7 March 2018
The trouble with predicting flu pandemics
On the morning of 25 January 1919, Sydney residents woke to read in the newspaper of a suspected case of influenza. A recently returned serviceman had been admitted to the military hospital in Randwick.
Three days later, four cases of influenza had been confirmed, and hospital staff began to fall ill. The deadly H1N1 strain, which would cause 40 million deaths worldwide, had arrived.
“All theatres, picture shows, and other places of indoor public resort … will be closed as from today and until further notice,” The Sydney Morning Herald reported. “The schools will not reopen today.”
Within weeks, what was called the Spanish flu had spread from Sydney into country towns. Eventually more than 6000 people died of flu across the state of NSW alone.
“Without doubt it was the worst and most destructive pandemic in recorded history,” Dr Alan Hampson, a virologist and former chair of the Influenza Specialist Group, says.
Unfortunately for public health, pandemic flu is about as difficult to predict as a volcanic eruption. Unlike seasonal flu, which breaks out like clockwork every winter, pandemics can occur at any time of year.
Pandemics are both rare and seemingly random; we can count on one hand the number of global outbreaks in the last hundred years.
There have been 15 flu pandemics since the first recorded event recorded in 1510, with the greatest gap between pandemics being 152 years.
Seasonal flu is caused by strains of influenza already circulating in the human population, which at present include type A H1N1 and type A H3N2.
The virus evades our immune system by wearing a different surface glycoprotein disguise, made possible through antigenic drift or gradual, accumulating mutations in the viral genome.
Pandemic flu is a very different beast. It can emerge when novel influenza strains spill over from the viral reservoirs in pigs or birds.
Or it can be forged anew when two or more influenza viruses release their genetic material into a single cell and the RNA recombines to form different viral progeny.
The human immune system is generally naïve to these novel influenza strains, and thus the infection and related complications can be more severe during a pandemic.
While seasonal flu epidemics are confined to local spikes in infection, pandemics, by definition, spread globally. To have pandemic potential, an influenza virus must transmit rapidly between human hosts, although it need not always be as severe as the 1918 strain.
The disease moves at the same rate of human travel, meaning that a pandemic that would have taken six months to circle the globe in the pre-industrial era now takes less than six weeks – and that releasing antiviral stockpiles and developing new vaccines is a race against time.
Both type A and B influenza cause seasonal flu, but only type A causes pandemics.
The pandemics in the past century have involved flu subtypes H1N1, H2N2, and H3N2. (These type A influenza strains are labelled on the basis of antigenic properties of the two surface glycoproteins, hemagglutinin (H) and neuraminidase (N).)
Ideally, we would spot a dangerous strain of influenza long before it took root in the human population and contain the disease.
But that is much easier said than done. There are simply too many animal populations carrying influenza and too many opportunities for human-animal interaction to catch an emerging pandemic strain before it takes off in humans.
“It’s like looking in a haystack but there are many, many haystacks and you don’t even know what the thing you are looking for looks like,” says Dr Richard Webby, the director of the WHO Collaborating Centre for Studies on the Ecology of Influenza in Animals and Birds.
“We would like to think we’re a bit smarter than that, but I don’t think we really are,” he says.
A great deal of work is being done to improve our predictive ability for pandemics, particularly by the CDC and WHO, however.
Looking for patterns
Much of this research involves looking for patterns in past pandemics that could help predict the next disaster.
A few of these projects were presented at the Immunisation Coalition Influenza Specialist Group annual scientific meeting in February.
One of the more intriguing elements of pandemic flu is that it often preys on healthy adults in the prime of their life.
Seasonal flu, by contrast, is most severe in vulnerable groups, namely the immunocompromised, chronically ill, flu naïve, pregnant women, Indigenous people, and the elderly.
The “enigma of 1918” was the W-shaped curve, where deaths peaked among those in their late 20s and dropped in the elderly and in children, Professor G. Dennis Shanks says.
Professor Shanks, who is the director of the Australian Defence Force Malaria and Infectious Disease Institute in Queensland, speculates that the W-shaped curve is due to a cohort effect.
Older populations may have built up immunity through protective exposure to a previous influenza pandemic, but the younger generation lack this immunological memory, Professor Shanks says.
“At the risk of having immunologists throw things at me, I think it wasn’t what [those in their late 20s] had, it was what they were missing,” he says.
“[Mortality] actually peaked at age 28. I can’t believe that it’s accidental that the previous pandemic just happened to have occurred just 28 years previously.”
A similar pattern emerged in the 2009 swine flu pandemic. The elderly were spared and the average age of death was 53 in Australia.
