The cut-off date for all data included in this analysis was 31 January, 2014. A total of 9919 (>99%) patients received ≥1 dose of study medication (Fig. 1). The ITT population for this integrated analysis comprised 7229 patients who were randomized to one of the six key treatment groups: FF/VI 200/25 (n = 956), FF/VI 100/25 (n = 2369), FF 200 (n = 608), FF 100 (n = 2010), VI (ICS) 25 (n = 216), and placebo (n = 1070). The majority of these patients received either FF/VI 100/25 or FF 100. The VI (ICS) 25 group contained the smallest number of patients, as this treatment was only included in two studies. Across the studies, 84% of patients in the key treatment groups completed the study and 16% withdrew prematurely. Lack of efficacy was the most common reason for withdrawal in all groups except for FF/VI 100/25; 23% of patients in the placebo group withdrew citing this reason. In the FF/VI 100/25 group, withdrawal of consent was the most common reason for withdrawal (reported by 4% of patients). Other reasons for withdrawal, such as protocol deviation and AEs, were reported by ≤2% of patients within each treatment group.
Demographic characteristics for the key treatment groups are provided in Table 1. The proportion of males to females and the mean age were similar across these groups. The majority of patients in each treatment group were aged 18–64 years and the majority were Caucasian. The mean duration of asthma ranged from 15 to 18 years.
Cumulative exposure varied widely across the key treatment groups, which is likely to reflect the differing duration of these studies (range of 4–76 weeks). The highest exposure was reported for the FF/VI 100/25 group (1537 patient years; key treatment groups displayed in Table 2). More than half (59%) of the FF/VI 100/25 patients were in studies of ≥24 weeks’ duration. As patients may not receive a long-term placebo, no patient was exposed to placebo for more than 28 weeks. Most (63%) of the patients who were randomized to receive placebo were in studies of <12 weeks’ duration.
The most frequently reported common AEs were headache, nasopharyngitis, and upper respiratory tract infection (key treatment groups displayed in Table 3). The incidence of some events was numerically higher in the FF/VI 100/25 group compared with the placebo group. However, there was no notable difference in exposure-adjusted incidence across treatment groups with the exception of oropharyngeal pain, for which the incidence was numerically greater in the FF/VI 200/25 group than the FF/VI 100/25 group, and back pain, which was reported at a numerically higher incidence in all active treatment groups compared with placebo.
Figure 2a presents the incidence of AEs reported by ≥3% of patients for FF/VI 100/25 compared with placebo (in those studies that included both FF/VI 100/25 and placebo) and shows that the incidence is similar (95% CI for the risk ratio included 1) for nasopharyngitis and upper respiratory tract infection, and favored placebo for headache. When the incidence of any AE reported by ≥3% of patients for FF/VI 100/25 was compared with FF 100, there was no evidence of a significant difference as the 95% CI for the risk ratio includes 1 (Fig. 2b). There is also no evidence of an increase in the incidence of any AEs reported by ≥3% of patients for FF/VI 200/25 compared with FF/VI 100/25 (Fig. 2c).
The incidence of any drug-related AE (based on investigator opinion) was 6% in the FF/VI 200/25, FF/VI 100/25, FF 200, and FF 100 groups, and 2% in the VI (ICS) 25 group, compared with 2% in the placebo group (Table 4). Among the drug-related AESIs, oral candidiasis had the highest incidence in the FF/VI 200/25 group (2%) compared with 0% in the placebo group, <1% in the FF/VI 100/25 and FF 100 groups, and 1% in the FF 200 group. The incidence of other drug-related AESIs was low across all groups [dysphonia: 1% with FF/VI 100/25, <1% with FF/VI 200/25, FF 100, FF 200, and placebo, 0% with VI (ICS) 25; oropharyngeal candidiasis: 1% with FF 200, <1% with FF/VI 200/25, FF/VI 100/25, FF 100, and placebo, 0% with VI (ICS) 25; oropharyngeal pain: <1% with FF/VI 200/25, FF/VI 100/25, FF 200, FF 100, and placebo, 0% with VI (ICS) 25; and bronchitis: <1% with FF/VI 100/25 and FF 100, 0% with FF/VI 200/25, FF 200, VI (ICS) 25, and placebo].
