April 10, 2014

expert reaction to Cochrane Review on Tamiflu and Relenza for treatment and prevention of influenza

The Cochrane Collaboration published an evidence review of Tamiflu (the antiviral drug oseltamivir) and Relenza (zanamivir) trials, reporting that Tamiflu shortens symptoms of influenza  but other claims made for the drugs were not well supported by evidence from clinical study reports. These comments accompanied a press briefing and a before the headlines analysis.

Prof Peter Openshaw, Director of the Centre for Respiratory Infection, Imperial College London, said:

“The latest analysis of antiviral trial data by the Cochrane Collaboration adds little to the author’s previous conclusions.

“Most of the antiviral trials they included looked at previously healthy adults infected with seasonal influenza strains. The biggest problem is the continuing failure of the group to recognise that we now face new challenges: since the 2009 pandemic, most of the flu we have seen has been caused by a different virus with unique features, especially its tendency to cause severe illness and death in younger adults and children.

“In planning for a pandemic we have to rely on findings from seasonal flu trial data. We have learnt so much from antiviral treatment during the recent pandemic and it would be dangerous to apply the findings of the latest Cochrane review of antiviral trials for seasonal flu to the current situation or to future outbreaks.

“Given the possibility that next time things might be worse, we have to ask ourselves: would we prefer to have no reserves of antivirals, when we have nothing else available with which to treat a novel pandemic flu virus? Since data from the 2009 pandemic clearly show that antivirals do save lives, it seems very sensible to have an adequate supply of these medicines.

“There were no randomised trials of antivirals for H1N1 flu, but observational data from the pandemic clearly show that antivirals are beneficial and safe in those who were hospitalised. We also know that the earlier treatment is given, the less likely someone will die from flu or require intensive care. Taking this into account, it is compelling to expect that antivirals given before admission to hospital might improve outcomes even more, although we don’t have – and are unlikely to have – studies to prove this. Therefore, we have to be pragmatic and work with the data we have. If you end up in hospital with severe flu, that’s a different matter; the overwhelming observational evidence for H1N1 is that antivirals reduce the risk of progressive illness and death and they should be given as soon as flu is suspected.

“Given all of the attention over antivirals, it is interesting that antibiotics – which have no activity whatsoever against the flu virus – are given out more frequently to people with mild flu-like illness, yet remarkably, no one has scrutinised their benefits or risks in flu patients with such dogged determination. In an average winter, the money we spend on flu antivirals will be far less that the money that’s spent on “best-guess” antibiotics for flu-like illness. We also know that side effects of antibiotics are common and can be significant – probably more so than with antivirals.

“In a previous review by Jefferson et al Vaccines for preventing influenza in the elderly (Cochrane Database Syst Rev 2010) confirmed vaccine safety, but found no convincing evidence for effectiveness against disease undermining campaigns to vaccinate the elderly.  In a re-review of the same evidence using sophisticated epidemiological and statistical methods, effectiveness against influenza-related fatal and non-fatal complications was found to be ∼30%, influenza-like illness was reduced by ∼40% and confirmed influenza infection was reduced by ∼50%. When influenza was circulating the overall vaccine efficacy against infection was ∼60% (Beyer et al Cochrane re-arranged: Support for policies to vaccinate elderly people against influenza, Vaccine 2013). This type of analysis is very technical, demanding and time-consuming. By not doing a thorough and detailed analysis it is possible (even probable) that important effects can be missed.  The new Cochrane review on antivirals should be subjected to a similar reanalysis, but this will take a lot of time and detailed work. At present, I cannot say how reliable it is.

“The conclusions of the Cochrane Collaboration and the publicity by the BMJ will be listened to by many, but their findings are open to misinterpretation. The release of previous Cochrane reviews during the midst of a pandemic was seen by some as being irresponsible. We now know that antivirals saved lives during the pandemic and we risk losing one of the few weapons we have because of overly negative publicity. We should demand better-designed clinical trials and greater transparency, but we should not put lives at risk by ignoring the overwhelming body of evidence accumulated over the past 5 years that supports the use of antivirals.”

