One of the biggest recent controversies in medicine involves the effectiveness – or otherwise – of the antiviral drug Tamiflu. Governments around the world have stockpiled the drug for use in severe influenza pandemics, but many have raised doubts about its effectiveness.
Influenza causes annual “seasonal” epidemics in temperate countries and circulates year-round in the tropics. Pandemics occur when there’s a relatively new flu virus containing components of bird or swine flu viruses, against which the human population has little protection.
Global pandemic preparedness efforts were spurred in the early 2000s by the emergence of SARS, and highly pathogenic H5N1 influenza in birds, which was associated with rare but often fatal infection in humans. The problem is that the severity of pandemics can vary markedly; from the Spanish flu of 1918-19, which is estimated to have killed 20-50 million people worldwide, to the much milder 2009 swine flu, which resulted in between 150,000 and 250,000 deaths (a similar number to the annual mortality of seasonal epidemics).
Governments tend to prepare for the worst because there’s no way of knowing the impact a flu pandemic will have on the population. And, as pandemics are rare, the evidence base for public health responses has to be largely taken from studies of seasonal influenza.
As part of pandemic preparation, in the early 2000s many countries amassed large stockpiles of the influenza neuraminidase inhibitor Tamiflu. A 2000 clinical trial had indicated modest benefits from this drug, but its ability to reduce disease severity and limit onward spread had potential for much greater benefits in reducing death and disease at the population level.
In 2006 the Cochrane Neuraminidase Inhibitors Review Team published a review on behalf of the Cochrane Collaboration, an independent network of researchers who review evidence for medical interventions to help improve health-care decision-making. Its overall conclusion was that the drug shouldn’t be used for treating seasonal flu, but was suitable for use as part of a suite of public health measures in pandemics.
But in the aftermath of the – fortunately – mild 2009 pandemic, the social and economic costs of the global public health response to this event were widely questioned. In particular, given the large amounts of money spent on Tamiflu, claims of conflict of interest and conspiracy theories, some more credible than others, abounded.
Then, in the process of updating their 2006 Cochrane review, and in response to questions raised through the Cochrane review’s feedback mechanism regarding prevention of complications and drug safety, the reviewing team requested access to Roche Tamiflu trials data in 2009. They were refused. In the years that followed, Roche came under particular scrutiny with claims that critical clinical trial information had been withheld from publication.
In 2012, the BMJ launched a website devoted to a public campaign, lobbying the company to release full clinical reports on all relevant studies conducted in support of the drug’s licence. The campaign was successful when, in 2013, Roche provided all the requested documentation.
The question mark
For the first time, full clinical study reports, which include tables of all study outcomes rather than those selected for publication, were made available. In addition, submissions to and correspondence with drug regulatory authorities were included. These reports, often many hundreds of pages long, formed the basis of a new meta-analysis conducted by the Cochrane Neuraminidase Inhibitors Review Team.
With the new information to hand, the authors concluded the risk of bias in several published studies was higher than had been previously assessed. This was due to missing or incomplete information, or deficiencies in study design.
Roche provided data on 83 studies, and regulatory authorities provided information on more than 200 trials. But only 46 studies (20 of Tamiflu and 26 of Relenza) were included in the final analysis as eligible and unbiased.
Still, the findings of this 2014 review were remarkably similar to the group’s previous reports. The review showed Tamiflu hastened flu recovery by about 17 hours in adults and 29 hours in kids. Anticipated side effects of nausea and vomiting were reported in less than 5% of treated people. And giving the drug to family members of an infected person prevented infections in about 15%.
The meta-analysis also looked at the risk of hospitalisations and secondary infections following flu. It concluded that antivirals were ineffective for reducing these adverse outcomes. But because the trials under consideration were mostly in generally healthy people with seasonal influenza infections, the number of hospitalisations was very small, affecting only 71 of 4,400 study participants. And while the main review finding was of a 1% absolute reduction in self-reported pneumonia, this figure represented a 56% relative risk reduction in the Tamiflu-treated group.
Based on this evidence, BMJ and Cochrane Collaboration questioned the usefulness of neuraminidase inhibitors in pandemics and called for governments to review their guidance for the drug’s use.
A different view
Meanwhile, another independent group, the Multiparty Group for Advice on Science, brought together four leading academics in the field of influenza to review and oversee re-analysis of Tamiflu trials data. To support this work, they obtained an unrestricted grant from Roche. Their aim was to resolve uncertainties regarding appropriate public health use of this drug.
The group negotiated with Roche to gain access to not just summary reports of treatment group outcomes (as previously analysed), but individually listed patient data from nine adult Tamiflu trials involving 4,328 participants. These provided much greater statistical power to assess differences. The trials, selected on the basis they assessed the currently recommended treatment dose, would have been included among those provided to the Cochrane reviewers.
This new analysis found, similarly to the Cochrane review, that the drugs hastened recovery from influenza infection by about a day, with the side effects of nausea and vomiting in a minority of patients. It said that in people with confirmed flu, Tamiflu reduced the risk of secondary respiratory infections requiring antibiotics by 44%. Hospitalisations in the Tamiflu group were down by 63%.
These findings were more in keeping with observational studies of “real world” Tamiflu use among patients hospitalised with a clinical or laboratory diagnosis of influenza during the 2009 pandemic. While this broader clinical definition of flu is less specific than in a randomised trial, it does represent the basis on which doctors make treatment decisions in everyday practice.
A meta-analysis of published data from 90 such studies published in 2012 demonstrated a 60% reduction in the odds of intensive care unit admission and death among hospitalised influenza patients who received prompt antiviral therapy. This study was also funded by an unrestricted grant from Roche.
And a follow-up 2014 analysis of individual patient data from nearly 30,000 participants who took part in studies identified through the 2012 meta-analysis found a 50% reduction in the odds of death among those treated within 48 hours of symptom onset, compared with no treatment. It was also funded by Roche.
Given that randomised controlled trials are logistically and ethically challenging to conduct in pandemic events, it’s unlikely there will ever be a consensus on Tamiflu’s effectiveness for use in pandemics. But observational studies conducted during the 2009 pandemic seem to reinforce the initial clinical trials evidence base on which Tamiflu was recommended for stockpiling against pandemic threats.
Indeed, the drug’s benefits appear to be greatest in severe influenza seasons and pandemics.