Abstract

Introduction

Neuraminidase inhibitors comprise nebulised zanamivir (Relenza; Glaxo Wellcome) and oral oseltamivir (Tamiflu; Gilead Sciences and F Hoffmann-La Roche), and others still under development for parenteral or long acting use.¹ Inhibiting neuraminidase—which, as with haemagglutin, is specific to influenza—blocks the exit of the influenza virus from the host cell, thereby preventing replication in other than a few host cells.²

The use of neuraminidase inhibitors has increased dramatically with the spread of the influenza A/H1N1 pandemic that began in April 2009, a novel and potentially serious infection. Partly because of the rise in resistance to amantadine and rimantadine and the lack of an effective vaccine, neuraminidase inhibitors became a widespread public health intervention. Their use for early containment and interruption was also recommended in many pandemic plans, and the World Health Organization had previously encouraged member countries to gain experience with them.³

Although several systematic reviews of the effects of neuraminidase inhibitors are published, none systematically investigated the potential harms of the drugs.^{4 5 6 7 8 9} In addition, our previous Cochrane review⁶ summary of the evidence on the effects of oseltamivir on lower respiratory tract complications was criticised by Hayashi through the public Cochrane reviews feedback mechanism (see web extra on bmj.com). This criticism centred on one paper in particular, a meta-analysis of the effects of oseltamivir on complications of influenza.¹⁰ Only two of 10 randomised sets of data had been published, and Hayashi was concerned that information was insufficient to assess methods, reliability, and applicability of the eight remaining datasets.

In updating our review we addressed these additional concerns while answering the original questions: what is the evidence on the effects of neuraminidase inhibitors in preventing or ameliorating influenza, transmission of the virus, and influenza related complications in otherwise healthy adults, and what is the frequency of adverse effects? Our original review had found positive evidence on all of these effects, and gastrointestinal harms.

Methods

We updated a search previously undertaken in any language for randomised or quasirandomised studies that compared oseltamivir or zanamivir in otherwise healthy people exposed to naturally occurring influenza, against placebo, control antivirals, or no intervention (or compared doses or schedules of the neuraminidase inhibitors) with the outcomes of influenza (efficacy) or influenza-like illness (effectiveness).⁶ We excluded experimental influenza challenge studies as their generalisability and comparability with field studies is uncertain. Studies had to include 75% or more of patients aged 14-60 (excluding older people at higher risk of complications). The updated search is summarised in the web extra. It included checking the references of other systematic reviews.^{4 5 7 8 9}

Two of us separately read all titles and studies retrieved in the search and applied inclusion criteria. Disagreements were resolved by discussion with a third reviewer. Data were extracted on to standard forms, checked, and recorded. Assessing for risk of bias (using established criteria)¹¹ became a major focus of this review.

Among complications we included a combined outcome of pneumonia, bronchitis, sinusitis, and otitis media requiring antibiotics in people with laboratory confirmed influenza. We were unable to meta-analyse the same outcomes reported by Kaiser et al¹⁰ because the data for those outcomes were not available to us for individual trials. We carried out a sensitivity analysis of complications by excluding the unpublished trials included in the Kaiser review criticised by Hayashi.

We used random effects methods to compare dichotomous outcomes (risk ratio for efficacy and odds ratio for safety); therefore estimates meta-analysed over multiple trials are average treatment effects. Where hazard ratios were not provided we converted the ratio of medians of treatment groups into (log) hazard ratios (estimating the variance of these)¹² to enable meta-analysis of time to event outcomes.

We carried out an additional search for evidence of harms, including submitting a Freedom of Information Act request to the US Food and Drug Administration for all data on the harms of oseltamivir and zanamivir.¹³ We also pursued authors of some papers and manufacturers in attempts to disentangle summarised or conflated data.

Results

Overall, 29 studies made up of 10 effectiveness and 10 safety studies (six were identified by both searches) were excluded, together with a further three effectiveness studies, excluded after more deliberation (figs 1​1 and 2​2).^{10 14 15} This left 20 included trials in 19 publications.^{16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34} Two studies that were excluded from the effectiveness outcome question were included in the safety data sources.^{14 15} (See web extra for details of the included and excluded studies.)

