Antiviral drugs for the treatment of influenza: a systematic review and economic evaluation

M2 membrane protein inhibitors

The anti-influenza action of amantadine (Lysovir^®, symmetrel syrup; Alliance Pharmaceuticals) was recognised in 1964, leading to the first licence for antiviral treatment for influenza. 29 Rimantadine (Flumadine^®; Forest Pharmaceuticals) is a derivative of amantadine, and was first approved for use in 1993. Both amantadine and rimantadine inhibit the M2 membrane protein ion-channel activity of influenza A, but have no effect on influenza B. 2,3 The blocking of the proton pump of the M2 protein stops the virus uncoating and inactivates newly synthesised viral haemagglutinin. Rimantadine is not licensed for use in the UK. 29 Amantadine is licensed for treatment of influenza A in people over the age of 10 years who are deemed to be at risk.

Neuraminidase inhibitors

The first sialic acid analogue, Neu5Ac2en, was developed in 1969. 27,30 However, this early analogue had low potency and poor specificity. 27 Subsequent development led to the production of zanamivir (first trialled in 1994) and oseltamivir (first described in 1997). 27 Zanamivir and oseltamivir are effective against both influenza A and influenza B. 2,3,27

Other neuraminidase inhibitors

Two neuraminidase inhibitors are currently under development: peramivir (BCX-1812, RWJ-270201; Biocryst Pharmaceuticals)34 and A-315675 (Abbott Laboratories). 35,36 As both of these drugs are still in the development phase, they are not considered in this review.

Symptoms

Seven trials of zanamivir in healthy adults reported the time to the alleviation of fever or all symptoms, or to the return to normal activity. 46,76,79,87–89,91–97 The follow-up period was 5 days in one trial,46 14 days in one trial,93–95 21 days in one trial,87 28 days in three trials,88,89,91,92,96,97 and between 21 and 28 days in one trial. 76 Three of the trials had a dropout rate greater than 5%, but none had one greater than 10%. 76,87,91,92

Treatment with zanamivir significantly reduced the time to the alleviation of all symptoms, but there was no clear evidence of an effect on the time to return to normal activity, in either the ITT (Table 6) or the ITTI (Table 7) population. Statistically significant heterogeneity was observed in the analyses of the time to return to normal activity in the ITT population. All the trials included in this analysis used the same outcome definition, and there are no consistent differences between the studies in terms of trial methodology or quality to account for the observed heterogeneity. One study reported the time to alleviation of fever, showing no difference between zanamivir and placebo in either the ITT or the ITTI population. 88,89 This same study reported a time to the alleviation of symptoms that was shorter than the other studies, although this was not reflected in the time to return to normal activity. This is most likely due to this study recruiting healthy young (17–29 years) men (99%) who were in the Finnish Defence Force, who may have recovered more quickly but may not have returned to full duties immediately after symptoms had been relieved.

Complications

Outcomes relating to complications in healthy adults were poorly reported, with only three zanamivir trials reporting any complication rates (Table 8 and 9). 76,87–89 Two of these had a dropout rate greater than 5%, but not greater than 10%. 76,87

The majority of the available information is from a study of primarily healthy young men in the Finnish Defence Forces. 88,89 This study reported high rates of hospitalisation and GP consultations, as those suffering with influenza were isolated from their colleagues; these data are unlikely to be representative of complication rates seen in a presenting population in general practice. The information available from the other two trials is limited, with the incidence of complications low. There is therefore insufficient relevant information from which to draw conclusions.

Adverse events

Four trials of healthy adults reported adverse events, all of which were in the ITT population (Table 10). 76,86–89 The pooled results showed no significant differences between zanamivir and placebo in overall, drug-related or serious adverse events. There was no statistically significant heterogeneity between studies for any of the pooled results.

Symptoms

Six trials of oseltamivir in healthy adults reported the time to the alleviation of fever or all symptoms, or to return to normal activity. 46,65,67–69,80,82,107 The follow-up period was 21 days in all six trials. Three of the trials had a dropout rate greater than 5%, but not greater than 10%. 65,67,82,83

Treatment with oseltamivir significantly reduced the time to the alleviation of all symptoms and to the time to return to normal activity in both the ITT and the ITTI populations (Tables 11 and 12 respectively). Statistically significant heterogeneity was observed only in the analysis of the alleviation of all symptoms in the ITTI population. This seems to be primarily driven by one small trial (Roche, WV15730). 48,82,83 When this trial was removed from the analysis, the impact of oseltamivir remained statistically significant (WMD –19.06; 95% CI –33.02 to –5.10; p = 0.007), but there was no longer statistically significant heterogeneity between the studies (χ² = 7.53; p = 0.11; I² = 46.9%). When there was sufficient information reported, oseltamivir also reduced the time to the alleviation of fever.

Complications

Six oseltamivir studies reported outcomes relating to complications in healthy adults (Tables 13 and 14),46,65,66,68,69,83,107 three of which were conducted in China66,69 or Japan,68 with two published in their respective languages. 66,68 Oseltamivir showed a statistically significantly reduction in the rate of antibiotic use in the ITT population. However, this result was based on only two studies, one of which was the trial by Deng et al. ,66 which was conducted in China and had a particularly high rate of antibiotic use in both arms, and which therefore has most weight in and drives the pooled result. In addition, the study by Deng et al. 66 was described as open label. Nonetheless, the Canadian trial by Nicholson,107 which used a matching placebo and appeared to be double blinded, showed a similar size of relative effect. The pooled result for the ITTI analysis has the same direction of effect, does not include the trial by Deng et al. and is of borderline statistical significance.

Overall, there was limited information on complications, their incidence was low, and no other statistically significant differences in complication rates between oseltamivir and placebo were observed. There was no statistically significant heterogeneity between studies in any of the analyses. Two trials had a dropout rate greater than 5%, but not greater than 10%. 65,82,83

Adverse events

Adverse events were reported in six oseltamivir studies of healthy adults, all of which were in the ITT population (Table 15). 46,66–69,80,82,83,107 Four of the studies were conducted in China66,67,69 or Japan,68 with three published in their respective languages. 66–68 Adverse event data were only reported for ITT populations. There were few events, and no clear evidence of differences in drug-related or serious adverse events was observed between oseltamivir and placebo. There was no statistically significant heterogeneity between studies in any of the pooled analyses. Two trials had dropout rates greater than 5% but not greater than 10%. 67,82,83

Symptoms

Nine trials of zanamivir in an at-risk population reported the time to alleviation of fever or symptoms, or to return to normal activity. Of these, six were in a general at-risk population,46,75,91–97,99 one in children, with results for those at risk provided separately by GlaxoSmithKline,46,100,101 one in the elderly,105,106 and one in adults with chronic obstructive airways disease (COAD) or asthma. 98,114,115 The follow-up period was 5 days in two trials,46,99 6 days in one trial,103 14 days in one trial,93–95 21 days in one trial,75 28 days in four trials,91,92,96–98,100,101,114,115 and 29 days (with a follow-up telephone call at 56 days) in one trial. 105,106 Three of the trials had a dropout rate greater than 5%, but not greater than 10%. 75,91,92,98,114,115

Treatment with zanamivir significantly reduced the time to the alleviation of all symptoms in the overall at-risk group in both the ITT and the ITTI populations. There was no clear evidence of an effect of zanamivir on the time to the return to normal activity in either population (Tables 16 and 17). There was statistically significant heterogeneity between studies for the time to return to normal activity when results for the overall at-risk groups were pooled.

