Dr No is not sure whether Sarah Montague on yesterday’s WATO almost had a Harry Commentator is your Carpenter moment. Alluding to the Secretary of State for Health, something sounding (at 14m 20s) rather like Hat Mancock got stiffled in a slew of splutters. Whether Montague said it or not, it is a fitting name for a man seen by many as a bit of a knob, a man who would rather pop cherries on cakes than deliver an effective test and trace system, and who happens, for the time being, to be the man with overall responsibility for covid–19 surveillance in England. This surveillance covers both how many people currently have covid–19 (swab tests for the virus), and the crucial question of how many people have had covid–19, with or without symptoms, in the community since the pandemic started (blood tests for antibodies).
The latter, also known as sero-prevalence, is crucial because the number of people with antibodies also tells us how many people don’t have antibodies, and so definitely are susceptible to infection. This is the number that really matters, which means we don’t need to fret about whether antibodies really do confer immunity, or for how long. If a certain percentage of the population don’t have antibodies(1), then the state of so called herd immunity — in which enough people in a population are immune to mean that community spread fizzles out — cannot exist. For covid–19, we need around 60% (or more) of the population to be immune for herd immunity to exist, or, to put it the other way round, if 40% (or more) of the population do not have antibodies, and so are susceptible, then herd immunity does not exist.
Immunity is undoubtedly one of the surest ways of shutting down an infectious disease, though not the only one, but the others either require intrusive restrictions (for example, taking handles off water pumps, and lockdowns), or slow general social trends over time (for example, better sanitation and housing), or labour intensive test and trace (with the constant merry go round of bolting stable doors after the horses have bolted). Immunity, on the other hand, ‘just works’, silently and in the background. Given immunity’s crucial importance, we have been rather slow to start measuring it — but maybe that’s just the way the cookie crumbles — you pay cherries, you get fruit-cakes.
So far, we have in England had two streams of antibody testing. The first, which Dr No doesn’t take too seriously, has been piggy-backed on to blood donations. Samples from donors are tested for antibodies, but by definition this is a selected sample, of healthy blood donors, and so not representative of the population at large. We can speculate until the cherries pop on whether this selected sample might have higher, the same or lower rates of sero-prevalence than the population at large. Perhaps it is higher (healthy survivors?) or lower (better general health, less susceptible to infection in the first place?). We simply don’t know. Nonetheless, for the record, the latest English sero-prevalence rate from this blood donor sample is around (typos in the report notwithstanding) 7.5% for the period 28th May 22nd June. Curiously this is down a little (from something just over 8%) on the previous period, perhaps merely sampling variation, or maybe a hint (just about allowable, as it is the same population being sampled) that immunity is short-lived.
The perhaps better survey, though numbers are painfully small, is the ONS Coronavirus (COVID-19) Infection Survey Pilot, conducted in partnership with a bewildering array of Universities, quangos and, who knows, other assorted cherry poppers. To date, some 3,298 individuals aged 16 and over in the community (and so excluding hospitals, care homes and other institutions) have been tested, with 153 testing positive, giving a weighted (using age, sex and various social factors) sero-prevalence rate of 6.3% (the unweighted rate is 4.6%, and there were some individuals who were tested more than once; 3,298 is the number of individuals tested, not the number of samples). The blood test sample is a 10% sample of a sample ‘drawn mainly from the Annual Population Survey (APS), which consists collectively of those who successfully completed the last wave of the Labour Force Survey (LFS) or local LFS boost, and who have consented to future contact regarding research’. Though likely to be more representative than the blood donor sample, this is still not a random sample, and furthermore it excludes those aged under 16, and those resident in institutions.
For an international comparison, we have a recent study from Spain, a country which was by no means bypassed by covid–19, which found an overall sero-prevalence rate of around 4.8%. Now, the notable thing is that in both the English studies, and the Spannish study, the sero-prevalence is always less than 10%, and may be as low as around 5%. Or, putting it the other way round, after the first wave of covid–19, somewhere between 90 to 95% of the population don’t have antibodies, and remain wholly susceptible to infection. Herd immunity isn’t the already visible light at the end of the tunnel, or even the distant glint of a cherry in Dido’s eye, it’s a dark star a long long way away, only visible through the most powerful of epidemiological telescopes.
In the good old days, when doctors didn’t have a clue what was going one with a patient presenting with all manner of symptoms, a good get out of jail response was to say confidently that ‘there was a lot of it about’, without specifying what ‘it’ was. Today, with covid–19, the situation appears reversed. Even in those with symptoms, it increasingly appears that actual infection rates conferring immunity remain remarkably low, and so we can say with some confidence, and indeed trepidation, that so far, there’s not a lot of it about.
Footnote (1): For the technically minded, the antibody tests used appear to have tolerable specificity, over 90% and perhaps approaching 100%, meaning not too many false negatives. Even so, with a sample of say 3000, a true prevalence rate of 5%, and specificity of say 95%, we will still get around 20 false negatives, ie around 20 people with a negative result despite having antibodies present. These figures do not substantially alter the sero-prevalence rates quoted later in the post, eg the unweighted 4.6% might become 5.4%.