A pragmatic single-blind randomised controlled trial and economic evaluation of the use of leukotriene receptor antagonists in primary care at steps 2 and 3 of the national asthma guidelines (ELEVATE study)

Article history

The research reported in this issue of the journal was commissioned by the HTA programme as project number 98/34/05. The contractual start date was in October 2001. The draft report began editorial review in May 2007 and was accepted for publication in October 2009. As the funder, by devising a commissioning brief, the HTA programme specified the research question and study design. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the referees for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

see Acknowledgements

Copyright statement

© Queen’s Printer and Controller of HMSO 2011. This work was produced by Price et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This journal is a member of and subscribes to the principles of the Committee on Publication Ethics (COPE) (http://www.publicationethics.org/). This journal may be freely reproduced for the purposes of private research and study and may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NETSCC, Health Technology Assessment, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2011 Queen’s Printer and Controller of HMSO

Scientific background

Asthma is a condition of the bronchial airways, characterised by airway hyper-responsiveness (or airway irritability) and reversible airway obstruction. In 2002, an estimated 3 million people, or 5% of the UK population, had asthma. 1,2 Asthma is caused by chronic inflammation of the small, or bronchial, airways. This inflammation causes the production of mucus, oedema formation, and nerve end exposure, and leads to an increase in airway hyper-responsiveness. Increased airway hyper-responsiveness causes narrowing of the airways (or bronchoconstriction), which may lead to coughing, wheezing, chest tightness and shortness of breath. In asthma, airway bronchoconstriction can be substantially reversed with a short-acting β₂-agonist or reliever medication.

Untreated chronic inflammation in the airways may lead, in some individuals, to structural changes (or airway remodelling), irreversible bronchoconstriction and persistent symptoms. The recognition that airway inflammation is present even in patients with mild asthma has led to a shift towards introducing anti-inflammatory therapy earlier in the management of asthma,2,3 with increased prescribing of inhaled corticosteroids4,5 in patients requiring daily use of a short-acting β₂-agonist [step 2 of British Thoracic Society (BTS) Asthma Guidelines]. 4,5 However, while there is some evidence of reduced morbidity, many patients with asthma still have considerable symptoms and lifestyle limitation. 6 Possible reasons for this include lack of disease recognition, poor adherence to inhaled steroids, poor inhaler technique, untreated rhinitis, smoking, and an inability of inhaled steroids alone to fully control asthma, with an increasing emphasis on the role of adding additional therapy to inhaled steroids rather than routinely increasing inhaled steroid dose. As a result these patients end up being treated at step 3 of the BTS Asthma Guidelines. 4

Efficacy studies of antiasthma therapies have traditionally used measures of airways function, such as spirometry [forced expiratory volume in 1 second (FEV₁)]7,8 and domiciliary peak expiratory flow (PEF),7 or measures of airway hyper-reactivity, such as methacholine bronchial challenge testing,9–12 to demonstrate therapeutic effectiveness. While these measures provide objective information on airway function, they provide no information on patient perceived effectiveness of an asthma treatment or asthma control. Indeed, the ‘real-life’ control of asthma is now regularly assessed in terms of changes in patient-reported quality of life (QOL), symptoms, exacerbations and rescue medication use. As the correlation is often poor between objective measures of airway function (e.g. domiciliary PEF) and measures of asthma control, international guidelines encourage the collection of measures of both airway function and disease control. 5

Anti-inflammatory treatments with the potential to treat mild to moderate asthma are inhaled corticosteroids and leukotriene receptor antagonists. Corticosteroids work by suppressing the production of inflammatory mediators by airway epithelial and smooth muscle cells, endothelial cells and fibroblasts. 13 However, inhaled steroids have been shown to have limited impact on suppressing the production or release of the cysteinyl leukotrienes LTC₄, LTD₄ and LTE₄, biologically active mediators derived from arachidonic acid, which collectively account for the biological activity known as slow-reacting substance of anaphylaxis. 14,15 These leukotrienes mediate many responses that are associated with asthma, including mucus production, decreased mucociliary clearance, changes in vascular permeability, inflammatory cell influx and smooth muscle contraction. 16 Thus, leukotriene receptor antagonists that act to reduce the production or block the action of leukotrienes may be important in asthma management and complementary to inhaled corticosteroids.

Hypotheses

In older children and adult patients with chronic asthma, initiation of a leukotriene receptor antagonist will provide, at no greater cost to the National Health Service (NHS) and patients, clinical improvements in QOL and other important asthma parameters that are at least equal to the alternative treatment options of inhaled corticosteroid at step 2 and adding a long-acting β₂-agonist at step 3. This study was designed as two separate, but concurrent, equivalence trials to determine whether a leukotriene antagonist is an equal choice to inhaled steroid as monotherapy, and to long-acting β₂-agonist as add-on therapy, for a real-world population of patients with perhaps milder asthma and who are less likely to adhere to therapy than those enrolled in classical clinical trials.

Rationale for this study

Participants

In the BTS British Guideline on the Management of Asthma4 the therapy of patients from the age of 6 years upwards follows the same strategy as for adults, except for alterations in dosage ranges to adjust for differences in body mass. Since exactly the same strategy is used across the age range of older children and adults, the findings of studies will have greater generalisability if they enrol patients from that entire range. Owing to limitations of validity of the MiniAQLQ and the Asthma Control Questionnaire (ACQ),62 we were unable to study children below the age of 12 but did allow children over this age, as well as adults of all ages, to be included to maximise generalisability of the study findings.

In the initial design of the study, participant recruitment was to be by primary care practice staff, as they conducted acute and routine respiratory care visits, identifying patients who met the entry criteria, informing them of the study and, if appropriate, consenting and enrolling them into the study. However, recruitment by this strategy was slower than originally anticipated owing to changes in clinical practice resulting from delays in study funding and changes in national asthma guidelines. The protocol and the process of identification of eligible patients were therefore modified, as described below, to allow prospective identification of possible study participants. All patients entering the study met the same eligibility criteria and follow-up was identical.

Further recruitment into the study was via a three-stage process.

Recruitment stage 1

Patients aged 12–80 years, attending 53 participating primary care (or general) practices in Norfolk, Suffolk, Essex, Cambridgeshire, Bedfordshire, Hampshire and Dorset, in the UK, who had received a prescription of short-acting β₂-agonist in the previous 2 years, were invited, by letter, to provide data allowing eligibility for the studies to be determined. Patients were asked to provide information on their current asthma status and inhaler usage. The case notes of patients whose asthma status was consistent with eligibility in the study were reviewed by practice and study staff against the following eligibility criteria.

Inclusion criteria

• Capable of understanding the study and study procedures (and parent/guardian’s capability of understanding the study and study procedures for patients aged under 16 years).

• Patient had a diagnosis of asthma [defined as (1) documented reversibility after inhaled short-acting β₂-agonist and/or (2) PEF variability on PEF diary and/or (3) physician-diagnosed asthma and/or (4) physician diagnosis of asthma plus history of response to treatment].

• Step 2 trial Patient was not currently receiving, and had not received, inhaled steroid or leukotriene antagonist within the previous 12 weeks.

• Step 3 trial (1) Patient had received inhaled steroid for at least the last 12 weeks, as ascertained from prescribing records and patient self-report and (2) had not received a long-acting β₂-agonist or leukotriene antagonist in the previous 12 weeks.

Exclusion criteria

• Patient had participated in a clinical trial involving an investigational or marketed drug within 90 days.

• Patients had received a substantial change in antiasthma medication within the previous 12 weeks.

• Patient was a current, or recent past, abuser (within past 3 years) of alcohol or illicit drugs.

• Patient had any other active, acute or chronic pulmonary disorder or unresolved respiratory infection within previous 12 weeks.

• Patient had a history of any illness that was considered to be immediately life threatening, would pose restriction on participation or successful completion of the study or would be put at risk by any study drugs (e.g. allergy to leukotriene antagonist).

