What “The Science” Really Says about the Coronavirus Pandemic

The answer is not much – at least, not yet.

While advocates of lockdowns and masking mandates claim to be invoking “the science,” science by its very nature can’t provide short-term answers to the efficacy of such measures. The scientific method demands extensive data gathering and testing, which generally take longer than the duration of a pandemic. An abundance of scientific evidence does exist for the effectiveness of vaccination, but whether vaccines can completely eradicate the coronavirus is an open question. Social distancing as a preventive measure is also on firm scientific ground.

Lockdowns have been used for centuries as a way to slow the spread of disease, including the Black Death plague in the 14th century and the Spanish Flu in 1918-1919. But all they do is initially reduce transmission of the virus, and to claim otherwise is scientifically disingenuous.

The primary purpose of slowing down the spread of a contagious and deadly disease is to prevent the healthcare system from becoming overwhelmed. If more people get sick enough to require hospitalization than the number of hospital beds available, some won’t get adequate treatment and deaths will increase. However, lockdowns also have a devastating effect on a society’s economic and mental health. Studies have shown that negative socioeconomic impacts greatly limit the effectiveness of lockdowns over time.

In some countries such as Taiwan and Australia, death rates from COVID-19 are very low so far after repeated lockdowns, causing lockdown supporters to link the two. But other nations with small populations such as Israel have much higher mortality rates despite continued shutdowns. So there’s no correlation and, in fact, many other factors influence the death rate.  

The science behind masking, shown below in use during the Spanish Flu pandemic, is muddier yet and has been badly contaminated by politics. Unfortunately, the gold standard in medical testing – the RCT (randomized controlled trial) – isn’t the basis for evaluating the benefit of mask-wearing by institutions like the U.S. CDC (Centers for Disease Control and Prevention) or the WHO (World Health Organization).

In an RCT or clinical trial, participants are divided randomly into two identical groups, with intervention in only one group and the other group used as a control. Neither the researchers nor the participants are told which group the participants are part of until the very end. Such double-blind trials are therefore able to establish causation.

For masks, just 14 RCTs have been carried out across the world to study how well masks guard against respiratory diseases, primarily influenza. Nearly all the trials tested so-called surgical, three-ply paper masks, rather than the N95 respirator style. Of the 14 trials, just two investigated the claim that wearing a mask benefits others who come in close contact with the mask wearer, while the other 12 tested the combination of benefit to others and protection for the wearer.

A recent analysis by a prominent statistician of all 14 RCTs, which include the only trial to test mask-wearing’s specific effectiveness against COVID-19, reveals that masks have no significant effect on either the wearer or those in close proximity, although some trials were ambiguous. There was no strong evidence that N95 masks performed any better than surgical or cloth masks. Exactly the same conclusions were reached in two independent analyses of 13 (see here) and 11 (here) of the same RCTs.

The CDC, however, relies on observational studies conducted since the start of the coronavirus pandemic, not RCTs, in issuing its masking guidance. An observational study is less scientific in being unable to assign a cause to an effect; it can only establish association.

Vaccination against infectious diseases, on the other hand, has a solid scientific basis. Pioneered by Edward Jenner at the end of the 18th century, vaccination has eradicated killer diseases such as smallpox and polio in many countries, and drastically curtailed others such as measles, mumps and pertussis (whooping cough).

Nevertheless, the science underlying vaccination against COVID-19 is incomplete. In the past it’s taken several years to develop a new vaccine, but the COVID-19 vaccines currently available were brought to market at lightning speed. Although such haste was seen as necessary to combat a rapidly proliferating virus, it meant shortening the RCTs designed to test vaccine efficacy, leaving questions such as long-term side effects and duration of effectiveness unresolved.

And barely understood yet is the greater protection against infection acquired though natural immunity – the result of having recovered from a previous COVID-19 infection – than from vaccination. This complicates calls for vaccine mandates, as those with natural immunity arguably don’t need to be vaccinated.

Moreover, the coronavirus is an RNA virus like influenza and so frequently mutates. This means that mandatory vaccination for the coronavirus is unlikely to be any more effective community-wide than a mandated flu vaccine would be. Regular COVID-19 booster shots will probably be needed, just like the flu.

That’s where science stands on the coronavirus. But rather than following the science, most decisions on lockdowns, masking and vaccination are ruled by politics.

