Budesonide/formoterol therapy: effective and appropriate use in asthma and chronic obstructive pulmonary disease

This review was structured based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for design and reporting of literature reviews. EMBASE was searched in November 2016, with a follow-up search conducted in August 2019, using keywords relating to asthma, COPD and BFC, alone or compared with similar products, and real-world efficacy or utility (Supplementary Table 1). Because BFC was first approved for use in asthma in 2006, the search was designed to capture only articles published on or after 1 January 2006.

Eligible studies involved patients with asthma or COPD and evaluated an aspect of real-world effectiveness, utility, health-related quality of life or appropriate use as an outcome and were based on data derived from insurance claims databases, healthcare administrative databases and population-based registries. In vitro studies, reviews, non-English publications and animal studies were excluded. Studies were also excluded if they were solely an abstract (e.g., conference proceedings).

A combined instrument to assess the quality, relevance and credibility of an observational study (Supplementary Table 2) was developed based on questions derived from three sources. Twenty-eight questions were extracted from the ISPOR-AMCP-NPC 2014 Good Practice Task Force Report on observational studies [23]. Twelve questions were adapted from the ISPOR 2009 Good Practice Task Force Report on the use of retrospective database studies for CE research [24]. In addition to overall reporting, questions from these two sources addressed methods and analyses used to minimize biases and confounding factors that could affect any measured difference in treatment outcomes. Eight questions were adapted from the Respiratory Effectiveness Group summary of quality standards for real-world research [25]. Respiratory Effectiveness Group standards stressed three points to ensure quality of included studies for analysis: preparation of research, analyses and reporting and discussion of results. The scoring instrument was modified for scoring appropriate use studies by either modifying items to make them valid for this application or eliminating those items (Supplementary Table 2).

Each included article was reviewed by ≥2 reviewers independently (Supplementary Table 3). The reviewers then conferred to determine an overall evaluation. To approximate an overall score, the percentage of checklist items that were satisfactorily addressed was converted to a 100-point score based on the number of items that applied to the study. The quality of each reviewed article was assumed to be related to the number of items on the checklist that were satisfactorily addressed in the published manuscript. No comparison was made on the relative importance of individual items, and no weighting was used.

The only CE study of BFC versus FSC use among patients with asthma conducted in the USA was published by Tunceli et al. [21]. This study compared patients with asthma newly initiated on ICS/LABA therapy with respect to exacerbations and healthcare resource utilization (HCRU). Patients with asthma were identified using the HealthCore Integrated Research Database between 1 June 2007 and 30 September 2010 to coincide with the approval of BFC for the treatment of asthma in the USA. Patients were aged 12–64 years and had a second prescription fill of the same ICS/LABA combination during the 12-month follow-up period (Table 3). Propensity score matching was used to balance the study cohorts by age, sex and asthma-related utilization at baseline. From 3122 BFC and 8177 FSC patients meeting study criteria, 3043 BFC initiators were matched to 3043 FSC initiators.

The exacerbation rate among BFC initiators, defined as the total number of exacerbations during the postindex period divided by the total follow-up time in person years, was lower than that of FSC initiators (0.83 vs 0.89 per patient-year, respectively). After adjustment for asthma severity and exacerbation risk, this difference remained statistically significant (0.85 vs 0.93, respectively; rate ratio [RR]: 0.92 [95% CI: 0.85–0.99; p = 0.0255]). The BFC cohort had significantly fewer oral corticosteroid (OCS) prescription fills (mean: BFC, 0.73; FSC, 0.79; p = 0.0299) and fewer emergency department (ED) visits (mean: BFC, 0.08; FSC, 0.10; p = 0.0486), but the numbers of patients with any asthma-related hospitalization were similar (BFC, n = 45; FSC, n = 46). An unadjusted, descriptive analysis of high-dosage (BFC, 160/4.5 μg [delivered dose]; FSC, 250/50 or 500/50 μg [metered dose]) or low-dosage (BFC, 80/4.5 μg; FSC, 100/50 μg) treatments found that exacerbation rates (events per year) were similar for low-dosage treatments (BFC, 0.68; FSC, 0.66) but were slightly lower for BFC initiators among high-dosage treatments (BFC, 0.88; FSC, 0.95 [0.86 for 250/50 μg and 1.33 for 500/50 μg]). The percentage of patients needing short-acting β₂-agonist (SABA) prescription fills was lower in the BFC cohort (66.4 vs 71.2%; p = 0.0027); the adjusted mean number of SABA fills per patient was also lower (2.4 vs 2.6; p = 0.0221). Treatment adherence (proportion of days covered), was minimally lower in the BFC cohort (adjusted mean proportion of days covered: 0.46 vs 0.49; p < 0.001), suggesting adherence differences did not explain the observed differences in health-related outcomes. In this study, patients had to have ≥2 BFC or FSC prescription fills for inclusion, making adherence figures biased toward increased usage compared with studies discussed later in this review.

