Testing Woes

Ayear ago​ I joined half a dozen academics at the University of California in Los Angeles, a group of physicians, geneticists and computational biologists, to try to help address the public health challenge of asymptomatic Covid-19 transmission. The need for testing was such that our small team was soon on Zoom calls with Silicon Valley startups, the California Covid-19 Testing Task Force, Admiral Brett Giroir (the US testing tsar) and the White House task force (two guys sharing a bedsit in San Francisco). We suggested to executives at Apple that we could regularly test all ten thousand employees at their UFO-shaped building near San Jose. Great idea, they replied. We talked to Anne Wojcicki, the CEO of the consumer genetics company 23andMe, who said her company could provide us with saliva collection kits. At one point it seemed as though the entire West Coast genomics community was on the same Slack workspace (I counted more than a thousand users from sixty companies and ninety universities). In the race to conquer Covid-19, barriers between scientists, government and business leaders seemed to make no odds, and – even more surprising – academics at different universities talked to one another. But somehow things didn’t quite work out as we hoped. Apple never took up our offer, we didn’t receive a single saliva tube from 23andMe and the Slack workspace went quiet.

The need to increase the rate of testing for Covid-19 became urgent when we realised that the virus was spreading between asymptomatic people. Until that point, epidemiologists and public health experts agreed that symptom-based case detection, quarantine and contact tracing would bring the virus under control. Sars-CoV-2 was thought to be like Sars-CoV-1: both are coronaviruses and both are animal diseases that crossed over to humans, with clinical features similar to respiratory tract infections. Sars-CoV-1 spread to 29 countries and was brought under control in eight months, with fewer than nine thousand cases and around eight hundred deaths.

I learned from colleagues in China about the disaster unfolding in Wuhan at the end of 2019. In early February 2020, I had a worrying conversation with an Oxford epidemiologist, Moritz Kraemer, who had been talking to doctors at the centre of the outbreak. After a visit to UCLA’s Ronald Reagan hospital in February, Kraemer told me that California hadn’t carried out a single virus test. Samples from the forty-odd suspected cases at UCLA had been sent to a laboratory of the Centres for Disease Control and Prevention in Atlanta. When he raised an eyebrow at this news, he was told not to worry: the hospital had ordered sufficient reagents to carry out two thousand tests. (This wasn’t atypical – only sixteen people a day were being tested in the US in February.)

There was published evidence of transmission occurring between people with no apparent symptoms as early as the end of January. If true, this meant that the strategy that had worked for Sars-CoV-1 wasn’t going to contain Sars-CoV-2. But many were unwilling to accept that infection was possible between asymptomatic carriers. Sweden’s public health agency stated that ‘there is no evidence that people are infectious during the incubation period.’ The French government’s response was even blunter: ‘No symptoms = no risk of being contagious.’ The UK’s Scientific Advisory Group for Emergencies took a similar view: ‘It was very much a hearsay study.’ A report in Science from 3 February, much cited by those who don’t believe in the use of masks, said the information ‘was wrong’.

Undeniable evidence of asymptomatic transmission emerged when the same genetic variant of the virus was found in two individuals in Germany. There is some debate as to whether we can talk of truly asymptomatic cases – symptomless individuals might be pre-symptomatic – but the point was that the carrier wouldn’t have attracted attention as someone who needed testing. ‘Their only encounter was a canteen visit, sitting back to back, when patient 5 turned to patient 4 to borrow the salt shaker from their table.’ The unfortunate passengers on the Diamond Princess cruise ship provided a natural experiment in tracking virus transmission. After the first diagnosis of Covid-19, no one was allowed to disembark. Everyone onboard was tested repeatedly for the virus. Three quarters of infections occurred between people without symptoms. The same was found in a nursing facility in Washington State: more than half the residents with a positive result were asymptomatic. Screening only people with coughs, high temperatures and aches and pains clearly wasn’t going to be sufficient to prevent transmission.

In April, Harvard’s Roadmap to Pandemic Resilience project estimated that testing could replace stay-at-home orders as the main tool for disease control once capacity reached between 2 and 6 per cent of the population a day (five to twenty million people). On 21 April, the Rockefeller Foundation called for three million national Covid-19 tests a week by July and thirty million by October. ‘To beat this virus we need a massive national effort … to get to thirty million and beyond with tests that are easy, fast and cheap. Only then can we keep the economy open and protect our most vulnerable.’