• Patient had received systemic, intramuscular or intra-articular corticosteroids within the previous 2 weeks (artificial baseline).

Patients who met those entry criteria that could be assessed by a records review in their general practice were invited for a screening visit (visit 1 – see Figure 1 and Table 1). All patients had at least 24 hours to review the patient information sheet prior to attending the visit. Patients attending for at least visit 1 will, from here on, be referred to as ‘participants’.

FIGURE 1.

Study flow charts, Patients at step 2 received initial controller therapy with leukotriene antagonist or inhaled steroid, Patients at step 3 received leukotriene antagonist or long-acting β₂-agonist as add-on to inhaled steroid, ICS, inhaled corticosteroids; LTRA, leukotriene receptor antagonist; LABA, long-acting β₂-agonist; PRN, ‘pro re nata’ – as needed; SABA, short-acting β₂-agonist.

TABLE 1 - Time lines for both step 2 and step 3 trials

GP, general practitioner.

Practices were encouraged, and participants were reminded to have follow-up visits on or near the dates as described, but where for pragmatic reasons participants had follow-up respiratory care visits only between or after the stated dates, information from those dates was utilised.

Visit	Baseline		Trial period
	1	2	3	4	5	6	7
Study timescale (weeks)	–2	0	8	26	52	78	104
Leeway allowed (days)a		± 7	± 21	± 21	± 21	± 21	± 21
GP and/or practice asthma nurse procedures
Assess inclusion/exclusion criteria	✓
Informed consent	✓
Record clinical/asthma history and prior medications		✓
Review clinical data and asthma therapy (per clinical need)	✓	✓	✓	✓	✓	✓	✓
Check patient has/can adequately use PEF meter	✓
Treatment arm randomisation by dial-up centre		✓
Review action plan for worsening asthma	✓	✓	✓	✓	✓	✓	✓
Review any adverse experiences		✓	✓	✓	✓	✓	✓
Record PEF (no inhaled β2-agonist for 4 hours if possible)	✓	✓	✓	✓	✓	✓	✓
Confirm patient resource utilisation		✓	✓	✓	✓	✓	✓
Blinded research assistant/study office
Collect completed patient symptom diary card		✓	✓	✓	✓	✓	✓
Collect data on patient costs		✓	✓	✓	✓	✓	✓
Asthma QOL and EQ-5D (quality of life) questionnaries		✓	✓	✓	✓	✓	✓
Rhinitis questionnaires		✓	✓	✓	✓	✓	✓
Dispense patient diary card for subsequent visit	✓	✓	✓	✓	✓	✓
Collect resource use data from practice records							✓

Recruitment stage 2

At visit 1, participants (and parent or guardian if appropriate) gave written informed consent and were allocated a unique study number. Participants were reviewed for the following additional entry criteria:

• Peak expiratory flow, while withholding β₂-agonist for at least 4 hours, of > 50% predicted.

• Females of child-bearing potential agreed to use adequate contraception throughout the study.

Participants meeting the above criteria completed a 2-week PEF diary,63 ACQ,64 and asthma-specific QOL questionnaire (MiniAQLQ)50 prior to returning for visit 2.

Recruitment stage 3

At visit 2, participants scoring ≥ 1 on the ACQ (range 0–6, with ≤ 0.75 being optimal65) and/or ≤ 6 (out of a maximum best score of 7) on the MiniAQLQ were registered and randomised within the step 2 or step 3 study by an automated ‘dial-up’ centre at the University of East Anglia, Norwich, UK. A computer responded to the telephone calls from practices by recording identification information. It then used input from the practice about the step at which the patient was to enter the study to perform a look-up into predefined tables of randomisation allocations (see Randomisation, below) and then inform the caller of the allocation for that participant.

Interventions

Using a pragmatic, randomised controlled trial design, leukotriene antagonist prescription was compared with (1) inhaled steroid prescription at step 2 of the guidelines and (2) long-acting β₂-agonist against a background of inhaled steroid at step 3 (Figure 1). Patients and health-care providers were aware of treatment allocations, while study data collection and statistical analyses were blinded.

• Leukotriene receptor antagonist montelukast 10 mg, once daily (as Singulair^®; Merck, Sharp & Dohme Ltd, Hoddesden, UK) or zafirlukast 20 mg, twice daily (as Accolate^™, AstraZeneca Ltd, Kings Langley, UK).

• Inhaled corticosteroid – step 2 study inhaled beclometasone dipropionate, budesonide or fluticasone propionate.

• Long-acting β₂-agonist – step 3 study salmeterol (as Serevent^®, GlaxoSmithKline, Uxbridge, UK) or formoterol (as Foradil^®, Novartis Pharmaceuticals UK Ltd, Camberley, UK; or Oxis^®, AstraZeneca Ltd, Kings Langley, UK); these are also available in fixed dose combinations with inhaled steroid (as Seretide^™, GlaxoSmithKline, Uxbridge, UK and Symbicort^®, AstraZeneca Ltd, Kings Langley, UK).

All individual drug and device choices within treatment allocations were made according to normal clinical practice by the health professional involved (and bearing in mind BTS guidelines), subject to the restrictions outlined below.

Objectives

Outcome measures

• Primary outcome measure: The primary outcome was a between-group comparison of disease-specific QOL (described in Health status measures, below) and cost to achieve this to the NHS and patient at 2 months (the primary time point) and 2 years (described in section Resource use assessment, below).

• Secondary outcome measures:

  1. – ACQ score

  2. – number of asthma exacerbations – defined as requiring at least one course of oral corticosteroids or hospitalisation for asthma; when a patient received more than one course of oral steroid during the course of the study, any two courses of oral steroid prescribed within a 14-day period were considered as a single exacerbation, irrespective of the fact the patient required ≥ 2 courses of oral steroid.

  3. – attendance at primary care practice for upper and/or lower respiratory tract infections (number of total respiratory tract infections and number of primary care practice attendances for those respiratory tract infections)

  4. – short-acting β₂-agonist prescriptions

  5. – change in inhaled steroid dose (for step 3 participants only)

  6. – clinic PEF, percentage of predicted normal values calculated using the Roberts equation66

  7. – Mini Rhinitis Quality of Life Questionnaire (mRQLQ) scores67

  8. – Royal College of Physicians – three (RCP3) asthma questions scores68,69

  9. – personal objectives scores

  10. – changes in treatment after randomisation

  11. – adherence with prescribed therapy.

Safety was evaluated by the analysis of the overall incidence of adverse experiences.

Sample size and power calculation

This was based on the published literature50,75 regarding sample sizes for assessing treatment differences in QOL. Treating this as an equivalence study, and assuming no true difference between the treatments in QOL for a two-tailed alpha of 0.05 and an upper limit of 0.3 for the 95% confidence interval (CI) for the difference between arms, a sample size of 142 participants was required. To allow for a 20% dropout rate, we aimed to recruit 178 participants to each study arm, resulting in a total of 356 participants at each of steps 2 and 3 (totalling 712 participants).

Randomisation

Participants were registered for entry into the study after giving written informed consent and returning completed QOL questionnaires. At visit 2, participants eligible for entry into step 2 or step 3 studies were randomised into the study. Randomisation into the study was stratified by practice, with a block size of 6. Practice nurses were informed of the randomised treatment to be given to their patient via an automated telephone centre (see Participants, above).

Blinding

This was a single-blind randomised controlled trial. General practitioners (GPs)/practice asthma nurses and participants were aware of the randomisation, while study research assistants were blinded to the randomisation. The role of the GPs/practice asthma nurses and research assistants in the conduct of the study is described below:

• General practitioners/practice asthma nurses GPs/practice asthma nurses had minimal involvement in data collection except baseline prior to randomisation, implementing the randomisation allocation and thereafter in administering the resource data collection sheet with participants. This allowed clinical freedom to change treatment as per normal management. Randomisation allocation was given directly to the GPs/practice asthma nurses by an independent automated telephone answering system.

