Posted 22 April 2020
…..SARS, COVID, IFR, CFR, PCR, IgM, IgG, RNA…..
The era of coronavirus is the era of the acronym. If the virus doesn’t get you the string of letters will!
I want to tackle some of those acronyms here and then ponder if there is anything about total exposure to the virus in the UK that we can deduce from the USS Roosevelt and the Diamond Princess.
The first three acronyms I am going to consider – PCR, IgM and IgG – are all about testing.
PCR testing is Polymerase Chain Reaction testing. For this test, a swab sample is taken from the patient, usually the nasopharynx (back of the throat) or sometimes, in a hospital environment, from the lower respiratory tract. The swab is then tested for presence of the viral RNA (refer to my piece yesterday for where that fits in). If the viral RNA is present in the swab material, this reveals that the virus is still active and therefore that the patient has a current infection. They may be with or without symptoms at the time the swab was taken, but the critical thing is the infection is current, has yet to complete its course in that patient and that patient has a potential to spread the virus to others.
When health systems report ‘new cases’ of COVID-19, they are generally working on PCR tests. To use the correct term: “each new case is when a patient has a positive PCR test”.
PCR tests can be unreliable. They may show false positives or false negatives, respectively suggesting infection where there is none or failing to show infection when it is in fact present. But if testing within one health service is consistently good or bad, then the timeline of positive cases gives us meaningful trend data.
One patient may be tested several times: If the first test is taken before they are infected, that test should be negative. If a subsequent test on the same patient is positive, it indicates the patient has become infected since the first test. Multiple testing of individual patients means that the number of PCR tests reported by a country does not mean this many patients have been tested. The number of people actually PCR tested will be less.
ANTIBODY TESTS (or serological tests) take a sample of blood and look for antibodies to the virus in that sample. If antibodies are present this is an indicator that the subject has immunity from the virus conferred from a now-completed infection. [You may remember from my article yesterday, there are a number of proteins in the coat of the virus. An antibody to a virus will stick to the proteins of that virus and so the isolated proteins from the virus are used in antibody tests].
There are two types of antibody that are most commonly looked for: IgM and IgG (both immunoglobulins) . The former appear very soon after the infection and are short-lived (usually a couple of weeks) and if present show the infection has recently occurred; IgG antibodies appear later and are more durable, lasting for months or often years protecting against further infection for that time period. No one yet knows how long they last for SARS CoV2.
PCR tests take time to assay as the RNA extracted in the swab has to be converted to DNA in a laboratory environment and then duplicated many times before testing. Antibody tests can be in ‘at home’ kit form and take 10-15 minutes.
Just to clarify: PCR testing answers the question “DO you have COVID-19?”. Antibody testing asks the question “DID you have COVID-19?”.
At the moment (here in the UK) PCR testing is performed as patients enter the healthcare system for treatment. As of 20 April, a cumulative total of 400,000 PCR tests had been carried out in hospitals; A little under 125,000 of those have been positive. The hospital reported deaths to this point is 16,509.
This leads me on to another acronym – CFR, the Case Fatality Rate. This is defined as the total number of confirmed fatalities divided by the total number of confirmed cases. So with the data in the previous paragraph: CFR = 16,509 deaths/125000 cases = 13.2%.
The problem with this number is that it is often confused with the death rate of the virus. If this were true it would imply that 13 per cent of those who get COVID-19 will die. This is most certainly not the case. Unfortunately the media often make this confusion and that has created much fear.
The UK predominantly only carries out PCR tests on people who are being hospitalised and as I indicated above only 400,000 PCR tests have so far been carried out. Even if I assume each person tested is only tested once (see above) this is only roughly 0.6% of the UK population. It is therefore a tiny sample and so hugely inflates the estimate of CFR as it assumes no other infections in the untested 99.4% of the population [This is blatantly not true as we know a high percentage of those with COVID-19 are without symptoms or too mild to need hospitalisation]. Sampling only the hospitalised group is also a sample skewed towards those who are experiencing a worse case of COVID-19 simply because they are entering the healthcare system. The tested 0.6% are therefore those more likely to die of COVID-19, skewing the CFR upwards.
Because of this last point, we cannot just extrapolate the 0.6% up: If 125,000 people have tested positive out of 400,000 tests then approximately a third of those tested were positive. What we cannot say is: “Therefore if we had tested all 66 million people in the UK, one third of those would have been positive too” (equating to 22 million). This has to be a significant over statement.
To be fair and complete, I should also say that there are also a couple of mechanisms by which CFR could be under-stated: There are confirmed COVID cases now who will unfortunately die but they haven’t died just yet. So, only once the whole pandemic is over will the full fatality numbers be known. Secondly, during the pandemic, deaths are almost certainly occurring due to COVID-19 but not being attributed to the virus [robust counter-arguments will also show that people who are dying of other causes will be listed as COVID deaths too]. But the scale of the under-attributing deaths to COVID is small compared to the massive under-reporting of total confirmed cases because we have only tested 0.6% of the population. The CFR is certainly therefore over-inflated.