The Tunceli et al. analysis rated highly in all categories of study quality (Table 4) [21]. The limitations were mostly related to inherent limitations of administrative claims data, including possible misclassification of asthma diagnosis or exacerbation events and questionable generalizability to populations not covered by commercial insurance or living outside of the USA. Formulary design and differential prescribing patterns between pulmonary specialists and primary care physicians have previously been found to result in a greater uptake of newly approved brand medications among patients with more severe asthma [26–28]. However, patients in each treatment group in the Tunceli et al. analysis were matched on asthma-related utilization at baseline. Thus, this analysis is less likely to have been affected by ‘novelty bias’, a form of indication or channeling bias in which patients with more severe disease have a greater likelihood of receiving prescriptions for novel or recently released medications [26].

After BFC was approved for the treatment of asthma in the USA, the National Asthma Education and Prevention Program (NAEPP) released updated clinical practice guidelines for asthma [29]. These guidelines stated that ICS/LABA combination products should be reserved for patients whose disease requires two controller therapies or who are not adequately controlled with other maintenance therapies. The following appropriate-use studies investigated whether these recommendations were followed for patients treated in the general population. Exclusion criteria are presented in Table 3.

Blanchette et al. examined initiation of ICS/LABA combination therapy (BFC or FSC) based on preinitiation HCRU [30]. This analysis used a large national commercial insurance database and included patients with asthma aged 12–64 years continuously enrolled for 12 months prior to initiation of therapy from 1 July 2007–30 June 2008. Initiation of ICS/LABA therapy was considered appropriate if patients had claims during the preinitiation period for an ICS or leukotriene receptor antagonist (LTRA), an asthma-related ED visit or hospitalization, ≥2 courses of OCS or ≥6 canisters of a SABA. These criteria were based on the NAEPP guidelines and a comprehensive literature review; HCRU criteria assessed risk of exacerbations and the SABA criterion was used as a measure of poor disease control.

During the study period, 16,205 patients initiating ICS/LABA therapy met the inclusion/exclusion criteria. Of these, 788 of 1417 patients (55.6%) in the BFC group and 5572 of 14,788 patients (37.7%) in the FSC group met ≥1 criterion for appropriate use (odds ratio [OR]: 1.79; 95% CI: 1.60–2.00; p < 0.001). Greater percentages of BFC than FSC users met the individual criteria for appropriate use: previous ICS use (31.1 vs 11.0%), previous LTRA use (29.6 vs 18.6%), ≥2 courses of OCS (18.2 and 11.2%), and high SABA use (9.7 vs 6.1%; p < 0.001 for all comparisons). There were no differences in percentages of patients with asthma-related ED visits or hospitalizations (BFC, 5.6%; FSC, 5.9%). Factors potentially associated with appropriate use included receipt of the initial ICS/LABA prescription from a pulmonologist or allergist and presence of specific comorbidities (allergic rhinitis, sinusitis, gastroesophageal reflux disease and acute respiratory infection; all p < 0.001).

