## Introduction

### Cost-effectiveness analysis

The composite outcome measures and cost-effectiveness analyses were prespecified according to the standard operating procedures of the research group.25 The analyses were carried out using IBM SPSS Statistics versions 19 and 20 (SPSS Statistics, IBM, Somers, NY, USA), SAS versions 9.2 and 9.3 (SAS Institute, Marlow, UK) and Microsoft Excel 2007 (Microsoft, Bellevue, WA, USA). Statistically significant results were defined as P0.05.

We matched patients sequentially on the following criteria: sex, age and criteria indicative of asthma severity at baseline, including number of acute oral corticosteroid prescriptions during the year; mean daily SABA dose; and (for the US only) number of asthma consultations with no acute oral corticosteroid prescription (details in Supplementary Appendix). Matching ratios, chosen to maximise patient numbers and statistical power for comparisons, were 1:1 for the UK and 1:3 for the US for extrafine-particle versus standard size-particle ICS cohorts.12

Summary statistics for baseline and unadjusted effectiveness variables were tabulated by matched treatment cohort and compared using conditional logistic regression. Total asthma-related and lower respiratory-related health-care costs (hereafter called asthma-related costs) were adjusted to 2007 (UK) or 2010 (US) prices, summed for baseline and outcome years and compared between cohorts using conditional logistic regression. Although the distributions were skewed, we report the arithmetic mean costs, as recommended by Thompson et al.,26 because means can be multiplied by a target population to estimate total costs, the main outcome of interest for policy-makers and providers. The adjusted odds of achieving risk-domain asthma control and overall control were compared between matched cohorts using conditional binary logistic regression models, with standard size-particle ICS as the reference treatment. Asthma control status was used as the dependent variable, with treatment and potential confounding factors (Supplementary Table S2) as explanatory variables.

Expected differences in adjusted mean costs were modelled using generalised linear models with gamma distributions to best fit the data; expected differences in adjusted proportions of patients achieving asthma control were modelled using binomial distributions.26 The 95% confidence intervals (CIs) were determined by bootstrapping methods, using 1,000 random samples taken, with replacement, from the data set27; differences (relative to standard size-particle ICS) were displayed graphically on a cost-effectiveness plane. Further information on the methods of analysis is detailed in the Supplementary Appendix.

## Results

### Patients

After matching, there were 865 patients in each UK treatment cohort (ages 12–60 years). The two cohorts were well matched for baseline demographic and asthma-related characteristics, with no clinically important differences (Table 1, Supplementary Table S3).

Baseline characteristics of the 2,578 and 7,734 patients in the US extrafine- and standard size-particle ICS cohorts were also well matched, as previously described (see Supplementary Table S4).12 The UK and US cohorts had similar sex and age distributions, with an additional 7% of US patients who were non-smokers 60–80 years old. During the baseline period, the US patients were more likely to have received oral corticosteroids and were less likely to have achieved either risk-domain or overall control than the UK patients. In the UK similar proportions of patients in the two cohorts were prescribed a spacer; in the US a larger proportion of patients prescribed standard size-particle ICS, compared with those prescribed extrafine beclomethasone, were prescribed spacers during baseline and/or outcome year (7% vs. 9%; P=0.006; Supplementary Table S5); however, the difference between cohorts was small and likely not clinically significant.

### Effectiveness

The percentage of patients achieving risk-domain asthma control in the UK study increased from 70% at baseline in both cohorts to 84% and 82% in extrafine-particle ICS and standard size-particle ICS cohorts, respectively, during the outcome year. There was no significant difference between cohorts in the odds of achieving risk-domain asthma control, with an adjusted odds ratio (OR) of 1.20 (95% CI, 0.93–1.57) for extrafine-particle ICS relative to standard size-particle ICS. The percentage of patients achieving overall control decreased from 65% at baseline in both cohorts to 63% and 58% during the outcome year in extrafine- and standard size-particle ICS cohorts, respectively. Patients in the extrafine-particle ICS cohort had significantly higher odds of achieving overall control (adjusted OR, 1.23; 95% CI, 1.01–1.50).

Effectiveness results for the US cohorts have been described in detail elsewhere.12 In brief, the adjusted odds of achieving risk-domain asthma control were similar for the two cohorts (adjusted OR, 1.05; 95% CI, 0.96–1.15), whereas the odds of overall control during the outcome year significantly favoured extrafine-particle ICS (adjusted OR, 1.19; 95% CI, 1.08–1.31).12

### Resource use and costs

In the UK as in the US,12 extrafine beclomethasone was prescribed at significantly lower doses than fluticasone on the index date (median (interquartile range), 200 (100–200) vs. 500 (200–500) μg/day in the UK, P<0.001). During the outcome year, the median (interquartile range) daily ICS dose exposure was also lower in extrafine beclomethasone compared with fluticasone cohorts at 110 (55–181) vs. 115 (41–247) μg/day in the UK and 44 (22–98) vs. 72 (36–195) μg/day in the US, respectively (P<0.001 for both), although the difference was likely not clinically significant in the UK.