“That’s 30 years younger than the average age of death for seasonal flu,” says Professor Raina MacIntyre, an infectious disease epidemiologist at University of NSW.
Acquired immunity through prior exposure was a crucial factor in determining who lived and who died during a pandemic.
Data from US, UK and ANZAC military sources showed that the single biggest risk factor for death from flu was how long soldiers had been in the army, Professor Shanks says.
“It was really very stark,” he says. “If you were a new recruit in any army, you were in deep trouble.”
The fate of medical staff on the frontlines was also determined by their amount of prior exposure to circulating pathogens. “Lots of doctors and nurses got ill,” says Professor Shanks. “Very few of them actually died unless they were new.”
Isolation from disease generally spelled bad news, unless you could establish an impervious quarantine.
Army data showed that large ships in the British, US and French navies (even those with companies of nearly 1,000 men) had either zero or one death from influenza.
But many of the ships on an isolated patrol in the Southern Hemisphere had enormous death rates.
Professor Geoffrey Rice, a historian at the University of Canterbury in New Zealand, saw a similar pattern on islands in the South Pacific, including Fiji, Tahiti and Tonga.
Island populations that had been largely isolated against influenza outbreaks saw some of the highest death rates in the world.
After the Spanish flu was brought to Western Samoa in November 1918 on the so-called “death ship”, the SS Talune, one-fifth of the population perished, with a total of 8,500 deaths.
“By contrast, just a few miles across the water, America Samoa imposed a very strict maritime quarantine,” Professor Rice says.
“They wouldn’t even let rowing boats or canoes from other islands ashore and they had no deaths from influenza.”
Australia managed to delay the arrival of pandemic influenza until January 1919. This was three months after the pandemic hit New Zealand, which lacked Australia’s more rigid border controls.
Tasmania had one of the lowest death rates in the world (0.81 deaths per 1000 population). The flu did not hit Tasmania until August 1919, due to a strict maritime quarantine imposed after NSW declared a pandemic in January of that year.
Pandemics are distinguished by their wave-like patterns. The past four pandemics had two or more waves of infection, separated by months. In 1918, the second wave was far more lethal than the first or the third.
The 2009 pandemic emerged from Mexico, but Russia and China have been the origins of most of history’s pandemics.
Another feature of pandemics is the presence of secondary bacterial infections. In 1918, most deaths occurred in the second week of infection, indicating secondary pneumonia.
The general belief was that bacterial co-infection wasn’t important in 2009, says Professor MacIntyre. But a review of fatal cases of influenza found that 23% had a bacterial infection.
Rates of secondary bacterial infection were higher in adults, in ICU patients and those with a fatal outcome. The most common co-infection was streptococcus pneumoniae.
While the rates of bacterial colonisation in the general adult population was quite low, a recent study by Professor MacIntyre found that almost 90% of 188 asymptomatic healthcare workers were carriers for bacteria, including streptococcus pneumoniae.
“To me, it’s a no-brainer that pneumococcal vaccination should be a standard part of pandemic planning, but it is really overlooked a lot of the time,” she says.
The genetic makeup of the 1918 flu was thought to be lost to history, but a decades-long effort has succeeded in resurrecting the virus.
In 1951, US virologist Johan Hultin recovered tissue samples from victims buried in the Alaskan permafrost, but he could not grow the virus in the laboratory.
Almost half a century later, in 1997, Hultin decided to return to Alaska and try again.
This time, he gave the samples to US molecular pathologist Jeffery Taubenberger who used them to piece together the entire genetic sequence and rebuild the virus from scratch.
The reconstruction of the 1918 virus, which Hultin described as “the most lethal organism in the history of man”, proved invaluable for assessing pandemic risk.
“Having a 1918-like virus in the laboratory can really help provide a reference and anchor for analysing viruses that are currently circulating today,” says Jessica Belser, a microbiologist at the CDC.
Using the 1918 virus and other pandemic strains, scientists can determine what molecular features make these strains so lethal and use that knowledge to predict whether an emerging strain has pandemic potential.
“With this diversity of viruses jumping from the animal reservoir to humans there is a need for assessing the pandemic risk of these viruses,” says Belser.
The genes contributing to the replication and virulence of the 1918 flu virus were discovered in 2008 by a team led by CDC scientist Terrence Tumpey. By switching genes from the 1918 virus with those from a more mild strain of seasonal flu, the team found that the hemagglutinin (H), neuraminidase (N), and polymerase subunit PB1 genes were responsible for the efficient replication of the virus in human airway cells.
