select search filters
roundups & rapid reactions
before the headlines
Fiona fox's blog

meta-analysis investigating effectiveness and side-effects of Tamiflu (oseltamivir) for seasonal influenza

Researchers have published in the Lancet journal a meta-analysis of the effects of Tamiflu, reporting beneficial effects on alleviation of symptoms and risk of complications, but also on an increase in side effects such as nausea. This analysis was accompanied by a roundup.


Title, Date of Publication & Journal

‘Oseltamivir treatment for influenza in adults: a meta-analysis of randomised controlled trials’  Joanna Dobson et al.

Published at 00:01 UK time on Friday 30 January 2015

The Lancet


Study’s main claims – and are they supported by the data

The data in paper do support the claim

  • that there was reduction in duration of symptoms in those on Tamiflu (by about one day- see figure 2), and
  • that there was a reduction in hospital admissions in those on Tamiflu (see p6 and Figure 4 of the paper, Relative risk=0.37, absolute risk reduction 1.1% or of 90 infected patients treated, there is one less hospital admission), and
  • that there was a reduction in lower respiratory tract complications in those on Tamiflu (see p5 and Figure 3 of the paper relative risk=0.56, absolute risk difference 3.8% or of 26 infected patients there is one less LRTC)

These are all compared to placebo.

The increase in nausea and vomiting in those on Tamiflu is also statistically significant and is not mentioned in the title of the press release, but is made clear in the paper itself. (3.7% and 4.7% risk difference, Table 2, p7)



The statistical analysis has been done well.

The claims in the press release reflect those in the paper.

The title of the press release does not reflect the nausea and vomiting side-effects.

The paper clearly presents the results with confidence intervals throughout.  These show the uncertainty about the possible effects.

The reduction in duration of symptoms was smaller in high risk than low risk patients (11 vs 28 hours for definition 1 and 18 vs 27 hours for definition 2, see Fig 3). Note that statisticians always advise subgroup analyses to be interpreted cautiously.

“The causes of admittance to hospital covered many disorders with no discernible pattern” (p 6). Therefore this finding should be interpreted cautiously (particularly as it is not quite statistically significant for the whole study population, p=0.066).

The paper includes all published and unpublished Roche-sponsored randomized placebo-controlled double-blind trials of Tamiflu in adult influenza.  This is clear, well defined, and is likely to form a reasonably homogeneous set of trials.  However “The authors excluded a Chinese treatment trial in adults because individual patient data were not available” (p 2); it had 451 Chinese adults but no detail is given. The authors do not comment on whether this Chinese trial was consistent with the trials included.

The meaning of the results depend on the clinical importance and clinical significance of the differences.



LRTC = Lower Respiratory Tract Complication


Before The Headlines is a service provided to the SMC by volunteer statisticians: members of the Royal Statistical Society (RSS), Statisticians in the Pharmaceutical Industry (PSI) and experienced statisticians in academia and research.  A list of contributors, including affiliations, is available here.

in this section

filter Headlines by year

search by tag