Day 71: BCG vs RCV1

Today's numbers are impressive: over 1.75 million cases worldwide, with well over 100,000 deaths. The US (with over 500,000 cases) has finally pulled ahead of Italy and exceeded 20,000 deaths. The UK probably won't hold on to its 7th place below China for long but remains there now, and Boris Johnson is out of the ICU.

New York State (at 180,000 cases) remains ahead of the major European hotspots. New Jersey is a distant second at 58,000 cases, and the remaining hotspots (Michigan, Pennsylvania, California, Massachusetts, Louisiana, Florida, and Illinois) are all hovering around 20,000 cases. Michigan has locked down for real this time, forbidding travel between homes and "public or private gatherings of any size."

Massachusetts' numbers are down again today, with 1886 new cases (up 9%) and 87 deaths. The state is now tracking deaths in long-term care facilities (304 total) along with infections (2645, which includes health care workers) and infected facilities (190). Also of note, the sidewalks of one street in Beverly have been declared one-way.

With Wyoming jumping on the bandwagon, all 50 states (plus 4 territories and the District of Columbia) have declared disasters simultaneously for the first time.

So, why us? You've heard this explanation here before, but since the media has picked up on another paper (apparently the fifth) about the correlation between tuberculosis vaccination and low case fatality rates from coronavirus, you're about to hear it four more times.

The Bacillus Calmette–Guérin (BCG) vaccine uses a live, attenuated strain of Mycobacterium bovis (yes, it's a real vaccine, perhaps even more so than vaccina). It protects against tuberculosis (Mycobacterium tuberculosis) and leprosy (Mycobacterium leprae and M. lepromatosis) to some extent, for a couple of decades. It has long been known to protect against other respiratory diseases indefinitely, by a still unknown mechanism; it is also sometimes used in the treatment of bladder cancer. Nevertheless, adoption of BCG vaccine varies by country, perceived risk of TB, and time; see the BCG World Atlas for lots of details.

1. Back in the day, PlagueBlog covered the first preprint on this topic, Correlation between universal BCG vaccination policy and reduced morbidity and mortality for COVID-19: an epidemiological study, by Aaron Miller, Mac Josh Reandelar, Kimberly Fasciglione, Violeta Roumenova, Yan Li, and Gonzalo H. Otazu, at several New York institutions. They said:

We compared large number of countries BCG vaccination policies with the morbidity and mortality for COVID-19. We found that countries without universal policies of BCG vaccination (Italy, Nederland, USA) have been more severely affected compared to countries with universal and long-standing BCG policies. Countries that have a late start of universal BCG policy (Iran, 1984) had high mortality, consistent with the idea that BCG protects the vaccinated elderly population. We also found that BCG vaccination also reduced the number of reported COVID-19 cases in a country.

They controlled for income and also charted start years of BCG vs. deaths for countries that do or did vaccinate.

2. Another early paper, BCG vaccination may be protective against Covid-19 by Paul K. Hegarty, Ashish Kamat, Helen Zafirakis, and Andrew DiNardo (two Brits and two Texans) quantified both the infection and death rates over a fifteen day period in March at almost 10 times lower with BCG vaccination. They did note that "[c]ountries that have a booster injection of BCG 7 to 14 years later had no better outcomes than those with a single inoculation only."

3. The story currently in the news is about a preprint by researchers at Johns Hopkins that estimated a death rate only 6 times lower with BCG: Differential COVID-19-attributable mortality and BCG vaccine use in countries, by Anita Shet, Debashree Ray, Neelika Malavige, Mathuram Santosham, and Naor Bar-Zeev. The mortality rate is 5.8 times lower in vaccinated populations than in the unvaccinated. The average death rate per 1 million people in vaccinated countries is 0.6, while in unvaccinated countries it is 8.6. (If you're curious about country by country data, take a look at this supplemental figure.)

