Effect Measure

Just as we are preparing to wind up our marathon series of posts on a mathematical model of antiviral resistance, a new paper has appeared in the Journal of the American Medical Association (JAMA) with data on antiviral resistance from Japan, the country that uses more Tamiflu and Relenza (the two available neuraminiase inhibitor antiviral drugs) than any other. It turns out the accompanying Editorial in JAMA specifically mentions the modeling paper and its results as a key to understanding the significance of this work. So all our labor has not been in vain. Here’s more.

Our eye was caught by a news story, “Flu Strain Resistant to Drugs” that appeared on an ABC-TV news outlet:

In the new study, appearing in Wednesday’s Journal of the American Medical Association, researchers collected virus samples from patients at four community hospitals in Japan.

In one part of the study, they took samples from 74 children before and after they were treated with Tamiflu.

They found drug-resistant virus in one of the children after treatment, indicating the resistance had emerged during treatment.

They also collected samples from 422 untreated children and adults with flu and found drug-resistant virus in seven of those patients.

The rate of resistance to this family of drugs, less than 2 percent, was lower than had been found previously in type A influenza.

Rates of drug-resistant type A virus have been reported as high as 18 percent. (ABC6)

This is a pretty good account of the paper, although the headline is inaccurate. Influenza B is a type of influenza, not a strain. Thus it is different than any influenza A, which is another type of influenza virus. Influenza A also has numerous subtypes (H3N2 and H5N1, for example) while influenza B doesn’t have subtypes. Each influenza A subtype and the influenza B type can have many variations or strains. The important thing is that influenza B is not just another strain of the influenza A virus. Influenza B is a completely different type, not just a different subtype or strain. Flu B causes (usually) milder disease and is not normally a cause of pandemics (although some evidence is emerging that there have been pandemics of influenza B we didn’t recognize).

So influenza A and B are different. But there are some important similarities. Both have a neuraminidase protein on their surfaces, and it is this protein, an enzyme that severs the connection between the virus and its cellular docking site, that is the target of the antivirals Tamiflu and Relenza. The drugs interfere with this releasing factor and the virus remains stuck to the surface of the infected cell that produced it. If you inhibit the ability of neuraminidase to sever the connection, you also interfere with the ability of the virus to find a new cell to infect. We know very little about mutations in the virus that produce a neuraminidase protein immune to the action of the drugs. We are unsure how often they arise, how easily they transmit and what their clinical relevance is. The new paper adds some important information but still leaves quite a lot to learn. We know the least about resistant mutants of influenza B, and that’s what this paper is about.

The researchers not only found resistant mutants but they sequenced the NA protein so we know what the mutants are. But we are still unsure about the clinical relevance. If we know the virus is resistant to the drug, how can we be unsure of the clinical relevance? The answer is that we know that viruses were both circulating in the community and on one occasion appeared during treatment with Tamiflu, that showed partial resistance to one or both of Tamiflu and Relenza. The resistance was not measured in clinical terms, however, but in the test tube by seeing whether the viral protein maintained its ability to sever the connection with the stuff of the docking site in the presence of the drug. As it turned out, there was no difference in duration of symptoms, amount of viral shedding or clinical outcome between patients infected with the sensitive or resistant virus after antiviral therapy. The numbers are small so it may turn out that some differences would be detectable with more patients, but so far, the clinical outcome seems unaffected by emergence of resistance. A pessimistic explanation would be that antiviral therapy doesn’t work at all, even for sensitive virus, but other work seems to contradict this.

