Covid 19 and the trade-off model of selection

One idea that I haven’t heard mentioned much in the last few weeks is the virulence-transmission trade-off model of selection in parasites.  This describes the evolution of virulence that we should expect in parasites, including viruses.  The model suggests that the benefits (to the virus, not to us!) of greater virulence – more rapid progression of the illness and shedding of viruses in greater numbers – need to be balanced against the disadvantages.  In particular, shedding in more virulent strains is likely to occur for less time, and the host is likely to move around less.  This is partly about timing.  At one extreme, if transmission is very fast compared to the development of the illness, selection becomes a race between the different strains to become infective as quickly as possible after exposure.  This sometimes happens when bird flu invades chicken barn.  The disease spreads very quickly, and every single bird may die within two or three days.  If, on the other hand, the virus has relatively few chances to jump to susceptible new hosts, milder strains will be favored so that shedding can take place for longer.  [See scientific reference 1, below.]  Note that the selection towards reduced virulence is not driven by the death of hosts – that takes too long.  Rather, it’s about how long hosts remain infective, and how many opportunities there are for reaching new hosts during that period.

There’s another way to look at this: the viruses within each host are, in evolutionary terms, competing with each other.  Selection within the host tends to favor more virulent sequences, because they tend to replicate themselves faster.  This is referred to as “short-sighted” selection, because it works well (from the point of view of the competing viruses) in the short term.  In order to survive, however, the virus needs to transmit itself to other hosts.  Here, less aggressive strains are favored, resulting in “far-sighted” selection.  The final result is a trade-off between short-sighted and far-sighted evolution.  That may be why coronavirus has a “proofreading” mechanism, which reduces copying errors and so prevents too much short-sighted evolution.

I created a simple model of CoV2 infection, which can be found here:

This view seems to be correct, because when viruses first appear in a new host – after jumping from one species to another – they are quite often extraordinarily virulent.  Ebola, SARS, myxomatosis in European rabbits, Lassa hemorrhagic fever, bird flu and Covid 19 are all examples of this.  In comparison to these illnesses, most “old” viruses are mild.  It seems to take time for evolution to select strains that are milder, so that an equilibrium can be reached in the new host species.  If all viruses were as virulent as Ebola, multicellular animals such as earthworms, insects, fish and mammals might well be impossible.  Over time, milder strains seem to come to the fore, and we end up with viruses similar to the 100 (largely unrelated) viral strains that cause the common cold.  They are active enough to make us sick, but they don’t usually kill us.

The point about reducing the mobility of hosts might be particularly important in humans.  For example, we are probably less likely to get on an airplane and travel to the far side of the world with a very active strain that makes us sick within a few hours, than with one that brews slowly and allows us to walk around shedding viruses for several days – simply because the incubation period of the milder illness is longer.  The transported, milder, strain can then start a new epidemic in a remote location.  If this process is repeated many times and on different scales, you can imagine that we will end up with intermediate strains – neither very passive nor very virulent.  An equilibrium is reached.  Nowadays, at least in richer societies, we have the luxury of going to bed when we get a fever, rather than being forced to work (which was often the case in the past and is still the case in some poorer communities).  This might also be expected to reduce virulence over time.

According to this view, the extreme virulence of Spanish influenza in 1918 was the result of the selective conditions that existed at that time.  People tended to “soldier on” when they had fevers, because they were determined to contribute to the war effort. An example was reported in the British Medical Journal in 1918 [ref. 2] by two doctors who shared a train compartment from London to York with a sick airman, who said he had influenza.  Two days later both doctors had bouts of influenza, and two days after that, the wife and two children of one of the doctors were also sick with the same illness.  That particular strain of influenza was very unusually virulent and fast-acting, and may have been subject to very unusual selective pressures – resulting in a dramatic increase in virulence and mortality.  Hope-Simpson [3, 4] noted in the 1970s that influenza infections in institutions tend to be unusually severe.  This puzzled him, but it may be that when residents live and sleep close together, selection favors virulent strains that develop fast, and are transmitted in spite of causing severe illness.  Examples include soldiers in barracks, refugees in crowded camps, children at boarding schools, and seniors in old people’s homes.  No matter how sick the patients become, the virus particles that they shed may be able to reach other nearby residents.  This is rather like the chicken barn with bird flu, where every host becomes infected.  We should also worry about hospitals, which can become the human equivalent of chicken barns.  Hospital epidemics can easily become very dangerous [1].  In the 1918 Spanish flu epidemic, it was reported that patients treated in an “open-air” (tented) hospital had a much higher survival rate than those in conventional hospitals [5].

