Let me state upfront that I am not an epidemiologist and nor do I work in public health; I’m a plant geneticist at UCT. However, as a data-driven scientist, I have a strong conviction that not only must government use an evidence-based approach to decide on the way forward in dealing with the Covid-19 pandemic, but that it must freely share the data with the South African public. Huge sacrifices have been asked of the entire population during this lockdown period and we have every right to see the data and models on which these decisions have been made.

My second motivation for writing this article is to try to counter some of the panic and fear that people may be feeling about Covid-19 – the disease caused by the virus Sars-CoV-2 – if they’ve read articles that are sensationalist and seem almost designed to cause fear (this piece is an excellent example of this), or have seen the outlandish claims about the percentage of the population that will succumb to Covid-19 that have appeared on Twitter, Facebook and elsewhere online (I’ve seen values up to 20%).

The reality is that at the current time we simply do not know how many Covid-19 cases there are in South Africa, and nor will we until there is testing of a large, randomly selected sample of the population.

However, during the past five weeks while we have been under lockdown, critical new information has come to light regarding the likely prevalence of the virus in the populations of various other countries. I say critical, as in order to calculate a realistic estimate of the Sars-CoV-2 infection fatality rate (IFR – the percentage of the population infected with the virus that will die), it is essential to know how many people have actually been infected with the virus in the first place.

Initial estimates of the IFR were close to 1% (far lower than the numbers circulating on social media), but on the basis of subsequent research, the Centre of Evidence Based Medicine at Oxford University currently estimates the IFR at 0.1-0.36%. To give some context, the IFR of the seasonal flu is 0.1%, but note that even if the IFRs of the two viruses were identical, this does not mean that they will necessarily kill the same absolute number of people, as one may infect a higher proportion of the population than the other. The decrease in the estimated IFR is due to the fact that the prevalence of the virus in the population is much greater than was previously thought.

There is increasing evidence that a significant proportion of people with Covid-19 are asymptomatic, or suffer only mild symptoms that do not cause them to seek medical assistance. For example, a study in Iceland where 13,080 people were randomly selected and tested for Sars-CoV-2 revealed that 43% of those who tested positive were asymptomatic, and a similar proportion of asymptomatic cases has been reported in the Vo municipality in Italy.

Even on the Diamond Princess, a cruise ship quarantined in Japan for four weeks early on during the pandemic, 18% of those infected with the virus displayed no symptoms at all – and this in a group with a large proportion of elderly people who are well known to be more susceptible to the virus. These studies were performed using polymerase chain reaction (PCR) based tests, which detect the presence of the virus’s genetic material in an individual.

Huge efforts have gone into developing serological tests to detect the presence of antibodies to Sars-CoV-2 in people, as these allow the detection of those who have already had and cleared the virus, and so would not test positive using the PCR method. Several preliminary (still to be peer reviewed) studies using such tests have suggested much higher proportions of asymptomatic cases in the population. For example, in a study of 3,300 people in Santa Clara, California, the infection rate as of 4 April was estimated at 2.5 to 4.2% of the population, which is 50 to 85 times higher than the official number of infected people in this region.

Criticisms have been levelled against such studies in terms of the use of non-representative samples and the effect of false positive results (a positive test result from a person who has actually never had the virus), given the small absolute number of people testing positive for anti-SARS-CoV-2 antibodies in these studies. These are valid criticisms, and a number of large-scale studies have now begun and new serological tests continue to be developed, which will be critical in determining accurate infection rates in the population. Nonetheless, initial serological studies from Los Angeles, Germany and the Netherlands all point in the same direction as the Santa Clara study and the PCR-based tests from Iceland and Italy – all indicating that many more people have been infected with Sars-CoV-2 than is realised.

A worst-case scenario, and one where we actually do have data for the whole “population”, comes from the Diamond Princess cruise ship mentioned earlier. Here more than 3,000 people on board were tested for the virus (some repeatedly); some 700 tested positive, of whom seven died (an IFR of 1%). Again, it is important to bear in mind that this cohort contained a large group of elderly people, who are known to be much more susceptible.

A recent paper, also awaiting peer review, has attempted to quantify the age-specific risk of dying from Covid-19 in several European countries and the US. Their analysis showed that people aged under 65 with no co-morbidities (such as diabetes, high blood pressure) made up an extremely small proportion of the fatalities; 0.3% in the Netherlands, 0.7% in Italy and 1.8% in New York at the time of publication. In this respect South Africa is fortunate in that only 6% of our population is over 65, versus 23% in Italy and 15% in the USA. On this basis we might thus expect a lower IFR rate than in those countries. However, this advantage may be countered by negative factors such as the higher incidence of HIV in South Africa versus these countries, resulting in a substantial cohort of immuno-compromised people, who are likely to be more susceptible to Covid-19.

