Thread by @petermbenglish

Diagnosis of SARS-CoV2 is far from straightforward. It includes-

Identifying cause of new, significant illness.

Identifying people who are potentially infectious.

Identifying people with late complications.

We are particularly concerned, now, about identifying people who could infect others.

Difficult, since most infections are in people with no or minor symptoms.

This may explain the clamour for tests.

But tests are very poor at identifying these people.

We know that - in sick adults - a single PCR tests will miss >=30% of cases. (Two tests will still miss >=20%.) In asymptomatic but infectious people it will miss more of the cases; and it cannot detect people who have been infected, but are not yet infectious.

This is slightly strange, because PCR testing is exquisitely sensitive - capable of detecting very few copies of the DNA* sequence it's looking for.

Presumably, for some reason, SARS-CoV-2 is less likely to make it onto swabs than, say, flu virus.

* Parenthetic tweet:

Yes I did write DNA. SARS-CoV-2 is an RNA virus, not a DNA virus. Which is why we use not plain PCR testing, but RT-PCR testing, using Reverse Transcriptase (RT) first, to convert the RNA to DNA).

For whatever reason, the test is lacks sensitivity (a measure of how likely the test is to detect true cases).

And of course, what do we mean by a "case"?

Usually, in medicine, a case is an individual with a disease.

But when we're looking for potentially infectious people, most of whom will be asymptomatic or paucisymptomatic, we often call them "cases" - for want of a better word.

Most of the time this slightly cavalier use of language doesn't matter, but occasionally it can trip us up.

Sticking with PCR testing for a bit longer, I am not clear if we know if the accuracy varies across population groups. Are children different from adults? If so, at what age does it change? Are older people different from younger people? Children at different ages?

Given that we can't test everybody every morning when they brush their teeth or have their first morning pee (now there's an idea...), we have to rely on identifying people with symptoms and testing them.

Parenthetic tweet - viral RNA is excreted when you go to the toilet, by people who are currently or have recently had SARS-CoV-2 infection. One approach to identifying communities where the infection is present is looking for the virus in sewage.

Using people with symptoms - as, sort of, sentinel cases - is a terrible idea. But, like democracy, its the best we've got.

Of course, most people with SARS-CoV-2 infection are a- or paucisymptomatic.

Parenthetic tweet.
The virus is called SARS-CoV-2. The disease is Covid-19. Or possibly COVID-19.

Jargon.
A symptom is "something a patient complains of"

"Asymptomatic" is medical speak for "without symptoms".

Paucisymptomatic is medical speak for "with few symptoms"

Of course, we would like everybody who is potentially infectious to self-isolate, so they don't infect anybody else.

And by identifying only people with symptoms, we are identifying only a fraction of infectious people.

Worse still, if they test negative (and >=30% of cases will test negative), they and their contacts can stop isolating - although the case is still infectious, and their contacts may well be or become so too.

(My apologies for the delay - I plan to continue this thread when I have time…)

I've had an enquiry from @icureiosity who has heard "that tests are extremely sensitive and also pick up dead cells. After case is no longer infectious?"

I'll respond to this quickly now...

@icureiosity I mentioned in a previous tweet

https://twitter.com/petermbenglish/status/1306873022003937280

that the test can detect minute quantities of viral RNA; but that (for reasons we don't fully understand) many swabs from infected people don't seem to have any viral particles on them.

@icureiosity As to her point about "dead cells" - the test is for viral RNA; and RNA from non-viable, "dead" viruses can persist - and be detected on swabs - for weeks after somebody is no longer infectious. See peterenglish.blogspot.com/2020/08/distin… for more details on this issue.

@icureiosity I should point out that the 70% sensitivity rates I've talked about for PCR testing (that they miss 30% of true cases) presume the test was done properly, by a well-trained person.

@icureiosity When swabs are taken by a parent with no experience of swabbing, or via self-swabbing as per @RMonkAudiology's tweet

https://twitter.com/RMonkAudiology/status/1306918712289656833

, poorer swabbing quality can be expected, and with it even lower sensitivity.

Returning to this thread after a long break…

I raised, earlier, the question of what we mean by a "case". There are many definitions!