Next: Sea Ice Update: No Evidence for Recent Ice Loss

Science on the Attack: The Vaccine Revolution Spurred by Messenger RNA

The lightning speed with which biotech companies Pfizer-BioNTech and Moderna developed a safe and effective COVID-19 vaccine is testament not only to scientific perseverance, but to the previously unrealized potential of messenger RNA (mRNA) to revolutionize medicine. Today’s blog post in my series showcasing science on the attack rather than under attack highlights the genetic breakthrough behind this transformational discovery.

Genetic vaccines are a relative newcomer to the immunization scene. Unlike traditional vaccines that use killed or weakened versions of the virus to stimulate the body’s immune system into action, genetic vaccines deliver a single virus gene or part of its genetic code into human cells. The genetic instructions induce the cells to make viral proteins that constitute only a small piece of the virus, but have the same effect on the immune system as the whole virus molecule.

But, until 2020, the only approved genetic vaccines – based on DNA, not RNA – were for animal diseases. It was the urgent need to come up with a vaccine to protect against COVID-19 in humans that triggered the worldwide quest to bring an mRNA vaccine to market.

The job of mRNA in the body is to transcribe the DNA code for one or more genes contained in a cell nucleus, and then deliver the encoded information to the protein factory in the cell’s outer reaches. There, the message is decoded and the requisite protein manufactured. DNA contains the blueprint for making nearly all the proteins in the body, while mRNA acts as a delivery service.

The concept of harnessing mRNA to fight disease goes back to the early 1990s, but hopes raised by promising early experiments on mice were dashed when multiple roadblocks arose to working with synthetic mRNA injected into the human body. The primary obstacle was the immune system’s overreaction to mRNA engineered to manufacture virus proteins. The immune system often destroyed the foreign mRNA altogether, as well as causing excessive inflammation in some people. Other problems were that the mRNA degraded quickly in the body and didn’t produce enough of the crucial virus protein for a vaccine to be effective.

So scientific attention switched instead to development of DNA vaccines, which cause fewer problems though are clunky compared to their mRNA cousins. Then, in a series of papers starting in 2005, two scientists at the University of Pennsylvania, Katalin Karikó and Drew Weissman, reported groundbreaking research that brought mRNA back into the limelight.

Karikó and Weissman found that tweaking the structure of the mRNA molecule could overcome most of the earlier obstacles. By exchanging one of mRNA’s four building blocks called nucleosides, they were able to create a hybrid mRNA that drastically suppressed the immune system’s reaction to the intruder and boosted production of the viral protein. In their own words, their monumental achievement was “the biological equivalent of swapping out a tire.”      

Their discovery, however, was initially received with a big yawn by many of their peers, who were still preoccupied with DNA. Karikó found herself snubbed by the research funding community and demoted from her university position. Eventually, in 2013 she was hired by the German company BioNTech to help oversee its mRNA research.

In the meantime, work proceeded on the final impediment to exploiting synthetic mRNA for vaccines: preventing its degradation in the human body. To reach the so-called cytoplasm of a cell where proteins are manufactured, the artificial mRNA needs to penetrate the lipid membrane barrier protecting the cell. Karikó, Weissman and others solved this problem by encasing the mRNA in small bubbles of fat known as lipid nanoparticles.

Armed with these leaps forward, researchers have now developed mRNA vaccines for at least four infectious diseases: rabies, influenza, cytomegalovirus and Zika. But testing in humans has been disappointing so far. The immune response has been weaker than expected from animal studies – just as with DNA vaccines – and serious side effects have occurred.

Nevertheless, COVID-19 mRNA vaccines have been a stunning success story. The major advantage of mRNA vaccines over their traditional counterparts is the relative ease and speed with which they can be produced. But until now, no mRNA vaccine or drug has ever won approval.

Maybe COVID-19 is the exception and synthetic coronavirus mRNA generates a stronger immune response with fewer adverse effects than the other viral mRNA vaccines investigated to date. Mass production of a beneficial and safely tolerated COVID-19 vaccine in less than 12 months is certainly an amazing accomplishment, considering that it’s taken several years to develop a new vaccine in the past. But whether the potential of mRNA vaccines to ward off other diseases or even cancer remains to be seen.

Next: Latest Computer Climate Models Run Almost as Hot as Before

Science on the Attack: The Hunt for a Coronavirus Vaccine (2)

In the previous post, we saw how three different types of coronavirus vaccine, all based on established technologies, are under development: virus (killed or attenuated live), viral vector, and protein-based vaccines. Here I review experimental genetic vaccines for SARS-CoV-2, which rely on protective antibody production – like the other three do; and T-cell-inducing vaccines, a newcomer approach which arouses a massive army of those immune system warriors, T cells.