There are other considerations here.  Although the evidence has been ignored by most virologists, for many decades scientists have noted that many “wild” strains (as opposed to most laboratory-bred strains) of respiratory viruses are temperature-sensitive [e.g. refs 6-8; coronaviruses were first isolated at 33°C, ref. 9.  For many other references see my paper, ref. 10.].  That is to say, they are more active at lower temperatures (typically 33 – 35°C) than at body temperature (around 37°C).  This makes sense because they live in the nose and throat, which are some of the coldest parts of the body.  The virus needs to stay out of the heart and lungs so that the host will continue to move around in order to maximize transmission, as discussed above.  Temperature-sensitivity is a potentially a very good way to make sure that this happens.  In fact there is one good bit of evidence that almost all respiratory viruses use temperature in this way: the simple observation that almost all respiratory viruses are more common in winter than summer.  This makes sense, because if they are more active at low temperatures, then as ambient temperature falls in the autumn, epidemics become more likely.  Conversely, in spring, when the temperature rises, all temperature-sensitive viruses will become less active, and cold epidemics rarer.  Note that respiratory viruses have a high mutation rate, and we can expect them to adjust their virulence quickly to suit their local climate and environment.  A 2011 scientific review acknowledged that there was no satisfactory scientific explanation for the seasonality of influenza [11].  After that article appeared, I published a review of seasonality in respiratory viruses in Medical Hypotheses, which focused on the natural temperature-sensitivity of respiratory viruses [10].  My blog gives a more informal description for the layperson [12].  Seasonality is clearly important, because we need to find the degree of temperature-sensitivity of Covid 19, so that we can predict the course of the epidemic when the weather warms up in the summer – and, even more important, when the temperature drops in the autumn.

If some strains of SARS-CoV-2 develop greater temperature-sensitivity several things may happen: (i) they may be more likely to colonize the nose and throat, because these are relatively cold parts of the body, rather than the lungs.  This can increase sneezing and runny noses, increasing transmission; (ii) they may cause a mild infection that the host doesn’t notice, allowing him or her to go to work, or get on a train or plane while already shedding virus; and (iii) they may be less likely to give rise to dangerous “systemic” infections where the virus spreads into the internal organs of the body.

Note for the technically-minded: a lot of scientific attention has focused in the last few weeks on the sequences of the proteins that the virus makes.  However, like influenza, coronaviruses have untranslated regions of RNA with secondary structure that is conserved [13].  This means that it must be doing something.  Secondary structure is inherently temperature-sensitive, and many organisms (both microorganisms and higher organisms) make use of ‘‘RNA thermometers” [14].  These are RNA segments that respond to temperature changes with three-dimensional conformational changes that alter gene expression.  Because many individual bases contribute to each conserved RNA configuration, the virus has the potential to fine-tune its temperature-sensitivity by changing bases in these regions.  These regions might be good places to look for mutations that give rise to milder strains.