Ye et al. conducted a retrospective cohort study of commercially insured US patients with asthma aged ≥12 years who initiated BFC or FSC therapy in 2007 and were continuously enrolled for 12 months prior to the initial prescription fill [31]. Use of ICS/LABA combination therapy was considered appropriate if patients met any of the following criteria preinitiation: use of ICS or an LTRA, an asthma-related ED visit or hospital admission, ≥2 prescription claims for OCS with a days’ supply of <21 days, or ≥6 SABA prescription fills.

During the study period, 203,348 patients with a claim for ICS/LABA therapy were identified, and 24,231 patients met inclusion criteria, with 993 initiated on BFC and 23,238 on FSC. The average age was 38.6 years, and 38.4% were male. Only 37.6% (9107 patients) in the total population were considered appropriate candidates for initiating ICS/LABA (BFC, 58.4%; FSC, 36.7%; p < 0.0001). After adjusting for other factors, the odds that the BFC cohort was appropriately prescribed ICS/LABA compared with the FSC cohort were twofold greater (OR: 2.01; 95% CI: 1.76–2.30; p < 0.001). Odds of appropriate use decreased with increased age, greater number of comorbidities, increased prescription fills for other medications and visiting pulmonary or allergy specialists.

Tunceli et al. also conducted a retrospective appropriate-use study using the same large national commercial insurance database analyzed in their CE analysis [32]. Patients with asthma aged 12–64 years who later initiated ICS/LABA therapy between 1 June 2007 and 31 December 2010 were included. Appropriate use was defined using the same criteria as the Blanchette and Ye studies [30,31].

A total of 50,541 BFC, FSC and mometasone/formoterol fumarate users satisfied all study inclusion/exclusion criteria, but the latter were excluded because there were only 126 patients, leaving 11,718 in the BFC cohort and 38,697 in the FSC cohort. A higher percentage of BFC patients (BFC, 57.3%; FSC, 51.6%; adjusted OR: 1.21; [95% CI: 1.16–1.27]) met ≥1 of the appropriate-use criteria. The largest difference in appropriate-use criteria was in prior asthma medication usage, particularly ICS (BFC, 24.9%; FSC, 14.8%), but also LTRA (BFC, 26.8%; FSC, 21.4%). Percentage differences for high-risk utilization during the preinitiation period were mixed across the various criteria, including asthma-related ED visits (BFC, 9.8%; FSC, 12.1%), hospitalizations (BFC, 4.9%; FSC, 6.3%), ≥2 OCS fills (BFC, 17.4%; FSC, 14.6%) and fills of ≥6 SABA canisters (BFC, 8.3%; FSC, 7.0%), with p < 0.001 for all comparisons. Greater percentages of BFC patients were diagnosed with allergic rhinitis (41.4 vs 31.6%) and sinusitis (28.9 vs 25.1%), which may have accounted for some of the OCS and LTRA differences. More allergists/immunologists and pulmonologists prescribed BFC (30.5%) versus FSC (17.5%), which was also noted in the other two appropriate-use studies. A unique finding in this study, however, was that appropriate use of both agents decreased every year of the study. It is important to note that concepts about ‘appropriate use’ of ICS/LABA, and particularly ICS/formoterol, have evolved substantially since the 2007 NAEPP guidelines [29] that were the foundation of these three analyses. Current Global Initiative for Asthma (GINA) guidelines [33] recommend much broader application of ICS/formoterol, including use as a first-line treatment and in patients with milder disease.

Limitations of the Blanchette and Ye studies included limited information about the study populations and analysis methods [30,31]. Tunceli rated higher in these areas due to more detailed reporting [32]. There was also limited discussion about the clinical relevance of the results. The authors acknowledged that claims data do not contain information on asthma severity in terms of clinical symptoms and pulmonary function testing, which would have been appropriate indications for ICS/LABA therapy. Claims data also do not capture prescriptions that are not filled or medication samples obtained during an office visit. In addition, factors such as time of approval may impact appropriate-use results when comparing two treatments.

