|Previous diaries have covered current plans, distribution, equity, insufficient stockpile, and related issues. I now move on to two central questions.
- How effective is tamiflu for reduction of mortality should the current H5N1 virus cause a pandemic?
- What would be the effect of widespread use, in the order of tens of millions of doses, on the evolution of the virus, specifically emergence of resistant strains?
Two recently published papers give some crucial and unfortunately worrying answers to these questions, further throwing into doubt the effectiveness of mass tamiflu usage in mitigating the effects of a pandemic.
Tamiflu belongs to the group of drugs called neuraminidase (NA) inhibitors.
First some background science: all influenza viruses, in order to cause disease, have to be able to enter the host's cells. This is facilitated by the binding of the virus to receptors on the cell surface by the HA (haemagglutinin) molecule. The virus then undergoes replication inside the cell. In the case of H5N1, for reasons that are not clear, this replication is hundreds often thousands of times more efficient than the regular seasonal flu viruses. In order for these newly generated virus particles to spread to other cells, they have to first leave the original cell. This again involves the receptors, but this time, the release of the virus from binding to the receptors is the work of the NA molecule.
NA inhibitors work by blocking this release. What this means is that essentially, tamiflu does not start blocking the virus until it has already gone through at least one round of replication.
ie No matter how early you take the drug, you are always going to be behind the curve which is particularly problematic since one reason why H5N1 is so lethal is its high replication rate.
Animal studies with H5N1 suggest that tamiflu is effective if given early enough, in high enough doses, and for longer than the normal duration of treatment. As a result of the very high fatality even in treated cases, several governments in Asia with current human cases have started using double the standard dose for H5N1 patients, in an effort to improve survival. The longer period of virus replication (see below) also means a longer course of treatment may be needed. Hence the use of standard dose, 75mg twice daily for 5 days only, as proposed by the government may carry a high risk of treatment failures.
A recent authoritative review Avian Influenza Virus (H5N1): a Threat to Human Health by Malik Peiris, de Jong, and Yi Guan, states
The limited clinical experience does not suggest a substantial impact of antiviral treatment on the mortality of human H5N1 influenza virus in the field setting.
The reasons given are:
- Late presentation of patients for treatment (median 4 days)
- Part of the disease process is caused by host immune dysfunction which cannot be mitigated by antivirals.
- Standard seasonal flu doses not sufficient for H5 cases
- Oral administration means absorption is compromised in patients with diarrhea (>70%)
Having said that, there are still encouraging results, eg from de Jong, 2005, where
even when oseltamivir treatment is started later in the course of infection, viral clearance can occur in some patients, and this is associated with a favorable clinical outcome (41). Thus, in contrast to human influenza virus, where treatment after the first 48 h provides little clinical benefit, there may be a wider therapeutic window of clinical benefit in H5N1 disease (and also in a pandemic situation).
Which makes the UK government's proposal to deny antivirals to those who have been sick for 48 hours even more unjustifiable, quite apart from the implementation problems discussed before.
Overall, it would seem that tamiflu is likely to be somewhat effective, if taken early enough, but there are serious doubts that current dosages are sufficient to provide enough benefit to reduce mortality in a pandemic. Given a fixed stockpile, doubling the dose would halve the number we can treat, and, of course, giving a longer course would reduce that even further.
This is the current status, ie before millions and millions of doses are used all over the world, as is likely to happen in a pandemic.
Emergence of resistance:
Antiviral drugs, like antibiotics, work only if the virus that you are targeting is still sensitive to the drug. In general, resistance is more likely to arise with increased usage and with partially treated infections, ie doses that are either too low or given for too short a time, or both. Although influenza viruses generally have a lower rate of resistance to NA inhibitors when compared to the other major group of influenza drugs, the adamantanes, there are signs that widespread use in seasonal flu eg in Japan, can cause substantial incidence of resistant strains to arise especially in children (up to 18% in one study). Since current proposal for antiviral use in a pandemic is still based on the seasonal flu dose (see above), such widespread use of suboptimal dosing would at least in theory substantially increase the risk of emergence of resistant strains.
Cases of H5N1 infections with tamiflu-resistant strains have been reported. The first one was in a Vietnamese girl who was on the prophylactic dose, and she subsequently recovered when given the standard dose, so it was not clear that there was actual clinical resistance. More disturbing are later cases reported by de Jong, when resistant strains were found in 2 out of 7 patients (a very high incidence, albeit in a very small sample) undergoing treatment, when falling viral titre rose again with the emergence of the resistant strain, and were associated with treatment failure and death.(see chart)
Concern was also raised when earlier this year 2 patients in a family cluster in Egypt who died also had resistant strains isolated, raising the possibility of either occurrence in the wild or h2h transmission of the resistant strain.
