Correlation Doesn’t Mean Causation (Not Even in a Pandemic)

A paper suggests BCG vaccines against tuberculosis, which are provided in Venezuela, can reduce COVID-19 deaths. But are they really the answer we’re looking for?

Photo: Carlos Jaimes.

A recent work from a research group at the New York Institute of Technology (NYIT) is showing a correlation between a tuberculosis vaccine and a reduced COVID-19 mortality rate, triggering a new ride on the hype train…

…and lots of people are coming on board without asking the right questions.

The paper, released on March 21st, shows that countries who currently take, or have taken the Bacillus Calmette-Guerin (BCG) vaccination (against tuberculosis) show less COVID-19 deaths than the rest of the population, and they’re even infected less. European countries, where the disease hits hard, don’t have universal BCG programs, although most had them at some point. The U.S. only uses the vaccine on recommended high-risk groups.

In case you’re wondering, Venezuela does apply the BCG vaccine to all newborns; you can actually see the characteristic scar it leaves in your arm, which stays with you forever. All countries in South America have had universal BCG application at some point, and Ecuador, where the disease has caused some of the most heartbreaking scenes in the region, is the only country which currently doesn’t.

So, is BCG protecting Venezuela from the most devastating effects of COVID-19? 

Well, although some studies certainly found a strong correlation, we can’t say that BCG vaccination causes a reduction in infections and death rates.

Well, although some studies certainly found a strong correlation, we can’t say that BCG vaccination causes a reduction in infections and death rates. Studies like these, where large sets of populations are compared to each other to see the effect that an individually applied intervention (like a vaccine) has on another variable (COVID-19) are called “ecological studies.”

Ecological studies are great to identify factors that can potentially protect against a disease, or increase the risk of getting it, but they’re terrible to prove actual causation. See, you could live in a country that applies the BCG vaccine, but that doesn’t necessarily mean that the individual inhabitants didn’t receive other interventions that could explain the result. This incapacity to link population results to an individual level is called “ecological fallacy” and it’s a well-described limitation of these studies.

Several factors beyond BCG could explain the differences, too. The vast majority of countries who still apply the BCG vaccine are low or middle-income countries, which generally have a larger proportion of youths compared to Europe or the U.S., and they might be testing far fewer people. Furthermore, many European countries used the vaccine until not too long ago, so elderly folks are probably vaccinated and should, therefore, be protected. Also, since the paper was published, many countries applying BCG, such as India or South Africa, saw considerable increases in cases. Ireland and Portugal, two high-income countries applying the BCG vaccine and who are definitely doing better than its neighbors in terms of COVID-19 control, also implemented social distancing measures very early in the epidemic, which probably explains the difference.

The take-home message is that ecological studies can give you a hypothesis to test, but it can’t confirm it. A well-known mantra in science is that correlation doesn’t mean causation, and there are numerous hilarious examples of this out there. 

To prove the hypothesis that BCG is causing the differences, we need a different kind of study: randomized trials, and they’re already underway.

As we reach more than 114,000 deaths and COVID-19 cases are expected to surpass the 2 million infected people mark this week, the need to find an effective treatment (or at least some good news) grows as fast as the pandemic. 

While it arrives, beware of hype and miracle cures.