In the UK, mean asthma-related costs during the baseline year were comparable between cohorts (Supplementary Table S6). Over the outcome year, the mean total adjusted asthma-related costs were significantly lower in the extrafine-particle ICS cohort both with and without the inclusion of ICS costs (Table 2); the unadjusted costs are depicted in Supplementary Table S7.

In the US, drug and consultation costs during the baseline year were significantly higher for patients in the extrafine-particle ICS cohort, whereas total annual asthma-related costs were comparable because of lower hospitalisation costs (Supplementary Table S6). During the outcome year, the unadjusted asthma-related costs were similar for both cohorts, as previously described (see Supplementary Table S8).12

### Cost-effectiveness

In the UK as in the US,12 total asthma-related costs during the outcome year, when adjusted for baseline confounders, were significantly lower for patients prescribed extrafine-particle ICS compared with standard size-particle ICS (Table 2). In the UK, the adjusted mean difference (95% CI) of £66 (£37–£93) was driven mainly by significantly lower ICS costs in the extrafine-particle ICS cohort (Tables 2 and 3).

The cost-effectiveness planes show the spread of the estimated differences in cost and effectiveness based on 1000 replicated samples (Figures 2 and 3). With regard to risk-domain asthma control, the primary effectiveness measure, there was a 92% (UK) or 84% (US) probability that extrafine-particle ICS was the preferred treatment strategy (less costly and more effective) and an 8% (UK) or 16% (US) probability that extrafine-particle ICS was less costly but less effective (a trade-off; Figures 2a and 3a). With regard to the overall control measure, after adjusting for potential confounders, in the UK there was a 98% probability that extrafine-particle ICS was the preferred treatment strategy, and a 2% probability that extrafine-particle ICS was less costly but less effective (Figure 2b). In the US, extrafine-particle ICS was less costly (100%) and more effective (100%) and thus the dominant treatment (Figure 3b).

## Discussion

### Main findings

The results of these cost-effectiveness analyses indicate that extrafine-particle ICS is likely to be the dominant treatment strategy (less costly and more effective), as compared with standard size-particle ICS, in both the UK and the US for patients prescribed ICS for the first time for asthma. In the UK, the probabilities of extrafine-particle ICS being the preferred strategy with regard to cost-effectiveness for risk-domain asthma control and overall control were 92% and 98%, respectively; in the US, the probabilities were 84% and 100%.

To our knowledge, this is the first comparative cost-effectiveness assessment of two ICSs in asthma performed using the same metrics in two countries with very different health-care systems. The direct asthma- and lower respiratory-related health-care costs, adjusted for baseline costs, were significantly lower for UK and US patients12 treated with extrafine-particle ICS with or without the inclusion of ICS costs. The health-care systems in the UK and US, however, are structured differently and, indeed, the absolute costs were very different between the two countries, notably lower for the UK primary care population than for US patients.

### Interpretation of findings in relation to previously published work

For the UK, the mean unadjusted total asthma-related cost of £155–218/year (depending on treatment cohort) was similar to the annual cost to the NHS of £166/year (or £332 over 2 years) reported for patients prescribed ICS in a recent UK pragmatic trial.28 With regard to effectiveness findings, as compared with our prior study that included children as young as 5 years,11 the proportions of patients achieving asthma control were roughly similar, whereas the adjusted odds of risk-domain asthma control were not significantly different between treatment cohorts in the present study (as we had found in the prior study11).

For the US, this study’s categorised costs were generally in line with those reported recently by Barnett and Nurmagambetov,5 with the exception of respiratory-related medication costs, which were lower by about $850/year in each cohort. Possible explanations for the lower medication costs in our analyses are (1) the exclusion of in-hospital medications from the study database and (2) the inclusion of patients 65 years of age (3% of all patients) who used health-care resources covered by Medicare that were not captured by the database. Our study’s total costs ($1952–2369, depending on cohort) are thus lower than the previously reported total per-patient annual incremental cost of $3259 for asthma in 2009 US$.5,12