Roberts et al. were the first to compare the effectiveness of BFC and FSC using data from the USA for healthcare costs, HCRU, adherence and COPD exacerbations [34]. For this study, administrative claims data from a database of >50 million patients were used to identify those with COPD newly initiated on ICS/LABA therapy. Propensity-matching techniques were used to select highly similar BFC and FSC cohorts based on age, sex, geographic region, treatment initiation year, comorbid conditions, months of follow-up and preindex COPD episode utilization. Cost and effectiveness outcomes were total healthcare expenditures, exacerbation events (defined as hospitalizations, ED visits or outpatient visits associated with OCS or antibiotic prescription fills), treatment medication adherence, and claims for rescue medications and outpatient encounters during the 3–6-month follow-up period.

After all propensity-matching methods were applied, 6770 patients (3385 BFC and 3385 FSC) who initiated treatment during the study period (1 January 2007–31 January 2009) were compared. Fewer BFC patients had claims for SABAs (34.7 vs 39.5%; p < 0.001) and ipratropium (7.8 vs 9.8%; p < 0.005) than FSC patients, suggesting that they had better overall symptom control. No significant differences were seen in other clinical outcomes, including exacerbation events (BFC, n = 2146 [63.4%]; FSC, n = 2116 [62.5%]; p = 0.45). BFC patients had slightly fewer office visits related to pneumonia (patients with ≥1 visits: BFC, n = 93 [2.7%]; FSC, n = 121 [3.6%]; p = 0.052). However, the percentage of patients hospitalized for pneumonia was comparable (patients with ≥1 hospitalizations: BFC, n = 60 [1.8%]; FSC, n = 65 [1.9%]; p = 0.652).

The major limitations of this study are common to propensity-matched CE studies when one treatment is used much less commonly than the other [35]. BFC had just been approved for COPD treatment at the beginning of the study period, whereas FSC had already been available for this indication for 6 years; therefore, the pool of BFC patients meeting inclusion criteria was 3390 versus 90,070 for FSC [34]. Before matching, there were subtle differences between the BFC and FSC populations (e.g., age). The large number of FSC patients allowed for near-complete matching of all BFC patients and a high degree of confidence about the results derived from the patients who were compared, but there is uncertainty about how well the results can be extrapolated to those with COPD who were not examined. In addition, the follow-up period for BFC was short (3–6 months), which does not measure 1-year exacerbation outcomes or the effect of seasonal issues, and also prevents comparison of potential longer-term differences in outcomes between BFC and FSC users.

In another comparison of BFC and FSC using the same database, Mapel et al. examined the medications’ effectiveness in reducing COPD exacerbations and SABA use [36]. Using propensity-score methods to create clinically similar cohorts, 3852 BFC patients from a large national database were matched 1:1 to FSC patients at initiation of treatment with either combination. Exacerbations resulting in an ED visit or hospitalization for COPD were captured during the periods 6 months before and after initiation of BFC or FSC. Patients were also compared by their SABA prescription fills before and after initiation of treatment and were stratified as no fills (no SABA prescription fills), low (one fill) or high numbers of fills (≥2).

Hospitalizations for COPD exacerbation were reduced by approximately 20% among initiators of either treatment (BFC = 6.2% patients with exacerbation prior to treatment vs 4.9% post; p = 0.008; FSC = 5.8% prior vs 4.7% post; p = 0.023). Increased SABA use was associated with higher numbers of exacerbations during the baseline period (9% in high fills, 6.6% in low fills and 4.9% in no fills), and increased SABA use at baseline was also a significant predictor of exacerbations during the follow-up period (high fill number vs no fills, hazard ratio [HR]: 1.38 [95% CI: 1.08–1.76]; low fills vs no fills, HR: 1.58 [95% CI: 1.21–2.06]). Most of the decrease in COPD-related ED visits or hospitalizations after initiation of ICS/LABA treatment was found among the patients with a high number of SABA fills (change in percentage of population with exacerbations: -4.1% [p < 0.01], +0.1% [not significant] and -0.6% [not significant] among high, low and no fill patients, respectively). SABA use was reduced overall after initiation of either ICS/LABA, but a higher percentage of BFC patients had no SABA fills after initiation (59.3 vs 56.2%; p < 0.01).

This study has a similar design as the Roberts analysis [34] and shares the same benefits and limitations introduced by the propensity score-matching technique. Adherence to treatment, a factor that potentially could have affected outcomes, was not examined, and the 6-month follow-up period was relatively short.