Genetic vaccines, sometimes called nucleic acid vaccines, utilize part of the coronavirus’s genetic code to deliver the genetic instructions for a coronavirus protein such as the spike protein, right into human cells. In this seemingly risky move, the cells read the instructions and crank out copies of the viral protein, but not of the whole virus as infected cells do – and thus don’t cause disease. The protein copies stimulate antibody generation, just like the viral protein fragments or shells in protein-based vaccines.

The genetic instructions can be in the form of either DNA or RNA. For DNA vaccines, an engineered loop of coronavirus protein DNA is inserted into cells, which then employ their own messenger RNA to assemble the viral proteins. RNA vaccines deliver synthetic viral messenger RNA directly into cells. An advantage of genetic vaccines is that they can be produced more rapidly than their traditional counterparts.

Other DNA vaccines have been approved for animal diseases such as West Nile virus in horses, and a DNA coronavirus vaccine based on the spike protein has been found to protect monkeys. But no DNA coronavirus vaccines so far have approval for human use. The same is true for RNA coronavirus vaccines, although biotech company Moderna recently obtained promising results in a small trial of coronavirus vaccine safety. 

T-cell vaccines have gained attention because of emerging evidence that many people may already have immune cells capable of recognizing the SARS-CoV-2 virus and warding it off. This extraordinary degree of protection is thought to come from T cells, not antibodies. Although studies have found that antibodies against the deadly coronavirus dissipate fairly quickly, T cells are able to remember past infections and kill pathogens if they reappear, even after long periods of time. A recent research paper reported that up to 50% of people who had never been exposed to the virus had high levels of SARS-CoV-2-specific T cells, a finding replicated in other studies.

Like many advances in science, this particular discovery was accidental. The paper’s authors were conducting an experiment with COVID-19 convalescent blood and needed a control blood sample for comparison. After choosing blood samples collected from healthy residents of San Diego between 2015 and 2018, several years before the current pandemic began, they found to their surprise that about half the samples showed strong T-cell reactivity against the virus.

The authors speculated that this T-cell recognition of the SARS-CoV-2 virus may come partly from previous exposure to one of the four known coronaviruses that cause the common cold and circulate widely among humans. If so, the discovery paves the way to a new type of vaccine, similar to those being used against certain cancers such as melanoma. However, the authors emphasized that the data hadn’t yet demonstrated the source of the T cells or whether they are actually memory T cells.

Memory T cells are the third type of T cell, in addition to helper T cells known as CD4+ cells that identify antigens, or viral protein fragments, and killer T cells that devour virus-infected cells. T-cell memory of past diseases is long lasting, up to decades. People who recovered from SARS, the disease most closely related to COVID-19, still show cellular immunity to that coronavirus after 17 years.

CREDIT: WIKIPEDIA COMMONS

An even more recent study appears to confirm the hypothesis that the observed T-cell response results from previous exposure to common cold coronaviruses. Should this turn out to be the case, it could explain the puzzle of why COVID-19 is much more severe in some people than in others: those who have recently wrestled with the common cold may have an easier time battling a more vicious member of the coronavirus family, and may get less sick. On the other hand, much is still unknown and pre-existing T cells could even interfere with other immune system responses.

As for a coronavirus vaccine, recent Phase III clinical trials have shown the efficacy of potential T-cell-inducing vaccines for diseases such as malaria and HIV. But nothing is yet licensed, so development of a coronavirus T-cell vaccine is unlikely in the short term.

Next: It’s Cold, Not Hot, Extremes That Are on the Rise

Science on the Attack: The Hunt for a Coronavirus Vaccine (1)

In my series of occasional posts showcasing science on the attack rather than under attack, this and the next blog post will review the current search for a vaccine against that unwelcome marauder, the coronavirus. This post examines vaccine approaches based on conventional, well-established techniques. The subsequent one will look at experimental technologies not yet approved for medical use.

The coronavirus (SARS-CoV-2) is a very large, bristly molecule – with a genome twice as large as that of influenza – studded with spiky flower-like proteins as seen in red in the figure below. It tricks cells in the body into letting it in through a cellular door: a cup-like protein called an ACE2 receptor, which forms part of the nervous system and regulates bodily processes such as blood pressure and inflammation. Latching on to the receptor enables the virus to penetrate the host cell membrane and hijack the cell’s replication machinery, making copies of itself that then wreak havoc throughout the body.