For comparability with the Turner et al. 46 review’s separate analyses of at-risk adults and children, a sensitivity analysis was undertaken in which the study in children was excluded from the overall pooled results (see Tables 15 and 16). 100,101 At-risk adults showed the same pattern of response as the whole at-risk population, with treatment with zanamivir resulting in a statistically significant reduction in the time to the alleviation of all symptoms in both the ITT and the ITTI populations. There was no clear evidence of an effect of treatment on the time to return to normal activity in either population. Statistically significant heterogeneity across studies was observed in the analyses of the time to return to normal activity in both the ITT and the ITTI populations. One study showed a significant reduction with zanamivir in the time to alleviation of fever in the ITT population. 99

Complications

Six zanamivir trials reported complications in an at-risk population. Of these, one was in patients with COAD or asthma,98,114,115 one in the elderly,105,106 and four in general at-risk populations. 91–97,99 Two had dropout rates of greater than 5%, but not greater than 10%. 91,92,98,114,115 Two trials had follow-up periods of only 5 days,99,103 and one of 14 days. 93–95

The incidence of complications was low, and there was no clear evidence of a difference in rates of overall complications between zanamivir and placebo in either the ITT or the ITTI populations (Tables 18 and 19). With a total of eight events, there was no evidence of a difference in the rates of pneumonia between groups in either population. Analysis of antibiotic use was also based on limited data, and there was no clear evidence of a difference in either population. The incidence of bronchitis was significantly lower with zanamivir in the ITT populations, which included one study restricted to an elderly population105 and another restricted to patients with COAD or asthma. 98,114,115 Statistically significant heterogeneity was observed in three analyses, overall complications in the ITT population, and antibiotic use in both the ITT and the ITTI populations.

Adverse events

Four zanamivir trials reported overall adverse events in an ITT, at-risk, population (Table 20). Of these, one was in patients with COAD or asthma,98,114,115 one in the elderly,105,106 and two in general at-risk populations. 91,92,99 There was no significant difference in the number of overall, drug-related or serious adverse events. The analysis of overall adverse events was subject to a high degree of statistical heterogeneity, seemingly caused by the study conducted in the elderly population.

Symptoms

Six trials of oseltamivir in an at-risk population reported the time to alleviation of fever, all symptoms, or return to normal activity. Of these, two were in a general at-risk population,46,73 one in children with asthma,78 and three in the elderly. 46,73 The follow-up period was 21 days in five trials46,73,109 and 28 days in two trials. 77,110 The dropout rates were unclear in six of the studies; only Whitley et al. 110 reported a dropout rate lower than 5%.

Although the direction of effect favoured oseltamivir, there was no clear evidence of a reduction in the time to the alleviation of symptoms or return to normal activity in the overall at-risk population with oseltamivir, in either the ITT or the ITTI populations (Tables 21 and 22 respectively). For comparability with the Turner et al. 46 review’s separate analyses of at-risk adults and children, sensitivity analyses were undertaken in which the study in children was excluded from the overall pooled results and analysed separately. 78,110 When this trial was removed, the time to return to normal activity was significantly less (by approximately 3 days) with oseltamivir in the ITTI at-risk adult population (Tables 21 and 22). There were no statistically significant differences between oseltamivir and placebo in at-risk children for either outcome in either population. There was no statistically significant heterogeneity across studies in any of the analyses.

Complications

Seven trials reported the rate of complications in at-risk populations in trials of oseltamivir, one in children with asthma,78 two in the elderly (combined),73 and four in a general at-risk population, three of which were combined71–73 (Tables 23 and 24). Treatment with oseltamivir significantly reduced antibiotic use in the ITTI population. However, the incidence of other outcomes was low, and there was no clear evidence of other differences between oseltamivir and the comparator. No statistically significant heterogeneity was observed in the pooled analyses.

Adverse events

Two trials reported the incidence of adverse events in at-risk populations in trials of oseltamivir, one in children with asthma78 and one in a general at-risk71,72 population (Table 25). There was no evidence of a difference in the rate of overall, drug-related or serious adverse events in the ITT or ITTI populations between oseltamivir and placebo. No statistically significant heterogeneity was observed in the pooled analysis.

Symptoms

Two studies reported the effectiveness of zanamivir on the time to the alleviation of symptoms and return to normal activity in children (Tables 26 and 27). 100,101,103 The follow-up period was 6 days in one trial103 and 28 days in the other. 100,101 Both trials had a dropout rate less than 5%. 100,101,103 Data were available for healthy and at-risk children separately for the trial by Hedrick et al. ,100,101 but not for the NAI30028 trial. 103

Zanamivir resulted in a significant reduction in the time to alleviation of symptoms, with no statistically significant heterogeneity observed in the pooled analysis of the ITT population. When the population recruited in Hedrick et al. 100,101 was split into healthy and at-risk children, this study showed a statistically significant reduction in the time to normal activity in the ITTI at-risk population; however, this analysis included only 22 children. The effect of zanamivir was less pronounced in healthy children, and the analysis was based upon 324 children. There were no significant differences between zanamivir and placebo for the time to return to normal activity.

Complications

Two studies of zanamivir in children reported data on complications (Tables 28 and 29);100–104 one of these had a follow-up period of only 5 days. 103 There was no clear evidence of differences between zanamivir and placebo for any complication in the ITT or ITTI populations.

Adverse events

The same two studies reported adverse event data;100,101,103,104 the study with a follow-up period of 5 days for previous outcomes had a follow-up period of 14 days for adverse events. 103 Adverse event data were reported only for the ITT population (Table 30). The pooled ORs revealed no significant differences between the zanamivir and placebo groups for overall, drug-related, or serious, adverse events. No statistically significant heterogeneity was observed in any analysis.