No differences were observed in the incidence of SAEs between the FF/VI 200/25, FF/VI 100/25, FF 200, FF 100, and VI (ICS) 25 groups, and the placebo group (Table 5). Across the FF/VI 200/25, FF/VI 100/25, FF 200, and FF 100 groups, seven patients experienced SAEs considered by the investigator to be related to study medication: one patient in the FF/VI 200/25 group (atrial fibrillation); two patients in the FF/VI 100/25 group (tachyarrhythmia and atrial fibrillation); and four patients in the FF 100 group (non-cardiac chest pain, asthma exacerbation, pleurisy, and pneumonia). Four deaths were reported. One patient died in the FF/VI 100/25 group (road accident) and two in the FF 100 group (stage IV lung cancer, and sepsis and pneumonia); none of these was determined to be asthma-related or considered to be related to the study medication. One patient in the placebo group died of an unknown cause.
The incidences of AESIs across the key treatment groups are presented in Table 6. The incidence of local steroid effects was numerically higher with FF-containing treatment arms (7–8%) compared with placebo (2%). The incidence of lower respiratory tract infection (LRTI) in the FF/VI 200/25 and FF/VI 100/25 groups was 3% and 4%, respectively, compared with 2% in the placebo group, and 3% and 6% in the FF 200 and FF 100 groups, respectively, with the rate of bronchitis events (5%) being the main factor in the FF 100 group. Symptoms that may be associated with hypersensitivity reactions were observed at a similar incidence across the groups [2% with FF/VI 200/25, FF/VI 100/25, FF 100, and placebo; <1% with FF 200 and VI (ICS) 25].
The exposure-adjusted incidence of local corticosteroid effects, pneumonia and LRTIs, cardiovascular effects, and hypersensitivity reactions was numerically higher within the FF/VI 200/25 group versus the FF/VI 100/25 group. A numerically greater exposure-adjusted incidence of local steroid effects was observed with FF 200 (283.8/1000 patient years) versus FF 100 (104.5/1000 patient years), and with FF/VI 200/25 (183.2/1000 patient years) versus FF/VI 100/25 (100.8/1000 patient years). A numerically greater exposure-adjusted incidence of cardiovascular events was experienced in the FF/VI 200/25 group (120.4/1000 patient years) compared with the FF/VI 100/25 group (66.3/1000 patient years). This was influenced by the higher number of extrasystoles observed in study NCT01018186 during Holter monitoring, although the events were not associated with any symptomatic AEs (investigators were instructed to record all events of extrasystoles as AEs). As the dose of VI is similar in both FF/VI 200/25 and FF/VI 100/25, the reason for this discrepancy is unclear.
Pneumonia was reported by <1% of patients in any of the key treatment groups. Investigators determined the diagnosis of pneumonia and were requested, but not mandated, to provide X-ray confirmation. Of the 40 patients in the key treatment groups who reported a pneumonia event, only 27 patients received a chest X-ray, and pneumonia was confirmed by X-ray in 25 of them (Table 7). The incidence of X-ray confirmed pneumonia was <1% in all the active treatment groups, compared with <0.1% in the placebo group (Table 7). The low number of events in each treatment group may mean any differences are difficult to identify. The exposure-adjusted incidence of X-ray confirmed pneumonia in the FF/VI 200/25 and FF/VI 100/25 groups was 5.2 and 7.2/1000 treatment years, respectively, compared with 4.7 in the placebo group, and 5.9 and 7.2 in the FF 200 and FF 100 groups, respectively (Table 7).
The safety of FF/VI in patients with asthma was evaluated in subpopulations based on age, gender, race, and region. The incidences of AEs in these subgroups were similar to the incidence in the ITT population across the key treatment groups.
The change from baseline in heart rate by treatment group is shown in Fig. 3. The mean baseline heart rates across the six key treatment groups ranged from 69.5 to 72.4 beats per minute. The maximum mean change from baseline at 24-h pre-dose was <1 beat per minute for the FF/VI and placebo groups.
Data from ten clinical studies (n = 2547) were used to assess urine free cortisol excretion (key treatment groups displayed in Table 8). At baseline, the geometric means for 24-h UC excretion ranged from 57.46 to 64.30 nmol/24 h across the five treatment groups. At the end of treatment, the 24-h UC excretion geometric means were numerically similar to baseline.