Prof Kevin McConway, Professor of Applied Statistics, The Open University, said:

“The Cochrane review is a very large and impressive piece of work, and I broadly agree with the conclusions the authors have drawn from their statistical analyses.  There are both strengths and limitations of the review.  The review is very comprehensive and the authors have looked at a great deal of data (far more data than would be used in a more typical review based on published journal articles) – this brings with it its own problems, such as how best to analyse and synthesise such a huge amount of data of different types and quality.

“One issue is that, when it comes to analysing data on complications due to influenza, and on possible adverse effects of the two drugs, in many cases there simply aren’t enough data to draw robust conclusions.  This is because many of the complications and adverse effects occurred so rarely in the dataset that robustly identifying any effects or patterns is difficult.  

“The study does have a limitation in its recommendations on public health policy (and, for example, on stockpiling the drugs) in relation to these drugs, in that the review itself makes no formal comparison of the harms and benefits of different policy choices. The general recommendation that public health decision makers should reconsider their policies in the light of all this extra information is unexceptionable – such policies should always be regularly reviewed in the light of emerging evidence. But in places the review seems to be going further and saying what the reviewers think the policy should be. This looks as if they are editorialising, in my view, in a way that is not completely justified by the data and analyses presented.

“Finally, it is a potential limitation of this study that the work has been carried out alongside campaigning on access to trial data. The writers of the review have a clear position in this controversy, and, although I personally do generally agree with their position, I feel it does at times lead to some confusion between reporting the results of the review of these particular drugs, and commenting on the general position on access to and use of unpublished data.”

Prof Wendy Barclay, Chair in Influenza Virology, Imperial College London, said:

“Treating influenza is a challenge. For most people it is an acute disease, lasting less than one week.  The virus replicates in the respiratory tract, peak virus replication occurs in the first days after a person gets infected, but they might not feel ill straight away. Severe influenza disease is associated with higher virus replication so it had always made sense that a drug that targets the virus should help to control symptoms. But getting the drug to the patient in time is difficult because often patients present late, some days after the virus replication has peaked. This difficulty in timing may explain why often these drugs have a bigger effect when given as prophylactics (before the person even catches the virus) than as treatments (Cochrane report says both Tamiflu and Relenza given as prophylaxis reduced the risk of symptomatic influenza).  In other words it’s not that the drugs don’t work, it’s just difficult to use them to their best effect.

“Most of the trials on the neuraminidase inhibitors drugs that have been re-analysed by Cochrane were conducted on previously healthy adults, infected with a seasonal influenza virus. Influenza viruses vary considerably in their severity year on year. The rates of influenza-like illness reported by GPs were low through the period (1990s and 2000s) when most of the drug trials were happening, and so the trials might only pick up ‘moderate’ effects of drug treatment, particularly when patients begin treatment several days into the infection and after virus replication has peaked.  Despite that the report concludes that the drugs did shorten the time to when adults felt better by about one day (16 hours) and in healthy children this was also the case (29 hours).  Although one day does not sound like a lot, in a disease that lasts only 6 days, it is. In the community this gets people back to work and school, and having the drugs available also serves as a safety net to treat people who get sick enough to go to hospital.

“Many of the criticisms in the new report seem to be levelled at trial design.  It may well be that trials can be done better in the future. But we should be careful about the message that is now sent to doctors and to the public. Influenza virus infections can lead to death. We have only two drugs with which we can currently treat influenza patients and there is some data to suggest they can save lives. It would be awful if, in trying to make a point about the way clinical trials are conducted and reported, the review ended up discouraging doctors from using the only effective anti-influenza drugs we currently have.  This might be particularly important in a pandemic before a vaccine is available.

“Most of the controversy surrounding these drugs is about whether they should be stockpiled for pandemics. The Cochrane group themselves cite this as a reason for their review, but it should be noted that most of the data they have re-analysed was not collected in a pandemic situation. The situation in a pandemic is very different to what we encounter each year with seasonal flu: many more people suffer severe influenza, and raised influenza awareness means they are more likely to present early. One can fairly extrapolate that the rather moderate beneficial effects of NAIs (neuraminidase inhibitors) during seasonal influenza in the 1990s and early 2000s would translate into more apparent positive benefits during a severe and widespread outbreak. Indeed retrospective observational studies published last month in Lancet Respiratory Medicine overwhelmingly support a beneficial effect for early treatment with NAIs in hospitalized patients, including pregnant women, during the swine flu pandemic in 2009/2010.  This suggests that stockpiling of these drugs was prudent.