Symptoms

Two studies reported the effectiveness of oseltamivir on the time to the alleviation of symptoms and return to normal activity in children;78,110 one was restricted to children with asthma. 78 The follow-up period was 21 days in one trial78 and 28 days in the other. 110 The dropout rate was less than 10% in both trials. The time to the alleviation of fever, all symptoms, and return to normal activity were significantly lower with oseltamivir than with placebo (Tables 31 and 32). The analysis of time to normal activity in the ITTI population was subject to statistically significant heterogeneity. One of the studies was restricted to children aged 6–12 years who had asthma,78 whereas the other included a larger number of children from a more representative population, aged between 1 and 12 years,110 which could explain the observed heterogeneity. 78

Complications

The same two studies reported on complications (Tables 33 and 34);78,110 one reported a significant reduction in antibiotic use and the incidence of otitis media in the ITTI population. 110 The other study, restricted to children with asthma, reported no statistically significant difference between oseltamivir and placebo for any complication, GP consultations or hospitalisation rates. 78 The trials had dropout rates of 3%110 and 9%. 78

Adverse events

The same two trials reported the numbers of overall or serious adverse events in the ITT population, with no evidence of differences between oseltamivir and placebo (Table 35). 78,110 The trial by Johnston et al. 78 had a particularly high rate of adverse events, which may have been due to the fact that the trial was restricted to children with asthma, and classed complications such as bronchitis and pneumonia as adverse events.

Symptoms

Five studies reported the effectiveness of zanamivir on the time to the alleviation of symptoms and return to normal activity in the elderly, although the numbers of patients in four of these were very small as the data were obtained from studies with mixed populations. 91,92,96–98,105,106,115,116 The follow-up period was 14 days in one trial,93–95 28 days in three trials,91,92,96–98,115,116 and 29 days with a further telephone call at 59 days in one trial. 105,106 The dropout rate was less than 5% in all trials. There was no clear evidence of differences between zanamivir and placebo for either outcome in any population assessed (Tables 36 and 37).

Complications

One study reported no significant differences between the treatment groups for any complication measured in the elderly (Tables 38 and 39). 105

Adverse events

One study reported no significant differences between zanamivir and placebo in the incidence of overall (zanamivir, 66/191; placebo, 50/166; OR 1.22, 95% CI 0.78 to 1.91) or serious (zanamivir, 5/191; placebo, 9/166; OR 0.47, 95% CI 0.15 to 1.43) adverse events. 105

Symptoms

Three trials evaluated the effectiveness of oseltamivir in the elderly with regards to the time to the alleviation of symptoms and return to normal activity. Results for these were reported either combined in the publication by Martin et al. ,73 or separately in the previous review by Turner et al. 46 The follow-up periods were 21 days, and the dropout rates unclear. There was no evidence that oseltamivir reduced the time to the alleviation of symptoms in either the ITT or the ITTI population. Oseltamivir significantly reduced the time to normal activities in the ITT population, but not in the ITTI population (Tables 40 and 41). There was no statistically significant heterogeneity across studies in either of the analyses.

Complications

Three trials with results combined reported complications in an elderly ITTI population (Table 42). 73 The numbers of patients and events were small, and there were no significant differences between oseltamivir and placebo.

Adverse events

No data were available for adverse events associated with oseltamivir in the elderly.

Symptoms

Thirteen trials evaluated the impact of zanamivir on the duration of symptoms and the time to return to normal activity. Of these, five recruited a mixed population (for which symptoms data for healthy and at-risk adults were available separately),46,75,79,91–97 three healthy adults,76,87–89 two those at risk,98,99,114,115 two children,100,101,103 and one the elderly. 88,89 The follow-up period was 5 days in two trials,46,99 6 days in one trial,103 14 days in one trial,93–95 21 days in two trials,75,87 28 days in five trials,88,89,91,92,96–98,100,101,114,115 between 21 and 28 days in one trial76 and 29 days with a follow-up telephone call at 56 days in one trial. 105,106 Four of the trials had a dropout rate greater than 5% but not greater than 10%. 75,76,91,92,98,114 Figures 3–6 illustrate the results for all trials of zanamivir versus placebo, grouped according to the participant subgroups described in previous sections.

FIGURE 3.

Time to the alleviation of all symptoms for zanamivir and placebo in the overall ITT population.

FIGURE 4.

Time to the alleviation of all symptoms for zanamivir and placebo in the overall ITTI population.

FIGURE 5.

Time to return to normal activity for zanamivir and placebo in the overall ITT population.

FIGURE 6.

Time to return to normal activity for zanamivir and placebo in the overall ITTI population.

For time to alleviation of symptoms, results in all patient subgroups favour the use of zanamivir and are statistically significant in both the ITT and the ITTI analyses. As might be expected, the size of treatment effects was larger in the ITTI analyses, reducing the median time to alleviation of symptoms by about 1 day in healthy adults (WMD –0.96; 95% CI –1.38 to –0.54) and healthy children (WMD –1.00; 95% CI –1.59 to –0.41), and by about 2 days in the at-risk population (WMD –1.83; 95% CI –2.81 to –0.86).

There is no overall statistically significant heterogeneity in either the ITT or the ITTI population. Therefore, although the focus of this review is on the individual subgroups that populate the decision model, on this basis it is not unreasonable to pool results across all trials to obtain overall ‘average’ estimates, with the use of zanamivir reducing the median time to alleviation of symptoms by 1 day (WMD –1.07; 95% CI –1.39 to –0.74) in the ITTI population.

A similar pattern of results is observed for time to return to normal activity where all subgroups in both the ITT and the ITTI populations favour the use of zanamivir, reducing the median time to return to normal activity by about 0.5 days in healthy adults and by about 2 days in at-risk adults with confirmed influenza. However, none of these differences were statistically significant. Statistically significant heterogeneity was observed within the healthy adult ITT and at-risk ITTI groups, and also for the overall population in both the ITT and the ITTI analyses.

There was no measure of variance available for two studies,75,105,106 therefore these could not be included in the pooled analyses. One trial reported a reduction in the median time to alleviation of symptoms of 1 day in the ITT population and 1.5 days in the ITTI population. The other reported a reduction of 2.75 days in the median time to return to normal activity in the ITTI population. These are therefore broadly consistent with the results of the trials included in the analyses.