The main function of the body’s immune system is to detect and annihilate invaders such as foreign bacteria and viruses like SARS-CoV-2. First, immune system scouts known as phagocytes – a type of white blood cell – recognize and digest intruder cells. The phagocyte surface then displays a flag or protein fragment of the bacteria or virus, called an antigen, that signals the foreigner’s identity. Other white blood cells called T cells identify the antigen, prompting the immune system arsenal to unleash one of two types of weapon against the assailant.

The two weapons are a different kind of T cell that homes in on infected cells and kills them, and yet another type of white blood cell called a B cell that produces disease fighting antibodies. Antibodies are specialized Y-shaped proteins with a search-and-destroy mission, either inactivating invasive cells directly or tagging them for elimination by phagocytes or other immune system killer cells. Coronavirus vaccines under development include both those that stimulate antibody production, and those that generate copious quantities of T cells.

Most of today’s vaccines utilize the virus itself. This can be in the form of a killed-virus vaccine,  which is produced by growing live virus and then inactivating it chemically, or an attenuated live-virus vaccine, in which live virus is weakened below the level where it can normally cause disease. Both types of vaccine induce the immune system to churn out antibodies.

The measles-mumps-rubella (MMR) vaccine is an example of a weakened virus vaccine; most flu shots are the inactivated type. An inactivated coronavirus vaccine is now in Phase III efficacy testing by Chinese company Sinovac.

Attracting more attention for SARS-CoV-2 are so-called viral vector vaccines. As indicated in the next figure, these are vaccines in which a “guest” virus such as measles (left) or adenovirus (right), which causes upper respiratory infections and related illnesses, is genetically engineered with the gene for the coronavirus spike protein. Key genes in the guest virus are usually disabled so it can’t replicate, but the piggybacking coronavirus gene is unloaded inside the body’s cells, generating antibodies that combat the coronavirus invasion.

CREDIT: SPRINGER NATURE

The only vaccine currently approved for Ebola is a viral vector vaccine manufactured by Johnson & Johnson, who also have a coronavirus vaccine in the works. But the most advanced coronavirus effort is that of the University of Oxford together with AstraZeneca, who have Phase III trials of a viral vector vaccine well underway.           

A third class of defense against the virus using established technologies is protein-based vaccines. Some protein-based vaccines contain fragments of the coronavirus spike protein, or of an important part of it known as the receptor binding domain. The fragments can’t cause disease because they’re not the actual virus, but the immune system is still able to recognize them as coronavirus proteins – triggering production of antibodies. Other protein-based vaccines contain a protein shell that mimics just the outer coat of the coronavirus, so again isn’t infectious but induces antibody production.

Current vaccines for shingles and human papillomavirus (HPV) are in this category. Several companies have Phase I or Phase II trials of a protein-based coronavirus vaccine in progress.

In the next post I’ll review experimental genetic vaccines for the coronavirus, which are based on antibodies, and newer candidates based on a strong T-cell response.

Next: Science on the Attack: The Hunt for a Coronavirus Vaccine (2)

How Science Is Being Misused in the Coronavirus Pandemic

Amidst the hysteria over the coronavirus pandemic, politicians constantly assure us that their COVID-19 policy decisions are founded on science. “Following the science” has become the mantra of national and local officials alike.

But the reality is that the various edicts and lockdown measures are based as much on political considerations as science.

One of the hallmarks of science is empirical evidence: true science depends on accumulated observations, not on models or anecdotal data. My previous post discussed the shortcomings of coronavirus models, which rely on assumptions about unknowns such as contagiousness and virus incubation period, and whose only observational data is from past flu epidemics or the current pandemic that the models are attempting to simulate.

Many governments thought they were being informed by science in employing models to forecast the epidemic’s course. But, as leaders discovered in places like Italy and New York where the healthcare system was rapidly overwhelmed, the models were of little use in predicting how many ventilators or how much other equipment they would need. It was their own on-the-spot observations and political experience, not science, that led the way.

Science is not a fountain of wisdom. As a UK sociologist remarks: “Scientists can provide evidence, but acting on that evidence requires political will.” Unfortunately, science can be subverted by the political process, politicians all too often choosing only the evidence that bolsters their existing beliefs. Because politics is more visceral than rational, the evidence and logic intrinsic to science rarely play a big role in political debate.

An example of how politics has trampled science in the coronavirus pandemic is the advice given by the UK government to its citizens on self-isolation (self-quarantine in the U.S.) for those with symptoms of COVID-19.

The UK NHS (National Health Service) says seven days after becoming sick is adequate self-isolation. Yet the WHO (World Health Organization), along with medical experts in many countries, recommends a self-quarantine period of 14 days, based on the observation that the incubation period after exposure to the virus ranges from 1 to 14 days. While scientists can and frequently do disagree, the difference between the NHS and WHO guidelines is purely the result of political interference with science.