It is very early days in terms of the present epidemic, but one potential glimmer of hope comes from looking at the resolved cases – those where a patient has either fully recovered or died – in South Africa versus the rest of the world. Taking the 28 April data, in South Africa there have been 93 deaths and 2,073 recoveries, giving a resolved case fatality rate of 4.3% versus 19% globally. Of course, this may increase as the virus spreads, particularly if our healthcare system were overwhelmed during the epidemic, but it is nonetheless a positive sign.

The infection fatality rate is a critical parameter in the epidemiological models used to predict the effects of an infection on a population. In their model, epidemiologists at Imperial College took into account the age-dependent susceptibility to Covid-19 by the use of age-group-specific IFRs ranging from 0.002% for 0 to 9 years to 9.3% for the over 80s, and then used an overall IFR of 0.9% to predict that 500,000 people in the UK and 2.2 million in the US could die if no preventative measures were employed. This was hugely influential, as the predictions of these models informed the decisions made by the UK and US governments including the imposition of a lockdown.

However, the outputs from these models necessarily depend on the parameters that are set, colloquially known in computer science as “garbage in, garbage out”. Re-running that model with an IFR of 0.23% (the mid-range of current estimates) without changing any of the other parameters – the reliability of which have also been questioned – would result in a massive decrease in the number of deaths predicted (while using a higher IFR would result in more predicted deaths). The other major parameters which remain highly uncertain are the basic reproduction rate (R0) which is a measure of the transmission potential of a disease and is the average number of new infections an infected individual will give rise to, and the percentage of the population that will be infected during the pandemic. Again, the higher these values are, the more deaths will tend to be predicted by the model.

Similar models have been developed in South Africa and used to inform government responses; however, in stark contrast to other countries these models and the parameters used in them have not been released for public scrutiny. We therefore have no idea what IFR value was used in them to derive the prediction of “tens of thousands of deaths” cited by President Ramaphosa, nor how this was adjusted for age or the proportion of the population that is immuno-compromised. Nor do we know what R0 value was used or what percentage of the population it was assumed would be infected.

Science is driven by transparency, and any scientist wishing to publish their work in the scientific literature is obliged to make their data available for scrutiny. I see absolutely no reason why this should be different now. This scrutiny is particularly important when the entire population has been told to remain in lockdown for many weeks on the basis of this modelling, and even more so given the concerns raised by Prof Shabhir Madhi (former head of the National Institute for Communicable Diseases) about the assumptions used in these models. We have a right to see the details of these analyses and they should be released immediately.

Details of the Sars-CoV-2 testing performed to date are similarly opaque. The Western Cape (1,737 cases as of 27 April), Gauteng (1,353) and KZN (902) have been labelled as “hotspots” that may need to be under extended lockdown. While these provinces do indeed have the highest absolute number of recorded cases, these numbers are meaningless without taking into account two other factors – the population of each province, and the testing rate (number of tests performed per million inhabitants) in each province. If we calculate the infection rate per million inhabitants using the 27 April Covid-19 figures from the National Institute for Communicable Diseases, and 2019 population estimates from StatsSA, (the three provinces with the highest infection rates are now the Western Cape (254/million), Gauteng (89/million) and Eastern Cape (88/million) with the others ranging from 5/million (Limpopo) to 80/million (KZN).

However, these infection rates cannot be directly compared unless the testing rates are similar across the provinces. Government provides a daily update of the number of tests performed nationally, but does not break this down by province. It must do so in order to allow us to know whether the infection rates are truly different between provinces, or simply reflect differential rates of testing between the provinces. This is particularly important given the announcement by President Ramaphosa that provincial-specific levels of restrictions may be imposed from 1 May.

The same logic would of course apply to any restrictions to be applied at a metro or district level. The Eastern Cape government announced that it had performed 9,618 tests as of 24 April while the Western Cape government had performed 26,666 tests as of 26 April. If we divide the number of confirmed Covid-19 cases on these dates by the number of tests that had been performed, we see infection rates in the test subjects of 5% for the Eastern Cape (480 cases/9,618 tests as of 24 April) versus 6% in the Western Cape (1,608 cases/26,666 tests as of 26 April). 

Hardly a dramatic difference despite the somewhat hysterical headlines proclaiming the Western Cape as the epicentre of the Covid-19 outbreak that have appeared on EWN, News24 and Business Day among others. One can only hope that the people advising government have a sounder grasp of numeracy.

These are unusual times, and we have all been under extreme levels of restriction regarding freedom of movement for the past five weeks, with enormous economic and social costs. I don’t doubt the government’s good intentions in wanting to protect South Africans, but we have every right to see the science behind these decisions and should all demand that the government make these data available with immediate effect. Asking us to accept these decisions on blind faith is simply not acceptable. DM