Very few data were available on time to resolution of fever, which was reported in only four trials. Only one of these trials, which reported a statistically significant reduction with zanamivir in the ITT population, provided a measure of variance. Therefore, pooled analyses could not be done (Table 43). 99

Complications

Of the 13 trials of zanamivir reporting complications, three recruited healthy adults,76,87–89 two children,100–104 one patients with COAD or asthma,98,114,115 one the elderly,105,106 one a general at-risk population,99 and five a mixed population of people aged 12 or 13 years and older including those with comorbid conditions. 74,75,79,91–97

The data for the incidence of complications were sparse, and are presented in Figures 7 and 8. The incidence of overall complications and the use of antibiotics showed a statistically significant reduction in both the ITT and the ITTI analyses for those treated with zanamivir. In the ITTI population, the OR for overall complications was 0.77 (95% CI 0.65 to 0.92). However, trials differed in which events were included in the composite of overall complications, and it is unclear whether this is a meaningful clinical outcome. Antibiotic use was significantly reduced in both the ITT (OR 0.76; 95% CI 0.62 to 0.94) and the ITTI (OR 0.79; 95% CI 0.63 to 0.99) populations.

FIGURE 7.

Complication rates associated with zanamivir and placebo in the overall ITT population.

FIGURE 8.

Complication rates associated with zanamivir and placebo in the overall ITTI population.

There was little information on bronchitis in the ITTI populations, but in the ITT population in which there were more data, zanamivir appeared to be associated with a statistically significant reduction in the incidence of bronchitis (OR 0.62; 95% CI 0.4 to 0.90).

Numbers of events were low and there was no clear evidence of differences in the incidence of pneumonia, GP consultations or hospitalisations. Statistically significant heterogeneity was observed only in the analysis of GP consultations in the ITTI population. This was based on only two studies, one of which, as discussed previously, was a study of young Finnish soldiers reporting a particularly high rate of consultations, which was unlikely to be representative of a civilian population. 88–90

There were no deaths in the seven trials that reported on mortality. 79,87–89,91–95,98,99,103,114,115

Adverse events

Of the 13 trials of zanamivir reporting adverse events in ITT populations, four recruited healthy adults,76,86–89 two children,100–104 one patients with COAD or asthma,98,114,115 one the elderly,105,106 one a general at-risk population99 and four a mixed population of people aged 12 or 13 years and older that included those with comorbid conditions. 75,79,91–97

Across all trials, treatment with zanamivir significantly reduced the incidence of overall adverse events (OR 0.85; 95% CI 0.75 to 0.96). There was no evidence of a difference in the incidence of drug-related adverse events between zanamivir and placebo. Even when pooled across all studies, there were very few serious adverse events and there was no evidence of a difference between zanamivir and placebo (Figure 9). There was no statistically significant heterogeneity in any of the analyses.

FIGURE 9.

Adverse event rates associated with zanamivir and placebo in the overall ITT population.

Summary for the effectiveness of zanamivir

The strength of evidence for the effectiveness of zanamivir in alleviating symptoms was greatest in the healthy adult population with analyses based on between 1811 and 3005 individuals. In those with confirmed influenza, reductions of approximately 1 day were seen in the median time to alleviation of symptoms and approximately 0.5 days in the median time taken to return to normal activity. The evidence base was more limited for the subgroups of healthy children and at-risk populations. Children were studied in only two trials (for one of these, data for healthy children and at-risk children were available separately100) and analyses were based on between 324 and 701 individuals. Results for healthy children showed similar reductions in time to alleviation of symptoms and time to return to normal activity as for healthy adults. Analyses for the at-risk group, which comprised adults and children with comorbid conditions and the elderly, were based on between 530 and 1136 individuals, and there is more uncertainty around the results for this group. In at-risk individuals with confirmed influenza, the results suggest that zanamivir reduces the median time to the alleviation of symptoms and the time to return to normal activity by approximately 2 days. Data for most individual complications were sparse and results largely inconclusive. The evidence is strongest for overall complications and use of antibiotics, which both show reduced incidence associated with zanamivir.

Symptoms

Fourteen trials evaluated the impact of oseltamivir on the duration of symptoms and the time to normal activity. Of these, six recruited healthy adults,46,65,67–69,80,82,83,107 one a mixed adult population,57 two those at risk,46,73 two children77,110 and three the elderly. 46,73 The follow-up period was 10 days in one trial,57 21 days in eleven,46,65,67–69,73,80,82,83,107 and 28 days in two trials. 77,110 The dropout rate was unclear in eight trials46,57,65,73,77 and was between 5% and 10% in two trials. 67,81,83 Figures 10–13 illustrate the results for all trials of oseltamivir versus placebo, grouped according to the participant subgroups described in previous sections together with a further ‘mixed’ subgroup which includes a large trial that could not be separated into subgroups. 81

FIGURE 10.

Time to the alleviation of all symptoms for oseltamivir and placebo in the overall ITT population.

FIGURE 11.

Time to the alleviation of all symptoms for oseltamivir and placebo in the overall ITTI population.

FIGURE 12.

Time to the return to normal activity for oseltamivir and placebo in the overall ITT population.

FIGURE 13.

Time to the return to normal activity for oseltamivir and placebo in the overall ITTI population.

For time to the alleviation of symptoms, results in each patient subgroup favoured the use of oseltamivir in both the ITT and the ITTI analyses. These were statistically significant for healthy adults, for healthy children and for the trial with a mix of participant types (data only for ITT). For the at-risk group, results were in the same direction and comparable with the other subgroups, but there was more uncertainty around these findings. Treatment effects were generally greater in the analyses of individuals with confirmed influenza, in which oseltamivir reduced the median time to alleviation of symptoms by about 1 day in healthy adults (WMD –22.19 hours; 95% CI –37.32 to –7.07), by about 1.5 days in healthy children (WMD –36.00 hours; 95% CI –53.51 to –18.49) and by approximately 0.5 days in the at-risk group (WMD –14.04 hours; 95% CI –36.34 to 8.26). With the exception of the ITT analyses of healthy adults, there was no statistically significant heterogeneity within any of the subgroups. There was no overall statistically significant heterogeneity across all trials in either the ITT or the ITTI population. Therefore, although the focus of this review is around the individual subgroups that populate the decision model, on this basis it is not unreasonable to pool results across all trials to obtain the overall ‘average’ estimate that the use of oseltamivir reduces the time to alleviation of symptoms by about 1 day (WMD –22.75 hours; 95% CI –33.39 to 12.11) in individuals with confirmed influenza.