Kern et al. conducted a CE study among patients with COPD initiating BFC versus FSC to examine exacerbation rates during a 12-month period [37]. A US national health insurance database of >31 million people from 1 March 2009 to 31 March 2012 was used to create the study cohort. New users of BFC or FSC aged ≥40 years with a diagnosis of COPD were included. A total of 3697 BFC initiators were matched 1:1 to FSC patients using propensity score-matching techniques.

There were no significant differences in the adjusted exacerbation rates between BFC (0.88) and FSC (0.86) during the follow-up period and no differences between treatments when stratified by type of exacerbation. Additionally, there were no significant differences between cohorts in a series of sensitivity and secondary analyses, including time to first exacerbation. During the follow-up period, 48% of BFC and 47% of FSC patients experienced ≥1 exacerbation event, and 6.2% of BFC and 6.9% of FSC patients had ≥1 COPD-related hospitalization. Adherence was poor in both groups: BFC patients averaged 3.8 prescription fills during the 12-month follow-up period versus 4.0 for FSC (p = 0.01), and 33.9% of BFC patients filled only their initial prescription versus 32.7% of FSC patients. In contrast to the previous two studies, there were no significant differences in SABA, short-acting muscarinic antagonist, or combination SABA/short-acting muscarinic antagonist inhaler utilization.

The Kern study was propensity score matched. The percentage of patients using BFC was more balanced than in the previous two studies, making the results more generalizable. Of the 10,227 patients with COPD who met all inclusion/exclusion criteria, 3788 (37.0%) initiated BFC, and among these, 3697 (97.6%) were matched to an FSC patient. This study also captured pneumonia-related events and found no significant difference in the percentage of patients having pneumonia after treatment initiation, either by total events (BFC, 17.3%; FSC, 19.0%; p = 0.1926) or only among those hospitalized (BFC, 8.9%; FSC, 10.3%; p = 0.0937).

In a subsequent analysis using the same database as Kern and the same matched patients, Davis et al. compared HCRU and costs among patients with COPD newly initiated on ICS/LABA therapy during the first 12 months after initiating BFC or FSC [38]. HCRU differences between the BFC and FSC populations did not reach statistical significance, but there were trends toward greater utilization among the FSC cohort. For example, the percentages of patients with all-cause hospitalizations (BFC, 30.6%; FSC, 31.9%), COPD-related hospitalizations (BFC, 6.2%; FSC, 6.9%), pneumonia-related hospitalizations (BFC, 8.9%; FSC, 10.3%) and all-cause intensive care unit admissions (BFC, 6.0%; FSC, 6.5%) were all higher among FSC patients. The 12-month mean all-cause hospitalization cost was $8519 per BFC patient and $10,314 per FSC patient (adjusted mean difference [AMD]: $1242 [95% CI: -2082 to -288; p = 0.01]), the mean COPD-related hospitalization cost was $966 per BFC patient and $1202 per FSC patient (AMD: $248 [95% CI: -377 to -147; p < 0.001]) and the mean pneumonia-related hospitalization cost was $2728 per BFC patient and $3409 per FSC patient (AMD: $640 [95% CI: -894 to -349; p < 0.001]). Significantly higher costs for FSC relative to BFC in outpatient visits and COPD-related pharmacy costs resulted in a total mean cost of $21,580 per BFC patient and $24,483 per FSC patient (AMD: $1884 [95% CI: -2685 to -1047; p < 0.001]).

The Kern and Davis studies rated high on the quality assessment tool (Table 4), with both having the major limitation of lacking detailed information about comorbidities other than asthma [38,39]. The quality assessment tool also emphasizes sensitivity analyses for testing key assumptions of study design or modeling, which can be difficult when using a propensity score-matching system because the purpose of matching is to neutralize the effects of the known confounding variables.

