Discussion of seroprevalence of COVID-19
SARS-CoV2 serology: Antibodies in the blood tell us that COVID-19 case numbers diagnosed by PCR swabs are the tip of the iceberg (click here). One thing that is clear with COVID-19 is that if you don’t test you won’t find cases. Another complexity is that the PCR testing of nasopharygeal swabs, the dominant form of testing used worldwide, only detects active infection. However, once a person recovers, this type of test becomes negative. Thus, these swabs are highly relevant to diagnose infected people and for contact tracing; however they miss a big group of people. We know that there are many people who have few symptoms when infected and may not present to be tested, particularly health young people. PCR testing capacity is also limited, another reason we underestimate the true COVID-19 burden. Enter the serosurvey, a test which measure the antibodies our bodies generate against the virus. These tests are not in widespread clinical use but they are excellent epidemiologic tools that tell us whether (not when) a person was infected with the virus. One caveat is that not every infection leads to production of antibodies.
I’m behind on reading the Economist (who isn’t) but their article on serosurveys from Sept 26-Oct 2nd (pg 21-24) merits review. The Economist summarizes the finding of 279 serosurveys from 19 countries. They conclude that the true rate of infection (based on positive serology) far exceeds the rate based on detection of virus by the PCR test (implying substantial underdiagnosis). Based on serology the Economist infers that infections were already running at over 1million a day by the end of January—at a time when we were barely aware of this virus’ existence. By May, serology suggested a global rate over 5 million cases a day. As of their article being written serology would suggest that 500m-730m people worldwide had been infected (i.e. would have positive antibodies) vs the actual number of cases diagnosed by late Sept (~30 million people)-See graphic below:
So many more people have antibodies than we know are infected??? As the kids say, “big if true”. The limited serology that is available would mean 6.4% to 9.3% of the world’s population had been infected prior to October. So is it true? Probably. While global data are messy because broad serologic testing is not being performed, there are interesting data on rates of infection vs rates of seropositivity form locales that have both pieces of data. Here is their data summary: “In Germany, where cases have been low and testing thorough, the seropositivity rate was 4.5 times the diagnosed rate in August. In Minnesota a survey carried out in July found a multiplier of seven. A survey completed on August 23rd found a 6.02% seropositivity rate in England, implying a multiplier of 12. A national serosurvey of India conducted from the middle of May to early June found that 0.73% were infected, suggesting a national total of 10m. The number of registered cases at that time was 226,713, giving a multiplier of 44. Such results suggest that a global multiplier of 20 or so is quite possible.” For those who are not number orient this means COVID-19 may be 1—20X as common as we realize.
If the virus is indeed causing many undiagnosed infections there is potential good and bad news. The good news would be that the advertised case fatality rates of ~5% would be an over estimate (i.e. most people are recovering without a diagnosis). Potential bad news could be that some people are dying from COVID-19 without a diagnosis! For example, The Economist has gathered all-cause mortality data from countries which report them weekly or monthly, a group which includes most of western Europe, some of Latin America, the United States, Russia and South Africa (see lower part of data panel). Between March and August these countries recorded 580,000 COVID-19 deaths versus a total of 900,000 excess deaths. They interpret this as meaning that the actual mortality from the pandemic was 55% greater than the official one (i.e. 900,000 excess deaths). In my view it is equally likely that the pandemic caused delays in health care for other conditions, like heart disease and cancer and that this accounts for the balance of the non-COVID-19 excess in mortality (click here).
What was the cause of the non-COVID-19 excess in mortality? Undiagnosed COVID-19 vs increased mortality due to delayed care for people with other diseases (heart disease and cancer for example).