Zanamivir in the Prevention of Influenza Among Healthy Adults: A Randomized Controlled Trial

Context The neuraminidase inhibitor zanamivir, a sialic acid analog administered directly to the respiratory tract, has been demonstrated in clinical studies to be effective in treatment of type A and B influenza. It has also been shown to prevent influenza infection and disease in an experimental model.

Objective To examine the efficacy of zanamivir, administered once daily, in the prevention of influenza infection and disease.

Design Double-blind, randomized, placebo-controlled trial.

Setting Two midwestern university communities.

Participants A total of 1107 healthy adults (mean age [range], 29 [18-69] years) were recruited in November 1997, before the influenza season.

Intervention At the start of the influenza outbreak, 554 subjects were randomized to receive placebo and 553 to receive zanamivir. The drug, 10 mg once per day, or identical placebo was administered by oral inhalation for a 4-week period.

Main Outcome Measures Illness occurrence was recorded by participants daily and records were evaluated weekly. Specimens were collected for viral isolation when symptoms were reported within 3 days of illness onset. Infection was also identified by testing paired serum samples for rise in antibody titer against the circulating influenza viruses.

Results Zanamivir was 67% efficacious (95% confidence interval [CI], 39%-83%; P<.001) in preventing laboratory-confirmed clinical influenza meeting the case definition and 84% efficacious (95% CI, 55%-94%; P=.001) in preventing laboratory-confirmed illnesses with fever. All influenza infections occurring during the season, with or without symptoms, were prevented with an efficacy of 31% (95% CI, 4%-50%; P=.03). The nature and incidence of adverse events in the zanamivir group did not differ from placebo. Compliance with the once-daily dosage was high.

Conclusions Zanamivir administered once daily is efficacious and well tolerated in the prevention of influenza for a 4-week period in healthy adults.

Twenty years ago, the antiviral drug amantadine was reported to be 70% efficacious in prevention of symptomatic influenza during an epidemic associated with the reappearance of type A(H1N1) virus.¹ A few years later, amantadine and the related compound rimantadine were demonstrated to be efficacious in preventing illnesses associated with type A(H3N2) and type A(H1N1) viruses.² However, while clearly effective, there has been a long-standing controversy about appropriate use of these drugs dating to the 1960s.^3,4 Concerns raised include the lack of effectiveness against type B viruses, occurrence of adverse effects mainly related to amantadine, and the rapid development of resistance to both agents.^5-7

A new class of antivirals, with a different mechanism of antiviral action, has been designed and developed to provide another approach to the control of influenza.^8,9 The compounds specifically inhibit the action of the viral neuraminidase enzyme, essential for replication of both type A and B viruses. The sites affected are conserved; thus, the antiviral action is independent of antigenic change.^8,9 Zanamivir (GG167), a sialic acid analog, has been demonstrated to be efficacious in clinical trials in shortening the duration and reducing the severity of type A and B influenza illness.¹⁰ Experimental studies have suggested promise in the prevention of influenza.¹¹ We report use of zanamivir administered once daily for a period of 4 weeks during the course of an influenza season in which the predominant circulating virus was A/Sydney/5/97 (H3N2), a variant not contained in the vaccine used in 1997-1998.¹²

The study was a randomized, double-blind, placebo-controlled trial conducted at 2 sites in the United States during the 1997-1998 influenza season. The study was approved by the institutional review boards at both participating universities. Prophylaxis began when an increase in occurrence of influenza was documented in each geographic area based on frequency of influenzalike illnesses and identification of influenza viruses circulating. As a result, randomization took place and active prophylaxis began at the University of Missouri site approximately 1 week before it began at the University of Michigan site. Otherwise, eligibility criteria, drug use, and data collection were identical at the 2 locations.

Zanamivir or an identically packaged placebo, both containing lactose as the principal base, was administered by self-activated inhalation once daily using a Diskhaler device (Glaxo Wellcome, Ware, England). The daily dosage of zanamivir used was 10 mg; the micronized drug powder was estimated to have an aerodynamic mass median diameter of approximately 3.0 µm. The first dose was taken under the supervision of the study staff. Participants returned weekly for study drug resupply, to have diary cards evaluated, and for inspection of the used medication to ensure appropriate use. If the drug or placebo had not been used for 6 days or more during the 28-day period, that individual was considered noncompliant with the protocol.