The daily number of cases reported at the dashboard is the number of positive tests.
coronavirus.data.gov.uk

A GP or hospital doctor might consider a Covid-19 case to be somebody who needs treatment, and who meets a case definition which may include a positive test.

What we appear to be using, mostly, to guide our response to Covid-19. is the number of confirmed cases per day in England.

But remember - the test misses at least 30% of cases in people tested (more if they're not ill); and not everybody d who could have the virus is tested.

This definition of a "case" - somebody who has had a PCR test for SARS-CoV-2, and the result was positive - is not entirely useless, but we have to understand its limitations.

In terms of controlling the spread of the infection, our ideal definition of a case might be along the lines of "somebody recently infected with the virus and still in their infectious period".

And we would want to know the true number of cases (as defined in the previous tweet) per unit population per unit time.

Parenthetic tweet:
A denominator commonly used by epidemiologists is per 100,000 popn. But you try to choose a denominator that gives an easy-to-understand numerator - between 1 and 100, ideally. I'd go for cases/100,000/day; but I think @IndependentSage uses case/million/week.

But as not everybody who might have been infected is tested, and as we don't have a perfect test, we can only estimate this true number of new cases.

Perhaps obviously, the more people you test, the greater the number of positive tests you can expect.

It's a truism in epidemiology that we don't usually have the data we'd ideally like to have; but imperfect data can be very useful, particularly if it is consistent in its imperfections.

If operation moonshot goes ahead, we may well test more people, but with a less accurate test (one with more false negatives and more false positives).

This will cause a change in the case numbers which is an artefact - a consequence of how we do the measuring, rather than a consequence of the true number of cases.

Epidemiologists and economists - and, one would hope, government scientists - are well used to distinguishing between artefacts and true changes (although it is not an exact science).

Indeed, nearly always, when you see a strange change to eg disease rates, the reason is a change in the way things are being measured.

This thread has referred to sensitivity and specificity of tests a lot.

Sensitivity is the proportion of true cases that are detected. A test with low sensitivity will have a lot of false negatives.

The sensitivity of PCR testing for SARS-CoV-2, in people who are ill, and who have the disease, is about 70%. The test is positive in about 70% of such people who have the disease; and is negative ("false-negative") in about 30% of them.

The specificity of PCR testing for SARS-CoV-2, in people who are ill, and who have the disease, is probably well over 99%.

I have described elsewhere (eg peterenglish.blogspot.com/2020/08/distin… ) how the test will detect viral RNA in people who have been infected, but are no longer infectious. Whether or not these results are "false-positives" depends on your point of view - what you expect from the test.

A technician might say that the test did what it was designed to do correctly - detect viral RNA. An epidemiologist, interested only in potentially infected people would prefer something else.

When you test eg care home staff every few weeks, recurrent positive results in people who probably just have leftover RNA are a nuisance - you have to assume they might be infectious and exclude them in case, as I discussed in peterenglish.blogspot.com/2020/08/distin…

It also becomes an issue if, for example, you test asymptomatic contacts of cases. Take two siblings. One is symptomatic, so they get a test. The other isn't, but somehow gets tested a few days later, and the test is positive. When did they become infectious?

The asymptomatic sibling was probably infected at about the same time as their symptomatic sibling. Or did they catch it from their sibling, in which case they probably got it 4-5 days later? So how long should they isolate? And when were they likely to have been infectious?

And if you test other asymptomatic people - we assume the test was done at the point at which they would have developed symptoms - but that is a complete guess.

I should add something about how we know who the "true" cases are, so we can evaluate tests.

It became apparent early in the pandemic that a lot of the cases in hospital did not test positive at first. But they did eventually.

I was told by @DocStrain that, in these people who eventually tested positive, about 67% had a positive test on first attempt, after two tests at least one was positive in about 68%, and in some, even after 4, 5 or more tests they still hadn't had a positive result.

In some cases the diagnosis remained a clinical diagnosis - on the basis of a clear cluster of clinical findings - or was eventually confirmed by subsequent antibody test results.

Which brings me on to antibody and other immunity testing, which I shan't go into in detail in this thread.

Not everybody produces antibodies after SARS-CoV-2 infection; and not all antibodies are "neutralising" antibodies that will protect them from future infections.