Open in a separate window

Fig 1Flow of studies identified from 2009 update of randomised controlled trials

Open in a separate window

Fig 2Flow of studies identified from search for evidence from post-marketing studies (excluding adverse event reporting system)

Evidence of benefits

Prophylaxis against influenza

We found no new studies of benefit fulfilling our criteria. Two trials compared a total of 697 adults treated with inhaled zanamivir 10 mg daily and 602 with placebo (followed for 22 days),^{21 30} and two trials compared 675 adults treated with oral oseltamivir 75 mg daily and 413 with placebo (followed for 49 days).^{22 36} Evidence was insufficient to support or refute the effect of neuraminidase inhibitors on prophylaxis of influenza-like illness: risk ratio 1.28 (95% confidence interval 0.45 to 3.66) for oseltamivir and 1.51 (0.77 to 2.95) for zanamivir.^{19 21 36} Higher dosages made no difference, although this conclusion is based on a single study with only nine events (fig 3​3).³⁶

Open in a separate window

Fig 3Effect of neuraminidase inhibitors compared with placebo on prophylaxis against influenza-like illness

Zanamivir reduced the chance of symptomatic laboratory confirmed influenza (0.38, 0.17 to 0.85 for 10 mg daily). Oseltamivir was similarly efficacious (0.39, 0.18 to 0.85 for 75 mg daily; fig 4​4).). Neither protected against asymptomatic influenza.^{22 30 36}

Open in a separate window

Fig 4Effect of neuraminidase inhibitors compared with placebo on prophylaxis against laboratory confirmed influenza

Postexposure prophylaxis against influenza

Postexposure prophylaxis against influenza entails giving exposed people neuraminidase inhibitors before symptoms develop. Two zanamivir trials reported significant protection for households (risk ratios 0.19³⁰ and 0.2¹⁹) and two oseltamivir trials reported similar results (0.16³⁴ and 0.42¹⁸).

Treatment

Eight trials of zanamivir treatment were included in the current review,^{16 17 20 25 26 27 28 32} of which two^{16 17} were linked to others^{27 28} (a total of 1878 participants in the treatment arm and 1310 controls, with a mean length of follow-up of 26 days). Five trials of oseltamivir were also included,^{10 24 31 33 38} one¹⁰ with supplementary outcome data from previous trials and another ²⁴ linked to a redundant publication³⁸ (totalling 1118 participants in the treatment arm and 679 controls, with a mean length of follow-up of 21 days).

There was evidence of benefit in shortening the duration of influenza-like illness for zanamivir (hazard ratio 1.24, 95% confidence interval 1.13 to 1.36) and for oseltamivir (1.20, 1.06 to 1.35; fig 5​5)) if taken within 48 hours of the onset of symptoms.

Open in a separate window

Fig 5Effect of neuraminidase inhibitors compared with placebo on alleviation of influenza symptoms (intention to treat analysis)

Data on the effectiveness of oseltamivir against complications of influenza principally came from one study,¹⁰ as pointed out by Hayashi (see web extra). This was a meta-analysis of 10 trials containing a mixture of published and unpublished data, two of which are reported in this update and the remainder inaccessible to proper scrutiny, so that we are now obliged to disregard them. The remaining data showed no benefit for oseltamivir against complications (fig 6​6).

Open in a separate window

Fig 6Effect of oseltamivir compared with placebo on complications (including pneumonia, bronchitis, or “other lower respiratory tract infections”) requiring antibiotics in laboratory confirmed influenza, based on study by Kaiser et al¹⁰ and three other studies (complications included pneumonia, bronchitis, otitis media, and sinusitis).^{24 31 33} Unpublished studies were excluded

Evidence on harms of neuraminidase inhibitors

Two of the studies excluded for addressing the benefits of neuraminidase inhibitors question nevertheless provided information on the harms of oseltamivir.^{14 15} Eighteen safety studies were included. This left 20 included trials in 19 publications.^{16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34} We also examined two sources of pharmacovigilance data describing reported adverse events from the FDA adverse event reporting system (AERS), one obtained by request using the Freedom of Information Act, “AERS-1”, and another accessed directly from two FDA websites, “AERS-2.”^{39 40} Another dataset containing reports of adverse events was located on a website by the Japanese Pharmaceuticals and Medical Devices Agency.⁴¹ We were unable to gain the same access to data from the European Medicines Agency.⁴² The table​table summarises the characteristics of the data sources.