4. An even more recent paper, Mandated Bacillus Calmette-Guérin (BCG) vaccination predicts flattened curves for the spread of COVID-19 by Martha K. Berg, Qinggang Yu, Cristina E. Salvador, Irene Melani, and Shinobu Kitayama at the University of Michigan, compared infection and death rates in the first 30 days of a country's outbreak to their BCG status. Countries with vaccination policies that did not last into this century (e.g., Spain, Germany) turned out to be equivalent to those who never vaccinated (e.g., the US, Italy). In the supplemental material, they "predict" the numbers for an alternate US with full BCG vaccination:

This analysis applied to the number of cases yielded a predicted value of 11.28, which translates to 79488.86 cases (compared to the actual 213372 cases reported in the US by April 1). This analysis applied to the number of deaths yielded a predicted value of 4.54, which translates to 93.97 deaths (compared to the actual 2467 deaths reported in the US by March 29).

5. There seem to be more than five such papers, so I'm picking a fifth from among the related preprints at medRxiv that contradicts the others: Association Between BCG Policy is Significantly Confounded by Age and is Unlikely to Alter Infection or Mortality Rates by Stefan Kirov of Bristol Myers Squibb. The title says quite a bit, but the author also notes a confounding correlation between BMI and BCG policy: countries with lower BMI also have BCG vaccination. (Obesity is a known risk factor for coronavirus.) He attempts to make a spurious connection between rubella vaccination and (poor) COVID-19 outcomes as a sort of cautionary example, but rubella vaccination dates back to 1964 at the earliest (while BCG is over a hundred years old) and thus is much more confounded by age than BCG coverage could possibly be. Rubella is also a poor choice of counterexample because rubella vaccine (RCV1) coverage has been fairly consistent across the western world (except, notably, in Italy), where the contrasts in outcome despite few cultural and biological confounders led to such a baroque theory as BCG in the first place.

One preprint from yesterday goes full on in the rubella direction with a study of the MMR vaccine: Homologous protein domains in SARS-CoV-2 and measles, mumps and rubella viruses: preliminary evidence that MMR vaccine might provide protection against COVID-19 by Robin Franklin, Adam Young, Bjoern Neumann, Rocio Fernandez, Alexis Joannides, Amir Reyahi, and Yorgo Modis (mostly in Cambridge, England). Though they accept the BCG theory at face value and postulate that adoption of the MMR vaccine might follow a pattern similar to that of BCG, for this paper the confounding element of age for BCG is the primary explanatory factor. Age is strongly correlated with MMR coverage because MMR only dates back to the early 70's, and its constituent vaccines only date to the late 60's. Because rubella vaccine was also given to women of childbearing age, immunity to COVID-19 would be increased differentially for women over 50 vs. men of the same age. (Unfortunately the figures for this in the paper are mostly unlabelled and inscrutable.)

They also found similarities in the viral genomes of SARS-CoV-2 and the measles, mumps, and rubella viruses; there are quite a few details there, but no wet results yet. However, they have collected rubella immunoglobulin data in severe vs. mild cases of COVID-19:

A further prediction of our hypothesis is that there should be a specific rise in rubella Immunoglobulin G (IgG) titres in COVID-19 patients, and that these should correlate with disease burden as a marker of immunogenicity against SARS-CoV2 [...]

Patients with a high severity illness had on average increased levels of rubella IgG (161.9+147.6 IU/ml) compared to patients with a moderate severity of disease (74.5+57.7 IU/ml) (Fig. 5). In comparison, Immunoglobulin M (IgM) levels were 0.21+0.16 IU/ml in severe disease and 0.26+0.21 IU/ml in moderate disease. Whilst we accept that it is possible that this trend could be representative of preinfection protection to rubella infection, it is not possible to determine this. In a study of 160 women of child bearing age, the IgG levels of non-infected patients measured between 24-143 IU/ml, suggesting that it is unlikely that those who developed severe symptoms of the disease had IgG levels far in excess of this prior to infection.