Trudo et al. published the only CE study of patients with COPD initiating BFC or tiotropium (TIO) in the USA [40]. Using the same database as the Kern and Davis studies and almost the same study period (1 March 2009–28 February 2012), 4051 patients with COPD without prolonged OCS use who were new users of BFC or TIO were identified. Of these, 1381 initiated BFC; 1198 were matched 1:1 to TIO initiators using propensity score-matching methods similar to the previous studies. Patients were followed for ≥12 months for the primary outcome of time to first exacerbation, with prespecified secondary outcomes including exacerbation rates, HCRU and costs, adherence and other COPD-related medication prescription fills.

The median time to first COPD exacerbation was 352 days in the BFC cohort and 243 days in the TIO cohort, with a significant HR of 0.78 (95% CI: 0.70–0.87; p < 0.001), and 51% of BFC patients had ≥1 exacerbation during the follow-up period compared with 59% of TIO patients. The sensitivity and subgroup analyses were consistent with these results, including an examination limited to patients without any prior asthma diagnosis, who had a median time to exacerbation of 343 days on BFC and 273 days on TIO (HR: 0.83; 95% CI: 0.72–0.96; p = 0.011). Exacerbation rates were correspondingly lower in the total BFC cohort (BFC, 1.23 per patient-year; TIO, 1.50 per patient-year; RR: 0.82 [95% CI: 0.73–0.91; p < 0.001]).

All-cause HCRU was not significantly different between BFC and TIO, but the number of patients with ≥1 COPD-related hospitalization (BFC, n = 90 [7.5%]; TIO, n = 113 [9.4%]; p = 0.01) or COPD-related outpatient visits (BFC, n = 889 [74.2%]; TIO, n = 1040 [86.8%]; p < 0.001) were significantly different. Substantially fewer BFC patients added TIO (n = 162, 13.5%) than TIO patients added an ICS/LABA to their treatment regimen (n = 441, 36.8%), accounting for much of the difference in COPD-related mean outpatient pharmacy costs per patient (BFC, $1310; TIO, $1750; p < 0.001). The COPD-related hospitalization and outpatient visit differences translated to significantly higher mean costs per patient for hospitalization (BFC, $982; TIO, $1433; p < 0.001) and outpatient visits (BFC, $1287; TIO, $1745; p < 0.001). The increased COPD-related medical and pharmacy costs accounted for slightly more than half of the difference in mean per-patient total all-cause costs between the BFC ($21,681) and TIO ($24,374) cohorts.

In terms of adherence, 41% of BFC patients filled their index medication once versus 31% of the TIO patients. The average fills for the BFC cohort (3.3 for 12 months) were also lower than TIO (4.3). However, treatment modifications among BFC patients were less common than for TIO patients (18 and 37% of each cohort, respectively), and the time to first treatment modification was later for BFC patients (HR: 0.41 [95% CI: 0.35–0.49]).

The Trudo study rated highly in all quality categories for the quality assessment tool (Table 4). However, the poor adherence in a retrospective analysis of intent-to-treat cohorts suggests that the results must be considered with caution. Unmeasured clinical factors related to outcomes, such as disability and comorbidities, may have influenced treatment choice, and residual bias by indication could exist. For example, the TIO patients meeting all inclusion/exclusion criteria for this study prior to propensity matching had substantially more COPD-related hospitalizations (28.4 vs 19.0%) and ED visits (31.9 vs 24.8%) during the baseline year.

The Global Initiative for Chronic Obstructive Lung Disease (GOLD) report for COPD treatment is updated annually, but in general the report has recommended starting with long-acting bronchodilators first and reserving ICS/LABA therapy for patients with high risk of COPD exacerbations (GOLD group C or D); long-acting muscarinic antagonists (LAMA)/LABAs are also recommended for patients with persistent symptoms or further exacerbations (GOLD group B, C or D) [5].

The only published comparison of BFC and long-acting bronchodilator initiation and appropriate use in COPD was conducted by Kern et al. [39]. Using the same database analyzed for the authors’ BFC versus TIO COPD CE study [37], the clinical and demographic characteristics and resource utilization in the 12-month preinitiation period were examined. However, patients who had used any controller therapy in the preinitiation period other than that cohort’s affiliated therapy were included. A total of 6940 BFC and 10,831 TIO patients met all inclusion/exclusion criteria.