There were no statistically significant differences in the asthma composite endpoint (defined as asthma-related hospitalizations, intubations, or death) analysis of all FF/VI doses versus all ICS doses (i.e., the 95% CIs for the difference all included zero; Fig. 4). The common odds ratio was 0.854 (95% CI 0.344–2.146). Within the subgroup analyses by race and age, there were no differences between all FF/VI and all ICS doses. For African Americans, there was a single composite endpoint event with FF/VI. In adolescents, there were four events in the FF/VI group and two in the ICS group; it is notable that three of the four events on FF/VI occurred at a single site in Eastern Europe, where the site protocol was to hospitalize patients with asthma exacerbations who required treatment with systemic corticosteroids. One patient was hospitalized due to an asthma exacerbation, but did not require treatment with systemic corticosteroids.
Data for trough and weighted mean FEV1 for the four pivotal studies are shown in Fig. 5. There was a statistically significant difference in trough FEV1 (p < 0.014; Fig. 5a) and 0–24 h weighted mean FEV1 (p < 0.048; Fig. 5b) between the FF/VI and FF treatment groups in all studies, with the exception of NCT01165138. It is notable that in the NCT01165138 study (which investigated FF/VI 100/25, FF 100, and placebo), the weighted mean FEV1 was assessed in just over half (51%) of the population, whereas all patients in the NCT01686633 study (which investigated FF/VI 200/25, FF/VI 100/25, and FF 100) underwent 24-h serial spirometry. In the NCT01134042 study, FF/VI 200/25 significantly improved trough and weighted mean FEV1, peak expiratory flow, and rescue- and symptom-free 24-h periods compared with FF 200. When rescue- and symptom-free 24-h periods were assessed, a statistically significant difference (p < 0.010) between FF/VI and FF treatment groups was observed in all studies, with the exception of study NCT01165138. The analysis of study NCT01165138 is descriptive only, as the statistical hierarchy stated that a positive outcome must be shown for the primary endpoints of FEV1 to allow interpretation of the secondary endpoints (Fig. 6) . To account for multiplicity across treatment comparisons and key endpoints, a specific step-down testing procedure was applied, whereby inference for a test in the predefined hierarchy was dependent upon statistical significance having been achieved for the previous tests in the hierarchy. For study NCT01165138, statistical significance was achieved at the 5% level for four of the six designated primary comparisons at the top of the hierarchy (FF/VI 100/25 versus placebo and FF 100 versus placebo for the co-primary endpoints of trough FEV1 and weighted mean FEV1), but was not achieved for the primary comparisons of FF/VI 100/25 versus FF 100 for these co-primary endpoints. Statistical significance was not demonstrated by all primary comparisons and no statistical inference could be made on the secondary endpoints (which included rescue- and symptom-free 24-h periods); therefore, these are only descriptive.
Integrated efficacy data were utilized to assess the effects of treatment on trough FEV1 at week 12 in subpopulations. Data from four key efficacy studies (NCT01165138, NCT01134042, NCT01086384, and NCT01686633) were integrated and the treatments were combined regardless of dose, so that treatment with FF/VI could be compared with FF for the subpopulations. Subpopulation analyses of trough FEV1 demonstrated that the treatment effect was directionally similar as in the total ITT population (Fig. 7). The widths of the CIs are a reflection of the sample size.
The benefit:risk ratio of FF/VI 200/25 versus FF/VI 100/25 and FF/VI 100/25 versus FF 100 is shown in Fig. 8. The combination of FF with VI showed increased efficacy compared with FF alone across a range of efficacy endpoints, including lung function, rescue-free 24-h periods, rate of asthma exacerbations, and asthma control [as measured by the Asthma Control Test™ (ACT) (GSK, Brentford, London)]. There was a relatively low incidence of drug-related AEs with FF/VI (6% for both doses) compared with 2% for placebo; no event occurred in >2% of patients at either dose of FF/VI, showing that the combination is associated with a low risk of events. As shown in Fig. 2b, there were no relevant differences with the addition of VI to FF on the incidence of asthma events and other safety outcomes. In the clinical setting, having an alternative dose of an ICS allows treatment to be modified with differences in disease severity. This ability to adjust the corticosteroid dose is in agreement with treatment guidelines for the ICS/LABA combination products. Comparing the efficacy of FF/VI 200/25 with FF/VI 100/25, there was a numerical benefit with the higher strength versus the lower strength on several efficacy variables, including lung function, rescue-free 24-h periods, and asthma control (as measured by ACT). As also shown, comparison of the safety profile of FF/VI 200/25 versus FF/VI 100/25 supports the positive benefit:risk ratio of FF/VI 200/25, as we did not observe any relevant increases in AEs with the higher strength; these include SAEs and other potential ICS-related effects. The efficacy and safety of FF/VI shows an overall positive benefit:risk profile for the combination therapy.