• Research assistants Research assistants were non-clinical personnel who worked with practice staff to ensure proper completion of the diaries and self-completed QOL and disease-related questionnaires. They collected resource use information from participants, data from prescribing records, and clinical resource utilisation data for the participants at the end of the study period. When collecting resource data the research assistants were blind to the randomised allocation of the participants.

Data and statistical analysis

All analyses were performed blind to study arm allocation. This section outlines the statistical analysis procedures that were performed.

Recruitment

All patients with any evidence of asthma from 53 participating practices were sent a postal asthma symptom questionnaire (adapted from the ACQ), which was used to evaluate initial eligibility. For patients meeting these initial criteria, a review of the their notes was undertaken to confirm eligibility. A further 80 patients from the practices were identified as meeting study entry criteria at the time of clinical visits by general practice staff. Of those patients considered potentially eligible, 449 (step 2) and 482 (step 3) responded positively to an invitation and were booked to attend a screening visit (Figures 2 and 3).

FIGURE 2.

Step 2, the Consolidated Standards of Reporting Trials (CONSORT) e-flowchart for primary end point.

FIGURE 3.

Step 3, the CONSORT e-flowchart for primary end point.

Work began on the study in October 2001, with initial piloting in one practice of study procedures completed by May 2002. Further practices were recruited from October 2001 through to September 2004. The first patient was enrolled on 3 May 2002 with the last step 3 patient being enrolled on 18 February 2004. The last step 2 patient was enrolled on 4 February 2005. The last clinical and QOL follow-up data were collected on 8 January 2007. The last resource data collection was in the same week.

Numbers analysed versus screened

For the step 2 trial, of the 449 screened, 123 participants were excluded (99 declined to participate and 24 were identified prior to randomisation as ineligible) and 326 participants were randomised (compared with the target of 356). No significant difference was found in mean age between excluded and analysed populations. There were more females among those excluded than those analysed (Table 2). For the step 3 trial, of the 482 screened, 121 participants were excluded (84 declined to participate and 37 were identified prior to randomisation as ineligible) and 361 participants were randomised (compared with the target of 356). No significant difference was found in either the sex distribution or mean age between excluded and analysed populations (Table 2).

Randomisation data

For the step 2 trial, this process resulted in an almost equal distribution of participants between leukotriene antagonist and inhaled steroid arms (162 and 164 participants, respectively). However, for the step 3 study, 9 fewer participants were randomised to leukotriene antagonist than to long-acting β₂-agonist (176 and 185 participants, respectively). This difference is likely to have arisen for two reasons. Firstly, a small number of practices had one or two more participants randomised to long-acting β₂-agonist than leukotriene antagonist. Secondly, because of an error in the way that the randomisation telephone calls were performed, practice 12 had four more participants randomised to long-acting β₂-agonist than to leukotriene antagonist (13 and 9 participants, respectively). There was never any prior or biasing knowledge of the allocation on the part of the nurse performing the randomisation, but the result was an excess of four participants receiving long-acting β₂-agonist at that practice.

Step 2 trial

Primary analyses

Change in QOL

Mini asthma quality of life questionnaire

Mean MiniAQLQ score increased (improved) from baseline in both leukotriene antagonist and inhaled steroid randomised groups (Table 7 and Figure 4).

TABLE 7 - Step 2 trial: mean (SD) and differences (95% CI) between means for MiniAQLQ scores at 2 months and 2 years

CI, confidence interval; ICS, inhaled corticosteroid; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Adjusted for baseline values.

Multiple imputation was used to impute missing data.

Treatment duration	Outcome measure	LTRA	ICS	Difference (95% CI) LTRA–ICS	Adjusted differencea (95% CI)
2 months (visit 3)	n	122	132	0.0 (–0.25 to 0.26)	–0.02 (–0.24 to 0.20)
	Mean	5.25	5.28
	SD	1.03	1.10
2 yearsb	n	145	155	–0.10 (–0.35 to 0.17)	–0.11 (–0.35 to 0.13)
	Mean	5.52	5.63
	SD	1.07	1.16

FIGURE 4.

Step 2: ACQ, MiniAQLQ and %PPEF over 2 years of treatment. Mean (standard deviation) ACQ (from 0 ‘best’ to 6 ‘worst’) and MiniAQLQ (from 1 ‘worst’ to 7 ‘best’) scores and median (interquartile range) %PPEF at baseline and over 2 years of treatment with leukotriene antagonist or inhaled steroid.

No statistically significant between-group differences in MiniAQLQ score were found at the 2-month time point, either unadjusted or adjusted for baseline values. At 2 months, the 95% CIs for both the unadjusted and adjusted differences in MiniAQLQ score were within the limits for equivalence: (–0.25 to 0.26) and –0.02 (–0.24 to 0.20), respectively, i.e. excluding 0.3. The limit of the one-sided 95% CI for the unadjusted difference was –0.25, and –0.18 for the adjusted difference.

At the 2-year visit, while the difference was again not statistically significant, and the estimated difference between groups was small, the 95% CI did include the equivalence value of 0.3, favouring inhaled steroid [imputed results, unadjusted difference (95% CI) –0.10 (–0.35 to 0.17), adjusted difference (95% CI) –0.11 (–0.35 to 0.13)]. The limit of the one-sided 95% CI for the unadjusted difference was –0.32 and was –0.31 for the adjusted difference, i.e. inferiority could not be excluded.

Asthma Control Questionnaire

Mean ACQ score decreased (improved) substantially from baseline in both leukotriene antagonist and inhaled steroid randomised groups (Table 8 and Figure 4). Again, no significant between-group differences in ACQ score were found at either the 2-month or 2-year time points, whether unadjusted or adjusted for baseline values. At 2 years (imputed results), the adjusted difference (95% CI) was 0.13 (–0.07 to 0.33). The CI is well within the minimum clinically important difference of 0.5.

TABLE 8 - Step 2 trial: mean (SD) and differences (95% CI) between means for ACQ scores at 2 months and 2 years

CI, confidence interval; ICS, inhaled corticosteroid; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Adjusted for baseline values.

Multiple imputation was used to impute missing data.

Treatment duration		LTRA	ICS	Difference (95% CI) LTRA–ICS	Adjusted differencea (95% CI)
2 months (visit 3)	n	123	132	–0.02 (–0.24 to 0.21)	0.01 (–0.20 to 0.22)
	Mean	1.54	1.53
	SD	0.93	1.00
2 yearsb	n	145	155	0.10 (–0.11 to 0.32)	0.13 (–0.07 to 0.33)
	Mean	1.23	1.15
	SD	0.95	0.92

Quality-adjusted life-years

Leukotriene antagonist participants experienced a mean of 0.122 QALYs over the 2-month period, compared with 0.132 in inhaled steroid participants, a mean (95% CI) difference of approximately 0.01 (–0.019 to 0.001) QALYs, falling to 0.001 (–0.005 to 0.002) QALYs after adjustment for baseline utility (Table 9). Over 2 years, leukotriene antagonist participants experienced 0.153 (–0.274 to –0.032) fewer QALYs than inhaled steroid participants. However, after adjusting for baseline utility, the difference falls to 0.050 (–0.126 to 0.026) QALYs, equivalent to 2.5 weeks of perfect health.

TABLE 9 - Step 2 trial: mean (SD) EQ-5D and differences between means for QALYs at 2 months and 2 years

CI, confidence interval; ICS, inhaled corticosteroid; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Adjusted for baseline utility.