Another area where science is being misused is antibody testing.

There’s been much fanfare about the possible use of antibody testing to determine whether someone who has recovered from COVID-19 is immune from reinfection by the virus, and can therefore circulate safely in society. That’s true for many other viruses, but hasn’t yet been established for the coronavirus. And if antibodies do confer protection against reinfection, it’s unknown how long the protection lasts – weeks, months or years.

Compounding these uncertainties is the unreliability of many currently available antibody tests, and the finding that some recovered individuals, as determined by an antibody test, still test positive for the coronavirus – meaning they could still possibly infect others. Recent research suggests these are false positives, arising from harmless fragments of the virus left in the body. However, until there’s evidence to resolve such questions, it’s a mistake for any politician or official to claim that science supports their policy position on antibody testing.

A third example of misuse of science during the pandemic is the debate over prescribing the malaria drug hydroxychloroquine as an early-stage treatment for COVID-19 patients.

It’s not unusual in medicine to prescribe a drug, originally developed to treat a particular illness, as an off-label remedy for another condition. Hydroxychloroquine has for many years been considered a safe and effective treatment for malaria, lupus and rheumatoid arthritis. At the beginning of the coronavirus pandemic, the drug was used successfully to treat COVID-19 in China, France and other countries.

But the use of hydroxychloroquine to treat coronavirus patients in the U.S. has been controversial. President Donald Trump, who took a course of the medication as a preventative measure and touted its potential benefits for sick patients, has been chastised by political opponents for his endorsement of the treatment. Several studies have appeared to show that the drug, not yet officially approved by the FDA (Food and Drug Administration), can cause serious heart problems. One of these studies has, however, been retracted because of doubts over the veracity of the data.

Nevertheless, what’s important about hydroxychloroquine from a scientific viewpoint is that all the studies so far have been epidemiological. As is well known, an epidemiological study can only show a correlation between the drug and certain outcomes, not a clear cause and effect. Epidemiological studies are notoriously misleading, as found in numerous nutritional studies. Delineation of cause and effect requires a clinical trial – a randomized controlled trial, in which the study population is divided randomly into two identical groups, with intervention in only one group and the other group used as a control. So far, no clinical trials of hydroxychloroquine have been completed.

Although science is a powerful tool for understanding the world around us, it has its limitations. It should not be used as an authority in policy making unless the science is firmly grounded in observational evidence.

Next: Absurd Attempt to Link Climate Change to Cancer Contradicted by Another Medical Study

Why Both Coronavirus and Climate Models Get It Wrong

Most coronavirus epidemiological models have been an utter failure in providing advance information on the spread and containment of the insidious virus. Computer climate models are no better, with a dismal track record in predicting the future.

This post compares the similarities and differences of the two types of model. But similarities and differences aside, the models are still just that – models. Although I remarked in an earlier post that epidemiological models are much simpler than climate models, this doesn’t mean they’re any more accurate.     

Both epidemiological and climate models start out, as they should, with what’s known. In the case of the COVID-19 pandemic the knowns include data on the progression of past flu epidemics, and demographics such as population size, age distribution, social contact patterns and school attendance. Among the knowns for climate models are present-day weather conditions, the global distribution of land and ice, atmospheric and ocean currents, and concentrations of greenhouse gases in the atmosphere.

But the major weakness of both types of model is that numerous assumptions must be made to incorporate the many variables that are not known. Coronavirus and climate models have little in common with the models used to design computer chips, or to simulate nuclear explosions as an alternative to actual testing of atomic bombs. In both these instances, the underlying science is understood so thoroughly that speculative assumptions in the models are unnecessary.

Epidemiological and climate models cope with the unknowns by creating simplified pictures of reality involving approximations. Approximations in the models take the form of adjustable numerical parameters, often derisively termed “fudge factors” by scientists and engineers. The famous mathematician John von Neumann once said, “With four [adjustable] parameters I can fit an elephant, and with five I can make him wiggle his trunk.”

One of the most important approximations in coronavirus models is the basic reproduction number R0 (“R naught”), which measures contagiousness. The numerical value of R0 signifies the number of other people that an infected individual can spread the disease to, in the absence of any intervention. As shown in the figure below, R0 for COVID-19 is thought to be in the range from 2 to 3, much higher than for a typical flu at about 1.3, though less than values for other infectious diseases such as measles.

