Tamiflu doesn’t work – at least not for anything important. In fact, because of its many side effects, it is probably harmful. For anyone following this evidence based debacle over the last decade, I think that conclusion is pretty clear.
You can read a little bit about the Tamiflu saga here. In short, Tamiflu was approved by licensing agencies and promoted by the WHO based on unpublished trials. None of those agencies had actually looked at the unpublished data. In fact, the primary authors of key Tamiflu trials had never been given access to the data – Roche just told them what the data supposedly said. Other papers were ghost-written. (Cohen 2009) The BMJ was involved in a legal battle with Roche for half a decade trying to get access to that information. When they finally got their hands on the data, the conclusions of the reviews suddenly changed. After countries had spent billions stockpiling the drug, it turned out that Tamiflu had no effect on influenza complications, was not effective in prophylaxis, and had significantly more harms than originally reported. (Jefferson 2014a; Jefferson 2014b) We had been duped. (There is currently a massive lawsuit claiming that Roche bilked the United States government out of 1.5 billion dollars through its false claims about Tamiflu.)
There is still some debate about Tamiflu’s benefit in terms of symptom duration. The Cochrane review reports that flu symptoms decrease by 17 hours with Tamiflu, but it is a subjective outcome and Cochrane reports a high risk of bias due to inadequate reporting, selective reporting, and attrition. (Jefferson 2014a)
On the other hand, it is pretty clear that objective outcomes are unchanged with tamiflu. Hospital admission was identical (1.7% vs 1.8%). There was no change in radiographic pneumonia. Somehow, despite including 10,000 patients in these RCTs, there was only a single death in the treatment trials (in the Tamiflu group) and 4 deaths in the prophylaxis trials. That gives you a sense of the significant selection bias that shapes these results. It also means that while there is no evidence that Tamiflu saves lives, we simply don’t know. One thing is clear in this data: Tamiflu causes harm, with more nausea (NNH 28), vomiting (NNH 22), neuropsychiatric events (NNH 94), and headaches (NNH32). (Jefferson 2014a)
The Tamiflu saga is also a great example of the problems with financial conflicts of interest in medical research. Dunn and colleagues looked at 37 assessments done in 26 systematic reviews and then compared their conclusions to the financial conflicts of interest of the authors. Among 8 assessments where the authors had conflicts, 7 (88%) had favourable conclusions about neuraminidase inhibitors. However, among the 29 assessments that were made by authors without conflicts, only 5 (17%) were positive. (Dunn 2014) Money distorts science. These conflicts need to be eliminated.
So hopefully you can see why I don’t prescribe Tamiflu; why I don’t think Tamiflu works. But wasn’t there just a big paper published in the Lancet that showed a benefit from Tamiflu?
Butler CC, van der Velden AW, Bongard E, et al. Oseltamivir plus usual care versus usual care for influenza-like illness in primary care: an open-label, pragmatic, randomised controlled trial. Lancet. 2020; 395(10217):42-52. PMID: 31839279 ISRCTN27908921
This is a pragmatic, open-label, multi-centre randomized trial.
Patients 1 year and older presenting to a primary care office with an influenza like illness (fever plus at least one respiratory symptom plus at least one systemic symptom), with less than 72 hours of symptoms, during a seasonal influenza epidemic.
- Exclusions: chronic renal failure, substantial impaired immunity, patients in whom the treating clinician thought Tamiflu or admission to hospital was required, allergy, planned general anesthesia in the next 2 weeks, life expectancy less than 6 months, severe hepatic impairment, requirement for any live viral vaccine in the next 7 days, and in some jurisdictions pregnant or lactating women.
- 75 mg by mouth twice daily for 5 days in adults and children more than 40 kg.
- For children, oral suspension was given according to weight (children weighing 10–15 kg received 30 mg, >15–23 kg received 45 mg, >23–40 kg received 60 mg, and >40 kg received 75 mg).
(All patients in both groups received “usual care”).
The primary outcome was patient reported time to recovery based on daily symptom journals.
The enrolled 3266 patients over 3 consecutive flu seasons. The patients were 56% female, and the vast majority were between the ages of 12 and 65 years old, with few comorbidities. Only a dismal 10% of the patients had had the flu shot, although the authors explain that influenza vaccination is not recommended for this healthy cohort in many of the participating European countries. Only 80% of the patients completed the full course of Tamiflu. They were able to get data on the primary outcome from 94% of the cohort (so 6% loss to follow-up).
For the primary outcome, they report a decrease in the length of illness for the Tamiflu group. The mean number of sick days was 6.7 with no treatment and 5.7 with Tamiflu. This was consistent across all their subgroups.