	LTRA		ICS		Difference (95% CI) LTRA–ICS	Adjusted differencea (95% CI)
	n	Mean (SD)	n	Mean (SD)
EQ-5D utility
Baseline	118	0.795 (0.245)	131	0.830 (0.195)	–	–
2 months	118	0.819 (0.261)	124	0.856 (0.212)	–	–
2 years	132	0.826 (0.261)	143	0.881 (0.218)	–	–
QALYs gained
2 months	95	0.122 (0.036)	106	0.132 (0.025)	–0.010 (–0.019 to 0.001)	–0.001 (–0.005 to 0.002)
2 years (discounted)	81	1.569 (0.458)	94	1.722 (0.327)	–0.153 (–0.274 to –0.032)	–0.050 (–0.126 to 0.026)

Resource use and cost

Point estimate costs and quantities of prescription medicines, over-the-counter medicines, NHS activity and indirect costs are reported in Tables 10–13. Total NHS costs are the sum of prescriptions and NHS activity. Total societal costs are NHS costs plus over-the-counter medications and indirect costs.

TABLE 10 - Step 2 trial: cost of prescription medicines at 2 months and 2 yearsa

ICS, inhaled corticosteroid; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Total and mean (SD) costs of asthma-related prescription medicines prescribed during the step 2 trial.

Treatment duration	LTRA					ICS
	n	Total items	Mean (SD) items	Total cost (£)	Mean (SD) cost (£)	n	Total items	Mean items (SD)	Total cost (£)	Mean (SD) cost (£)
2 months	159	463	2.88 (1.87)	10,055.61	62.46 (31.30)	164	330	2.01 (1.38)	2969.78	18.11 (14.92)
2 years (discounted)	154	3489	22.7 (16.1)	60,509.00	393 (299)	157	2797	17.8 (15.2)	28,882.00	184 (237)

TABLE 11 - Step 2 trial: cost of over-the-counter medicines at 2 months and 2 yearsa

ICS, inhaled corticosteroid; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Total and mean (SD) costs of asthma-related over-the-counter medicines prescribed during the step 2 trial.

Treatment duration	LTRA					ICS
	n	Total items	Mean (SD) items	Total cost (£)	Mean (SD) cost (£)	n	Total items	Mean items (SD)	Total cost (£)	Mean (SD) cost (£)
2 months	109	49	0.45 (0.73)	211.26	1.94 (6.96)	108	79	0.73 (1.16)	1041.07	9.64 (46.97)
2 years (discounted)	109	225	2.06 (1.52)	756.39	6.94 (11.83)	108	268	2.48 (2.33)	1637.09	15.16 (49.66)

TABLE 12 - Step 2 trial: cost of NHS activity at 2 months and 2 yearsa

ICS, inhaled corticosteroid; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Total and mean (SD) costs of asthma NHS health care-related activity during the step 2 trial.

Treatment	LTRA					ICS
	n	Total contacts	Mean (SD) contacts	Total cost (£)	Mean (SD) cost (£)	n	Total contacts	Mean (SD) contacts	Total cost (£)	Mean (SD) cost (£)
2 months
Total routine care	156	41	0.26 (0.63)	331.67	2.13 (5.52)	158	25	0.16 (0.56)	255.00	1.61 (5.98)
Total patient initiated	156	50	0.32 (0.75)	1153.05	7.39 (22.79)	158	22	0.14 (0.46)	371.00	2.35 (8.08)
Total secondary care	156	4	0.03 (0.25)	635.95	4.08 (41.93)	158	3	0.02 (0.18)	381.00	2.41 (22.53)
Total NHS activity	156	95	0.61 (1.09)	2120.67	13.59 (61.00)	158	50	0.32 (0.71)	1007.00	6.37 (24.16)
2 years (discounted)
Total routine care	151	474	3.14 (2.53)	4055.67	26.86 (24.92)	151	420	2.78 (2.32)	3827.35	25.35 (22.92)
Total patient initiated	151	397	2.63 (3.19)	8761.64	58.02 (78.46)	151	416	2.75 (3.46)	9181.14	60.80 (81.95)
Total secondary care	151	40	0.26 (0.87)	12,045.23	79.77 (601.74)	151	55	0.36 (1.06)	8446.08	55.93 (268.62)
Total NHS activity	151	911	6.03 (4.89)	24,862.54	164.65 (633.99)	151	891	5.90 (5.25)	21,454.57	142.08 (310.13)

TABLE 13 - Step 2 trial: indirect costs at 2 months and 2 yearsa

ICS, inhaled corticosteroid; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Total and mean (SD) indirect costs during the step 2 trial.

Treatment duration	LTRA					ICS
	n	Total hours	Mean (SD) hours	Total cost (£)	Mean (SD) cost (£)	n	Total hours	Mean hours (SD)	Total cost (£)	Mean (SD) cost (£)
2 months	80	335	4.19 (18.54)	4398.55	54.98 (243.43)	71	142	2.00 (10.37)	1864.46	26.26 (136.16)
2 years (discounted)	80	1309.5	16.37 (62.61)	14,938.39	186.73 (740.80)	71	1075	15.14 (86.24)	14,114.75	198.80 (1132.29)

Over 2 months and 2 years, patients who were prescribed leukotriene antagonists were significantly more expensive than patients who were prescribed inhaled steroid participants in terms of both NHS (NHS activity plus prescription medicines) and societal costs (all costs – Table 14). At 2 months, leukotriene antagonist participants cost on average £70 more, or 2.5 times the cost, of inhaled steroid participants (95% CI £29 to £111), and at 2 years, £294 more, or 1.8 times the cost of inhaled steroid participants (95% CI £107 to £481). This may be driven principally by higher prescription costs and possibly compliance (including leukotriene antagonists themselves) and/or higher NHS activity in the leukotriene antagonist group.

TABLE 14 - Step 2 trial: total NHS and societal costs at 2 months and 2 yearsa

CI, confidence interval; ICS, inhaled corticosteroid; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Total and mean (SD) NHS and societal costs during the step 2 trial.

Treatment duration	LTRA			ICS			Cost (£) difference (95% CI)
	n	Total cost (£)	Mean (SD) cost (£)	n	Total cost (£)	Mean (SD) cost (£)
Total NHS costs (prescription medicines and NHS activity combined)
2 months	156	12,029.60	77.11 (66.38)	158	3839.82	24.30 (30.56)	52.81 (41.29 to 64.33)
2 years (discounted)	147	84,309.00	573.00 (764)	148	49,090.00	332.00 (435)	242.00 (100 to 384)
Societal costs (prescription and over-the-counter medicines – NHS activity and indirect costs combined)
2 months	74	8704.96	117.63 (152.94)	65	3097.17	47.65 (81.52)	69. 99 (29.47 to 110.50)
2 years (discounted)	70	46,628.00	666.00 (664)	60	22,341.00	372.00 (374)	294.00 (107 to 481)

Cost-effectiveness analyses

Mini Asthma Quality of Life Questionnaire

At both 2 months (Table 15) and 2 years (Table 16), inhaled steroid is, on average, a dominant strategy over leukotriene antagonist. Mean INB is negative, irrespective of willingness to pay for a point improvement in MiniAQLQ.

TABLE 15 - Step 2: incremental cost-effectiveness ratio at 2 monthsa

ICS, inhaled corticosteroid; inc., incremental; LTRA, leukotriene receptor antagonist.

Inc. cost, inc. MiniAQLQ and ICERs are reported for patient groups with complete societal and NHS cost and MiniAQLQ data, following imputation for missing data and adjustment for baseline MiniAQLQ.

	n		Cost (£)		MiniAQLQ		Source data only			Including imputed data			Including imputed data (adjusted for baseline MiniAQLQ)
	LTRA	ICS	LTRA	ICS	LTRA	ICS	Inc. cost (£)	Inc. AQLQ	ICER (£)	Inc. cost (£)	Inc. AQLQ	ICER (£)	Inc. cost (£)	Inc. AQLQ	ICER (£)
NHS	120	127	80	25	5.242	5.278	55	–0.036	(ICS dominant)	52	–0.019	(ICS dominant)	53	–0.081	(ICS dominant)
Societal	63	57	117	52	5.257	5.410	65	–0.153	(ICS dominant)	37	–0.019	(ICS dominant)	38	–0.081	(ICS dominant)

TABLE 16 - Step 2: incremental cost-effectiveness ratio at 2 yearsa

ICS, inhaled corticosteroid; inc., incremental; LTRA, leukotriene receptor antagonist.