My impression, from seeing (if not reading them all thoroughly - I don't have the time) is that people who get neutralising antibodies are likely to be immune and protected for at least six months; but the jury is still out on this.

And testing for T-cell responses - cellular immunity - is difficult to do and expensive. Without a breakthrough in technology, it is unlikely to be widely available (although samples of vaccinated people will have this sort of testing done so we can understand vaccine response).

By this point in this long and rambling thread (which I started about 4 days ago), you might be asking yourself "well, why are we so hung up about testing, then?"

If that's what you're asking yourself, you've been paying attention!

Testing is of very limited value.

It is good for telling you where* in the population the disease is.

It is almost useless for screening individuals.

*Geographically, by age group, by occupation or…

The "Test, test test" imperative, and the constant attention on how many tests we are doing, and why we aren't doing more, is a distraction.

Yes, we need to know how much disease is out there; and testing, using an appropriate, systematic sampling frame is useful for that.

But what matters in "test, trace, isolate" is ISOLATE.

Testing is not an intervention that, in itself, reduces transmission.

If we are to control this virus, we need to reduce transmission.

Basic principles continue to apply. The virus spreads via droplets (of various sizes), and via fomites

The more people mix with other people, the greater the opportunity for the virus to spread (and the increase is, at least until gatherings become very large, non-linear, near exponential). threadreaderapp.com/thread/1269257…

So…
•Keep mixing – the number of people you meet – to a minimum
oSo don’t go out if you don’t have to; encourage support people to work from home wherever; think about how travel can be minimised…

The rule of six has the virtue of being easy to enforce; but it doesn't go far enough. Remember, the chances the person you meet will be infectious depends how many people they have met.

How much of a risk are you prepared to take?

How much of a risk are you prepared to put the people you have contact with to?

Remember - you might be infectious without knowing it.

As the disease spreads in the community, further restrictions will be imposed. They are a direct consequence of how careful people are about how much they spread the infection.

Parenthetic tweet.
This is not (entirely) about blaming individuals for spreading the disease. Policies which force people out to work, especially without adequate PPE; policies which cause or exacerbate deprivation and overcrowding - these are the result of political decisions.

When you can't avoid mixing with other people, do what you can to reduce (mitigate) the risks.

Wear a facemask (not a useless visor – and there should be no exemptions for staff!) where you can’t avoid getting closer than 2m to others.

Meet up outdoors wherever possible: infectious droplets are dispersed more quickly where there's better ventilation; sunlight rapidly kills the virus; humid, poorly ventilated spaces are particularly risky.

The virus also survives better in the cold - but better cold and ventilated than warm with poor ventilation.

Washing your hands (or, if your hands are clean, using alcohol based hand gel) will kill any virus on your hands; and clean surfaces the same way.

Parenthetic tweet:
Early studies of how long the virus will "survive" on surfaces mainly used PCR testing. Remember - PCR tests will detect degraded viral RNA long after the virus was inactivated. Don't worry too much about surfaces, other than those recently touched by others.

We need to protect the most vulnerable - people with pre-existing conditions, older people. If you're going to visit older or vulnerable people, try to minimise your chance of getting infected before you do so - self-isolate as far as possible for a week or more beforehand.

Keep an eye on the local risk measures. Whatever measure of current case levels is available - is it high? Is it rising (indicating it's probably higher than reported but the data hasn't caught up yet)? As risk levels rise, take more care to minimise your contact with others.

And to WEAR A MASK whenever you go out. Masks aren't 100% effective; but they can make a big difference to transmission rates.

If you are vulnerable, wear a medical mask (FFP2 or equivalent). This provides you with extra protection.

If you are particularly vulnerable - especially if your eyes might get splashed - consider adding eye protection such as a visor; but remember, visors are for eye protection only, and you should never wear one without also wearing a mask!

Parenthetic tweet:
Feel free to scream at the telly or - better still - write to complain - every time you see somebody wearing a visor but no mask.

Avoid shops etc where the staff (and their managers) care so little about their customers' health that they don't wear masks.

The more that individual people can do to reduce the spread of infection, the less need there will be for government to introduce and enforce restrictions.

@threadrerapp unroll

@threadreaderapp unroll