“If another pandemic came tomorrow, and the government had no drug with which to treat thousands of influenza infected patients, I imagine there would be a public outcry. It may be that we can do better at designing clinical trials, and at licensing the next era of drugs but this will take time and meanwhile, this new report, taken alongside a lot of other data collected in different settings, does not convince me that the risks of taking Tamiflu or Relenza would outweigh the benefits.”

Prof David Spiegelhalter, Winton Professor of the Public Understanding of Risk, University of Cambridge, said:

“This is a ground-breaking review.  Since important studies have never been published, the reviewers have had to go back to clinical trial reports comprising over 100,000 pages: the effort to obtain these is a saga in itself. The poor quality of these reports clearly made extracting relevant data a massive struggle, with many pragmatic assumptions having to be made, but the final statistical methods are standard and have been used in hundreds of Cochrane reviews.  Let’s hope that in future high-quality data can be routinely obtained and this type of review becomes unnecessary.”

Dr John McCauley, Director of the WHO Collaborating Centre for Influenza, MRC National Institute for Medical Research (NIMR), said:

“Recent work^ǂ by Prof Nguygen-Van-Tam at the University of Nottingham, using data collected from collaborators around the world, reached the conclusion that during the 2009-2010 influenza pandemic (H1N1, swine flu) early initiation of treatment with neuraminidase inhibitors reduced the likelihood of severe outcomes from infection compared with no treatment or delayed treatment. The work done by Jefferson et al had a different design and was based on clinical trials. It is clear then that the differing conclusions reached by the two groups need to be carefully examined.”

^ǂ MUTHURI, S.G., MYLES, P.R., VENKATESAN, S., LEONARDI-BEE, J. and NGUYEN-VAN-TAM, J.S., 2013. ‘Impact of neuraminidase inhibitor treatment on outcomes of public health importance during the 2009-2010 influenza A(H1N1) pandemic: a systematic review and meta-analysis in hospitalized patients’, The Journal of Infectious Diseases. 207(4), 553-563

Prof Patrick Wolfe, Professor of Statistics, UCL, said:

“This review attempts to assess the evidence from 99 manufacturer-sponsored trials of the drugs oseltamivir (Tamiflu) and zanamivir (Relenza) for the treatment and prevention of seasonal and pandemic influenza in adults and children.  The authors discuss their decision not to include published trial reports – i.e., data from journal articles – but instead to use manufacturers’ reports to regulators (which they term “clinical study reports”) and regulator comments as part of their assessment.  They also discuss why an additional 8 trials were in their view unsuitable for inclusion in the analysis, and they describe a method for rating the risk of included trials for selection bias or other problems.

“The report concurs that there is evidence “both oseltamivir and zanamivir reduce the time to symptomatic improvement in adults (but not asthmatic children) with influenza-like illness” and that “Using either drug as prophylaxis reduces the risk of developing symptomatic influenza”.  However, the authors posit that “these drugs do not have an influenza-specific effect” and that based on their findings, “there is little support for their use as prophylactic agents, for example, during influenza epidemics”.  Overall, the authors note that their analysis “agrees with the conservative conclusions on both drugs drawn by the US Food and Drug Administration (FDA)”, which “described the overall performance of both drugs as ’modest’.”

“Caveats to bear in mind: As the authors state, “Our decision not to use published evidence as a basis for trial appraisal and data extraction meant that we had to reconcile and synthesise information from multiple unpublished sources”.  Moreover, “We are still uncertain whether the complete study reports represent an exhaustive and coherent source of trial narrative and data”.

“If nothing else, this report shows the importance of making full clinical trial, methods, and procedural data available for independent evaluation.”

‘Neuraminidase inhibitors for preventing and treating influenza in healthy adults and children (Review)’ by T Jefferson et al. published in The Cochrane Library, Wiley on Thursday 10 April 2014.