COVID and the moral panic over obesity

It’s now part of the conventional wisdom surrounding COVID that fat people are especially vulnerable to being hospitalized after being infected by the virus, and of dying from it.  For example, In just the first two weeks of September, 2021, more than 2,300 stories in the LEXIS news data base mentioned both COVID-19 and obesity, usually in terms of how the latter condition was a major risk factor for suffering severe medical complications from the novel coronavirus.

In the United States, the evidence most often cited for such claims is a huge CDC (Centers for Disease Control and Prevention) study published this spring, surveying outcomes for more than three million patients seen in emergency rooms and admitted to hospitals between March and December of last year.  148,494 of these admissions were COVID patients.

The study’s authors found, in their words, “an increased risk for severe COVID-19–associated illness among persons with excess weight” and a “relationship between higher BMI and risk for hospitalization, ICU admission, invasive mechanical ventilation, and death.”

This study is cited constantly by the media, in the thousands of stories published every month that emphasize the supposed extent to which severe COVID illness is a product of those who contract the virus being too fat.

Yet, as is so often the case when it comes to the subject of weight and health, the data from the study itself either seriously undercut or simply flat-out contradict the study’s much-heralded conclusion, especially as that conclusion is recharacterized and amplified in the popular media.  In the more than twenty years that I’ve been following the medical literature on weight and health, I’ve seen this pattern repeated countless times.

To see exactly how this happened in the context of this particular study, we first need to cover some basic definitions. 

The CDC study uses what have become the standard metrics for defining “healthy” as opposed to “excess” weight.  The study uses the Body Mass Index, which is a simple mathematical formula for putting the weights of people of different heights into a single linear scale.  Following current CDC standards, it defines “underweight” as a BMI below 18.5, “healthy weight” as a BMI of between 18.5 and 24.9, and “overweight” as BMI 25 to 29.9.  It then stratifies “obesity” into four categories:  those with BMIs of 30-34.9, 35-39.9, 40-44.9, and 45 and above.  The last two categories are defined as “severe obesity.”

It then calculates the risk ratios for various outcomes from severe COVID illness (hospitalization, intensive care admission, invasive mechanical ventilation, death) for people in different BMI ranges.  A risk ratio is determined by comparing outcomes for patients in various groups with the outcomes in some reference group. 

In this study, the reference group is made up of “healthy weight” (BMI 18.5-24.9) patients, who by definition manifest a risk ratio of exactly one.  So for example if the death rate in a particular group is 10% higher than that in the reference group, this represents a risk ratio of 1.1.  If the death rate is 50% higher this is a risk ratio of 1.5.  If the death rate is double that seen in the reference group, the risk ratio is 2.0., and so forth.

When interpreting risk ratios, it’s important to keep three factors in mind. First, if the baseline risk in the reference group is very low, then even tiny increases in the absolute number of various outcomes can produce large changes in relative risk in regard to those outcomes.  For example, if five out of every 15,000 people in the reference group die from a particular cause in a given year, ten deaths among 15,000 people in the comparison group in that year represents a 100% increase in relative risk in the comparison group.

Second, and relatedly, what sound in the abstract like impressive increases in relative risk for a certain group can turn out to be trivial in comparison to vastly larger increases in relative risk for other groups, that are not part of the current impressive-sounding comparison.

Third, observational studies that find correlations between risk factors and various health outcomes do not thereby demonstrate the extent, if any, to which such correlations indicate a causal relationship between the risk factor and the outcome (Of course if the correlation is very strong this is highly suggestive of a causal relationship of some sort, but weak correlations tend to be ignored by epidemiologists, because so many unaccounted-for confounding factors may actually be producing the observed correlations, rather than the risk factor itself).

Now let’s take a look at the study’s actual data, which in fact turn put to illustrate all three of these points in particularly vivid ways.

Perhaps the single most striking aspect of these data is that, at the most basic level, they don’t really support the authors’ claim that “excess weight” is “associated with an increased risk for severe COVID-19 illness.”  In fact the vast majority of people in the study who were, per the study’s definitions, unhealthily fat, did not have any statistically significant increased risk for either hospitalization or death from COVID, relative to people at what the study defines as a “healthy weight.”

All but the most careful readers of the study will miss this fact, however, because of the way the authors present their data.  Only readers who delve into that data with some care have a chance of recognizing what the study’s numbers actually reveal.

Consider, as an initial matter, the relative risks faced by the “overweight” (BMI 25-29.9) COVID patients the study.  The relative risk for death in this cohort was 0.95, while the risks for hospitalization and admission to intensive care were both 0.99.  In other words, the relative COVID risk faced by this group – which at present includes nearly one-third of all adult Americans, and a solid plurality of the members of the study – was lower than that faced by the “healthy weight” reference group. 

To those of us who have come to recognize the extent of the moral panic over fatness in American society, this result is in no way surprising.   Many large-scale epidemiological studies have found exactly the same thing in recent decades:: in these studies, both the life expectancy and the overall health found among the “overweight” turns out to be better than that of the “healthy weight” reference group.  (If you are wondering why the “overweight” group isn’t then defined as the “healthy weight” group for referencing relative risk, you are asking the right question).

It’s true that the relative risk in the overweight group is only slightly lower than that in the healthy weight group.  Normally, epidemiologists would consider a 5% difference in observed relative risk between two cohorts to be so trivial that it ought to be for all practical purposes ignored.  This is in part because of the first two reasons mentioned above: a 5% difference in relative risk compared to a tiny baseline risk is still a tiny risk, and it will be dwarfed by vastly larger relative risks when making comparisons to other groups.

