Share

In the first of our posts in this short series, we looked at the initial phase of the Covid pandemic. Despite the hype, we showed that the initial rise in infections lasted 2-3 weeks in China and Northern Italy. However, deaths kept adding up for some time afterwards.

Here, we look at the science underlying two often quoted rates: the infection fatality ratio (IFR) and the case fatality ratio (CFR).

The Hallett Inquiry leading KC, Mr Keith, put it to Boris Johnson that “BSE did not have a 2% fatality rate, swine flu did not have a 2% fatality rate, so when you say there was an institutional failure to realise the seriousness of the position because of Asiatic, prior Asiatic, epidemics, or because of BSE or swine flu, the difference, and it was known to government, was that Covid had a 2% fatality rate and BSE and swine flu had not.”

Mr Keith made it clear he was referring to the IFR in his question's preamble. However, his estimate was twice the one per cent figure that modellers often cite. Nevertheless, the actual IFR is much lower. A pooled analysis of 40 studies found that the IFR is as low as 0.0003% for children and adolescents, while it can reach 0.5% for people aged 60-69.

This was one of many wrong assumptions. But the main mistake is assuming that the current way the IFR is calculated makes epidemiological sense.

We have previously noted that estimating the IFR in the early stage of outbreaks is so error-prone that it should come with a warning, as it overpredicts the number of deaths, influencing the political decisions without considering their long-term harm and well-being effects.  

Mr Keith told Sunak that the IFR amongst the hospitalisation data was as “plain as a pikestaff” to tip you into lockdown. 

The WHO defines the IFR as an estimate of the proportion of deaths among all infected individuals. This definition is unclear, as we shall explain, as it is not defined by the term “infected”. For now, let us state that to measure the IFR accurately, a complete picture of the number of infections and deaths caused by SARs-COV-2 must be known. 

The other often quoted measure is the CFR. This can be estimated if only PCR cases and deaths are considered. However, CFRs are highly variable, depending on who is tested for what reasons. Early on, they are subject to selection bias as more severe cases are tested – generally those in the hospital setting or those with more severe symptoms. 

The population age structure also influences both the prevalence of disease and mortality, often making comparisons worthless. Furthermore, analyses not knowing the time between the PCR positive test and death are often undertaken, leading to often scarily high estimates.  We must remember that qualitative PCR on its own (i.e. positive/negative) is not proof of active infection; it is not proof of the presence of an actively replicating virus in your body, even in the absence of test contamination.

A properly used PCR, following a person during the evolution of the infectious disease, is necessary for any correct inference to be drawn. 

We have shown that the higher the viral load, the more likely the person is to have active disease and be capable of passing on the bug. However, this is not a one-test static situation; as most people get better, fewer die. Infections, especially acute ones, evolve, yes? So, each case needs follow-up, multiple PCR tests, and clinical examinations to be labelled as “infected and infectious”. “Infected” could mean either having been infectous weeks before and having recovered or having an active infection, as qualitative PCR cannot distinguish the two unless you have an estimation of the viral load and a clinical examination. This was one of the reasons why mass testing was a science-free colossal fund drain.

Such a clinical approach is as far away from models and spreadsheets as you can be. This is one of the main things lacking in the whole saga.

We seem incapable of learning this simple point - the CFR is affected by substantial biases and is consistently overinflated early in pandemics.

By 11 February, the CFR in China was reported as 2.3%. In Wuhan, it was as high as 4.3%. Reviews of influenza H1N1 and SARs also show early estimates of the CFR were grossly inflated.  

The anxiety the CFR creates is hard to overcome; it carries considerable weight in influencing early pandemic decision-making. 

Because the IFR takes account of all those infected, it is significantly lower than the CFR. To calculate all those infected, antibody testing is used to try to provide a more accurate understanding of how many people have been infected.  However, antibodies decay with time, and further complicating matters, endemic human coronaviruses are cross-reactive to SARs-CoV-2 antibodies.

The considerable uncertainty over how many people are infectious means that reported IFRs are often overestimated. In Swine influenza, the IFR ended up at  0.02%, fivefold less than the lowest estimate during the outbreak (the lowest estimate was 0.1% in the 1st ten weeks of the epidemic). 

Population demographics vary the IFR  significantly: The IFR will be lower if younger populations are infected. Conversely, if the elderly or those with comorbidities are infected, this will increase the IFR.  Therefore, a single IFR estimate across all age groups is meaningless. These limits are often ignored. 

It is also essential to understand whether individuals are dying with or from the disease. Cause of death information from death certificates is often inaccurate and incomplete, particularly for conditions such as pneumonia. There is no accepted definition of COVID-19-related death, and as we have shown, deaths can be recorded in one of 14 different ways.

So, we have shown that the system for calculating the IFR is a mess. Sloppy science has infected the academic method. But is there a way forward? 

Yes, there is, if the issues of infectiousness and deaths are assessed accurately.

First, accurate information on the cause of death is required. To do this would require a clinical review of the case notes and a clear causal chain from the point of infection. Italy has carried out one such review of a sample of case notes.

Second, we need accurate information on the nature of the infection. Was the person infectious at the point of death, and was that infection SARs-COV-2, or was it another acute respiratory infection?

Third, an accurate assessment of the number of people actively infected at the time of observation is required. To achieve this requires a prospective population-based study in a specific setting. 

So, what can we conclude?  

The current system suits the academic agenda and the narrative for intervening in pandemics. Yet, no one has a clue what is going on - it's like flying blind in the era before radar.  There are few incentives to improve the current situation. Given the enormous impact on society, the reasons for this are perplexing, but commercial forces are indeed at work. 

Until we get accurate estimates of what people died of, what they were infected with and when, then it’ll be more of the same - confusion, contradictions and calls for more interventions.