This leads me to another one of the acronyms: IFR, the Infection Fatality Rate. The infection fatality rate is the total number of confirmed fatalities divided by the total number of actual infections (not just the ones sampled with the PCR testing). Putting that another way, if we had been able to PCR test or antibody test the whole population of the UK to know who has or has had the infection, and then divided that into the 16,509 so far confirmed deaths that would give us the true IFR. This number is a much more meaningful answer to the question: “if I get COVID-19 what are my chances of dying”.
Just one problem: We do not have the PCR and antibody test data for the whole U.K. population.
The Centre for Evidence Based Medicine suggests that globally the IFR for COVID-19 is somewhere between 0.5% and 0.1%. This is based on all country data (some of whom are coming to the end of there infection curves) and paying particular attention to those countries that have PCR and antibody tested random samples of their populations, not just the hospitalised population. For comparison, seasonal flu has an IFR of around 0.1% to 0.2%.
This is of course an average over the whole population. So if we looked at age groups for example, we would find the IFR for younger people to be lower than that of older people. All we can say is that over the whole population the estimated average IFR is between 0.1 and 0.5%.
This gives us a back-door estimate of how many people in the UK may have been infected. If the true IFR is 0.1-0.5% and in the UK there has been 16,509 deaths to date, then that implies (to date) 3.3-16.5 million people have already been infected. Compare that to the 125,000 ‘confirmed cases’.
The IFR will vary from country to country due to demographics – age profile, levels of obesity and diabetes – but even at lower case estimate, that is a far higher case number than reported. But we all know the inadequate level of testing is a big issue here in the UK so maybe this should be no surprise.
The only way we can know how many people in the UK have previously been infected with COVID-19 but recovered (as 60-85% of all infections are completely symptom free, we cannot guesstimate from ’self-reporting’) is to go out into the population with antibody tests and randomly sample from within whole population, not just the hospitalised population.
One study published in the last few days did exactly this in one county of America. For this study they selected a representative slice of the population and performed the antibody test. They then extrapolated the results to the whole population in the county and compared that to the total confirmed cases in the county (derived from a similar method to here in the UK). Their study suggested that the total confirmed cases from hospital testing underestimated the total infection numbers by a factor of between 50 and 85 times. 50-85 times!
If I use the lower of these two under estimate factors on the UK data, that would suggest 6.2 million actual infections, not that far from my estimate using the estimated global IFR. At the upper end of the under estimate factors, that would equate to just over 10 million cases in the U.K. so far.
[For sake of balance, as to today 23 April, I have seen another serological study that estimates under-reporting of COVID-19 cases by a factor of between 28-55. Applying this range to U.K. data would suggest a range of total cases between 3.5 and 6.9 million]
OK, so this is a study of one county of the U.S. but it is a start to answering the key question of how many people have actually had the COVID-19 infection.
And that leads me to one other aspect. Whilst it is key to know how many of us have already had COVID-19, not everyone who is exposed to SARS CoV2 develops the infection. This is the concept of Susceptibility: When a population is exposed to a pathogen, not everyone is going to succumb to the associated illness even if everyone is exposed. This is known as the susceptibility. A high susceptibility means more of us get the infection; a lower susceptibility means more of us will be without infection.
So follow me on this one. If we know how many of us have or have had COVID-19 and we know what percentage of the population develop COVID if exposed, we can back-estimate how many people have already been exposed to the virus. And if we have a handle on how many of us have been exposed to the virus then we have a clearer of model for what happens after lockdown.
Two closed environments might give us a clue to our susceptibility to COVID-19: The USS Roosevelt and the Diamond Princess
The 4,800 crew of the USS Roosevelt were all exposed to the virus in close quarters for a number of weeks. 94% of the crew have now been PCR tested for the virus. 678 have been confirmed PCR positive. (For complete stats, 8 have been hospitalised, 1 has died giving a CFR of 0.15%.) So despite confined exposure to the virus, only 14% (678/4800) of the crew developed COVID-19.
A similar analysis for the Diamond Princess shows that 18% of the 3.700 on board developed COVID-19. The age profile of the passengers on this cruise liner would be older than the aircraft carrier.
So is it possible that the overall susceptibility of the population to COVID-19 is around 14-18%? That is, only 14-18% of the population are going to become infected even if we are all exposed? You could do the sum either way: If (lower end estimate) 6.2 million UK citizens have had COVID-19 and 18% is the susceptibility that means 33 million of us have been exposed already. Half of us.
Or the numbers the other way: If 14-18% of the UK population are going to get COVID-19, this equates to 9-12 million (of 66 million). Compare that to my range of estimates a few paragraphs up: 6.2 million -10 million. Maybe we are getting near to the total that will get COVID-19 and maybe this is why new cases levelled-off a week or so ago.
Yes, the closed environment on a seafaring vessel is a unique environment for virus transmission and so maybe we have to be careful extrapolating from there to the UK population. But if anything, the ease of transmission on an aircraft carrier and cruise liner will be so much greater than an island of 66 million people. Maybe the Roosevelt and Diamond Princess just give us the lower end estimate of how many have been exposed!!