Summary of databases on adverse effects of neuraminidase inhibitors

Data source	Who reports	What is reported	Date range of contents	Comments
US FDA AERS-1 (obtained through Freedom of Information Act request)	Manufacturers under mandatory reporting requirements in accordance with FDA Code of Federal Regulations title 21 section 314.80; health professionals and consumers report voluntarily	Post-marketing surveillance adverse event reports occurring in US and abroad. For adverse events occurring in US, manufacturers must report to FDA all adverse events except for labelled non-serious events (assuming companies have been granted such a waiver). For non-US events, manufacturers are only required to report adverse events that are “both serious and unexpected.” For first three years after approval, FDA enters all adverse events into AERS database. After three years, all electronically submitted and all “serious” adverse events are entered into AERS	Aug 1999 (zanamivir) and Dec 1999 (oseltamivir) to Jul 2009	Date of event, timing in relation to drug, country of report unclear
US FDA AERS-239 40	Manufacturers under mandatory reporting requirements in accordance with FDA Code of Federal Regulations title 21 section 314.80; health professionals and consumers report voluntarily	Post-marketing surveillance adverse event reports occurring in US and abroad. For adverse events occurring in US, manufacturers must report to FDA all adverse events except for labelled non-serious events (assuming companies have been granted such a waiver). For non-US events, manufacturers are only required to report adverse events that are “both serious and unexpected.” For first three years after approval, FDA enters all adverse events into AERS database. After three years, all electronically submitted and all “serious” adverse events are entered into AERS	Jan 2004 to Mar 2009	More detailed reports indicating reporting country consistently from Jul 2005. Reporting of neuropsychiatric adverse events more common among children and adolescents taking oseltamivir, but relation to drug is unclear48
Japanese Pharmaceuticals and Medical Devices Agency41	Documents reviewed were submitted by Chugai Pharmaceutical to Japanese regulators in conjunction with new drug application for prophylactic use of Tamiflu 75mg capsules	Summary adverse events occurring in treatment trial protocol numbers WV15670, WV15671, and WV15730. Individual patient level description of serious adverse events occurring during these three trial protocols while receiving treatment (six participants), not receiving treatment (two participants), and resulting in early withdrawal (12 participants)	Event dates not given; using trial date ranges, events occurred between Dec 1997 and Sep 1998	No cases of neuropsychiatric adverse events were reported in trials of otherwise healthy adults

Open in a separate window

FDA=Food and Drug Administration; AERS=adverse events reporting system.

Quality of evidence

In total, 1416 titles were found, most for oseltamivir, and we discarded all except for 20 studies of different design, the two sources of pharmacovigilance data from the AERS, and the Japanese data.^{13 39 40 41} The Japanese data contained tables of adverse events occurring in manufacturer funded clinical trials, and included individual patient level data.

Both FDA datasets include post-licensure adverse event reports (voluntary and mandatory) from a variety of sources, including consumers, healthcare providers, and manufacturer. Reports in the FDA AERS database are of three types: “expedited (15-day)” and “periodic,” which are mandatory reports from manufacturers, and “direct,” which are reports voluntarily submitted by non-manufacturers. The two manufacturer channels comprised 2062/2275 (90.6%) of all reports on oseltamivir from the AERS. All adverse events that are “both serious and unexpected” must, according to FDA rules, be reported by manufacturers within 15 days of initial receipt of information about the event. “Serious,” by the FDA’s definition, are those events that result in “Death, a life-threatening adverse drug experience, inpatient hospitalization or prolongation of existing hospitalization, a persistent or significant disability/incapacity, or a congenital anomaly/birth defect.”⁴³ For all other adverse events, reporting requirements differ on the basis of the origin of the information. Not all notices of adverse events received by the FDA are necessarily entered into the AERS.

The Japanese data contained tables of adverse events, including individual patient level data for serious adverse events.⁴¹ However, these were discarded because of irreconcilable differences in serious adverse events reported here but apparently not reported in two published trials.^{31 33}

Evidence of harms

The trials identified only one serious adverse event³¹ (so labelled in the Japanese data, a patient with neutropenia), and, in particular, no neuropsychiatric events. Oseltamivir induced nausea (odds ratio 1.79, 95% confidence interval 1.10 to 2.93), especially at the higher dose of 150 mg daily (2.29, 1.34 to 3.92; fig 7​7).). No statistically significant adverse event was found for zanamivir from trials.^{26 27 28 32}

Open in a separate window

Fig 7Effect of oseltamivir compared with placebo on nausea (intention to treat analysis)

Two published studies reported additional retrospective comparative safety data on oseltamivir.^{14 15} Their data suggest an incidence of neuropsychiatric adverse events per 1000 adults aged 18-49 of 20-27 at 14 days and 30-40 at 30 days,¹⁴ and for neuropsychiatric adverse events in prospective clinical trials, an incidence of 0.5%.¹⁵

AERS-1includes 2275 adverse event reports for oseltamivir and 453 for zanamivir (excluding follow-up reports on the same individual event) generated worldwide between December 1999 and July 2009 (the month our request was answered). Unfortunately it indicates neither reporting country nor how long the event occurred before receipt of the report by the FDA. The period from 2004 onwards overlaps with AERS-2, which has reports from January 2004 to March 2009, indicating both initial and follow-up reports, and reporting the date of the adverse event.^{39 40} From July 2005 it indicates the reporting country. From July 2005 to March 2009, 1205 initial adverse events occurred. Most (681, 56.5%) were reported from Japan, followed by the United States (390, 32.4%). Most (1109, 92.0%) were for oseltamivir (perhaps reflecting its higher use). A disproportionate amount of reports are for people aged less than 20 (with data on age missing for many).