BFC patients tended to be younger (mean age, 64.3 vs 66.9 years), female (54.2 vs 50.6%), with more having respiratory comorbidities (particularly asthma [(24.7 vs 12.7%]), and had ≥1 OCS prescription fill during the preinitiation year (45.8 vs 34.6%), with p < 0.0001 for all comparisons. However, similar percentages of BFC and TIO patients had exacerbations requiring hospitalization (7.1 vs 7.0%) or ED visits (7.7 vs 8.6%). The prevalence of most chronic nonpulmonary comorbidities was generally similar except for cardiovascular diseases, which were less common in the BFC cohort.

As a descriptive observational study, this study had difficulty meeting some criteria of the quality assessment tool because most of the analyses were descriptive rather than hypothesis testing. HCRU databases also do not collect pulmonary function testing data, although guidelines have recently been amended to focus more on symptoms and exacerbations as the basis for COPD severity assessments [5]. The demographic differences in the groups may reflect appropriate use since the guidelines support use of ICS/LABA in patients with a history of exacerbations and features of asthma [5]. Nevertheless, this study does reveal some of the important factors associated with prescribers’ choices of initial COPD treatment.

The study of BFC versus FSC use published by Tunceli found that BFC users had fewer exacerbations, a difference that remained significant after adjustment for asthma severity and exacerbation risk [21]. Most of the difference was found among patients using high-dosage treatments. A Cochrane analysis of five RCTs comparing BFC and FSC in adults and children with asthma (total N = 5537 adults) was unable to make definitive conclusions about the superiority of either product due to uncertainty about asthma exacerbation measures [20]. Additional CE analysis of retrospective databases or prospective pragmatic clinical trials of real-world asthma populations would be useful to examine the relative benefits of BFC versus other asthma treatments in the USA.

Perhaps the most remarkable finding among the appropriate-use studies in asthma is the consistency of the findings across studies. A slight majority of patients with asthma initiating BFC [21,30,31] met the definition of appropriate use by having utilization levels that indicated they were at greater risk of exacerbations based on ED visits or hospitalizations for asthma, or increased OCS use, or had tried other controllers first (ICS alone and/or LTRAs + SABA) before stepping up treatment. Also, each study found that patients with asthma initiating FSC during the study periods were less likely to have met these objective measures based on utilization. Clinically, asthma severity is based primarily on subjective measures that are not likely to be captured in healthcare administrative data, such as the presence of wheezing on physical exam or patient-reported incidence of nighttime symptoms. The fact that retrospective data demonstrate such reproducible objective evidence of increased asthma-related HCRU supports the validity of this approach as a proxy for estimating asthma severity in healthcare–based populations. It is also remarkable that factors associated with appropriate use were consistent among these studies, such as provider type (i.e., pulmonary or allergy specialist vs primary care provider) and regional differences across the USA. The finding that asthma specialists were more likely than primary care providers to prescribe BFC suggests the potential for the novelty bias discussed previously.

It is important to note that the definition of ‘appropriate use’ for ICS/LABA treatment in asthma has changed substantially since the 2007 NAEPP guidelines, which were heavily influenced by the FDA ‘black box’ warning about LABA use in asthma, and recommended use only after failing more conservative therapy [26]. Since that time, studies have demonstrated that ICS/LABA treatment does not result in significantly more serious adverse events than ICS monotherapy [41,42]. The black box warning for LABA-containing treatments was removed in 2017 [43]. The current GINA guidelines now recommend ICS/LABA as a first-line asthma treatment. Furthermore, GINA recommends low-dose ICS/formoterol for those with infrequent symptoms, and as both a maintenance and reliever inhaler for patients with asthma who have moderate persistent symptoms. These new applications are considered ‘off label’ use per US FDA approved prescribing information. Nevertheless, our findings suggest that guidelines do impact asthma treatment utilization and that this research approach can provide useful and reproducible insights into clinical decisions about asthma treatment.

