Stream 4: Review Summary
Antivirals for influenza: a summary of a systematic review and meta-analysis of observational studies
Correspondence: Holger J. Schünemann, MD, PhD, MSc, FRCPC, Chair, Department of Clinical Epidemiology & Biostatistics, Michael Gent Chair in Healthcare Research, Professor of Clinical Epidemiology, Biostatistics and Medicine. Department of Clinical Epidemiology & Biostatistics, McMaster University Health Sciences Centre, Room 2C10B, 1200 Main Street West, Hamilton, ON L8N 3Z5, Canada. E-mail: email@example.com
Despite the use of antivirals to treat patients with severe influenza, questions remain with respect to effects and safety. Although a recent systematic review has provided some indication of benefit, the analysis is limited by the quality of the available evidence from randomized controlled trials. To supplement the existing information, the authors conducted a systematic review of observational studies of antiviral treatment for influenza. This report summarises the findings of that review. Similar to the randomised trials, the confidence in the estimates of the effects for decision-making is low to very low primarily due to the risk of selection and publication bias in the observational studies. From these observational studies, the summary estimates suggest that oseltamivir may reduce mortality, hospitalisation and duration of symptoms compared with no treatment. Inhaled zanamivir may also reduce symptom duration and hospitalisations, but patients may experience more complications compared with no treatment. Earlier treatment with antivirals is generally associated with better outcomes than later treatment. Further high-quality evidence is needed to inform treatment guidelines because of the overall low to very low quality of evidence.
Influenza remains a major global health concern. Most cases of influenza are self-limiting, and prevention through annual influenza vaccination may be an effective strategy. However, antivirals, such as neuraminidase inhibitors (oseltamivir and zanamivir) and M2 ion channel blockers (amantadine and rimantadine), are used to reduce symptoms and prevent hospitalisation or death in patients with severe illness.
Evidence about the effects and safety of antivirals continues to accumulate, and although a recently updated systematic review of randomised control trials (RCTs) offered some indication of treatment benefit, concerns remain over the quality of evidence from published and unpublished randomized trials. The evidence from randomized trials is further limited by the lack of high-quality evidence for patient-important outcomes and lingering questions about the treatment of specific groups such as hospitalised or immunocompromised patients. Observational studies may provide additional information or higher quality evidence than the currently available RCTs for certain elements of antiviral treatment. This report summarises the results of a recently published systematic review of observational studies of antiviral treatment. This review was intended to inform WHO guidelines, and the WHO essential medicine list about the antiretroviral treatment of influenza.
Electronic databases and grey literature were searched up to 16 November, 2010, using pre-defined eligibility criteria, without restriction on publication language or study type. Observational studies examining the effects of non-intravenous antiviral treatment with oseltamivir, zanamivir, amantadine or rimantadine compared with no treatment, and studies comparing the antivirals with one another or early administration of a particular antiviral (<48 hours) compared with later administration of the same antiviral (>48 hours) were included. Investigators independently screened all citations by title and abstract and by full text (920 articles) for inclusion. Several patient-important outcomes were assessed, including patient death, hospitalisation, duration of signs and symptoms, complications and adverse effects (see original report for full list of outcomes).
Two investigators independently extracted data from the included studies using a pretested electronic form. Established methods to assess the risk of bias were used. The quality of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach. Following GRADE guidelines, confidence in the effect estimates were categorised into four levels from very low to high.
Meta-analyses were conducted by using either the odds ratio (dichotomous outcomes) or mean differences or standardised mean differences (continuous outcomes). Whenever possible, studies were pooled for analysis according to whether or not results were adjusted for confounding. Although greater emphasis should be placed on data from the adjusted meta-analyses, the pooled results from unadjusted studies were also reported to provide a comprehensive view of the available evidence. Studies and other unpublished observational data that could not be pooled were synthesised narratively. When data were available, subgroup analyses were performed by age, risk of complications, laboratory-confirmed influenza versus influenza-like illness as well as other variables.
Twelve thousand one hundred and eighty-eight database citations and 27 articles from the grey literature were identified, and 74 articles met the eligibility criteria. The majority of studies reported comparisons of oral oseltamivir with no antiviral therapy. Table 1 summarises the effects of oseltamivir over no treatment and the quality of evidence for each of the outcomes. Additional tables summarising the effects of the other antivirals are in the original publication.
Table 1. GRADE evidence profile for oral oseltamivir versus no antiviral therapy
| Mortality |
|681 (three studies)||⊕⊕⊝⊝ LOW*||59/242 (24·4%)||31/439 (7·1%)||Adjusted OR 0·23 (0·13–0·43)||240 deaths per 1000||172 fewer deaths per 1000 (from 120 to 201 fewer)|
|1557 (nine studies)||⊕⊝⊝⊝ VERY LOW*,** due to risk of bias||61/320 (19·1%)||228/1237 (18·4%)||OR 0·51 (0·23–1·14)***||240 deaths per 1000||101 fewer deaths per 1000 (from 172 fewer to 25 more)|
| Hospitalisation |
|150710 (four studies)||⊕⊕⊝⊝ LOWf||1238/100585 (1·2%)||431/50125 (0·86%)||Adjusted OR 0·75 (0·66–0·89)||12 hospitalisations per 1000||3 fewer hospitalisations per 1000 (from 1 to 4 fewer)|
|242762 (six studies)||⊕⊝⊝⊝ VERY LOW**,f due to risk of bias||1738/146410 (1·2%)||1086/96352 (1·1%)||OR 0·75 (0·66–0·86)||12 hospitalisations per 1000||3 fewer hospitalisations per 1000 (from 2 to 4 fewer)|
| ICU admissions/mechanicalventilation/respiratory failure |
|1032 (six studiese)||⊕⊝⊝⊝ VERY LOW*,f due to risk of bias, inconsistency||–||200/1032 (19·4%) Pooled Risk 13·0% (95% CI 11–15%)||–||–|| |
| Duration of hospitalisation (days) |
|832 (five studies)||⊕⊝⊝⊝ VERY LOW*,e,f due to risk of bias, inconsistency||–||832||–||–||The mean duration of hospital stay was 5·16 days (5·02–5·29)|
| Duration of signs and symptoms (measured from onset of symptoms or treatment) (time to return to normal activity was not measured) |
|5842 (six studies)||⊕⊝⊝⊝ VERY LOW*,f due to inconsistency||449||5393||–||–||The mean time was 0·91 standard deviations lower (1·25–0·57 lower)g|
| Complications – Pneumonia |
|150466 (three studies)||⊕⊝⊝⊝ VERY LOWf,f due to inconsistency||2111/100449 (2·1%)||647/50017 (1·3%)||Adjusted OR 0·83 (0·59–1·16)||21 pneumonias per 1000||4 fewer pneumonias per 1000 (from 9 fewer to 3 more)|
|265276 (six studies)||⊕⊝⊝⊝ VERY LOW**,f,f due to risk of bias, inconsistency||3244/166256 (2%)||1273/99020 (1·3%)||OR 0·64 (0·46–0·88)||20 pneumonias per 1000||7 fewer pneumonias per 1000 (from 2 to 10 fewer)|
| Complications – Otitis media |
|78407 (two studies)||⊕⊕⊝⊝ LOWf||546/40022 (1·4%)||285/38385 (0·74%)||Adjusted OR 0·75 (0·64–0·87)||14 otitis media per 1000||3 fewer otitis media per 1000 (from 2 to 5 fewer)|
|193105 (four studies)||⊕⊝⊝⊝ VERY LOW**,f,f due to risk of bias, i nconsistency||2053/105758 (1·9%)||1381/87347 (1·6%)||OR 0·77 (0·63–0·94)||19 otitis media per 1000||4 fewer otitis media per 1000 (from 1 to 7 fewer)|
| Complications – Cardiovascular outcomes |
|100830 (two studies)||⊕⊕⊝⊝ LOWf||62385||38445||Adjusted OR 0·58 (0·31–1·1)||200 cardiac events per 1000||73 fewer cardiac events per 1000 (from 128 fewer to 16 more)|
|60678 (two studies)||⊕⊝⊝⊝ VERY LOW**,f,f due to risk of bias, i nconsistency||6814/50696 (13·4%)||606/9982 (6·1%)||OR 0·45 (0·25–0·81)||110 cardiac events per 1000||57 fewer cardiac events per 1000 (from 19 to 80 fewer)|
| Critical Adverse Events |
|104930 (five studies)||⊕⊕⊝⊝ LOWf||60817||44113||Rate Ratio 0·76 (0·7–0·81)||420 adverse events per 1000 patient years||101 fewer adverse events per 1000 patient years (from 80 to 126 fewer)|
The evidence suggests that oseltamivir may reduce mortality in high-risk populations, such as hospitalised patients, compared with no antiviral treatment, but this effect is less pronounced when pooling data from unadjusted studies.[4-14] Meta-analyses indicate that oseltamivir may also reduce both hospitalisation in outpatients[15-18] as well as duration of symptoms.[19-24] Data from several studies also suggest that oseltamivir may reduce complications such as pneumonia,[15, 17, 18] otitis media[15, 18] or any recurrent cardiovascular outcome.[25, 26] Pre-planned subgroup analyses showed statistically significant effects in children compared with adults for both pneumonia and otitis media.
When comparing zanamivir to no treatment, there was moderate quality evidence that zanamivir reduced the duration of symptoms[24, 27, 28] (23 hours, CI 95% 17–28 hours; SMD −0·94, CI 95% −1·21 to −0·66); and very low-quality evidence that hospitalisations may be reduced[11, 29] (OR 0·66, CI 95% 0·37–1·18). However, patients with influenza-like illness may experience more complications with inhaled zanamivir, including otitis media, respiratory disease or other complications.
In direct comparisons of oseltamivir with zanamivir, inconsistent evidence suggests that zanamivir may have a slight advantage in reducing the duration of symptoms[20, 23, 28, 30, 31] (by 7 hours, CI 95% 2–12 hours; SMD 0·26, CI 95% 0·07–0·45). Very low to low-quality evidence showed that the two treatments may not differ with respect to mortality, hospitalisation, ICU admission, critical adverse events or viral shedding (as measured 5 days after treatment).[11, 20, 24]
Results from studies evaluating the effects of initiating treatment with oral oseltamivir within 48 hours of symptom onset versus initiation of treatment after 48 hours of symptoms suggest that mortality (OR 0·33, CI 95% 0·12–0·86), hospitalisations (OR 0·52, CI 95% 0·33–0·81) and ICU admission (OR 0·22, CI 95% 0·15–0·33) may be reduced when oseltamivir treatment is initiated earlier.[32-39] Similarly, there was low-quality evidence showing a reduction in symptom duration and risk of complications,[41, 42] when treatment was initiated earlier rather than later with amantadine or rimantadine, respectively.
The results of this review underscore the need for additional, high-quality studies to inform guidelines about the use of antiviral treatments for influenza. Many of the studies indentified in this review had a high risk of bias resulting from the lack of control for confounding variables and possible selection bias. In addition, meta-analysis was limited by the fact that few studies reported appropriately adjusted effect measures. These limitations make it difficult to draw firm conclusions from the studies, and care must be used when using this information to make healthcare decisions. Based on this evidence, however, oral oseltamivir and inhaled zanamivir may provide net benefit over no treatment, although treatment with zanamivir may result in more complications than no treatment. In direct comparisons, zanamivir may be slightly more effective in reducing the duration of symptoms than oseltamivir. The evidence further suggests that administering antivirals within 48 hours of symptoms may be of greater benefit than initiating treatment at a later time.
Although the overall confidence in the effect estimates is low to very low, this review has nevertheless provided important evidence supporting a role for antivirals in the treatment of influenza and must be viewed in the context of the information available from RCTs and the substantial burden of influenza worldwide. There remains a need for high-quality evidence from RCTs that address patient-important outcomes and include hospitalised patients with influenza. Observational studies can continue to supplement this evidence by contributing data about special populations, adverse effects and rare harms.
The authors thank Rasha Khatib and Meg Fukuzawa for translation of Japanese articles, Wojtek Wiercioch for help with reference management, DianeHeels-Ansdell for statistical support, Glaxo Smith Kline and James Smith (Roche) for responding to our requests to identify studies, and Dr. Tom Jefferson for his help in interpreting the recent Cochrane reviews that he authored summarising the RCTs on this topic. This article is a summary from Hsu J, et al. Antivirals for treatment of influenza: a systematic review and meta-analysis of observational studies. Ann Intern Med 2012; 156:512–524.
Conflict of interest
Ms. Santesso; Drs. Mustafa, Brozek, Hopkins, Flottorp, and Schünemann; Mr. Chen; Ms. Cheung; Mr. Wong; and Mr. Tian report the following: Grant (money to institution): World Health Organization.
World Health Organization and McMaster University.