Discussion

The data suggest that neuraminidase inhibitors are effective at reducing the symptoms of influenza. The evidence is of modest benefit—reduction of illness by about one day. This benefit has been generalised to assume benefits for very ill people in hospital. This seems reasonable, although it is worth remembering that we have no data to support this, and it is unlikely that ethics committees would allow a trial of no treatment for people with influenza who have life threatening disease.

One important caveat to these results arises from concerns about the difference between efficacy (treatment response to pure influenza virus infection) and effectiveness (the real life response to influenza-like illness, when real cases of influenza are indistinguishable from other causative agents not responsive to neuraminidase inhibitors⁴⁴). Understanding the proportion of influenza-like illness caused by both seasonal and epidemic influenza is critical to generalising the results of this review to clinical practice. The finding of treatment effectiveness for the neuraminidase inhibitors may have been enhanced by the high percentage of influenza-like illness caused by influenza in most of the included trials—for example, up to 80%.²³

Data on the effectiveness of oseltamivir against the complications of influenza are confusing. Hayashi pointed out that the original data, which led to the 2005 version of this original Cochrane review reporting benefits for oseltamivir on reduction of complications from lower respiratory tract infection, principally came from one meta-analysis that summarised 10 trials containing a mixture of published and unpublished data.¹⁰ Only two of the trials it contains are published (and are reported in this Cochrane review update), the remainder were offered to us under conditions we thought unacceptable, and what was offered to us was insufficient to analyse properly. (Comments on the Kaiser et al paper are in the web extra). This means we are now obliged to exclude the meta-analysis.¹⁰ The remaining published evidence is insufficient to answer the question about the effectiveness of either neuraminidase inhibitor on reducing the complications of lower respiratory tract infection, antibiotic use, or admissions to hospital. It is possible that there is a publication bias, especially as we know of eight trials that are unpublished and inaccessible. We have not undertaken a funnel plot because there are only three trials (fig 6), and so the issue of publication bias remains unresolved. Its direction might be in favour of exaggerating the treatment effect. Hayashi’s comments point out a serious problem with our original review, which we now address.

The results from the meta-analyses involving hazard ratios should be viewed with caution because of the approximate methods used to extract estimates for each study. Hazard ratios were rarely reported directly, so we had to use the ratio of the observed median duration of symptoms in each group as an approximation to the hazard ratio. This approach may be overly simplistic, as it makes a comparison at only one time point and assumes a constant survival difference over time, and thus may produce different meta-analysis results than if the actual hazard ratio estimates were available.⁴⁵

Notes

We thank Liz Dooley, Dr Aoki, Jon Deeks, Carlo Di Pietrantonj, and Vittorio Demicheli for input and advice on earlier versions of this review; Liz Dooley, Janet Wale, John Bartlett, Sree Nair, Tom Fahey, Ruth Foxlee, and Alex Rivetti for help with the searches; and the Australian National Health and Medical Research Council and the UK NHS Research and Development fund for grants to enable this international rapid update.

Contributors: TJ reapplied inclusion criteria for the 2009 update. All authors reappraised and investigated extracted data, while CDM supervised the process and arbitrated when necessary. MJ and PD checked and transformed data and supervised the revised meta-analysis. TJ, CDM, MJ, and PD edited the text. All authors contributed to the final draft. CDM and TJ are the guarantors.

Funding: This study was supported by grants from the Australian National Health and Medical Research Council and the UK NHS Research and Development fund.

Competing interests: All authors have completed the unified competing interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare that (1) CDM and TJ have support from the UK National Institute for Health Research and the Australian National Health and Medical Research Council for the submitted work; (2) CDM has provided expert advice to GlaxoSmithKline about vaccination against acute otitis media and TJ was a paid ad hoc consultant to Roche from 1997-9; (3) their spouses, partners, or children have no financial relationships that may be relevant to the submitted work; and (4) no authors have non-financial interests that may be relevant to the submitted work.

Ethical approval: Not required.

Notes

Cite this as: BMJ 2009;339:b5106

References