A similar pattern of results is observed for time to return to normal activity where all subgroups in both the ITT and the ITTI populations favour the use of oseltamivir. Results are statistically significant in all groups except the at-risk subgroup in the ITTI analyses. In individuals with confirmed influenza, oseltamivir reduced the median time to return to normal activity by approximately 2.5 days (WMD –63.17 hours; 95% CI –99.08 to –27.22) in healthy adults, approximately 1 day in the at-risk subgroup (WMD –19.20 hours; 95% CI –41.42 to 3.01) and approximately 2 days (WMD –44.57 hours; 95% CI –63.75 to –25.39) in the trial of healthy children. There was no consistency across subgroups as to whether the size of reduction was greater in the ITT or ITTI analysis. Again, there is no overall statistically significant heterogeneity in the results within any of the subgroups or over all trials in either the ITT or the ITTI analyses. Pooling results across all trials gives an overall estimate that ‘on average’ the use of oseltamivir reduces the median time to return to normal activity by about 1.5 days (WMD –36.31 hours; 95% CI –48.44 to –24.17) in individuals with confirmed influenza.

Some studies did not provide a measure of variance, and could therefore not be included in the meta-analyses (Table 44), but results are broadly consistent with the trials that could be analysed.

Figures 14 and 15 illustrate the data that were available on time to alleviation of fever. As data were given for only a small number of trials, and primarily in healthy or mixed populations, they are not shown by subgroup. All trials show a reduction in the time to alleviation of fever associated with oseltamivir and, with the exception of one small trial, individually show statistically significant results. There is no statistically significant heterogeneity. Overall, oseltamivir reduced the median time to alleviation of fever by about 1 day in the ITTI population (WMD –24.41 hours, 95% CI –31.64 to –17.17).

FIGURE 14.

Time to the alleviation of fever for oseltamivir and placebo in the overall ITT population.

FIGURE 15.

Time to the alleviation of fever for oseltamivir and placebo in the overall ITTI population.

Complications

Of the 16 trials of oseltamivir reporting complications, two were in mixed populations,57,81 two in children,77,78,110 three in a general at-risk population,71–73,108 three in the elderly,73 and six in healthy adults. 46,65,66,68,69,83,107 Three trials were conducted in China66,69 or Japan,68 with two published in their respective languages. 66,68 One trial had a dropout rate greater than 5% but not greater than 10%. 82,83

As for zanamivir, data on complications were sparse, and results are presented for all trials combined rather than split into different patient subgroups (Figures 16 and 17). Very little data were available for the ITT analysis; for example, across all trials a total of only 46 overall complications were reported. There is therefore considerable uncertainty in these results and there was no clear evidence of differences in the incidence of bronchitis, pneumonia, GP consultation, hospitalisation or overall complications.

FIGURE 16.

Complication rates associated with oseltamivir and placebo in the overall ITT population.

FIGURE 17.

Complication rates associated with oseltamivir and placebo in the overall ITTI population.

More data were available for the ITTI analyses in which the incidence of the use of antibiotics showed a statistically significant reduction for those treated with oseltamivir with an OR of 0.62 (95% CI 0.46 to 0.83).

With the exception of the ITT analysis of antibiotic use and the ITTI analysis of overall complications, there was no statistically significant heterogeneity observed in the results when analysed in this way.

Of the nine trials that reported on mortality,65,69,73,78,80,83,85,107,109,110 a single death was reported in the placebo arm of a trial in an at-risk population; it was unclear whether this was related to influenza. 73

Adverse events

Of the nine oseltamivir trials that reported adverse events, six were in healthy adults,46,66–69,80,83,107 (of which four were conducted in China66,67,69 or Japan68), two in children77,78,110 and one in a general at-risk population. 71,72 Two trials had a dropout rate of greater than 5% but not greater than 10%. 67,82,83

Analyses of adverse event data are shown in Figures 18 and 19. Data were limited and there was no evidence of a difference in the incidence of overall, serious or drug-related adverse effects in either the ITT or the ITTI analyses. There was no statistically significant heterogeneity in any of the pooled analyses.

FIGURE 18.

Adverse event rates associated with oseltamivir and placebo in the overall ITT population.

FIGURE 19.

Adverse event rates associated with oseltamivir and placebo in the overall ITTI population.

Summary for the effectiveness of oseltamivir

Although analyses are based on similar numbers of individuals (between 609 and 1410 in the healthy adult subgroup and between 907 and 1472 in the at-risk subgroup), the evidence for the effectiveness of oseltamivir in alleviating symptoms was stronger in healthy adults and children. Results in healthy adults and children showed statistically significant reductions in the median time to the alleviation of symptoms and the time to return to normal activity, in both the ITT and the ITTI analyses. For healthy adults with confirmed influenza, reductions of approximately 1 day were seen in median duration of symptoms and approximately 2.5 days in the median time taken to return to normal activity. Healthy children were studied in only one trial, which showed a reduction in median time to alleviation of symptoms of 1.5 days, and median time to return to normal activity of around 2 days. Results for at-risk individuals were broadly similar, but there was more uncertainty around these results, and there was no clear evidence that oseltamivir reduced median time to alleviation of symptoms or return to normal activity. In at-risk individuals with confirmed influenza, the median time to alleviation of symptoms was reduced by about half a day and the median time to return to normal activity reduced by about 1 day (reduction in the median time to return to normal activity in the ITT analysis was approximately 2.5 days and of borderline statistical significance). There was little statistical heterogeneity observed in any of these analyses. Data for most individual complications were sparse, and results largely inconclusive. The evidence is strongest for the use of antibiotics in the ITTI population, which showed reduced incidence associated with oseltamivir use.

Key findings

This review showed that treatment with zanamivir or oseltamivir, when compared with placebo, generally reduced the median duration of symptoms and median time to return to normal activity.

In healthy adults, zanamivir reduced the median duration of symptoms by between approximately 0.5 and 1 days, and oseltamivir by between 0.5 and 1 days. The median reduction in the time taken to return to normal activity was about 0.5 days with zanamivir and approximately 1–2.5 days with oseltamivir. Although these results were statistically significant, the absolute effects were small and their clinical significance questionable, particularly in an otherwise healthy population. As might be expected, estimates derived from the analyses of those with confirmed influenza generally produced greater effect sizes than those derived from all randomised patients.

For the at-risk subgroups, estimates of difference in symptom duration often failed to reach statistical significance, although the direction of effect remained in favour of the NI treatments. Treatment reduced the median duration of symptoms in at-risk populations by approximately 1–3.75 days (ITTI high-risk children) with zanamivir, and by 0.5–1 days (elderly ITTI population) with oseltamivir. A similar pattern was seen in the time taken to return to normal activity, with the median time being reduced by between 1 and 2.5 days (ITTI high-risk children) with zanamivir and by 0.5–4 days (elderly ITT population) with oseltamivir in at-risk populations. These effect sizes were generally larger, and potentially more clinically significant, than those seen in healthy adults. However, there was greater uncertainty around these results. Very limited data were available for the at-risk subgroups of children and the elderly. The ranges for the overall at-risk population may therefore be more representative of the response in these subgroups. In the overall at-risk population, treatment reduced the median duration of symptoms by approximately 1–2 days with zanamivir, and by 0.5–0.75 days with oseltamivir, and the median time to return to normal activity by between 1 and 2 days with zanamivir and by 0.75–2.5 (data for at-risk adults only) days with oseltamivir.