It’s COVID-19’s high R0 that causes the virus to spread so easily, but its precise value is still uncertain. What determines how quickly the virus multiplies, however, is the incubation period, during which an infected individual can’t infect others. Both R0 and the incubation period define the epidemic growth rate. They’re adjustable parameters in coronavirus models, along with factors such as the rate at which susceptible individuals become infectious in the first place, travel patterns and any intervention measures taken.

In climate models, hundreds of adjustable parameters are needed to account for deficiencies in our knowledge of the earth’s climate. Some of the biggest inadequacies are in the representation of clouds and their response to global warming. This is partly because we just don’t know much about the inner workings of clouds, and partly because actual clouds are much smaller than the finest grid scale that even the largest computers can accommodate – so clouds are simulated in the models by average values of size, altitude, number and geographic location. Approximations like these are a major weakness of climate models, especially in the important area of feedbacks from water vapor and clouds.

An even greater weakness in climate models is unknowns that aren’t approximated at all and are simply omitted from simulations because modelers don’t know how to model them. These unknowns include natural variability such as ocean oscillations and indirect solar effects. While climate models do endeavor to simulate various ocean cycles, the models are unable to predict the timing and climatic influence of cycles such as El Niño and La Niña, both of which cause drastic shifts in global climate, or the Pacific Decadal Oscillation. And the models make no attempt whatsoever to include indirect effects of the sun like those involving solar UV radiation or cosmic rays from deep space.

As a result of all these shortcomings, the predictions of coronavirus and climate models are wrong again and again. Climate models are known even by modelers to run hot, by 0.35 degrees Celsius (0.6 degrees Fahrenheit) or more above observed temperatures. Coronavirus models, when fed data from this week, can probably make a reasonably accurate forecast about the course of the pandemic next week – but not a month, two months or a year from now. Dr. Anthony Fauci of the U.S. White House Coronavirus Task Force recently admitted as much.

Computer models have a role to play in science, but we need to remember that most of them depend on a certain amount of guesswork. It’s a mistake, therefore, to base scientific policy decisions on models alone. There’s no substitute for actual, empirical evidence.

Next: How Science Is Being Misused in the Coronavirus Pandemic

Coronavirus Epidemiological Models: (3) How Inadequate Testing Limits the Evidence

Hampering the debate over what action to take on the coronavirus, and over which epidemiological model is the most accurate, is a shortage of evidence. Evidence includes the infectiousness of the virus, how readily it’s transmitted, whether infection confers immunity and, if so, for how long. The answers to such questions can only be obtained from individual testing. But testing has been conspicuously inadequate in most countries, being largely limited to those showing symptoms.

We know the number of deaths, those recorded at least, but a big unknown is the total number of people infected. This “evidence fiasco,” as eminent Stanford medical researcher and epidemiologist John Ioannidis describes it, creates great uncertainty about the lethality of COVID-19 and means that reported case fatality rates are meaningless. In Ioannidis’ words, “We don’t know if we are failing to capture infections by a factor of three or 300.”

The following table lists the death rate, expressed as a percentage of known infections, for the countries with the largest number of reported cases as of April 16, and the most recent data for testing rates (per 1,000 people).

As Ioannidis emphasizes, the death rate calculated as a percentage of the number of cases is highly uncertain because of variations in testing rate. And the number of fatalities is likely an undercount, since most countries don’t include those who die at home or in nursing facilities, as opposed to hospitals.

Nevertheless, the data does reveal some stark differences from country to country. Two nations with two of the highest testing rates in the table above – Italy and Germany – show markedly distinct death rates – 13.1% and 2.9%, respectively – despite having not very different numbers of COVID-19 cases. The disparity has been attributed to different demographics and levels of health in Italy and Germany. And two countries with two of the lowest testing rates, France and Turkey, also differ widely in mortality, though Turkey has a lower number of cases to date.

Most countries, including the U.S., lack the ability to test a large number of people and no countries have reliable data on the prevalence of the virus in the population as a whole. Clearly, more testing is needed before we can get a good handle on COVID-19 and be able to make sound policy decisions about the disease.

Two different types of test are necessary. The first is a test to discover how many people are currently infected or not infected, apart from those already diagnosed. A major problem in predicting the spread of the coronavirus has been the existence of asymptomatic individuals, possibly 25% or more of the population, who unknowingly have the disease and transmit the virus to those they come in contact with.

A rapid diagnostic test for infection has recently been developed by U.S. medical device manufacturer Abbott Laboratories. The compact, portable Abbott device, which recently received emergency use authorization from the FDA (U.S. Food and Drug Administration), can deliver a positive (infected) result for COVID-19 in as little as five minutes and a negative (uninfected) result in 13 minutes. Together with a more sophisticated device for use in large laboratories, Abbott expects to provide about 5 million tests in April alone. Public health laboratories using other devices will augment this number by several hundred thousand.

That’s not the whole testing story, however. A negative result in the first test includes both those who have never been infected and those who have been infected but are now recovered. To distinguish between these two groups requires a second test – an antibody test that indicates which members of the community are immune to the virus as a result of previous infection.

A large number of 15-minute rapid antibody tests have been developed around the world. In the U.S., more than 70 companies have sought approval to sell antibody tests in recent weeks, say regulators, although only one so far has received FDA emergency use authorization. It’s not known how reliable the other tests are; some countries have purchased millions of antibody tests only to discover they were inaccurate. And among other unknowns are the level of antibodies it takes to actually become immune and how long antibody protection against the coronavirus actually lasts.       

But there’s no question that both types of test are essential if we’re to accumulate enough evidence to conquer this deadly disease. Empirical evidence is one of the hallmarks of genuine science, and that’s as true of epidemiology as of other disciplines.

Next: Does Planting Trees Slow Global Warming? The Evidence

Coronavirus Epidemiological Models: (2) How Completely Different the Models Can Be

Two of the most crucial predictions of any epidemiological model are how fast the disease in question will spread, and how many people will die from it. For the COVID-19 pandemic, the various models differ dramatically in their projections.

A prominent model, developed by an Imperial College, London research team and described in the previous post, assesses the effect of mitigation and suppression measures on spreading of the pandemic in the UK and U.S. Without any intervention at all, the model predicts that a whopping 500,000 people would die from COVID-19 in the UK and 2.2 million in the more populous U.S. These are the numbers that so alarmed the governments of the two countries.

Initially, the Imperial researchers claimed their numbers could be halved (to 250,000 and 1.1 million deaths, respectively) by implementing a nationwide lockdown of individuals and nonessential businesses. Lead scientist Neil Ferguson later revised the UK estimate drastically downward to 20,000 deaths. But it appears this estimate would require repeating the lockdown periodically for a year or longer, until a vaccine becomes available. Ferguson didn’t give a corresponding reduced estimate for the U.S., but it would be approximately 90,000 deaths if the same scaling applies.

This reduced Imperial estimate for the U.S. is somewhat above the latest projection of a U.S. model, developed by the Institute for Health Metrics and Evaluations at the University of Washington in Seattle. The Washington model estimates the total number of American deaths at about 60,000, assuming national adherence to stringent stay-at-home and social distancing measures. The figure below shows the predicted number of daily deaths as the U.S. epidemic peaks over the coming months, as estimated this week. The peak of 2,212 deaths on April 12 could be as high as 5,115 or as low as 894, the Washington team says.

The Washington model is based on data from local and national governments in areas of the globe where the pandemic is well advanced, whereas the Imperial model primarily relies on data from China and Italy alone.  Peaks in each U.S. state are expected to range from the second week of April through the last week of May.

Meanwhile, a rival University of Oxford team has put forward an entirely different model, which suggests that up to 68% of the UK population may have already been infected. The virus may have been spreading its tentacles, they say, for a month or more before the first death was reported. If so, the UK crisis would be over in two to three months, and the total number of deaths would be below the 250,000 Imperial estimate, due to a high level of herd immunity among the populace. No second wave of infection would occur, unlike the predictions of the Imperial and Washington models.

Nevertheless, that’s not the only possible interpretation of the Oxford results. In a series of tweets, Harvard public health postdoc James Hay has explained that the proportion of the UK population already infected could be anywhere between 0.71% and 56%, according to his calculations using the Oxford model. The higher the percentage infected and therefore immune before the disease began to escalate, the lower the percentage of people still at risk of contracting severe disease, and vice versa.

The Oxford model shares some assumptions with the Imperial and Washington models, but differs slightly in others. For example, it assumes a shorter period during which an infected individual is infectious, and a later date when the first infection occurred. However, as mathematician and infectious disease specialist Jasmina Panovska-Griffiths explains, the two models actually ask different questions. The question asked by the Imperial and Washington groups is: What strategies will flatten the epidemic curve for COVID-19? The Oxford researchers ask the question: Has COVID-19 already spread widely?  

Without the use of any model, Stanford biophysicist and Nobel laureate Michael Levitt has come to essentially the same conclusion as the Oxford team, based simply on an analysis of the available data. Levitt’s analysis focuses on the rate of increase in the daily number of new cases: once this rate slows down, so does the death rate and the end of the outbreak is in sight.