There is a great deal of fascinating material in this paper I have not discussed. But I will close with a quote from the accompanying Editorial by Moscona and McKimm-Breschkin, two noted authorities on antiviral resistance in influenza:

Part of the complacency about neuraminidase inhibitor-resistant influenza has been fueled by experiments in vitro and in animal models that have generally found neuraminidase inhibitor-resistant influenza viruses to be compromised in infectivity and transmissibility. This has led to the belief that significant transmission is unlikely to occur among humans. A 2003 model of the spread of resistant viruses in a pandemic concluded that community spread of such variants would be negligible because of decreases in biological fitness and transmissibility of drug-resistant viruses.[cite omitted] However, a recent report predicted that, while use of antiviral drugs could significantly slow or stop transmission in the absence of drug resistance, the emergence of resistance could seriously diminish this benefit,[citation to the paper of Lipsitch et al.] depending on the cost of biological fitness: if there were a modest biological fitness cost and transmissibility were high, the effectiveness of antiviral use would plummet. Even if resistant strains emerge de novo at extremely low frequencies in individuals receiving antiviral drugs for treatment or prophylaxis, these strains may well make a significant contribution in an epidemic or pandemic setting. These concepts, together with the findings reported by Hatakeyama et al [the JAMA paper] suggesting that partially-resistant influenza B viruses are circulating, mean it is no longer possible to be confident that resistant strains will have little effect on epidemic or pandemic influenza. (Moscona and McKimm-Breshkin, JAMA, subscription required, alas)

I hope you recognize that the linchpin of the argument these authorities are making is the very Lipsitch et al. paper we have been examining in our exhaustive series (to be concluded tomorrow).


  1. #1 Marissa
    April 4, 2007

    Not seeing a clinical difference bothers me a lot and certainly does imply that antiviral treatment is not very effective. If I recall, however, from the vast majority of Tamiflu studies, the clinical symptoms do not greatly diminish. For example, duration is reduced by average by only one day, so with small numbers, and poor clinical description, the difference would pass unnoticed. It should make a big difference though with H5N1. To me, what’s lacking is a definitive H5N1 Tamiflu RCT. Of course, that would be highly unethical, wouldn’t it (never mind the design problems)?

  2. #2 revere
    April 4, 2007

    Marissa: I agree with all your points. Tamiflu is now the accepted therapy so an RCT against a placebo is not possible, but observational studies still are, but that would require what we don’t want: a lot more cases. The other possibility is another antiviral working by a different mechanism that has shown some clinical efficacy. Then we could do an RCT against the NIs.

  3. #3 Melanie
    April 4, 2007

    This adds to the “I don’t know” factor.

  4. #4 birdflunewsflash
    April 4, 2007

    Hope the following is of interest.

    A new study, carried out by mathematical modellers from Imperial College London and the London School of Hygiene and Tropical Medicine, is published today.

    The study focuses on how H7N3, H7N7 and H7N1 strains of the virus were transmitted between different farms in three recent outbreaks.

    It is hoped that the study could help analyse the success of different control measures in the event of future outbreaks.

    Here is a link to the article from the Imperial College London:


  5. #5 marquer
    April 6, 2007

    …we know that viruses were both circulating in the community and on one occasion appeared during treatment with Tamiflu, that showed partial resistance to one or both of Tamiflu and Relenza.

    I suggested quietly in a comment a few posts back that because Tamiflu and Relenza are going after the same structural target by differing chemical means, there was a risk of Tamiflu overuse potentially generating resistance to both antivirals.

    And several people jumped straight down my throat with both feet for having dared to suggest such a thing.

  6. #6 revere
    April 7, 2007

    marquer: I hope I wasn’t one who jumped down your throat, but in response this comment, it isn’t clear this is from “overuse.” Probably most of it was administered to treat flu, which is the prescribed use.

  7. #7 CPG
    April 8, 2007


    Notice you didn’t repond at the time because, I add to say, you were plainly wrong. Now when it suits you you do. Hindsight is a wonderful thing. Pretty weak.

    Re that study you are referring too. It would pay if you read it fully and understand what you are saying. Can you answer me this Marquer. Of all the people that developed resistance ( 422 ) in the study on Influenza B. How many were resistant to Relenza.

    PS remember also the comment from above by Revere if it suits you. Look forward to your reply !!!!

    ”The important thing is that influenza B is not just another strain of the influenza A virus. Influenza B is a completely different type, not just a different subtype or strain. Flu B causes (usually) milder disease and is not normally a cause of pandemics”