The study was conducted in the midwestern US university communities of Ann Arbor, Mich, and Columbia, Mo. Participants were mainly students or community volunteers. Persons aged 18 to 64 years were eligible for participation if they did not have chronic conditions that put them in the groups at risk for complications and for which influenza vaccination was recommended. Therefore, individuals with hypertension were eligible. Eligibility was not restricted to unvaccinated participants. Pregnant and lactating women were excluded, as were those unable to return to the study office at the scheduled times.

To ensure that prophylaxis would begin when the influenza outbreak began, participants were recruited during November 1997. At enrollment, the study design was reviewed and informed consent obtained. Blood samples were collected for initial safety evaluations, which included blood counts, electrolyte and liver enzyme determinations, and a metabolic profile. An aliquot of serum was frozen at −70°C for serologic testing.

Surveillance of influenza virus occurrence was conducted in the enrolled study panel, in the surrounding community, and in university-associated student health services. Specimens at the Michigan site were processed for viral isolation using standard techniques at the World Health Organization Collaborating Center for Influenza in the School of Public Health. At the Missouri site, a commercial laboratory was used. In January, the number of specimens increased. After influenza occurrence was detected, the active phase of the study began. Participants were called to the study offices, where exclusions were reevaluated and consent to participate was reviewed. Any subject presenting with a respiratory illness at this visit was excluded. Subjects were sequentially assigned treatment numbers corresponding to a box of disks containing prerandomized drug or identical placebo to maintain the blind. Randomization was stratified for vaccination states in blocks of 10 at each site. All participants were blindly randomized to drug or placebo in a 1:1 ratio and prophylaxis began.

Participants were given a diary card to record twice daily any symptoms occurring during the following week, using a severity scale of 0 to 3 (representing none, mild, moderate, and severe). At the same time, subjects obtained and recorded their temperatures by tympanic methods. Participants returned to the centers, where each 7-day diary card was checked and any other symptoms were sought. Subjects then received a new diary card and a new supply of study drug or placebo. The procedure was repeated during the 28 days of prophylaxis, after which time drug administration ended. At day 35, participants returned for the final visit, at which a second blood specimen was obtained for safety determinations and viral serology.

Throughout the study period, the subjects were asked to report to the study center should they have any respiratory symptoms. At that point, a culture for virus isolation was obtained. Collected serum samples from both sites were treated with receptor-destroying enzyme and tested as pairs in the laboratory at the University of Michigan.¹³ Antigens used were obtained from the Centers for Disease Control and Prevention (CDC) or produced in the laboratory from seed viruses obtained from the same source. The hemagglutination-inhibition tests were carried out using antigens of A/Sydney/5/97 (H3N2), A/Wuhan/359/95 (H3N2), and A/Johannesburg/82/96 (H1N1). A 20% subset of serum samples was tested with B/Guangdong/8/93. A 4-fold rise in titer or appearance of inhibition 2 dilutions higher than the initial dilution, when negative, was considered to represent infection.

Individuals were classified as infected with influenza if they had a rise in titer with 1 or more of the antigens used and/or isolation of an influenza virus during the period of prophylaxis. An influenza-associated illness (laboratory-confirmed clinical influenza) was considered to have occurred in those with documented infection who also had 2 or more of the following signs or symptoms recorded concurrently on 3 or more successive diary card entries: cough, headache, sore throat, myalgia, feverishness, or a temperature of at least 37.8°C. Febrile laboratory-confirmed influenza required that a temperature of at least 37.8°C be recorded in individuals with rise in antibody titer and/or isolation of virus. Total febrile illness required only the occurrence of a temperature of at least 37.8°C.

The primary efficacy end point was the proportion of randomized subjects who, during prophylaxis, developed laboratory-confirmed clinical influenza. The sample size was based on the assumption that the attack rate in the placebo group would be 10% and the effectiveness of zanamivir would be 70%. On this basis, 270 subjects were required in each group to give 90% power at the 2-sided P=.05 level of significance.¹⁴ Because an influenza outbreak was declared at both study sites, the planned sample size was exceeded. The analysis was performed using an exact test^15,16 for stratified 2 × 2 tables. Analyses were conducted on the intent-to-treat population (all randomized subjects, regardless of whether study drug was taken or if the subject completed the planned duration of the study) with stratification by study sites and vaccination status and on the nonvaccinated population (all nonvaccinated, randomized subjects who took at least 1 dose of study medication), stratified by study site. Statistical tests were performed at the 2-sided P=.05 level of significance. Corresponding estimates of the odds ratios (ORs) of laboratory-confirmed clinical influenza were obtained, together with 95% confidence intervals (CIs).¹⁷ Approximate risk ratios (RRs) were also calculated based on stratified analyses, using Mantel-Haenszel estimates with test-based CIs. Efficacy was expressed as 1 − RR.