By examining data from 78 of the countries reporting more than 50 new cases of COVID-19 each day, Levitt was able to correctly predict the trajectory of the epidemic in most countries. In China, once the number of newly confirmed infections began to fall, he predicted that the total number of COVID-19 cases would be around 80,000, with about 3,250 deaths – a remarkably accurate forecast, though doubts exist about the reliability of the Chinese numbers. In Italy, where the caseload was still rising, his analysis indicated that the outbreak wasn’t yet under control, as turned out to be tragically true.

Levitt, however, agrees with the need for strong measures to contain the pandemic, as well as earlier detection of the disease through more widespread testing.

Next: Coronavirus Epidemiological Models: (3) How Inadequate Testing Limits the Evidence

Coronavirus Epidemiological Models: (1) What the Models Predict

Amid all the brouhaha over COVID-19 – the biggest respiratory virus threat globally since the 1918 influenza pandemic – confusion reigns over exactly what epidemiological models of the disease are predicting. That’s important as the world begins restricting everyday activities and effectively shutting down national economies, based on model predictions.

In this and subsequent blog posts, I’ll examine some of the models being used to simulate the spread of COVID-19 within a population. As readers will know, I’ve commented at length in this blog on the shortcomings of computer climate models and their failure to accurately predict the magnitude of global warming. 

Epidemiological models, however, are far simpler than climate models and involve far fewer assumptions. The propagation of disease from person to person is much better understood than the vagaries of global climate. A well-designed disease model can help predict the likely course of an epidemic, and can be used to evaluate the most realistic strategies for containing it.

Following the initial coronavirus episode that began in Wuhan, China, various attempts have been made to model the outbreak. One of the most comprehensive studies is a report published last week, by a research team at Imperial College in London, that models the effect of mitigation and suppression control measures on the pandemic spreading in the UK and U.S.

Mitigation focuses on slowing the insidious spread of COVID-19, by taking steps such as requiring home quarantine of infected individuals and their families, and imposing social distancing of the elderly; suppression aims to stop the epidemic in its tracks, by adding more drastic measures such as social distancing of everyone and the closing of nonessential businesses and schools. Both tactics are currently being used not only in the UK and U.S., but also in many other countries – especially in Italy, hit hard by the epidemic.

The model results for the UK are illustrated in the figure below, which shows how the different strategies are expected to affect demand for critical care beds in UK hospitals over the next few months. You can see the much-cited “flattening of the curve,” referring to the bell-shaped curve that portrays the peaking of critical care cases, and related deaths, as the disease progresses. The Imperial College model assumes that 50% of those in critical care will die, based on expert clinical opinion. In the U.S., the epidemic is predicted to be more widespread than in the UK and to peak slightly later.

What set alarm bells ringing was the model’s conclusion that, without any intervention at all, approximately 0.5 million people would die from COVID-19 in the UK and 2.2 million in the more populous U.S. But these numbers could be halved (to 250,000 and 1.1-1.2 million deaths, respectively) if all the proposed mitigation and suppression measures are put into effect, say the researchers.

Nevertheless, the question then arises of how long such interventions can or should be maintained. The blue shading in the figure above shows the 3-month period during which the interventions are assumed to be enforced. But because there is no cure for the disease at present, it’s possible that a second wave of infection will occur once interventions are lifted. This is depicted in the next figure, assuming a somewhat longer 5-month period of initial intervention.

The advantage of such a delayed peaking of the disease’s impact would be a lessening of pressure on an overloaded healthcare system, allowing more time to build up necessary supplies of equipment and reducing critical care demand – in turn reducing overall mortality. In addition, stretching out the timeline for a sufficiently long time could help bolster herd immunity. Herd immunity from an infectious disease results when enough people become immune to the disease through either recovery or vaccination, both of which reduce disease transmission. A vaccine, however, probably won’t be available until 2021, even with the currently accelerated pace of development.

Whether the assumptions behind the Imperial College model are accurate is an issue we’ll look at in a later post. The model is highly granular, reaching down to the level of the individual and based on high-resolution population data, including census data, data from school districts, and data on the distribution of workplace size and commuting distance. Contacts between people are examined within a household, at school, at work and in social settings.

The dilemma posed by the model’s predictions is obvious. It’s necessary to balance minimizing the death rate from COVID-19 with the social and economic disruption caused by the various interventions, and with the likely period over which the interventions can be maintained.

Next: Coronavirus Epidemiological Models: (2) How Completely Different the Models Can Be