And again, an even more crucial caveat is that these are merely crude observational correlations: no reputable epidemiologist or statistician would venture to argue that the fact that “overweight” people had a 5% lower risk of death from COVID relative to “healthy weight” people means that difference in risk was caused by the fact that the “overweight” people weighed more.  That would be ridiculous, given the almost unlimited number of confounding factors that might individually and collectively account for that difference.

Ridiculous it would be; yet similarly insignificant differences in relative risk are, in the context of the moral panic over obesity, treated as irrefutable proof that fat people are dying of this or that illness, that has been caused supposedly by their fatness.

This is all the more reason that similarly insignificant differences in relative risk are important to keep in mind when we consider the rest of the study’s data.  Consider that, among what the study defines as Class I obesity (BMI 30-34.9), the relative risks for hospitalization and death, relative to the “healthy weight” group, were 1.07 and 1.08 respectively.  Again, just as in the case of the very slightly lower  hospitalization and death rates among the “overweight” relative to the “healthy weight,” these are, statistically speaking, trivial differences. 

How trivial?  Consider a group, such as people in their late 50s and early 60s, who in 2020 had an overall mortality rate of one death from COVID for every one thousand people in that particular age group.  (Note that .1% wasn’t the case or infection fatality rate from COVID in this cohort: it was the overall mortality rate from COVID for everyone in the cohort, whether they contracted the virus or not).  What the statistics in the CDC study indicate is that in a group of 12,000 “healthy weight” people in that age group, you could expect 12 COVID deaths, while in a group of 12,000 Class I obese people (BMI 30-34.4) in that age group you could expect one more annual death from COVID than the total number of deaths from COVID in the healthy weight group.  And again, this in no way can be taken as any kind of real evidence that this one extra annual death (among 12,000 people) was caused by the “excess” weight of those in the obese cohort.

In sum, if we combine the overweight and Class I obese patients into a single cohort, we get a relative risk for death from COVID that is essentially identical to that seen in the “healthy weight” category – slightly lower in the overweight category, and slightly higher in the Class I obese category.  

In fact the relative risk in the combined cohort is so close to that in the “healthy weight” cohort that there is no statistically significant difference in risk for COVID death between, on the one hand, the “healthy weight” group, and, on the other, everyone in the overweight and Class I obese categories.

The overweight and Class I obese categories together currently include 53% of all American adults, and 73% of all adults that the CDC classifies as weighing too much.  In other words, for the vast majority of adults that our public health agencies claim are too fat, their weight does not even correlate with any increased COVID risk.

And even among the fattest adults, the associations between their weight and COVID risk remain quite modest compared to many other risk factors.  People with BMIs between 35 and 39.9 in the study had a 14% increase in relative risk of death from COVID; those with BMIs between 40 and 44.9 had a 33% increased risk; and the very fattest Americans – the 3% or so of the population with a BMI of 45 and above – had a 61% increased risk.

61% may sound like a lot, but consider that, for instance, men in general face a 58% increased risk of death from COVID relative to women.  Meanwhile, even small difference in age produce vastly greater increases in relative risk than any of these figures.

For example, 30-year-olds face a 293% increased risk of death from COVID compared to 20-year-olds.  These numbers suggest that, for example, a 26-year-old woman of average height (5’4”) who weighs 263 pounds (BMI 45) faces a lower mortality risk from COVID than a 29-year-old 127-pound woman of the same height (this is a BMI of 21.8, i.e., in the middle of the “healthy” range).

Here is another way of seeing how relatively insignificant the association between weight and COVID risk actually is.  We can for illustrative purposes construct a hypothetical world that makes two equally fantastic assumptions: that everyone who was not currently in that weight range had in February of 2020 their current weight magically transformed to a “healthy weight” of between 18.5 and 24.9 BMI, and that this transformation gave people who were then overweight (or, more properly, “overweight”) the same health characteristics as people who did not require any such magical transformation (The reason the latter assumption is fantastic is that there is no scientific basis for assuming that turning fat people into thin people would produce the same health profile in the newly-thin people as that of the people who were thin all along).

If we make these two assumptions, how many fewer COVID deaths would have seen in the USA since February of last year, when the epidemic began?

The answer is that this fantastical hypothetical, in which every person in America is ideally thin, reduces COVID deaths in the USA over the course of the pandemic by just 5.4%, from 662,200 to 626,495.

The specific math: our fantasy epidemiology produces a reduction of 11,708 deaths among the Class I obese, 9,734 fewer deaths among the Class II obese, 11,582 fewer deaths among people with BMIs between 40 and 44.9, and 12,118 fewer deaths among people with BMIs of 45 and up.  (Also included are 993 fewer deaths among the relative handful of Americans who are technically “underweight,” with BMIs below 18.5).  These lives saved, however, must then be offset by the 10,430 more deaths generated among the “overweight,” who, as we have seen, have a 5% lower death rate from COVID than their “healthy weight” comparators.

Now let’s run the same fantastic exercise with a couple of other risk factors.  What sort of reduction in COVID deaths would we have seen if, via the magic of epidemiological thought experiments, we had a year ago last February turned all American men into women?  The answer is 244,200, or 37%.

A far more effective mitigation strategy would have been to make everyone 20 years old: this would have reduced COVID deaths by an impressive 99.45%, thus saving roughly 659,000 lives.

Of course those fighting the never-ending war against “overweight” and “obesity” will be quick to point out that it’s impossible to make everybody a woman or 20 years old.  The response to this is that, if you are not part of the weight loss industrial complex – a Blob every bit as self-interested and intransigent as the Blob that screamed bloody murder when the US government finally refused to keep fighting and endless and pointless war in Afghanistan – it should be self-evident that trying to turn people with BMIs of 35 or 40 or 45 into people who maintain BMIs between 18.5 and 24.9 is just as irrational as trying to make everybody a woman, or forever young.