When data were available for adverse events and complication rates, there was little overall difference associated with the use of either zanamivir or oseltamivir when compared individually with placebo. However, data were reported for few trials, studies were not designed to detect changes in these outcomes, and the numbers of events were generally very small. There is therefore considerable remaining uncertainty in all subgroups with respect to these outcomes. The majority of available data were for ‘overall complications’ and antibiotic use. However, the definition of overall complications varied considerably between studies, and the clinical value of this composite outcome is unclear. The most consistent data and strongest evidence related to antibiotic use. Both zanamivir and oseltamivir appeared to reduce antibiotic use, which could be regarded as a surrogate measure for reduced incidence of bacterial infections.

Comparison with the review by Turner et al.46

The results reported here expand on those reported by Turner et al. ,46 which helped inform the current NICE guidelines (TA58). This review adds 14 trials to those used by Turner, six in zanamivir (three in healthy adults, one in the elderly, one in at-risk adults and one in children which included a minority of at-risk participants) and eight in oseltamivir (four in healthy adults, one in an at-risk population of undefined age, one in at-risk children and two in adult populations which included both healthy and at-risk individuals). It additionally analyses incidence of complications and adverse events. This review also excluded one trial (NA130010)84 that was primarily of prophylaxis and found to be ineligible here, but had been included in the Turner et al. 46 review. Despite the inclusion of additional studies, the results for median time to resolution of symptoms and median time to return to normal activity remain similar to those reported by Turner. The results of both reviews are summarised in Tables 45 and 46. For ease of comparison, the results for oseltamivir have been converted from hours to days.

Zanamivir

When comparing the results of the two reviews, the new data for zanamivir resulted in a slightly smaller pooled effect estimate for the time to alleviation of symptoms in healthy adults; nevertheless, this remained statistically significant in favour of zanamivir for both the ITT and the ITTI populations, and the conclusion drawn stayed the same. Likewise, there was a slightly reduced pooled effect estimate for the impact of zanamivir on the time to return to normal activities in the healthy adult ITTI population such that the confidence interval now includes zero effect and there is no longer a statistically significant reduction with zanamivir. However, the overall direction of effect was the same, and the difference in the actual pooled estimates between the two reviews was very small.

The overall at-risk group in the current review included children with comorbid conditions. Therefore, in order to be able to make a direct comparison with the previous report, we also analysed data for at-risk children and at-risk adults separately. There were no differences in effectiveness in either the ITT or the ITTI populations for time to alleviation of symptoms between the two reviews. There were no new data relating to high-risk children. The current review included approximately twice as many patients in the analyses of time to normal activity for at-risk adults, making these analyses less dominated by one study, NAIB2007,46 that reported a zero effect with very narrow CIs. As a result, the WMD was increased in favour of zanamivir compared with the previous report, for both the ITT and the ITTI populations, although there remained no statistically significant difference when zanamivir was compared with placebo.

Oseltamivir

The new data available for oseltamivir in this review resulted in smaller effect estimates for the time to alleviation of symptoms in otherwise healthy adults,67–69 and the results were statistically significant in both reviews. There were no new trials included in this analysis; the increase in the effect size in the ITTI population appears to result from the use of N rather than r, as the studies included in this particular analysis had high numbers of patients who had not returned to normal activity (the number censored is unclear); for the studies combined, the rate was substantially higher in patients receiving placebo than in those receiving oseltamivir (35% and 20% respectively).

The overall at-risk group in the current review included children with comorbid conditions. Therefore, in order to be able to make a direct comparison with the previous report, we also analysed data for at-risk children and at-risk adults separately. There is a difference between the two reviews in the results for the time to the alleviation of symptoms in at-risk ITTI adults, despite data from the same studies being used. The reported WMDs were –10.91 hours (0.45 days) and –20.09 hours (0.84 days) from the previous and current review respectively. We repeated the analysis using the data we extracted, SDs calculated from the SEs reported in the previous review, and r instead of N. All these analyses resulted in a WMD of approximately –20 hours. Despite this discrepancy, none of these analyses reported a significant difference between oseltamivir and placebo, and therefore the conclusions drawn from these data would not be altered. No differences in effectiveness were seen in either ITT or ITTI populations for time to return to normal activity.

Generally, zanamivir and oseltamivir reduced the median duration of symptoms and median time to return to normal activity when compared with placebo, in both the current and the Turner et al. 46 review, and in terms of symptom relief the conclusions drawn from the two reviews are similar. There was no comparison of complication or adverse event data undertaken by Turner et al.

Overview

The study was designed to compare alternative options for the treatment and prophylaxis of influenza in four separate patient groups: (1) otherwise healthy adults, (2) high-risk adults, (3) children, and (4) the elderly in residential care. The treatment model compared four alternative strategies in each of the patient groups: usual care (i.e. no use of antivirals) and treatment with antivirals (amantadine, zanamivir, oseltamivir). The analysis adopted an NHS perspective, although the impact of incorporating productivity costs was addressed in the sensitivity analysis. A range of scenarios was undertaken to reflect alternative assumptions for key inputs. In addition, probabilistic sensitivity analysis using second-order Monte Carlo simulation was used to reflect parameter uncertainty in the model. Results were expressed as a series of pair-wise estimates in terms of the incremental cost per QALY gained for each antiviral relative to usual care. A lifetime time horizon was adopted, although all influenza-related events were assumed to take place within a single season, and a lifetime horizon was employed solely to capture QALYs lost due to influenza mortality.

The treatment model took the form of a decision tree, outlined in Figure 20. The model considered all patients in each group with ILI during periods when influenza was reported by surveillance schemes to be ‘circulating in the community’ (i.e. when the rate of ILI referrals exceeded 50 per 100,000 of the population, based on the definition in use at the time of the study). Progression through the model was dependent upon a series of separate considerations reflecting a broad range of policy, behavioural and epidemiological issues. These included: (1) the policy decision (e.g. whether antivirals were routinely available to health-care providers and the potential impact this might have on subsequent consultation rates and other model parameters such as the probability that an ILI is influenza); (2) the disease status of the patient (i.e. whether patient had influenza or other ILI; the proportion of influenza cases that were due to influenza type A or B); (3) behavioural aspects related to the individual with an ILI (i.e. whether they present to a health-care provider, and timing of presentation) and the health-care provider themselves (e.g. compliance with recommendations on the use of antivirals, inappropriate prescription of antivirals and antibiotics, etc.); and (4) the outcome of the illness and the impact of alternative treatments (in terms of symptomatic relief and also subsequent influenza-related complications comprising both minor and major events and potentially mortality).