The main testing laboratories in the US were falling behind as demand outstripped supply. To encourage more laboratories to provide tests, the reimbursement rate that insurance companies paid was increased from $51 to $100 (our UCLA testing group estimates that the true cost of the test is between $30 and $40, a point that entrepreneurs in Silicon Valley and elsewhere weren’t slow to notice). The governor of California, Gavin Newsom, issued an executive order suspending certification and licensing requirements for any researcher with relevant skills and an authorised laboratory to run Covid-19 tests.

Geneticists love this sort of challenge. We’ve been processing genome-sized units – all 3.6 billion base pairs of a mammalian genome – for decades now. The itsy-bitsy Sars-CoV genome has a mere thirty thousand base pairs. We have machines that sequence entire animal genomes; we have automated systems that track hundreds of thousands of samples of nucleic acid; we analyse petabytes of data. The technology to get ahead of the virus was already here, we thought. We just needed to deploy it.

By May there were more than fifty new testing methods either published in journals or deposited on the BioRxiv pre-print server, an open access repository for papers in the biological sciences. People came up with some great ideas. A group in San Francisco developed a rapid portable test that takes about thirty minutes to run. Especially cool: they wrote an app that repurposes your mobile phone camera to measure fluorescence, and tells you whether you have Covid-19 or not. I heard a proposal to give everyone a polymerase chain reaction (PCR) machine for self-testing at home. Others embraced instrument-free nucleic acid detection, which replaces the complicated boiling and cooling of a PCR machine with a single reaction at 65°C. Since beeswax melts at the same temperature, the idea was that everyone could be given a lump of beeswax to keep the reaction going. My favourite proposal came from a group at the Institute of Molecular Pathology in Vienna and requires the use of a sous vide (which you can buy on Amazon for $100 or less). The Viennese institute praised the sous vide method as a ‘game-changer for population-wide screening, especially in disadvantaged environments, such as developing countries’ (presumably because of developing countries’ preferential access to sous vide equipment).

The UCLA group’s contribution to novel testing was SwabSeq, a cheap and highly accurate method with the potential to scale up to millions a day. It’s the invention of Sri Kosuri, an associate professor now running a start-up in the Bay Area. The DNA sequencing technology he uses makes it possible to process tens of thousands of samples at a time (and in principle hundreds of thousands, though we haven’t done that yet). By April we were detecting the virus; by May we could detect a single molecule, and, with some loss of sensitivity, process saliva samples. This meant that with a small number of sequencing machines and a staff of a dozen or so we could test most of the city of Los Angeles in a single day.

It didn’t happen. As far as I know, apart from our group, which is currently testing at a rate of around ten thousand a week, and another at the University of Illinois at Urbana-Champaign, which developed a test that works directly on saliva and was deployed across the state, the inventiveness on display at the BioRxiv pre-print server did nothing to ease the country’s testing burden. The National Institute of Health’s Rapid Acceleration of Diagnostics awarded around $140 million to four companies to provide high throughput diagnostics, none of which delivered testing at the scale required. By the end of January this year, a breakdown of tests carried out in California showed that commercial laboratories had delivered 1.7 million of the 2.3 million tests carried out in the state since the start of the pandemic (we’d need to be doing this much testing every day to replace stay-at-home orders for disease control). Half a million tests came from medical centre-affiliated laboratories and around 200,000 from public health laboratories. Almost all the tests used standard methods.

Some universities scaled up existing molecular tests to screen their staff and students. Berkeley’s Innovative Genomics Institute, staffed with volunteer researchers, began providing on-campus virus testing. At UCSF, Joe DeRisi, supported by the Chan-Zuckerberg BioHub, turned his laboratory into a Covid-19 testing centre. The biggest academic contributor in the US has been the Broad Institute, the large genomics research centre for Harvard and MIT, which ramped up testing to almost 100,000 samples a day in November. Its efforts accounted for almost 5 per cent of the 186 million Covid-19 tests carried out in the US by the end of November. Aside from these exceptions, academic genomics hasn’t done much to alter the course of the pandemic.