In 2009, a team led by another CDC scientist, Neal Van Hoeven, found that the 1918 PB2 protein and the 1918 hemagglutinin gene were necessary for efficient transmission using the ferret model. These findings could be extended to other flu subtypes, including the H5N1 avian flu, which pose an immediate pandemic threat, the researchers say.
The proteins in modern avian flu viruses only differ from the 1918 flu strain by a few amino acids, according to research published in 2014 by the University of Wisconsin. The researchers found one avian influenza PB2 protein that differed from 1918 PB2 by only eight amino acids. This suggests that 1918-like pandemic viruses may emerge as pandemics in the future, the researchers say.
The 2009 flu virus initially appeared to have no similarities with other pandemic strains. However, some patients with severe flu (around 1% of cases) were found to carry a virus with a particular marker that was also found in the 1918 virus, the D222G mutation in the hemagglutinin.
This side-by-side comparison of different pandemic flu viruses gives scientists hints about which flu strains are going to be severe and spread rapidly among humans.
Humans are surrounded by a sea of constantly evolving zoonotic influenza viruses and the spill over of animal viruses into humans is inevitable.
“Evolution has a great ability to do things if given long enough,” says Professor Edward Holmes, an evolutionary biologist and virologist at the University of Sydney.
But it’s hard to know when or where dangerous influenza will emerge, says Dr Webby.
“Which of these viruses present in the animal population should we be worried about? With limited resources what should we put our money on?” he asks.
Animal influenza is constantly moving into human hosts. The pandemic of 2009 was named “swine flu” because the influenza strain was most recently seen circulating in pigs before it jumped across to humans.
The CDC believes swine flu was created through genetic reassortment when North American and Eurasian pig herds came into contact through international trade.
Similarly, the Asian flu of 1957 was caused by a H2N2 virus that was created through reassortment of avian and human influenza.
On other occasions, animal influenza has breached the species gap and infected humans without causing a pandemic.
Several hundred human cases of H5N1 have been reported since the virus was identified in Chinese poultry in 1997.
Since 2013, there have been around 1500 human cases of the Chinese avian influenza strain H7N9. Avian influenza strains H7N7 and H9N2, and swine subtypes H1 and H3 have also caused sporadic human infections.
Sometimes the response is swift. In 1997 in Hong Kong around 1.5 million birds were culled in three days to contain the influenza virus. In 2003, nearly 30 million birds were killed in one week in the Netherlands to stop the spread of A (H7N7).
“Rapid action in both of these situations potentially averted an influenza pandemic in humans,” says the European Centre for Disease Prevention and Control.
“However, as a reality check it has to be remembered that there are many parts of the world where the surveillance … is weak or non-existent. In those countries, mass culling cannot be done.”
The challenge of predicting the evolutionary trends in influenza is magnified by the diversity of influenza in animals. There are 18 H subtypes and 11 N subtypes of flu, with a theoretical total of 198 strain variations.
Bird populations alone harbour a large amount of influenza, with 35% birds infected in some regions. Ducks, geese and other wild waterfowl are the top reservoir for many influenza A viruses.
There is evidence that we are just scratching the surface in our understanding of influenza, says Professor Holmes. It was only recently that researchers looked for the flu virus across the animal kingdom.
“It turns out there are flu-like viruses in worms, in mosquitoes,” says Professor Holmes. “We’ve got flu in fish, in toads and in hagfish, jawless vertebrates. Basically, the flu virus has been around for the entire history of animals.”
The patterns of transmission between animals and humans also bewilder scientists.
“One of the great puzzles is that dogs certainly carry H3N2 virus but there is no evidence of any dogs infecting humans,” says Dr Webby.
Similarly, horses carry H3N8 and yet there is no evidence of cross-species transmission into humans. “[This] makes me believe that our view of the barriers to infection and the receptor binding interactions are oversimplified,” he says.
Influenza subtypes also differ in their evolutionary pattern in ways that we do not yet understand, says Professor Holmes.
For instance, the phylogenetic tree of H3N2 shows a sharp boom and bust cycle.
“H3N2 gets diverse and crashes, gets diverse and crashes. And that’s flu seasonality,” he says. However, the phylogenetic tree of the 2009 pandemic flu is completely flat, with no boom and bust cycle.
“There are different dynamics … but we don’t really know why,” he says. “It varies but there is no consistent accepted explanation.”
These obscurities make Professor Holmes wary of any prediction framework for pandemic flu.
“I’m very anti-prediction,” he says. “I don’t really trust the risk-assessment tools. It worries me that we think in a static way.”
Influenza, it seems, doesn’t give up its secrets easily.
The images used are © State of New South Wales through the State Records Authority of NSW 2016