Within the last couple of weeks, a number of articles have come out on various factors which might change one’s risk of illness. One of the surprising things about COVID-19 is the degree to which the burden seems to vary across people. We have seen nursing homes where 60% of cases are completely asymptomatic, and yet mortality rates among nursing home residents are extremely high. It is crucial to understand who is most susceptible to the virus, and who is most likely to get the sickest and need to be hospitalized.
There are some very basic correlations we understand well. The virus is worse for older people and for those with pre-existing conditions. Death rates are higher for groups with worse insurance coverage. Now, people are starting to look for other features — genetic variants for example — which might relate to illness risk.
Two things came out last week which relate here.
The first was some evidence on COVID-19 and baldness. Here is an example media article on this set of studies, which suggested bald men were more likely to experience severe COVID-19 (as measured by hospitalization). The studies underlying this, though, are underwhelming. Researchers showed that most men hospitalized for COVID-19 were bald. Unfortunately, they didn’t really control for age. And since we know that older people are both more susceptible to serious illness and more likely to be bald, it seems much more likely that the results are driven by age. Conclusion: there is no reason to think hair protects you.
The second (more serious) evidence was on genetic variants and COVID-19, specifically related to blood types. Genetic variants are specific sequences within our DNA that make us unique from each other, such as the sequences that determine our eye color or our blood type. The underlying study — accessible here — is an example of a GWAS (“genome-wide association study”) which exploits genetic sequence data to correlate particular genetic markers with disease.
In this case, the researchers studied two groups of patients with serious COVID-19 illness (one group in Spain, the other in Italy) and compared them with two groups of matched control patients from the same area without COVID-19. All of the participants — patients and controls — were genotyped, meaning their DNA was analyzed and individual genetic variants were isolated. The basic idea is to look at which genetic variants are more common among COVID-19 patients than the controls.
(For the more technically minded: they use a particular off-the-shelf sequencing and focus on single nucleotide polymorphism (SNPs). Each of the samples – Spain and Italy – results in about 9,000,000 SNPs).
What did they find? In their words:
“We found two loci to be associated with Covid-19 induced respiratory failure with genome-wide significance (P<5×10-8) in the meta-analysis (analysis I), the rs11385942 insertion-deletion GA/A SNP at chromosome 3p21.31, OR meta 1.77 (95% CI, 1.48 to 2.11), P=1.14×10-10and the rs657152 A/C SNP at 9q34.2, OR meta 1.32 (95% CI, 1.20 to 1.47), P=4.95×10-8.”
In our words: they found two areas of the genome in which the variant of the gene someone had at that location predicted having serious COVID-19 disease. One of these is on chromosome 3, and the other on chromosome 9. People with a particular version of the gene at the chromosome 3 location were 1.8 times as likely to have serious disease, and those with the variation at chromosome 9 were 1.3 times as likely.
The chromosome 3 variation shows the stronger link here, but it isn’t associated with any particular individual characteristic (which is to say, you wouldn’t know what variant you had unless you’d had a genetic test, and as a result this is of somewhat less media excitement!). The variant on chromosome 9, though, is related to blood groups.
So, the authors do a further analysis focused specifically on blood groups and find that people with A-positive blood are more likely to show up in the serious disease group, and those with O-type blood are less likely to have a serious disease. These effects are moderate. There is a 50% increase in risk for the A-positive group and a 35% reduction in risk for the O-type blood group.
Another way to see these magnitudes: about 35% of people have A-positive blood type in the unaffected population, but 46% of the COVID-19 group have this blood type. Conversely, about 47% of the overall population has O-type blood, but they make up only 37% of the COVID-19 population.
This type of analysis is interesting and more work in this direction is likely to be valuable for learning about the virus. There are some caveats, though. First, studies like this do a tremendous number of statistical comparisons, and as a result, have to be very careful about false positives. This study uses standard, conservative approaches but it’s still worth keeping in mind this might be a statistical accident.
The other caveat is effect size and what we learn from this. Yes, the results do suggest some difference in risk by blood type.Butit’s not like those with O-type blood cannot get the virus or that those with A-type blood definitely will. These effects are at best moderate and much smaller than (for example) variation by age or other illnesses.
Bottom line: your blood may protect you a bit, but we suggest you wear a mask.