Debate Regarding Oseltamivir use for Seasonal and Pandemic Influenza

Aeron C. Hurt; Heath Kelly


Emerging Infectious Diseases. 2016;22(6):949-955. 

In This Article

Oseltamivir Treatment of Severe Influenza

Although necessary to consider oseltamivir's effect on more serious infections, no randomized controlled trials exist that can be included in a meta-analysis. The Cochrane group chooses only to conduct meta-analyses of randomized controlled trials, which are generally accepted to be the highest level of evidence. Thus, the Cochrane group could not review severe outcomes of laboratory-confirmed influenza. Evidence is instead derived from observational studies on the use of oseltamivir to treat complications of influenza virus infection, as in hospitalized patients or in those who died. These studies are subject to uncontrolled bias. For instance, sicker patients may be more (or less) likely to be treated, thus attenuating (or exaggerating) the effect of the intervention. Also, a serious outcome may occur soon after the treatment was initiated in a severely ill patient, so that the treatment has not had a chance to succeed. Similarly, patients who receive early treatment are more likely to benefit from treatment than patients who receive late treatment. To minimize bias, researchers conducting observational studies have attempted to adjust for time from disease onset to treatment and time from treatment to outcome. Some observational studies have also adjusted for propensity to be treated as well as patient coexisting conditions and disease severity, which may affect treatment decisions and outcomes.

Observational studies that enrolled adults have often used death as an outcome, given its ease of definition. However, many observational studies fail to control completely for potential biases, including time biases. Among studies that have attempted to control for these biases, a decreased risk for death after oseltamivir treatment has been reported, and early treatment appears to be critical.[4–16] For instance, in a retrospective cohort study from Israel of 449 patients hospitalized with influenza A(H1N1)pdm09 infection, all patients were treated with oseltamivir, and 189 (42%) were treated within 48 hours. This observational study controlled for propensity to treat and patient coexisting conditions and demonstrated the odds of death increased by 2.2 times (95% CI 1.4–3.5 times) if treatment was started late (>48 hours after symptom onset).[16]

In an attempt to overcome the criticisms of design and analysis of the observational studies, Roche chose to fund a patient level meta-analysis of individual data from 78 different observational studies, which included >29,000 patients. By adjusting for time, propensity to treat, and patient coexisting conditions, and comparing the effect of treatment with no treatment in patients infected with influenza A(H1N1)pdm09, researchers found that the odds of death were reported as 0.50 (95% CI 0.37–0.67) for adults whose treatment was initiated within 48 hours, compared with the odds of death for untreated adults.[4] However, in a series of exchanges published in the British Medical Journal and The Lancet Respiratory Medicine, even this carefully designed study has been criticized on methodologic grounds.[17–20] A more recent meta-analysis of individual patient data from the same group of investigators examined the potential effect of oseltamivir on influenza-related pneumonia among >20,000 patients with laboratory-confirmed influenza A(H1N1)pdm09.[21] However, this study has many of the predictable methodologic problems associated with retrospective reviews and does not add to the evidence base. There is scant other evidence for the benefit of oseltamivir on reducing the risk for death to help resolve residual uncertainty. The few potentially informative observational studies report on human infection with avian influenza strains[22] and seasonal influenza,[15,23] including an unpublished review sponsored by Roche.[24,25]

In a review of individual patient data for 308 patients from observational studies conducted in 12 countries, based on data from a patient register funded by Roche, oseltamivir treatment was reported to decrease the risk for death from influenza A(H5N1) virus by 49% (95% CI 23%–66%). The analysis of risk for death was restricted to 258 patients from 7 countries; mean values were substituted for missing data.[22]

Two prospective observational studies of hospitalized patients with laboratory-confirmed seasonal influenza have shown oseltamivir treatment decreases the risk for death. In a prospective observational cohort study from Hong Kong, which enrolled 754, mostly elderly, hospitalized patients with co-existing conditions during 2007–2008, oseltamivir treatment was associated with a reduced risk for death (adjusted hazard ratio 0.27, 95% CI 0.13–0.55; p<0.001), with a further reduction associated with earlier treatment.[15] A small prospective observational study of patients hospitalized with laboratory-confirmed influenza in the 2005–06 season in Ontario, Canada, found the adjusted odds ratio of death among oseltamivir-treated patients was 0.21 (95% CI 0.06–0.80; p = 0.03), based on 22 (10%) of 219 deaths in the untreated group compared with 4 of 103 deaths in the treated group.[23]

Also supporting the conclusion that oseltamivir use has a beneficial effect on reducing the risk for death were findings from a large review from the Ingenix Research Data Mart,[24] apparently sponsored by Roche. We have only been able to find an abstract of the study with an associated commentary.[25] The observational study found that oseltamivir use decreased the risk for death in patients of all ages with influenza (1 death/39,202 patients) compared with untreated patients (56 deaths/136,799 patients; p = 0.02). However, the commentary raised several issues related to study design, which could not be resolved without further detail.[25]

We have not been able to find an analysis of oseltamivir effectiveness for treating infections with avian influenza A(H7N9) virus in China, a virus that, since its emergence in 2013, has caused a greater number of annual cases and deaths than influenza A(H5N1) virus. Such a study might also contribute to the evidence base.