Inc. cost, inc. MiniAQLQ and ICERs are reported for patient groups with complete societal and NHS cost and MiniAQLQ data, following imputation for missing data and adjustment for baseline MiniAQLQ.

	n		Cost (£)		MiniAQLQ		Source data only			Including imputed data			Including imputed data (adjusted for baseline MiniAQLQ)
	LTRA	ICS	LTRA	ICS	LTRA	ICS	Inc. cost (£)	Inc. AQLQ	ICER (£)	Inc. cost (£)	Inc. AQLQ	ICER (£)	Inc. cost (£)	Inc. AQLQ	ICER (£)
NHS	133	140	587	327	5.529	5.633	260	–0.104	(ICS dominant)	233	–0.086	(ICS dominant)	242	–0.151	(ICS dominant)
Societal	64	60	666	372	5.549	5.681	294	–0.132	(ICS dominant)	209	–0.086	(ICS dominant)	217	–0.151	(ICS dominant)

Incremental net benefit details (± 95% CI) are shown in Figure 5a–d. Note, however, that there is a great deal of uncertainty around these estimates (95% CI in Figure 5).

FIGURE 5.

Step 2: INB, MiniAQLQ, INB and 95% CI are shown for (a) 2-month NHS; (b) 2-month societal; (c) 2-year NHS; and (d), results based on imputed data adjusted for baseline MiniAQLQ.

Even with a very high willingness to pay for a point improvement in MiniAQLQ, there is, at most, only a 25% probability of leukotriene antagonist being cost-effective compared with inhaled steroid in step 2 patients (Figure 6).

FIGURE 6.

Step 2: MiniAQLQ CEACs. Based on imputed, adjusted results of MiniAQLQ, societal/NHS indicates that results were from a societal or NHS cost perspective, respectively, and these outcomes took place at 2 months or 2 years. This figure indicates that in all cases the probability of leukotriene antagonist being cost-effective compared with inhaled steroid is < 50%, irrespective of the willingness to pay for a point improvement in MiniAQLQ.

Asthma Control Questionnaire

At both 2 months (Table 17) and 2 years (Table 18), inhaled steroid is on average, a dominant strategy over leukotriene antagonist. Mean INB is negative, irrespective of willingness to pay for a point improvement in ACQ. However, there is great deal of uncertainty around this estimate (95% CIs in Figure 7).

TABLE 17 - Step 2: ICER at 2 monthsa

ICS, inhaled corticosteroid; inc., incremental; LTRA, leukotriene receptor antagonist.

Inc. cost, inc. ACQ and ICERs are reported for patient groups with complete societal and NHS cost and ACQ data, following imputation for missing data and adjustment for baseline ACQ.

	n		Cost (£)		ACQ		Source data only			Including imputed data			Including imputed data (adjusted for baseline ACQ)
	LTRA	ICS	LTRA	ICS	LTRA	ICS	Inc. cost (£)	Inc. ACQ	ICER (£)	Inc. cost (£)	Inc. ACQ	ICER (£)	Inc. cost (£)	Inc. ACQ	ICER (£)
NHS	121	127	80	25	1.550	1.541	55	–0.009	(ICS dominant)	52	–0.015	(ICS dominant)	53	–0.083	(ICS dominant)
Societal	64	57	116	52	1.544	1.489	64	–0.055	(ICS dominant)	37	–0.015	(ICS dominant)	38	–0.083	(ICS dominant)

TABLE 18 - Step 2: ICER at 2 yearsa

ICS, inhaled corticosteroid; inc., incremental; LTRA, leukotriene receptor antagonist.

Inc. cost, inc. ACQ (95% CI) and ICERs are reported for patient groups with complete societal and NHS cost and ACQ data and following imputation for missing data.

	n		Cost (£)		ACQ		Source data only			Including imputed data			Including imputed data (adjusted for baseline ACQ)
	LTRA	ICS	LTRA	ICS	LTRA	ICS	Inc. cost (£)	Inc. ACQ	ICER (£)	Inc. cost (£)	Inc. ACQ	ICER (£)	Inc. cost (£)	Inc. ACQ	ICER (£)
NHS	133	140	587	327	1.223	1.142	260	–0.081	(ICS dominant)	233	–0.062	(ICS dominant)	249	–0.116	(ICS dominant)
Societal	64	60	666	372	1.086	1.094	294	0.009	34546	209	–0.062	(ICS dominant)	220	–0.116	(ICS dominant)

FIGURE 7.

Step 2: INB, ACQ, INB and 95% CI are shown for (a) 2-month NHS; (b) 2-month societal; (c) 2-year NHS; and (d) 2-year societal perspectives. Results based on imputed data adjusted for baseline ACQ.

Incremental net benefit (± 95% CI) is shown in Figure 7a–d. A higher ACQ score indicates worsening of asthma control, and negative incremental ACQ indicates that leukotriene antagonist is less effective than inhaled steroid.

Even with a very high willingness to pay for a point improvement in ACQ, there is only at most a 22% probability of leukotriene antagonist being cost-effective compared with inhaled steroid in step 2 patients (Figure 8).

FIGURE 8.

Step 2: ACQ CEACs. Based on imputed, adjusted results of ACQ, societal/NHS indicates that results were from a societal or NHS cost perspective, respectively, and these outcomes took place at 2 months or 2 years. This figure indicates that in all cases the probability of leukotriene antagonist being cost-effective compared with inhaled steroid is < 50%, irrespective of the willingness to pay for a point improvement in ACQ.

Quality-adjusted life-years (cost–utility analysis)

At both 2 months (Table 19) and 2 years (Table 20) inhaled steroid is, on average, a dominant strategy over leukotriene antagonist. Mean INB is negative, irrespective of willingness to pay for QALY (Figure 9).

TABLE 19 - Step 2: ICER at 2 months

CI, confidence interval; ICS, inhaled corticosteroid; inc., incremental; LTRA, leukotriene receptor antagonist.

Inc. cost, inc. QALY (95% CI), ICERs and probability that the intervention is cost-effective given a willingness to pay of £30,000 for QALY (P£30 k, see Figure 6) are reported for patient groups with complete societal and NHS cost and QALY data and following imputation for missing data.

	n		Cost (£)		QALYs		Source data only					Including imputed data					Including imputed data (adjusted for baseline utility)
	LTRA	ICS	LTRA	ICS	LTRA	ICS	Inc. cost (£)	Inc. QALYs	ICER (£)	P£30 k (%)	INB/£30 k (95% CI)	Inc. cost (£)	Inc. QALYs	ICER (£)	P£30 k (%)	INB/£30 k (95% CI)	Inc. cost (£)	Inc. QALYs	ICER (£)	P£30 k (%)	INB/£30 k (95% CI)
NHS	94	102	85	27	0.122	0.133	58	–0.011	(ICS dominant)	0.10	–277 (–454 to 97)	52	–0.006	(ICS dominant)	1.24	–177(–151 to 112)	52	–0.004	(ICS dominant)	2.82	–155 (–305 to 0)
Societal	52	45	115	59	0.118	0.135	56	–0.017	(ICS dominant)	0.20	–406 (–677 to 159)	37	–0.006	(ICS dominant)	2.12	–128 (–212 to 142)	36	–0.004	(ICS dominant)	3.50	–145 (–305 to 8)

TABLE 20 - Step 2: ICER at 2 yearsa

CI, confidence interval; ICS, inhaled corticosteroid; inc., incremental; LTRA, leukotriene receptor antagonist.