These ideas have practical implications for dealing with the current epidemic.  They suggest that milder strains of CoV-2 will arise spontaneously and, in principle, their selection can be encouraged.  In the next few weeks I expect to hear a lot more in the media and from scientists about less  virulent CoV-2 strains.  For example, the German strain seems to be less virulent – although of course this may just be a reflection of the way infections are counted.  One conclusion is that it may be helpful to ask people not to self-isolate for very long periods after exposure to Covid.  Illnesses that arise after say 7 days may be less severe (on average) than those that appear in the first two or three days, so if we self-isolate for 7 days rather than 14, such strains may emerge.  (If we do get a severe form of the disease we should remain out of circulation until it’s really clear that we’re no longer infectious.)  If mild strains can spread, the population may develop protective herd immunity more quickly, which would result in far fewer deaths.  At the other extreme, we should take great care to prevent strains that arise in institutions such as old people’s homes from escaping into the general population, because these strains may acquire greater virulence as they spread in those settings.  Confining large numbers of patients in beds that are physically close together may be a recipe for generating more virulent strains.  Hospital epidemics are a particular worry, and there are reports that they have already contributed to mortality in Italy.  The policy of China and South Korea of preventing patients with Covid symptoms from returning home, and instead confining them in large sanatoriums may increase the death toll in patients and exacerbate the crisis in the community if virulent strains escape.

All this needs to be taken into account in designing measures to reduce mortality in the current epidemic.  In the long-run it opens up new opportunities to understand viral respiratory illnesses at a fundamental level.

Patrick Shaw Stewart, Sunday 15 March

Scientific references –

[1] Frank, Steven A. “Models of parasite virulence.” The Quarterly review of biology 71.1 (1996): 37-78.

[2] Macdonald, Peter, and J. C. Lyth. “Incubation period of influenza.” British medical journal 2.3018 (1918): 488.

[3] Hope-Simpson, Robert Edgar. “Epidemic mechanisms of type A influenza.” Epidemiology & Infection 83.1 (1979): 11-26.

[4] Hope-Simpson, R. E. “Age and secular distributions of virus-proven influenza patients in successive epidemics 1961–1976 in Cirencester: epidemiological significance discussed.” Epidemiology & Infection 92.3 (1984): 303-336.

[5] Hobday, Richard A., and John W. Cason. “The open-air treatment of pandemic influenza.” American journal of public health 99.S2 (2009): S236-S242.

[6] Richman, Douglas D., and Brian R. Murphy. “The association of the temperature-sensitive phenotype with viral attenuation in animals and humans: implications for the development and use of live virus vaccines.” Reviews of infectious diseases 1.3 (1979): 413-433.

[7] Stern, H., and K. C. Tippett. “Primary Isolation of Influenza Viruses at 33° C.” Lancet (1963): 1301-2.

[8] Kung, H. C., et al. “Influenza in China in 1977: recurrence of influenzavirus A subtype H1N1.” Bulletin of the World Health Organization 56.6 (1978): 913.

[9] Tyrrell, D. A. J., and Parsons, R. “Some virus isolations from common colds. III. Cytopathic effects in tissue cultures.” Lancet (1960): 239-42.

[10] Shaw Stewart, Patrick D. “Seasonality and selective trends in viral acute respiratory tract infections.” Medical hypotheses 86 (2016): 104-119.

[11] Tamerius, James, et al. “Global influenza seasonality: reconciling patterns across temperate and tropical regions.” Environmental health perspectives 119.4 (2011): 439-445.


[13] Yang, Dong, and Julian L. Leibowitz. “The structure and functions of coronavirus genomic 3′ and 5′ ends.” Virus research 206 (2015): 120-133.

[14] Narberhaus, Franz, Torsten Waldminghaus, and Saheli Chowdhury. “RNA thermometers.” FEMS microbiology reviews 30.1 (2006): 3-16.


For information about the natural temperature-sensitivity of most respiratory viruses see

For discussion of the probable seasonality of Covid-19, and whether we can expect it to become rarer in the summer, or reappear in the fall, please see

A simple model of CoV2 infection can be found here:

For practical tips on avoiding respiratory illness see

For detailed scientific information including discussion of the trade-off model, viral dormancy and much else, see my 2016 paper:

Shaw Stewart, PD.  Seasonality and selective trends in viral acute respiratory tract infections. Medical Hypotheses 2016; 86 104–119.