There have been regulatory hurdles. Before the pandemic, you needed a licence from the US Clinical Laboratory Improvement Amendments programme to set up a testing laboratory. Once the laboratory was approved you had to convince the FDA that your test had the necessary accuracy and sensitivity. These rules were relaxed when the pandemic arrived: by the end of February 2020, the FDA was allowing any suitably qualified laboratory to run unauthorised tests as soon as an application had been submitted. But final California state and federal approval was still required, as well as emergency use authorisation from the FDA. The process was extremely time-consuming. It took us until October to get FDA authorisation and even longer to establish an approved laboratory.

There are different stories about regulation. Last February, the immunologist Helen Chu, a professor at the University of Washington in Seattle, tested swabs sent to her laboratory as part of the Seattle flu study. She found 25 Covid-19 positives and was the first to demonstrate community transmission. This had huge public health implications, and the sooner the Department of Health knew and acted the better. But the University of Washington only had permission to test for flu, not Covid-19, so technically Chu was transgressing. The FDA sent a cease and desist order, preventing any further testing.

Elsewhere, the problem was that there wasn’t enough regulation. By the end of June, the US government had committed $16 billion to controlling the spread of infection. Many of the companies receiving this cash had no experience in sourcing or providing medical supplies. Paul Wexler, an ex-telemarketer, had never worked in medical logistics, but he did have multiple previous accusations of fraud to his name. He set up Fillakit and signed a contract with the Federal Emergency Management Agency to provide more than three million tubes for Covid-19 test kits. Two months later the company collapsed after it emerged that there were problems with the quality of the tubes. One million Chinese-made diagnostic tests ordered by the Trump administration turned out to be ‘contaminated and unusable’. Poor quality testing hasn’t been unique to the US: Slovakia purchased 1.2 million faulty antibody tests from China; authorities in the Czech Republic said that around a third of a batch of 300,000 tests were defective. Spain returned thousands of faulty rapid tests purchased from a Chinese firm. Similar problems afflicted the UK, where contracts worth billions of pounds were awarded to companies with no relevant experience, and to companies with histories of fraud, tax evasion and worse.

The discussion around testing on social media has frequently descended into abuse. Papers were discredited by commentators with no expertise. Fights between scientists broke out in public, for instance over the use of lateral flow devices in the UK. A piece in the BMJ ran with the standfirst: ‘The more certain someone is about Covid-19, the less you should trust them.’ The author, Bristol epidemiologist George Davey Smith, argued that there are ‘many rational people with scientific credentials making assertive public pronouncements on Covid-19 who seem to suggest there can be no legitimate grounds for disagreeing with them’. He had a point. Even those trained in relevant disciplines have behaved strangely: last year a petition was submitted to the European Medicines Agency demanding that clinical trials be halted on the grounds that vaccines could cause infertility in women. One of the petition’s co-authors was Michael Yeadon, a former vice president of Pfizer, where he spent sixteen years as an allergy and respiratory researcher.

Perhaps the real problem is hubris. There have been so many things we thought we knew but didn’t. How many people reassured us Covid-19 would be just like flu? Or insisted that the only viable tests were naso-pharyngeal swabs, preferably administered by a trained clinician? Is that really the only way? After all, if Covid-19 is only detectable by sticking a piece of plastic practically into your brain, how can it be so infectious? We still don’t understand the dynamics of virus transmission. We still don’t know why around 80 per cent of transmissions are caused by just 10 per cent of cases, or why 2 per cent of individuals carry 90 per cent of the virus. If you live with someone diagnosed with Covid-19, the chances are that you won’t be infected (60 to 90 per cent of cohabitees don’t contract the virus). Yet in the right setting, a crowded bar for example, one person can infect scores of others. What makes a superspreader? How do we detect them? And what can we learn from the relatively low death rates in African countries, despite their meagre testing and limited access to hospitals?

That we are still scrambling to answer these questions is deeply worrying, not just because it shows we aren’t ready for the next pandemic. The virus has revealed the depth of our ignorance when it comes to the biology of genomes. I’ve written too many grant applications where I’ve stated confidently that we will be able to determine the function of a gene with a DNA sequence much bigger than that of Sars-CoV-2. If we can’t even work out how Sars-CoV-2 works, what chance do we have with the mammalian genome? Let’s hope none of my grant reviewers reads this.