Malik's paper raises a very important point on antiviral resistance. Commenting on the much higher incidence of resistance in children, the suggestion is that in primary influenza infections ie in those with no prior immunity, such as young children for seasonal flu and all patients with H5N1 infections, the mechanisms are different from our normal experience of seasonal flu in adults, and the rate of virus replication and mutations are extremely high, giving rise to higher chance of resistance emerging. If this reasoning is correct, then we may expect far higher incidence of drug resistance in a H5N1 pandemic compared to seasonal flu
ie the efficacy of tamiflu in seasonal flu infections cannot be extrapolated to a pandemic with H5N1 or indeed any other completely novel virus for which the population have zero immunity!
So far, the mutations seen with the resistant strains appear to be associated with moderate `fitness cost', ie the selection of the resistant virus is at the expense of a reduction in transmissibility. However, we know that, like antibiotics, widespread use of antivirals does encourage the emergence of resistance. Tamiflu, despite being so famous by now, is still not very widely used (except in Japan), so the full implications of extremely widespread use eg in the order of tens or hundreds or million doses worldwide in a pandemic, are not clear.
A recent paper in PLOS Medicine by Lipsitch et al Antiviral Resistance and the Control of Pandemic Influenza uses mathematical modeling to attempt to find some answers to this and related questions. This was very elegantly explained in a 16-part series by revere ;-). Here I will only draw upon some of their conclusions and discuss their implications for pandemic policy.
Even though, so far, the resistant viruses isolated appear to have high to moderate fitness costs, in principle, such fitness trade-off must be seen as a continuum, ie at the 2 ends of the range, there would be those viruses with such high fitness-costs that they do not survive at all (and hence we will not know anything about them!), as well as, at the other end of the range, those with low fitness-costs. Both these would be rare, compared to the more commonly seen ones with moderate fitness-costs, but since the incidence of resistance emergence is proportionate to the number of treatment courses used, at very high rates of use, there would be a small but significant chance of emergence of a resistant strain that can transmit in a comparable way to the original sensitive virus.
The effects of widespread antiviral use would include:
- Emergence of resistant strain
- Blocked or slowed transmission of wild type, favoring selection for transmission of the resistant virus.
- Reduction in AR causes slower rate of build-up of herd immunity before resistant strain appears, resulting in a higher proportion of cases due to the resistant strain.
It's important to remember, though, that the overall epidemic size with antiviral use is still generally smaller than without. However, and this gets really interesting, the researchers found that this effect ie increasing antiviral use reducing epidemic size, is only true at the lower end of usage, and that, beyond a certain point, the advantages are reversed, and the overall AR rises again. This is because at very high rates of use, the build-up of immunity due to infection is much slower, but the blockage of transmission of the sensitive strain is ongoing, so that more people are susceptible to and become infected by any resistant strain that can be transmitted h2h. Since these infections cannot be treated by tamiflu, these patients will transmit the virus to others, resulting in higher attack rates again, only this time the virus would be resistant to treatment!
In summary, even though animal studies suggest that tamiflu may be effective in reducing mortality in H5N1 infections if given early enough, in high enough doses, and for long enough durations, limited clinical experiences have so far been unable to reproduce such success in humans. If in future it is found that efficacy is improved with higher doses and longer duration of treatment, then a fixed 25% stockpile would mean being able to treat a substantially smaller proportion of patients, or treating at suboptimal doses with the increased risk of drug resistance emerging. In addition, the effects of tens of millions of doses being used in UK alone over a short period of time are entirely unknown, although modeling would suggest at least a small but significant risk of resistant strains arising and spreading as part of the epidemic.
As part of an overall pandemic mitigation plan, antivirals are indeed indispensable. However, in the absence of other additional effective mitigation strategies, the overall outcome in terms of mortality and epidemic size are at best uncertain. Should drug resistant viruses appear and cause even a small number of fatal cases, the panic and social instability that can arise in a society conditioned to the idea that we have enough antivirals stockpiled to treat all those who get sick cannot be underestimated.
Even though the addition of other antiviral drug stockpiles may improve the prospects somewhat, and should be undertaken in any case, it would be far more prudent to create and strengthen other non-pharmaceutical interventions in addition to drug treatment. Whether or not antivirals work, the ability to use effective social distancing measures (in a co-ordinated manner under the guidance of government) to reduce the impact of a pandemic would reduce the sense of helplessness that would otherwise pervade society.
In a mass casualty possibly mass fatality situation lasing many weeks and months, self-sufficiency, community resilience, trust, and continuing dialogue may mean the difference between societies that succeed or fail to emerge intact.