Inc. cost, inc. QALY (95% CI), ICERs and probability that the intervention is cost-effective given a willingness to pay of £30,000 for QALY (PE30 k, see Figure 6) are reported for patient groups with complete societal and NHS cost and QALY data and following imputation for missing data.

	n		Cost (£)		QALYs		Source data only					Including imputed data					Including imputed data (adjusted for baseline utility)
	LTRA	ICS	LTRA	ICS	LTRA	ICS	Inc. cost (£)	Inc. QALYs	ICER (£)	P£30 k (%)	INB/£30 k (95% CI)	Inc. cost (£)	Inc. QALYs	ICER (£)	P£30 k (%)	INB|£30 k (95% CI)	Inc. cost (£)	Inc. QALYs	ICER (£)	P£30 k (%)	INB/£30 k (95% CI)
NHS	78	89	616	369	1.571	1.736	247	–0.165	(ICS dominant)	0.00	–3551(–5986 to 1270)	233	–0.091	(ICS dominant)	0.98	–698	228	–0.073	(ICS dominant)	3.30	–2255 (–4550 to 211)
Societal	43	39	711	433	1.524	1.731	278	–0.207	(ICS dominant)	1.00	–4349(–8157 to 762)	209	–0.091	(ICS dominant)	1.26	–87	204	–0.073	(ICS dominant)	3.14	–2241 (–4542 to 118)

FIGURE 9.

Step 2: INB, QALYs, INB and 95% CI are shown for (a) 2-month NHS; (b) 2-month societal; (c) 2-year NHS; and (d) 2-year societal perspectives. Results based on imputed data adjusted for baseline utility.

At a typical willingness to pay of £30,000 per QALY gained, mean (95% CI) INB at 2 months is –£155 (–£305 to £0) or –£145 (–£305 to £8) from the NHS and societal perspectives, respectively, deteriorating to –£2255 (–£4450 to £211) and –£2241 (–£4542 to £118) at 2 years. At the £30,000 threshold, there is a very low probability (2.8% to 3.5%) of leukotriene antagonist being cost-effective compared with inhaled steroid in patients at step 2 of the national asthma guidelines (Tables 19 and 20, and Figure 10).

FIGURE 10.

Step 2: Cost–utility analysis – CEACs. Based on imputed, adjusted results of cost–utility analysis, societal/NHS indicates that results were from a societal or NHS cost perspective, respectively, and these outcomes took place at 2 months or 2 years. This figure indicates that the probability of leukotriene antagonist being cost-effective compared with inhaled corticosteroid is approximately 2.8–3.5%, given a typical willingness to pay of £30,000 per QALY gained.

Figure 10 indicates that in all cases the probability of leukotriene antagonist being cost-effective compared with inhaled steroid is below 50%, irrespective of the willingness to pay for a QALY.

Summary of cost-effectiveness analyses

The results of the cost-effectiveness analyses suggest that leukotriene antagonists are both more expensive and less, although not statistically significantly, effective than inhaled steroids, with a very low probability of cost-effectiveness compared with inhaled steroids for patients initiating controller therapy at step 2 of the national asthma guidelines.

Whether an intervention is deemed cost-effective is dependent upon the willingness-to-pay threshold for a unit of outcome. For QALYs, a ‘reasonable’ willingness to pay is thought to be in the order of £30,000. The threshold for a point improvement in MiniAQLQ or ACQ is less well established. However, even given an infinite willingness to pay for a point improvement in MiniAQLQ or ACQ, there is at best a 22–25% probability that leukotriene antagonists are a cost-effective alternative to inhaled steroids in patients at step 2 of the national asthma guidelines (depending on time horizon and perspective).

Step 3 trial

Primary analyses

Change in QOL

Mini Asthma Quality of Life Questionnaire

Mean MiniAQLQ score increased from baseline in both leukotriene antagonist and long-acting β₂-agonist randomised groups (Table 30 and Figure 11). No statistically significant between-group differences in MiniAQLQ score were found at the 2-month time point either unadjusted or adjusted for baseline values. In each case, the point estimate of the mean difference between groups was small. The 95% CIs for the unadjusted difference in MiniAQLQ score was just at the limits for equivalence: (–0.18 to 0.30). However, the 95% CI for the adjusted difference was within the value of 0.3, i.e. (–0.29 to 0.10). The limit of the one-sided 95% CI for the unadjusted difference was –0.16 and was –0.28 for the adjusted difference.

TABLE 30 - Step 3 trial: MiniAQLQ scoresa

CI, confidence interval; LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Mean (SD) and differences (95% CI) between means for MiniAQLQ scores at 2 months and 2 years.

Adjusted for baseline values.

Multiple imputation was used to impute missing data.

Treatment duration	Outcome measure	LTRA	LABA	Difference (95% CI) LTRA–LABA	Adjusted differenceb (95% CI)
2 months (visit 3)	n Mean (SD)	153 5.09 (1.15)	156 5.04 (1.11)	0.06 (–0.18 to 0.30)	–0.10 (–0.29 to 0.10)
2 yearsc (visit 7)	n Mean (SD)	169 5.43 (1.14)	181 5.42 (1.08)	0.01 (–0.22 to 0.25)	–0.11 (–0.32 to 0.11)

FIGURE 11.

Step 3: ACQ, MiniAQLQ and %PPEF over 2 years of treatment. Mean (standard deviation) ACQ (from 0 ‘best’ to 6 ‘worst’) and MiniAQLQ (from 1 ‘worst’ to 7 ‘best’) scores and median (interquartile range) %PPEF at baseline and over 2 years of treatment with leukotriene receptor antagonist or long-acting β₂-agonist as add-on to ICS.

At the 2-year visit, the estimated between-group mean difference was again small and not statistically significant. The 95% CI for the unadjusted difference did not include 0.3 (–0.22 to 0.25). The 95% CI for the adjusted difference did, although marginally, include 0.3 in favour of long-acting β₂-agonist (–0.32 to 0.11). The limit of the one-sided 95% CI for the unadjusted difference was –0.30 and was –0.28 for the adjusted difference, i.e. inferiority of leukotriene antagonist could be excluded.

Asthma Control Questionnaire

Mean ACQ score decreased (improved) from baseline in both leukotriene antagonist and long-acting β₂-agonist randomised groups (Table 31 and Figure 11). No significant differences in ACQ score were found at either the 2-month or 2-year time points. At 2 years, the adjusted difference (95% CI) was 0.04 (–0.15 to 0.22). The CI is well within the minimum clinically important difference of 0.5.

TABLE 31 - Step 3 trial: ACQ scores. Mean (SD) and differences (95% CI) between means for ACQ scores at 2 months and 2 years

CI, confidence interval; LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Adjusted for baseline values.

Multiple imputation was used to impute missing data.

Treatment duration		LTRA	LABA	Difference (95% CI) LTRA–LABA	Adjusted differencea (95% CI)
2 months (visit 3)	n Mean (SD)	153 1.62 (1.00)	156 1.60 (0.98)	0.01 (–0.20 to 0.22)	0.12 (–0.06 to 0.30)
2 yearsb (visit 7)	n Mean (SD)	169 1.31 (0.96)	181 1.34 (0.92)	–0.04 (–0.24 to 0.16)	0.04 (–0.15 to 0.22)

Quality-adjusted life-years

Leukotriene antagonist patients experienced a mean 0.122 QALYs over the 2-month period, compared with 0.120 in long-acting β₂-agonist patients, a mean (95% CI) difference of +0.002 (–0.007 to 0.010) QALYs, falling to –0.001 (–0.004 to 0.002) QALYs after adjustment for baseline utility (Table 32). Over 2 years, leukotriene antagonist patients gained 0.041 (–0.072 to 0.156) more QALYs than long-acting β₂-agonist patients. However, after adjusting for baseline utility, the difference drops to 0.012 (–0.064 to 0.088) QALYs, equivalent to 4.4 days of perfect health gained.