FIGURE 20.

Schematic of model (Turner et al. )46

The antivirals (amantadine, zanamivir or oseltamivir) were assumed to provide additional benefits relative to usual care alone only for those individuals who: (1) have influenza A or B (only influenza A in the case of amantadine), rather than an alternative ILI; and (2) present to a GP within 48 hours of the onset of symptoms (with some exceptions, see below).

In each scenario, it was assumed that the patient would either recover directly from treatment without experiencing complications (in which case the benefits of antivirals were based on QoL gains associated with symptomatic relief alone) or encounter complications (either minor or severe) and subsequently recover (non-fatal) or not (fatal). Minor complications were assumed to result in additional consultations with a GP only (and potentially additional use of antibiotics), whereas more severe complications resulted in hospitalisation and potentially mortality.

In the base-case analysis, the authors assumed that antivirals conferred additional benefits, relative to usual care, only in terms of symptomatic relief and reductions in minor complications requiring a revisit to a GP (and subsequent use of antibiotics). Hence, the base-case analysis assumed that the antivirals had no additional impact on major complications associated with influenza requiring hospitalisation or leading to premature mortality. The sensitivity analysis considered an alternative scenario (termed the ‘extrapolated’ model) in which the use of NIs only (oseltamivir and zanamivir) also resulted in reductions in major complications requiring hospitalisation and an associated reduction in influenza-related mortality. The adverse events associated with the use of NIs were considered to be sufficiently minor not to impact the model and hence were not included. However, an additional probability of adverse events was assigned to amantadine due to the more serious nature of these events. Issues of resistance to the use of antivirals were not considered.

The model assumed that all patients with ILI presenting to a GP within 48 hours of the onset of symptoms would be prescribed an antiviral in the relevant treatment strategies. However, due to difficulties in accurately gauging the timing of symptom onset, a small proportion of these were assumed to actually have experienced symptom duration in excess of 48 hours (i.e. in those presenting within the last 12 hours of the 48-hour period and with ‘insidious onset of influenza symptoms’ as opposed to an abrupt onset). It was argued that such patients ‘would be presenting within 48 hours of apparently developing symptoms but would actually be outside 48 hours from onset’. For this proportion of individuals, it was assumed that, although antivirals would be prescribed, they would confer no additional benefits compared to usual care alone. Meanwhile, it was assumed that those presenting to a GP after 48 hours would not be prescribed any antiviral treatment except in cases in which the patient falsely reports the time of onset of symptoms or cannot recollect the time of onset accurately (more likely when the onset is ‘insidious’ rather than ‘abrupt’), or the GP inappropriately prescribes an antiviral treatment as a means of satisfying the patient’s wish to leave with a prescription.

Summary of effectiveness data

Estimates of the proportion of all ILIs that is influenza in individuals presenting to a health-care provider (47.5% in the children group, 46% in all other groups) were based on national surveillance data over a 10-year period (1992–3 to 2000–1) from the RCGP obtained through personal communication. Estimates for the children group were based on data for the under 15 years age group, all other groups were derived from the population aged over 15 years for periods in which the threshold for ‘circulating in the community’ had been reached. For amantadine treatment it was necessary to distinguish between influenza A and B; the probabilities that influenza is type A (68.4% for healthy adults, 79.9% for high-risk adults and the elderly, 70.5% for children) were estimated from surveillance data for the period 1992–2001. The distinction between influenza A and B was only important in estimating the proportion of individuals in whom amatandine was assumed to be effective. No distinction was made in terms of subsequent differences in prognosis, costs or QoL for the different influenza types, and the NIs were assumed to be equally effective in both types.

The probability that a patient with ILI presents to a GP (28.2% for healthy adults, 32.5% for high-risk adults and the elderly, 15.5% for children) was derived from estimates of the size of the population, excess influenza consultations and attack rates obtained from a variety of separate sources. The probability that this presentation occurs within 48 hours of the onset of symptoms (17.8% for healthy adults, 12.2% for high-risk adults and the elderly, 57.2% for children) was derived from Ross et al. ,118 adjusted according to the proportion of persons experiencing abrupt onset of influenza using data derived from a meta-analysis of studies reporting rates of abrupt onset in subjects with influenza A. Estimates of the proportion of those presenting within 48 hours of onset of symptoms who do not receive a drug treatment (4.8% for all groups) and the proportion of those presenting after 48 hours who do receive a drug treatment (2.8% for healthy adults, 1.1% for high-risk adults and the elderly, 12.1% for children) were also derived from Ross et al.

The beneficial health effects of drug treatments were assumed to arise from improved QoL resulting from shorter duration of illness and, in the sensitivity analysis, a reduction in probability of severe complications and associated mortality with the NIs. The probability of severe complications was proxied by the probability of pneumonia and, in the children group only, otitis media. For usual care and amantadine treatment, the probability of pneumonia (1.3% for healthy adults and children, 2.87% for high-risk adults) was based on the pooled control arms of the oseltamivir and zanamivir trials, while for oseltamivir (0.2% for healthy adults and children, 2.19% for high-risk adults) and zanamivir (0.4% for healthy adults and children, 2.0% for high-risk adults), it was based on the treatment arms of the respective trials. Using data from Roche trials, the probability of otitis media complications in children was assumed to be 21.3% for usual care and 12.0% for oseltamivir treatment; for zanamivir or amantadine treatment the probability associated with usual care was adopted due to lack of data.

Hospitalisations were only considered as part of the ‘extrapolated’ model. The probability of hospitalisation for usual care was based on UK data reporting excess hospital admissions, combined with estimates of influenza attack rates and population estimates to derive the estimated probability of hospital admission. The relative risk of pneumonia was applied to these estimates to obtain the hospitalisation rates associated with each NI. The probability of influenza-related mortality was derived from a large UK observational study15 for each patient group, and was assumed to be constant (0.039% for healthy adults, 0.7% for high-risk adults and the elderly, 0.0048% for children) across all treatment strategies in the base case. In the sensitivity analysis, it was assumed that the estimated relative risk reductions in pneumonia associated with oseltamivir or zanamivir treatment also reduced mortality by the same proportion.