A total of 1107 individuals were randomized to receive either placebo or zanamivir (Figure 1). As shown in Table 1, there were 554 and 553 subjects in each of those groups, respectively, with mean ages of 28.6 and 29.0 years, with an age range of 18-69 years, reflecting the nature of the pool from which the participants were recruited. (Although the initial age range was 18-64 years, 1 individual older than 64 years was recruited in each group and retained in the analysis.) (Although the initial age range was 18-64 years, 1 individual older than 64 years was recruited in each group and retained in the analysis.) Women predominated, and 14% of each group had received vaccine prior to randomization. Slightly more than half of the population was at the Michigan site.

Reports of the sporadic occurrence of influenza began to appear in both areas in November. Virus was first isolated in December, but illness frequency did not start to increase until January 6, 1998, with the wave of illnesses continuing until February 12, 1998. There were 19 viral isolates at the Michigan site. All were type A(H3N2); with the assistance of the Influenza Branch of the CDC, 15 were further identified: 12 (80%) as A/Sydney/5/97 (H3N2) and 3 (20%) as A/Wuhan/359/95 (H3N2). At the Missouri site, there was only 1 type A(H3N2) influenza isolate, but the frequency of rises in titer among the type A subtypes and variants was similar to that seen in Michigan.

The frequency of illnesses of various characteristics and of infection in the overall randomized, or intent-to-treat, population is shown in Table 2. The primary end point was prevention of laboratory-confirmed clinical influenza meeting the case definition. The number of events observed demonstrated that this was an outbreak of moderate size. The OR of the illness frequencies between study groups was 0.31. As shown, this approximates an RR of 0.33 (95% CI, 0.17-0.61) or an efficacy of 67% (95% CI, 39%-83%). Fever is one of the best predictors of influenza positivity when respiratory illnesses among adults are tested for presence of virus.¹⁸ When those illnesses considered were restricted to those with temperatures of at least 37.8°C that were laboratory-confirmed as influenza, the efficacy was 84% (95% CI, 55%-94%). When all febrile episodes were considered with or without confirmation of influenza etiology, a significant efficacy could still be demonstrated and was 43% (95% CI, 14%-62%). Finally, influenza infections, with or without illness, were prevented with a lower efficacy than symptomatic infections of 31% (95% CI, 4%-50%). This is in agreement with the lower efficacy in the prevention of influenza infection characteristic of amantadine and rimantadine.^1,2

The results were consistent between the 2 sites, with the exact test for homogeneity of ORs between centers demonstrating P=.73. The viruses circulating at the 2 sites also appeared to be similar based on similar serologic data. There were 52 rises in titer detected for A/Sydney/5/97 (H3N2), 50 rises for A/Wuhan/395/95 (H3N2), and 6 for A/Johannesburg/82/96 (H1N1) in serum samples collected from Michigan; results for Missouri were 59, 40, and 5, respectively. Many individuals had rises in titer for both A(H3N2) strains, indicating cross-reactivity. There were no rises in titer detected in specimens from either site for type B virus.

Only 14% of the participants were vaccinated, and the vaccine minimally protected against the A/Sydney/5/97 (H3N2) virus.¹² When the analysis was restricted to unvaccinated persons, the OR for laboratory-confirmed clinical influenza was 0.38 (95% CI, 0.17-0.80), with an estimated RR of 0.40, which translates to an efficacy of 60% (95% CI, 24%-80%; P=.009). For laboratory-confirmed influenza with fever, the efficacy was 81% (95% CI, 46%-96%; P=.004). The number of vaccinated individuals was too small to draw conclusions about the interaction of the antiviral with even a marginally protective vaccine.