There are few economic evaluations comparing ICSs of different particle sizes as monotherapy for asthma.2931 In a 1-year pragmatic, open-label study of adults with asthma, extrafine beclomethasone was a cost-effective alternative when administered at half the dose of the larger-particle chlorofluorocarbon beclomethasone in the US, UK, Belgium and the Netherlands.32 More recently, in a 1-year US database study, significantly lower drug and total medical costs and fewer emergency department visits were reported with extrafine beclomethasone compared with fluticasone for adults and children (5–64 years old) with persistent asthma.33 Prior studies comparing extrafine beclomethasone and fluticasone were short-term (3–12 weeks) randomised trials that did not evaluate cost-effectiveness.34

### Strengths and limitations of this study

Our study has several potential limitations. As for any observational study using a large database, the analyses relied on the accuracy and completeness of data recording. Some data were not available for all patients, including socioeconomic status and body mass index, both factors linked to asthma control; thus, we cannot rule out the possibility of unrecognised confounding factors. However, our matching process paired patients in each cohort using several criteria reflective of asthma severity and control at baseline, with subsequent adjustments for minor residual confounding.

We used composite measures of asthma control to capture all key elements of control (impairment and risk) that were available from the database. The risk-domain asthma control measure included the absence of criteria defining severe exacerbations,18 which are important drivers of asthma-related costs. However, a study limitation is that the databases, and thus our outcome measures, do not capture symptoms or lung function measurements. We included SABA prescriptions as a proxy for symptoms in the ‘overall control’ measure, because SABA use reflects symptom control. The control cutoff point of a mean SABA use of 2 puffs/day corresponds to the level 2 category (2 of 4, with 1 being the best controlled) of the validated approach of Schatz et al. 20 for SABA canister dispensing to assess asthma symptom control. The definitions of the outcome measures are inherently related to the use of health-care resources; therefore, it is unsurprising that costs and outcome in these analyses work in the same direction. For the US analyses, the later index dates of patients in the extrafine-particle ICS cohort12 might have biased comparative costs, increasing the cost of extrafine-particle ICS and decreasing the cost–benefit relative to standard size-particle ICS, despite the fact that costs were adjusted to 2010 dollars, because of the general increases in costs over time.

A strength of these analyses is the consistency of results across effectiveness measures and between the two countries. Moreover, the large sample sizes reassure us of the robustness of the findings. We evaluated outcomes over 12 months to eliminate seasonal variations and evaluate long-term effects, as recommended by expert group consensus.18,19 An important strength is the fact that cost and effectiveness data were drawn from large numbers of patients in real-life clinical practice, as opposed to economic modelling based on non-representative clinical trial data, which is the most common approach. Studied patients thus included smokers, patients with comorbidities and potentially those with poor adherence or poor inhaler technique, enhancing the applicability of the findings to real-life practice.

### Implications for future research, policy and practice

Calculations of indirect costs would have been of interest, as patients with asthma have significantly higher work loss than those without,35 and recent estimates put lost productivity because of asthma at 8–12% of the total societal burden of asthma in the US.5 Moreover, continuing research is needed to better characterise the role of small airway disease in asthma and the benefits of extrafine-particle ICS for treating both large and small airway inflammation.36 Comparisons of other extrafine- and larger-particle ICS are needed, and our findings should not be extended beyond this comparison of extrafine beclomethasone and fluticasone.

A direct comparison between clinical outcomes in the US and UK would be inappropriate because of differences between the databases and health-care systems. However, several observations are intriguing. The percentage of patients meeting the asthma control measures was much lower, for both control measures, in the US12 than in the UK during baseline as well as outcome. We do not have an obvious explanation for this observation but have marked it as a topic for further study. In the US there was a greater disparity than in the UK between dose prescribed at the index date and dose exposure during the outcome year, possibly indicating poor adherence to therapy. A change in therapy in the form of an increase in ICS dose or additional therapy was more frequent in the US than in the UK (see Supplementary Appendix). The percentages of patients achieving risk-domain asthma control increased from baseline during the outcome year in both treatment cohorts in the UK and US, although the increase was small (1%) in the US.12 However, the percentages achieving overall control decreased, likely driven by increased SABA prescriptions during the outcome year, as noted in our prior studies,11,12 for patients newly diagnosed with persistent asthma who were prescribed additional SABA together with ICS.

### Conclusions

Real-life comparative effectiveness and cost-effectiveness analyses are needed to guide practical decision making with regard to asthma therapies. These analyses serve as preliminary evidence of the relative cost-effectiveness of extrafine-particle ICS and suggest that initiating ICS therapy for asthma as extrafine-particle ICS is the dominant treatment option (less costly and more effective) as compared with standard size-particle ICS for adults and children 12 years old in both the UK and US. Further studies, including prospective pragmatic trials, are warranted to investigate the comparative effectiveness and cost-effectiveness of other examples of extrafine- and larger-particle ICS.