAC use across all subgroups. That reduction just failed to meet the conventional level of statistical significance (p=0.06) but nevertheless indicates an important trend.

    You wrote that “The RCTs of novel anticoagulants show a risk of *brain* bleeds of less than 1% per year, but the risk of major bleeds that will land you in the hospital with blood gushing out of someplace you'd rather it wasn't is much higher”. What’s relevant here is how the absolute risk reduction for hemorrhagic stroke compares with the absolute risk increase in GI bleeds when a NOAC is used instead of warfarin. The meta-analysis indicates that hemorrhagic stroke was reduced by 133/29,200 while GI bleeds were increased by 140/29,200. That’s a wash. It is noteworthy that hemorrhagic strokes are much more likely to be deadly or debilitating than GI bleeds, so it should be no surprise that NOACs are life saving drugs compared to warfarin.

    When low-dose regimens of dabigatran and edoxaban were examined, no increase in GI bleeding was observed while the risk of stroke and death remained lower compared to warfarin, so I share your disappointment that the FDA did not approve the low dose of dabigatran in America.