Several measures were examined in an attempt to identify any adverse effects associated with use of zanamivir in the 4-week period of administration. Because this was a prophylactic study, any symptoms occurring during the period could possibly be related to drug use and, thus, were recorded. In both placebo and zanamivir groups, some symptoms were reported by 75% of individuals. When examined as a whole or by system, there were no significant differences observed. The systems examined were ear, nose, and throat; endocrine and metabolic; eye; gastrointestinal tract; lower respiratory tract; musculoskeletal; neurologic; reproductive; and non–site specific. Adverse effects thought by the investigators to be potentially drug-related were observed in 27 (5%) of the placebo group and 30 (5%) of the zanamivir group. Potential adverse effects that were considered severe were seen in 1 (<1%) of the placebo recipients and 1 (<1%) of the zanamivir recipients. Withdrawals from the trial for any reason occurred in 17 subjects (3%) from the placebo group and 10 (2%) from the zanamivir group; those thought by the participants to be potentially drug related occurred among 7 (1%) and 4 (<1%) subjects in the 2 groups, respectively. The remaining 10 and 6 withdrawals, respectively, were related to voluntary withdrawal of consent or by loss to follow-up (Figure 1). Thus, adverse events were not associated with drug prophylaxis. The study staff assessed compliance by the number of doses that each subject used from the disks. In the placebo group, 525 (95%) were recorded as taking doses for 23 to 28 or more days and in the zanamivir group, the number was 536 (97%).

There were a number of paradoxes associated with the 1997-1998 influenza year in the United States. The predominant strain was A/Sydney/5/97 (H3N2), a virus significantly different from that in the vaccine. This was the first time in many years that such an event happened, especially since it began almost at the start of the season. Not surprisingly, there was excess mortality documented during 9 consecutive weeks, as well as major community outbreaks in the western states in early winter.¹² However, much of the rest of the country had more modest outbreaks, which sometimes occurred over a long period. Thus, the frequencies in the placebo group of laboratory-confirmed clinical influenza of 6% and of influenza infection of 14% is lower than might have been expected in many influenza seasons. Still, the demonstrated clinical protective efficacy of 67% was consistent between the 2 sites.

When laboratory-confirmed episodes with documented fever were considered, the efficacy estimate rose to 84%. Fever is a major predictor of the ability to isolate influenza virus from respiratory illnesses with cough, and this value may be a better estimate of the ability of zanamivir to prevent typical influenza.¹⁸ The difference in efficacy estimates might also suggest that zanamivir is better in preventing disease—in particular, more severe disease—than infection. Supporting this observation is the lower efficacy of 31% of the drug in preventing all influenza infections, clinical and inapparent. In any event, because asymptomatic infection still occurs in those receiving prophylaxis, antibodies will be produced that could protect individuals against a later infection. Interestingly, the same pattern of a better ability to prevent symptomatic than asymptomatic infection has been previously observed with amantadine and rimantadine.^1,2

Zanamivir is a neuraminidase inhibitor and, thus, has a totally different mechanism of action than amantadine or rimantadine, compounds that are limited in activity to type A influenza. While only type A influenza circulated during the conduct of the current trial, zanamivir's demonstrated activity in therapy of type B infections plus in vitro data suggest that it would be equally efficacious against type B infections.^8,19 This broader spectrum would be an advantage over the current influenza antiviral drugs, especially in years with mixed or overlapping type A and B outbreaks.

Viral resistance to amantadine and rimantadine develops rapidly and can be managed by careful use, but concern regarding this issue has been widely expressed.^4,20,21 Resistance to zanamivir has been sought in specimens resulting from trials in therapy, a situation in which resistant strains would have been routinely identified if amantadine had been used, but none has yet been found. Obviously, continued surveillance for resistance is required as use increases. The 1 instance in which resistance has been demonstrated occurred when the drug was given for a prolonged period to an immunosuppressed individual.²¹ The leading advantage of zanamivir in prophylaxis is its safety and its once-daily use. This study did not demonstrate any excess adverse events vs placebo in the healthy adults who received the drug, and the observed compliance was high. It should be remembered that this drug will be used in practice in other populations and in less-than-ideal settings, where it might function differently; however, initial results in nursing home residents have been encouraging.²²

Antiviral prophylaxis of influenza has been considered an adjunct to vaccination, useful in specific situations, and this will continue to be the case.²³ One of these situations was exhibited in 1997-1998, when a change in the circulating virus limited the efficacy of the vaccine. Another would be when a person in the risk groups recommended for vaccination is found to be unvaccinated by medical personnel after influenza transmission has started. In this case, vaccine could be administered and the drug given and continued for at least the limited period while antibody develops. Also, outside the high-risk group, a decision may sometimes be made between a patient and medical personnel to use drug prophylaxis for a short period of exposure, such as when traveling, or for longer periods, when virus is known to be transmitting. These approaches have been used in the past with amantadine and rimantadine, and they should serve as guidelines for the use of zanamivir, recognizing its advantages over the older compounds.