TABLE 32 - Step 3 trial: mean (SD) EQ-5D and differences between means for QALYs at 2 months and 2 years

CI, confidence interval; LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Adjusted for baseline utility.

	LTRA		LABA		Difference (95% CI) LTRA–LABA	Adjusted differencea (95% CI)
	n	Mean (SD)	n	Mean (SD)
EQ-5D utility
Baseline	148	0.780 (0.237)	159	0.772 (0.234)	–	–
2 months	127	0.796 (0.267)	130	0.794 (0.225)	–	–
2 years	160	0.807 (0.255)	170	0.798 (0.268)	–	–
QALYs gained
2 months	120	0.122 (0.35)	120	0.120 (0.032)	0.002 (–0.007 to 0.010)	–0.001 (–0.004 to 0.002)
2 years (discounted)	112	1.597 (0.418)	115	1.556 (0.451)	0.041 (–0.072 to 0.156)	0.012 (–0.064 to 0.088)

Resource use and cost

Point estimate costs and quantities of prescription medicines, over-the-counter medicines, NHS activity and indirect costs are reported in Tables 33–36.

TABLE 33 - Step 3 trial: cost of prescription medicines at 2 months and 2 yearsa

LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Total and mean (SD) costs of asthma-related prescription medicines prescribed during the step 3 trial.

Treatment duration	LTRA					LABA
	n	Total items	Mean (SD) items	Total cost (£)	Mean (SD) cost (£)	n	Total items	Mean items (SD)	Total cost (£)	Mean (SD) cost (£)
2 months	175	707	4.02 (2.30)	12,984	73.77 (33.22)	184	682	3.69 (2.61)	13,403	73 (38)
2 years (discounted)	169	7129	42.2 (27.3)	116,477	689 (404)	177	6262	35.4 (31.1)	103,599	585 (437)

TABLE 34 - Step 3 trial: cost of over-the-counter medicines at 2 months and 2 yearsa

LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Total and mean (SD) costs of asthma-related over-the-counter medicines prescribed during the step 3 trial.

Treatment duration	LTRA					LABA
	n	Total items	Mean (SD) items	Total cost (£)	Mean (SD) cost (£)	n	Total items	Mean items (SD)	Total cost (£)	Mean (SD) cost (£)	Cost (£) difference
2 months	131	130	0.99 (0.96)	453.34	3.46 (11.32)	142	133	0.94 (0.88)	362.80	2.55 (7.02)	0.91
2 years (discounted)	131	293	2.24 (2.11)	1024.69	7.82 (18.82)	142	311	2.19 (1.46)	850.38	5.99 (10.54)	1.83

TABLE 35 - Step 3 trial: cost of NHS activity at 2 months and 2 yearsa

LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Total and mean (SD) costs of asthma NHS health care-related activity during the step 3 trial.

Treatment	LTRA					LABA
	n	Total items	Mean (SD) items	Total cost (£)	Mean (SD) cost (£)	n	Total items	Mean items (SD)	Total cost (£)	Mean (SD) cost (£)
2 months
Total routine care	175	38	0.22 (0.59)	403.34	2.30 (6.32)	182	47	0.26 (0.73)	398.34	2.19 (6.61)
Total patient initiated	175	59	0.34 (0.79)	1154	6.59 (17.17)	182	56	0.31 (0.75)	1129	6.20 (15.54)
Total secondary care	175	1	0.01 (0.08)	191	1.09 (17.44)	182	4	0.02 (0.18)	76	0.42 (3.43)
Total NHS activity	175	98	0.56 (0.99)	1748.34	9.99 (25.00)	182	107	0.59 (1.05)	1603.34	8.81 (17.64)
2 years (discounted)
Total routine care	169	646	3.82 (2.63)	5710.33	33.79 (26.53)	176	699	3.97 (2.56)	5934.30	33.72 (24.49)
Total patient initiated	169	654	3.87 (4.20)	13,333.99	78.90 (89.75)	176	561	3.19 (2.94)	12,187.43	69.25 (71.42)
Total secondary care	169	74	0.44 (1.16)	12,873.62	76.18 (370.05)	176	47	0.27 (0.91)	13,091.82	74.39 (511.48)
Total NHS activity	169	1374	8.13 (5.94)	31,917.94	188.86 (398.89)	176	1307	7.43 (4.19)	31,213.56	177.35 (541.93)

TABLE 36 - Step 3 Trial: indirect costs at 2 months and 2 yearsa

LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Total and mean (SD) indirect costs during the step 3 trial.

Treatment duration	LTRA					LABA
	n	Total hours	Mean (SD) hours	Total cost (£)	Mean (SD) cost (£)	n	Total hours	Mean hours (SD)	Total cost (£)	Mean (SD) cost (£)
2 months	110	292	2.65 (13.94)	3833.96	34.85 (183.08)	115	295	2.57 (13.57)	3873.35	33.68 (178.14)
2 years (discounted)	110	1468	13.35 (51.96)	19,274.84	175.23 (682.17)	115	1077	9.37 (53.84)	13,368.11	116.24 (702.80)

Total NHS costs are the sum of prescriptions and NHS activity. Total societal costs are NHS costs plus over-the-counter medications and indirect costs. At 2 months, costs were similar between the two groups, although point estimate was higher in leukotriene antagonist patients [+£2 (95% CI –£7 to +£11 from NHS perspective), +£15 (95% CI –£35 to +£65 from societal perspective – Table 37)]. At 2 years, leukotriene antagonist patients cost £115 (–£46 to £276) or £263 (–£3 to £529) more than long-acting β₂-agonist patients from the NHS or societal perspectives, respectively.

TABLE 37 - Step 3 trial: total NHS and societal costs at 2 months and 2 yearsa

CI, confidence interval; LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist; SD, standard deviation.

Total and mean (SD) NHS and societal costs during the step 3 trial.

Treatment	LTRA			LABA
	n	Total cost (£)	Mean (SD) cost (£)	n	Total cost (£)	Mean (SD) cost (£)	Cost (£) difference (95% CI)
Total NHS costs (prescription medicines and NHS activity combined)
2 months	175	14,732.57	84.19 (41.71)	262.83 (–3 to 529)	14,911.91	81.93 (45.84)	2.25 (–6.88 to 11.39)
2 years (discounted)	166	147,299.54	887.35 (617.05)		132,052.97	772.24 (862.96)	115.11 (–46 to 276)
Societal costs (prescription and over-the-counter medicines – NHS activity and indirect costs combined)
2 months	101	12,818.31	126.91 (202.11)		11,655.96	112.08 (158.73)	14.84 (–35.13 to 64.80)
2 years (discounted)	97	111,196.77	1146.36 (965.55)		87,469.45	883.53 (923.89)	262.83 (–3 to 529)

Cost-effectiveness analyses

MiniAQLQ

The incremental cost per point improvement in MiniAQLQ is £48 and £115, respectively, from the NHS and societal perspectives at 2 months (Table 38). At 2 years this has deteriorated to £3366 and £6267 (Table 39). Mean INB is positive so long as the willingness to pay for a point improvement in MiniAQLQ is greater than these mean estimates. However, we did not detect a statistically significant result at any threshold (Figure 12).

TABLE 38 - Step 3: ICER at 2 monthsa

Inc., incremental; LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist.

Inc. cost, inc. MiniAQLQ (95% CI) and ICERs are reported for patient groups with complete societal and NHS cost and MiniAQLQ data and following imputation for missing data (*).

	n		Cost (£)		MiniAQLQ		Source data only			Including imputed data			Including imputed data (adjusted for baseline MiniAQLQ)
	LTRA	LABA	LTRA	LABA	LTRA	LABA	Inc. cost (£)	Inc. AQLQ	ICER (£)	Inc. cost (£)	Inc. AQLQ	ICER (£)	Inc. cost (£)	Inc. AQLQ	ICER (£)
NHS	153	155	84	82	5.092	5.038	3	0.055	47	3	0.040	67	3	0.058	48
Societal	95	90	129	115	5.110	4.978	14	0.132	109	6	0.040	162	7	0.058	115

TABLE 39 - Step 3: ICER at 2 yearsa

Inc., incremental; LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist.