COPD exacerbations were an outcome evaluated by the RCTs summarized in Table 2 and were also examined by all five of the BFC studies. In the four studies that compared patients who initiated BFC or FSC, exacerbation rates during the follow-up periods were similar, an expected result given that both are ICS/LABA combinations with subtle differences in pharmacology and delivery. There were trends favoring BFC in exacerbations defined as outpatient visits associated with OCS or antibiotic prescription fills in the Mapel study and BFC in exacerbations giving rise to hospitalizations for COPD in the Kern and Davis studies [36–38]. Although the comparison groups were closely matched, these are retrospective observational studies, so it is possible that the observed differences are due to selection bias or residual confounding. Nevertheless, the economic study by Davis demonstrates that even small differences in COPD exacerbation rates could result in substantial cost impacts [38].

Comparing BFC to TIO, the Trudo study found that BFC patients had substantially lower exacerbation rates and longer time to first exacerbation [40]. These differences are noteworthy given that they confirm, using real-world data, results that have been demonstrated in clinical trials. After initial therapy, far more TIO patients added an ICS/LABA than BFC patients added TIO, a phenomenon also observed in a recent comparison of FSC and TIO [44]. Similar reductions in COPD exacerbations were reported in BFC versus TIO effectiveness in a study from Canada [45]; however, the study was conducted using a BFC device not approved in the USA. It is possible that the patient identification and matching methods applied biased selection toward populations that were predisposed to having a more favorable response to ICS/LABA treatment. The observation of such large differences in FSC and TIO CE suggests that a tailored approach to COPD treatment could yield improvements in exacerbation outcomes.

Several other useful observations were made in these studies. The rate of exacerbations decreased by 20–25% after initiation of ICS/LABA treatment in the Mapel study, which is consistent with the reduction in exacerbations noted in RCTs (Table 2). Davis noted reductions in the percentage of patients with ≥1 COPD-related hospitalization (11.4% before BFC treatment, 6.2% after) and ≥1 COPD-related ED visit (14.1% before BFC, 12.3% after), which was similar to the reductions among matched FSC patients [38]. Roberts demonstrated that SABA use decreased after initiation of therapy, with the reduction being significantly greater among BFC users [34]. However, Mapel demonstrated that reduced SABA use was mostly driven by patients with higher baseline SABA use [36]. The reduction in SABA use suggests that patients being treated with BFC have less need for ‘rescue’ bronchodilator use to treat acute respiratory symptoms [38]. Increased SABA use was correlated with increased exacerbation risk at baseline and was also predictive of future exacerbations [36], confirming that SABA fills are a valid indicator of COPD severity. Pneumonia-related events were uncommon, but Roberts found a slightly lower rate among BFC users which was not seen in the Kern study [34,37]. A systematic review addressing the risk of pneumonia with ICS versus placebo or LABA in patients with COPD found no increased risk of pneumonia with budesonide-containing inhaled therapies, whereas therapies containing fluticasone propionate were associated with a 43–78% increased risk of pneumonia [46].

The most notable finding in the Kern study is that two inhaled treatments with the same indication can have substantially different HCRU based on patient demographic characteristics, comorbidities and prior respiratory medication utilization [37]. It is interesting to note that the utilization observed during the 2009–2012 study period more closely resembles current GOLD COPD treatment recommendations than the recommendations for ‘appropriate use’ of that time. These observational data support the concept that COPD is a heterogeneous syndrome; patients with COPD may benefit from a personalized approach to treatment.

Limitations of research in this area include lack of information about comorbidities, particularly concomitant asthma and pneumonia in patients with COPD. Poor adherence to any of these long-acting respiratory medications casts some doubt about the causality of the relationship between the initiated treatment and observed outcomes. At the time these studies were conducted, use of FSC was more widespread than use of BFC pMDI, which may have introduced a ‘novelty bias’, wherein more patients with more severe disease were more likely to be prescribed the newer medication and thus could have affected appropriate use and CE measures. Although an intent-to-treat design provides information on the CE of the initial treatment decision, future analyses may benefit from implementing newer approaches that incorporate adherence and treatment switching into the analysis, such as marginal structural models and microsimulation techniques.