There was no difference in any of the objective secondary outcomes. Hospital admissions were the same (3.6% vs 2.9%). Repeat healthcare visits were also identical (52% vs 53%), as was the use of over the counter medications (82% vs 82%).
Let’s skip right to the point: this was an unblinded trial measuring a subjective outcome. With that combination, we expect significant bias. We expect that the patients given the fancy pill to think they are getting better (placebo effect), while the patient who were given nothing will feel worse.
In my mind, there is really no reason to design the trial this way. The authors say that they “deliberately chose to do an open-label trial in the context of everyday practice, because effect sizes identified by placebo-controlled, efficacy studies with tight inclusion criteria might not be reproduced in routine care. We also wished to estimate time to patient reported recovery from the addition of an antiviral agent to usual care rather than benefit from oseltamivir treatment compared with placebo.” The logic here seems to be completely backwards. There is certainly a role for real world trials, because treatments often look worse in the real world, when medications are not always taken and patients are not so tightly selected. However, the existing evidence for Tamiflu is dismal, so a trial designed to see a worse outcome in a real world setting doesn’t make a lot of sense. More importantly, the desire to study Tamiflu combined with usual care has nothing to do with using a placebo or properly blinding a trial. There are many trials that compare usual care plus a treatment to usual care plus placebo. Deciding to make the trial unblinded simply introduces unnecessary bias.
Interestingly, they present data that proves the small effect they are seeing is entirely due to the placebo effect, although they fail to recognize it as such. Neuraminidase inhibitors work specifically on the enzyme neuraminidase, which is found on influenza A and B viruses. The enzyme is not found on other respiratory viruses, which is why we don’t use Tamiflu to treat other viral illness. However, these authors found that Tamiflu was equally effective whether or not the patient actually had the flu. The HR for patients with proven influenza was 1.27 as compared to an HR of 1.31 for patients with negative swabs. There may have been some false negative swabs in there, but this number primarily demonstrates the placebo effect. There is no biological mechanism for Tamiflu to make patients without influenza better. They felt subjectively better because they were given a placebo, just like the patients who had influenza.
On the other hand, placebos don’t tend to affect objective outcomes. In this trial, none of the objective outcomes were improved. Tamiflu did not reduce hospitalizations, x-ray confirmed pneumonia, or the use of over the counter medications. (It is interesting that placebo will make you claim that you feel better, but you will still use just as much ibuprofen.) The only thing that was truly changed in this trial was an increase in nausea and vomiting because of Tamiflu.(Although this could be nocebo, the side effects of Tamiflu are well established in prior, properly done, placebo controlled trials.)
Overall, this study is a wonderful lesson on the value of blinding, but of no value to the practicing clinician. We have multiple blinded studies, and we already know: Tamiflu doesn’t work.
The positive findings in this unblinded trial are completely unconvincing. Based on plaebo controlled studies, we know that tamiflu doesn’t change any important, objective outcomes. It may decrease the length of illness slightly, although the studies that report than benefit have a high risk of bias. However, it causes side effects that patients tend to find worse than the flu itself. Personally, I do not and will not prescribe Tamiflu.
The Tamiflu Debacle – REBEL EM
If you like this content, consider subscribing to the blog:
Butler CC, van der Velden AW, Bongard E, et al. Oseltamivir plus usual care versus usual care for influenza-like illness in primary care: an open-label, pragmatic, randomised controlled trial. Lancet (London, England). 2020; 395(10217):42-52. [pubmed]
Cohen D. Complications: tracking down the data on oseltamivir. BMJ2009;339:b5387.
Dunn AG, Arachi D, Hudgins J, Tsafnat G, Coiera E, Bourgeois FT. Financial conflicts of interest and conclusions about neuraminidase inhibitors for influenza: an analysis of systematic reviews. Annals of internal medicine. 2014; 161(7):513-8. [pubmed]
Jefferson T, Jones MA, Doshi P, Del Mar CB, Hama R, Thompson MJ, Spencer EA, Onakpoya IJ, Mahtani KR, Nunan D, Howick J, Heneghan CJ. Neuraminidase inhibitors for preventing and treating influenza in adults and children. Cochrane Database of Systematic Reviews 2014, Issue 4. Art. No.: CD008965. DOI: 10.1002/14651858.CD008965.pub4. (Jefferson 2014a)
Jefferson T, Jones M, Doshi P, Spencer EA, Onakpoya I, Heneghan CJ. Oseltamivir for influenza in adults and children: systematic review of clinical study reports and summary of regulatory comments BMJ. 2014; 348(apr09 2):g2545-g2545.
Justin Morgenstern. Tamiflu doesn’t work, First10EM, 2020. Available at: