tetano
Editor, Senior Moderator
Heath Kelly, Epidemiologist
Benjamin J. Cowling (School of Public Health, The University of Hong Kong, Hong Kong)
Victorian Infectious Diseases Reference Laboratory, North Melbourne, Victoria, Australia
Influenza is a common, almost ubiquitous, disease caused by influenza virus infection. Annual risks of infection can exceed 20% in some years [1,2]. However the great majority of influenza virus infections do not present as the classical triad of fever, cough and fatigue [3-5], and a substantial proportion of infections, perhaps even more than half, are asymptomatic [1,2]. Even symptomatic illnesses are generally self-limiting but a small proportion of persons with influenza virus infections will require admission to hospital, intensive care and a smaller proportion will die [6]. These outcomes are uncommon and are influenced by age, with risk increased at the two extremes of life, and the presence of co-morbidities [7]. For instance, unadjusted annual risk estimates of laboratory confirmed influenza hospitalisation in hospitals from the Emerging Infections Program in the United States between 2005 and 2011 ranged from 20-72 per 100,000 for children aged 0-4 years, 16-64 per 100,000 for adults aged at least 65 years but only 5-14 per 100,000 for adults aged 20-64 years, although higher in the first year after H1N1pdm09 emerged [8]. About 10-30% of people hospitalised with influenza will require intensive care [9-11], and about 3-10% of patients hospitalised with laboratory confirmed influenza will die [10-12].
Because serious outcomes are relatively rare, randomised controlled trials (RCTs) for the prevention of influenza by vaccination or the treatment of influenza with anti-viral drugs in ambulatory settings have not been designed with sufficient power to examine them. RCTs of vaccines [13] and anti-viral drugs [14] have shown efficacy against suspected and laboratory-confirmed influenza acquired and managed in the community but there are no RCTs with outcomes of hospitalisation or death due to laboratory-confirmed influenza. Indeed it is generally acknowledged that when outcomes are rare, the RCT is not necessarily the study design of choice. The classic case control study, in which cases and controls are ascertained retrospectively, has often been the preferred alternative design. Recently, a variation of the classic design has become popular for studying vaccine effectiveness against specific outcomes, including hospitalisation due to laboratory confirmed influenza. Referred to as the case test-negative design, patients with respiratory symptoms are ascertained prospectively and vaccine coverage is compared between those who test positive and those who test negative for influenza, adjusting for potential confounders [15]. These studies have shown that inactivated influenza vaccines are associated with around a 50% lower risk of hospital admission for laboratory confirmed influenza [16,17]. This is similar to effectiveness estimates from community observational studies using the same design [18,19], and efficacy estimates from meta-analyses of community-based trials [13].
For information on the effectiveness of anti-viral medications among hospitalised patients, we likewise need to rely on observational studies. A recent review critically examined published cohort studies assessing oseltamivir treatment for laboratory-confirmed influenza and found evidence suggesting protection against mortality in four studies, all of which were judged to be of reasonable quality, and between which there was no statistical heterogeneity [20].
However, even the best designed observational studies may be subject to residual bias, and we can all agree that further placebo-controlled trials should be conducted to improve the evidence base [14,20]. These studies would require only a small fraction of the resources invested in stockpiling for pandemic preparedness. However RCTs of anti-viral medication in outpatients with increased risk of complications, and in patients hospitalised soon after onset of symptoms, may no longer be feasible because oseltamivir is the accepted front-line treatment against these groups of patients with suspected or confirmed influenza [21-23] and such trials may no longer be granted ethical approval. The same argument applies to influenza vaccination for people aged 65 years and over. In the absence of data from RCTs, observational studies provide the next best level of evidence. Case test-negative studies of influenza vaccines have confirmed their effectiveness (not efficacy, which is the measure of effect from trials) against laboratory confirmed influenza requiring hospitalisation. Cohort studies of antiviral medication used to treat laboratory confirmed influenza have suggested protection against death [20,24]. Ignoring imperfect evidence from observational studies, while pointing to the lack of better quality (but also imperfect) evidence from trials, can lead to conclusions based on the absence of evidence, rather than the evidence of absence.
It is almost trivial to note that policy should be based on the best level of available evidence, but a policy vacuum should not be left when there are no RCTs that address a specific policy question. There have been recent discussions about the efficacy of influenza vaccines against serious outcomes in the elderly because of the absence of trial data and the policy implications of this missing evidence [25]. There is a similar current discussion about the efficacy of anti-viral medication [14,26]. Because of the absence of RCTs, contemporary policies related to influenza vaccination and treatment with anti-viral medications for serious laboratory-confirmed outcomes, including hospitalisation and death, need to incorporate findings from observational studies.
It is being increasingly recognised that influenza infection in the community is common, and that infections are associated with a wide clinical spectrum, but the serious consequences of infection are generally uncommon [2]. An improved understanding of this apparent contradiction, coupled with better quality data on the management of the serious outcomes of influenza virus infection, should lead to improved evidence-based policies for the control of influenza.
http://www.bmj.com/content/348/bmj.g2548/rr/695808
Benjamin J. Cowling (School of Public Health, The University of Hong Kong, Hong Kong)
Victorian Infectious Diseases Reference Laboratory, North Melbourne, Victoria, Australia
Influenza is a common, almost ubiquitous, disease caused by influenza virus infection. Annual risks of infection can exceed 20% in some years [1,2]. However the great majority of influenza virus infections do not present as the classical triad of fever, cough and fatigue [3-5], and a substantial proportion of infections, perhaps even more than half, are asymptomatic [1,2]. Even symptomatic illnesses are generally self-limiting but a small proportion of persons with influenza virus infections will require admission to hospital, intensive care and a smaller proportion will die [6]. These outcomes are uncommon and are influenced by age, with risk increased at the two extremes of life, and the presence of co-morbidities [7]. For instance, unadjusted annual risk estimates of laboratory confirmed influenza hospitalisation in hospitals from the Emerging Infections Program in the United States between 2005 and 2011 ranged from 20-72 per 100,000 for children aged 0-4 years, 16-64 per 100,000 for adults aged at least 65 years but only 5-14 per 100,000 for adults aged 20-64 years, although higher in the first year after H1N1pdm09 emerged [8]. About 10-30% of people hospitalised with influenza will require intensive care [9-11], and about 3-10% of patients hospitalised with laboratory confirmed influenza will die [10-12].
Because serious outcomes are relatively rare, randomised controlled trials (RCTs) for the prevention of influenza by vaccination or the treatment of influenza with anti-viral drugs in ambulatory settings have not been designed with sufficient power to examine them. RCTs of vaccines [13] and anti-viral drugs [14] have shown efficacy against suspected and laboratory-confirmed influenza acquired and managed in the community but there are no RCTs with outcomes of hospitalisation or death due to laboratory-confirmed influenza. Indeed it is generally acknowledged that when outcomes are rare, the RCT is not necessarily the study design of choice. The classic case control study, in which cases and controls are ascertained retrospectively, has often been the preferred alternative design. Recently, a variation of the classic design has become popular for studying vaccine effectiveness against specific outcomes, including hospitalisation due to laboratory confirmed influenza. Referred to as the case test-negative design, patients with respiratory symptoms are ascertained prospectively and vaccine coverage is compared between those who test positive and those who test negative for influenza, adjusting for potential confounders [15]. These studies have shown that inactivated influenza vaccines are associated with around a 50% lower risk of hospital admission for laboratory confirmed influenza [16,17]. This is similar to effectiveness estimates from community observational studies using the same design [18,19], and efficacy estimates from meta-analyses of community-based trials [13].
For information on the effectiveness of anti-viral medications among hospitalised patients, we likewise need to rely on observational studies. A recent review critically examined published cohort studies assessing oseltamivir treatment for laboratory-confirmed influenza and found evidence suggesting protection against mortality in four studies, all of which were judged to be of reasonable quality, and between which there was no statistical heterogeneity [20].
However, even the best designed observational studies may be subject to residual bias, and we can all agree that further placebo-controlled trials should be conducted to improve the evidence base [14,20]. These studies would require only a small fraction of the resources invested in stockpiling for pandemic preparedness. However RCTs of anti-viral medication in outpatients with increased risk of complications, and in patients hospitalised soon after onset of symptoms, may no longer be feasible because oseltamivir is the accepted front-line treatment against these groups of patients with suspected or confirmed influenza [21-23] and such trials may no longer be granted ethical approval. The same argument applies to influenza vaccination for people aged 65 years and over. In the absence of data from RCTs, observational studies provide the next best level of evidence. Case test-negative studies of influenza vaccines have confirmed their effectiveness (not efficacy, which is the measure of effect from trials) against laboratory confirmed influenza requiring hospitalisation. Cohort studies of antiviral medication used to treat laboratory confirmed influenza have suggested protection against death [20,24]. Ignoring imperfect evidence from observational studies, while pointing to the lack of better quality (but also imperfect) evidence from trials, can lead to conclusions based on the absence of evidence, rather than the evidence of absence.
It is almost trivial to note that policy should be based on the best level of available evidence, but a policy vacuum should not be left when there are no RCTs that address a specific policy question. There have been recent discussions about the efficacy of influenza vaccines against serious outcomes in the elderly because of the absence of trial data and the policy implications of this missing evidence [25]. There is a similar current discussion about the efficacy of anti-viral medication [14,26]. Because of the absence of RCTs, contemporary policies related to influenza vaccination and treatment with anti-viral medications for serious laboratory-confirmed outcomes, including hospitalisation and death, need to incorporate findings from observational studies.
It is being increasingly recognised that influenza infection in the community is common, and that infections are associated with a wide clinical spectrum, but the serious consequences of infection are generally uncommon [2]. An improved understanding of this apparent contradiction, coupled with better quality data on the management of the serious outcomes of influenza virus infection, should lead to improved evidence-based policies for the control of influenza.
http://www.bmj.com/content/348/bmj.g2548/rr/695808