Summary of resource utilisation and cost data

The mean costs of GP visits (£21.38 for healthy adults and children, £27.43 for high-risk adults were derived from UK published sources, weighted by the proportion of patients receiving home visits. 118,149 The cost of zanamivir (£24.98) was taken from published sources, the expected cost of oseltamivir (£19.16) was obtained from Roche, while the cost of amantadine (£3.38) was taken from the 2001 Alliance Pharmaceuticals submission to NICE. The cost of antibiotics (£4.05) was derived from Davey et al. 150 The drug costs were increased to take into account pharmacy prescribing fees and container allowances.

The cost of inpatient stays was based on estimates of mean length of stay and mean cost per diem. The mean duration of stay was assumed to be 2.3 days for children, 11.9 days for adults and 15 days for those older than 65 years based on national hospital statistics. It was assumed that 4.9% of adults admitted with pneumonia received assisted ventilation in an intensive care unit; the length of intensive therapy unit (ITU) stay and mechanical ventilation were 28 days and 21.5 days respectively.

The unit costs of hospitalisation were calculated with reference to the National Schedule for Reference Costs: the mean cost per day for adults was taken from Healthcare Resource Group (HRG) code D14 (£189), HRG code D13 (£164) for the elderly and HRG code P04 (£309) for children. The mean cost per diem for an intensive care unit was assumed to be £1193.

Summary of quality of life data

The estimation of QoL benefits was expressed in terms of mean QALYs. The estimates of QALYs comprised a number of separate elements: (1) symptoms – based on the illness duration; (2) minor complications; (3) adverse events (amantadine only); (4) major complications; and (5) premature mortality and associated loss of remaining quality-adjusted life expectancy (QALE). The base-case analysis evaluated the differential QALYs between the antiviral treatments and usual care in terms of elements (1), (2) and (3). As part of the sensitivity analysis, termed the ‘extrapolated model’, differential QALYs were estimated for all separate elements; only the NIs were assumed to result in differential QALYs, compared with usual care, for elements (4) and (5).

The QALY estimates representing symptomatic benefits were derived from the mean duration of the illness for each treatment, and the associated QoL during this period. The base-case analysis derived estimates based on the time to resolution of symptoms; sensitivity analysis explored the robustness of the results to the inclusion of productivity costs based on time to return to normal activity. Given that estimates of the symptom duration were mainly reported in terms of median estimates in the clinical trials, a parametric survival analysis approach was employed in order to estimate the mean values required for the economic model. The authors assumed that the survival curve for time to recovery followed an exponential distribution (i.e. assuming a constant hazard). Median estimates were thus extrapolated, assuming an exponential distribution, in order to obtain the required mean estimates, converted into days, for input into the model. Estimates for the length of influenza illness for patients receiving usual care and oseltamivir and zanamivir were obtained directly from the relevant trials, while the length of illness for those receiving amantadine was based on the observed length of fever in the Cochrane review of amantadine (modified to include only those trials using doses of 100 mg per day, extrapolated to give length of influenza illness using meta-regression results based on the observed lengths of fever and length of influenza illness observed in the oseltamivir trials). This alternative approach to estimating the duration of illness for amantadine was due to the lack of comparable outcomes to data reported for the NIs (i.e. outcomes were expressed in terms of the duration of fever as opposed to resolution of symptoms or return to normal activities).

QALYs based on symptoms estimated for each of the four treatment alternatives were derived from the mean duration of illness with respect to data reported for influenza positive patient health valuations conducted alongside a number of randomised trials of oseltamivir. In these trials, a 10-point Likert scale was completed daily over a 21-day period; this was recalibrated to mean VAS scores from the Measurement and Valuation of Health (MVH) study, which were then converted to time trade-off (TTO) equivalent utility scores using an equation. 151 Separate scores were used for healthy and at-risk populations. The TTO values for those receiving placebo and oseltamivir were used directly in the model as the QoL values for the usual care and oseltamivir treatment strategies respectively. To derive the QoL values for the zanamivir and the amantadine treatment strategies, the QoL values for the oseltamivir treatment strategy were adjusted according to the respective length of influenza illness associated with zanamivir and amantadine treatment.

The QoL associated with minor complications was assumed to be reflected in the patient health valuations given alongside the trials, while the QoL associated with severe complications was estimated separately based on WHO disability weights for lower respiratory conditions. 152 These were 0.724 in the healthy adult population and 0.72 in the other populations. In the children’s model, the QoL associated with otitis media was taken to be 0.977 from the same source. The valuation of adverse events resulting from amantadine (0.81 for healthy adults and children, 0.74 for high-risk adults) was based on an assumed EQ-5D status, and the length of illness from amantadine adverse events was assumed to be 5 days.

QALYs lost due to premature mortality were estimated according to the QALE for the separate patient groups based on estimation of the mean age of influenza deaths, remaining life expectancy (based on standard lifetables), the discount rate (1.5%) and quality adjustment according to age (based on the mean EQ-5D values derived from UK population norms). This approach assumed that individuals who die from influenza would have otherwise had a normal life expectancy with an associated QoL similar to that of the general population. The authors noted that this approach may overstate the benefit of preventing influenza deaths, as those who die of influenza may be a less healthy subgroup. As previously stated, the relative risk reduction reported for pneumonia for the separate NIs was employed as proxy for the relative risk reduction for mortality.

Summary of cost-effectiveness data

Sensitivity analysis

In the healthy adult sensitivity analyses, ‘extrapolating’ the base case to include the impact of treatment on hospitalisations and deaths substantially improves the ICERs of oseltamivir and zanamivir for each of the subgroups. This effect is most evident in the high-risk and residential care elderly populations in which the baseline risk of hospitalisation and mortality is highest (and hence the absolute gain associated with treatment is potentially greater). The ICER for amantadine remains unaltered from the base case as no additional effect was assumed, relative to usual care, for hospitalisations and mortality. The ICERs for each of the antivirals across the subgroups was below £20,000 per QALY. In the residential care model, both oseltamivir and zanamivir dominated (i.e. less costly and more effective) usual care. The results are summarised in Table 57.

A number of additional scenarios (presented as one-way and two-way sensitivity analyses) were explored to examine the robustness of the model assumptions. The results indicated that the results were particularly sensitive to the probability that patients presenting with ILI have influenza. The results of the one-way sensitivity analyses are summarised in Table 58. As expected, lower probabilities (i.e. akin to a higher false positive rate) resulted in marked deteriorations in the ICERs of antiviral treatment and vice versa for higher probabilities.