Friday, April 11, 2014

Tamiflu, the Bell Tolls for Thee

Have you read the revised Cochrane Review regarding neuraminidase inhibitors for preventing and treating influenza?  If you have, I’m impressed; it’s 559 pages long.   There are almost 250 pages of outcomes and treatment comparisons.  It is serious business.

And, it needs to be – because these folks from the Cochrane Collaboration are going up against Roche and a spin machine supporting of billions of dollars in revenue and pandemic flu strategic stockpiling.  The question:  do oseltamivir and zanamivir provide symptomatic relief and prevent serious complications from influenza infection with a reasonable safety margin?

Covering 20 oseltamivir (9,623 participants) and 26 zanamivir (14,628 participants) trials under a new data-sharing agreement, these authors ultimately conclude any faith in neuraminidase inhibitors is unwarranted.  Each treatment provided an advantage of less than a day with regard to symptom resolution – reducing the average duration of symptoms from nearly 7 days to slightly over 6 days.  Neither treatment had any effect on hospitalizations.  Pneumonia definitions were not standardized across trials; a 1% relative risk reduction was seen in investigator-reported pneumonia, but no reduction was seen in studies with radiologic objective measures of pneumonia.  Importantly, both neuraminidase inhibitors – osletamivir especially – were associated with adverse effects.  Nausea & vomiting were the most prominent symptoms, but neuropsychiatric disturbances were increased in a dose-dependent fashion.

Even more damning, there was universal risk of bias across trials – particularly for osletamivir.  The quality of the outcome follow-up for post-treatment complications was poor, allocation was not always blinded, and there was evidence of selective reporting of results.  That any government or guideline organization would base massive public health expenditures on a pharmaceutical corporation’s incomplete reporting of low-quality trials is simply indefensible.

So, there is a small, debatably unimportant effect on symptom duration.  There is zero evidence supporting the routine use of these agents to reduce subsequent infection.  This evidence does not exclude an important effect in a selected sub-population, but the authors of this review feel the underlying mechanism of action does not support such use until an effect is verified.  Furthermore, with regard to follow-up trials, the overall incidence of bacterial complications from influenza is low enough a sample size of over 20,000 patients would be required to detect a clinically meaningful difference.

It must be said – there is likewise potential bias on the side of these authors, who have long been vilifying Roche in this battle over open drug data.  The next step, we hope, will be opening the data to all to review and verify their findings.  We can hope, through repeated battles such as this, the open data day will come where we are able to properly protect patients from inadequate trial evidence.

In the meantime – let us purge ourselves of mandates and patient expectations for Tamiflu before the next influenza season rears its ugly head.

“Neuraminidase inhibitors for preventing and treating influenza in healthy adults and children”


  1. I'm not a huge proponent of Tamiflu by any means, and feel it is grossly overused, however I think that this whole issue does demonstrate some limitations of large-scale RCTs and meta-analyses when studying treatment of an acute, most often benign and self-limited illness, with significant heterogeneity in presentation factors, especially the timing of treatment initiation.

    There is a plethora of anecdotal "evidence" that very early treatment can result in dramatic symptom reduction. This kind of treatment often occurs only in patients who have a known exposure risk (ie, a close contact of an infected patient, or a healthcare worker during flu season), who presents at the immediate onset of symptoms, and may experience dramatic improvement with Tamiflu. Such events are comparably rare, and not likely to be included in RCTs, and if they are, will be masked by the vast majority of lay patients who are often uneducated about the symptoms of influenza (which may often be mild and undifferentiable from a "viral URI"), don't go to the doctor until they have been sick for 3 or 4 days, at which time treatment is probably of minimal utility. Aside from the anecdotes, there is the rather impressive 55% reduction in symptomatic influenza when Tamiflu is used as prophylaxis. The fact that it at the same time failed to produce a decrease in asymptomatic influenza suggests that the drug prevented the clinical syndrome even when administered in the incubation phase of the illness. If the drug doesn't work, how can you account for this fairly dramatic effect? I think the most coherent interpretation of this data and anecdotal experience is that, like many other antiviral drugs, the drug can have a dramatic effect when given early (ie, in the incubation period, or very soon after symptom onset) that gradually wanes as the illness progresses, by which time the host immune system is already hard at work clearing the now high viral burden, making the patient feel like crap, and the drug adds very little to the mix. The effect on secondary complications is likely also very small or non-existent, though again, it is plausible that very early treatment could prevent these events, with a very high NNT given their comparative rarity.

    1. Great thoughts - I agree, the main limitation is the trial design and outcome follow-up, which was, regardless, horribly poor across trials. Time-dependent symptom alleviation is a very reasonable thought ... although, in practical terms, it would be nearly impossible to treat someone until their disease/viremia had progressed sufficiently to cause the symptoms ... at which point, as you suggest, the potential benefit is rapidly decreasing.

      This review also only concerns the outpatient population; there's no generalizability of this ambulatory population to hospitalized patients with severe illness. The risk/benefit discussion changes dramatically at that point, considering the still-high societal morbidity of seasonal influenza.


Comments on posts > 10 days old will be moderated; blame spammers.