Inc. cost, inc. MiniAQLQ (95% CI) and ICERs are reported for patient groups with complete societal and NHS cost and MiniAQLQ data and following imputation for missing data(*).

	n		Cost (£)		MiniAQLQ		Source data only			Including imputed data			Including imputed data (adjusted for baseline MiniAQLQ)
	LTRA	LABA	LTRA	LABA	LTRA	LABA	Inc. cost (£)	Inc. AQLQ	ICER (£)	Inc. cost (£)	Inc. AQLQ	ICER (£)	Inc. cost (£)	Inc. AQLQ	ICER (£)
NHS	155	164	891	791	5.452	5.416	100	0.037	2734	111	0.016	6886	114	0.034	3366
Societal	94	96	1152	893	5.352	5.438	259	–0.086	(LABA dominant)	213	0.016	13,235	215	0.034	6367

FIGURE 12.

Step 3: INB, MiniAQLQ, INB and 95% CI for (a) 2-month NHS; (b) 2-month societal; (c) 2-year NHS; and (d) 2-year societal perspectives. Results based on imputed data adjusted for baseline MiniAQLQ.

At higher thresholds, there is between a 61% and 68% probability of leukotriene antagonist being cost-effective compared with long-acting β₂-agonist in step 3 patients (Figure 13). However, from a societal perspective over 2 years, there is a 50% or greater probability of leukotriene antagonists being cost-effective only when the threshold is above £385 per point improvement in MiniAQLQ.

FIGURE 13.

Step 3: MiniAQLQ CEACs. Based on imputed, adjusted results of MiniAQLQ, societal/NHS indicates that results were from a societal or NHS cost perspective, respectively, and that these outcomes took place at 2 months or 2 years. The figure indicates that in all cases the probability of leukotriene antagonist being cost-effective compared with long-acting β₂-agonist is above 50%, irrespective of the willingness to pay for a point improvement in MiniAQLQ. The exception is the 2-year analysis from a societal perspective, where the probability is above 50% as long as the willingness to pay for a point improvement in MiniAQLQ is > £432.

Asthma Control Questionnaire

Over 2 months (Table 40) and 2 years (Table 41), long-acting β₂-agonist is, on average, a dominant strategy over leukotriene antagonist. Mean INB is negative, irrespective of willingness to pay for a point improvement in ACQ. However, there is great deal of uncertainty around this estimate (Figure 14).

TABLE 40 - Step 3: ICER at 2 monthsa

Inc., incremental; LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist.

Inc. cost, inc. ACQ (95% CI) and ICERs are reported for patient groups with complete societal and NHS cost and ACQ data and following imputation for missing data.

	n		Cost (£)		ACQ		Source data only			Including imputed data			Including imputed data (adjusted for baseline ACQ)
	LTRA	LABA	LTRA	LABA	LTRA	LABA	Inc. cost (£)	Inc. ACQ	ICER (£)	Inc. cost (£)	Inc. ACQ	ICER (£)	Inc. cost (£)	Inc. ACQ	ICER (£)
NHS	153	155	84	82	1.616	1.606	3	0.010	(LABA dominant)	3	0.020	(LABA dominant)	4	0.106	(LABA dominant)
Societal	95	90	129	115	1.642	1.619	14	0.024	(LABA dominant)	6	0.020	(LABA dominant)	7	0.106	(LABA dominant)

TABLE 41 - Step 3: ICER at 2 yearsa

Inc., incremental; LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist.

Inc. cost, inc. ACQ (95% CI) and ICERs are reported for patient groups with complete societal and NHS cost and ACQ data and following imputation for missing data.

	n		Cost (£)		ACQ		Source data only			Including imputed data			Including imputed data (adjusted for baseline ACQ)
	LTRA	LABA	LTRA	LABA	LTRA	LABA	Inc. cost (£)	Inc. ACQ	ICER (£)	Inc. cost (£)	Inc. ACQ	ICER (£)	Inc. cost (£)	Inc. ACQ	ICER (£)
NHS	154	164	886	791	1.272	1.350	96	–0.077	1241	111	–0.063	1768	133	0.002	(LABA dominant)
Societal	94	96	1152	893	1.349	1.306	259	0.044	(LABA dominant)	213	–0.063	3398	226	0.002	(LABA dominant)

FIGURE 14.

Step 3: INB, ACQ, INB and 95% CI are shown for (a) 2-month NHS; (b) 2-month societal; (c) 2-year NHS; and (d) 2-year societal perspectives. Results based on imputed data adjusted for baseline ACQ.

Even with a very high willingness to pay for a point improvement in ACQ, there is up to a 49% probability of leukotriene antagonist being cost-effective compared with long-acting β₂-agonist in step 3 patients (depending on perspective and time horizon – Figure 15).

FIGURE 15.

Step 3: ACQ CEACs. Based on imputed, adjusted results of ACQ, societal/NHS indicates that results were from a societal or NHS cost perspective, respectively, and that these outcomes took place at 2 months or 2 years. This figure indicates that over a 2-year time horizon the probability of leukotriene antagonist being cost-effective compared with long-acting β₂-agonist approaches 50% as the willingness to pay for a point improvement in ACQ rises.

Quality-adjusted life-years (cost–utility analysis)

The incremental cost per QALY gained is between £5521 and £22,589, dependent upon perspective and time horizon (Tables 42 and 43).

TABLE 42 - Step 3: ICER at 2 monthsa

CI, confidence interval; inc., incremental; LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist.

Inc. cost, inc. QALY (95% CI), ICERs and probability that the intervention is cost-effective given a willingness to pay of £30,000 for QALY (P£30k – see Figure 8) are reported for patient groups with complete societal and NHS cost and QALY data and following imputation for missing data (*).

	n		Cost (£)		QALYs		Source data only					Including imputed data					Including imputed data (adjusted for BL utility)
	LTRA	LABA	LTRA	LABA	LTRA	LABA	Inc. cost (£)	Inc. QALYs	ICER (£)	P£30k (%)	INB/£30k (95% CI)	Inc. cost (£)	Inc. QALYs	ICER (£)	P£30k (%)	INB/£30k (95% CI)	Inc. cost (£)	Inc. QALYs	ICER (£)	P£30k (%)	INB/£30k (95% CI)
NHS	119	119	87	87	0.122	0.120	0	0.001	(LTRA dominant)	60.00	24 (–143 to 204)	3	0.001	2470	60.00	57 (–15 to –44)	3	0.001	5521	54.64	14 (–161 to 194)
Societal	77	74	117	126	0.122	0.119	–9	0.002	(LTRA dominant)	67.80	54 (–181 to 276)	6	0.001	6018	58.56	36 (7 to –36)	6	0.001	12,290	54.10	9 (–155 to 191)

TABLE 43 - Step 3: ICER at 2 yearsa

CI, confidence interval; inc., incremental; LABA, long-acting β₂-agonist; LTRA, leukotriene receptor antagonist.

Inc. cost, inc. QALY (95% CI), ICERs and probability that the intervention is cost-effective given a willingness to pay of £30,000 for QALY (P£30k – see Figure 8) are reported for patient groups with complete societal and NHS cost and QALY data and following imputation for missing data (*).

	n		Cost (£)		QALYs		Source data only					Including imputed data					Including imputed data (adjusted for BL utility)
	LTRA	LABA	LTRA	LABA	LTRA	LABA	Inc. cost (£)	Inc. QALYs	ICER (£)	P£30k (%)	INB/£30k (95% CI)	Inc. cost (£)	Inc. QALYs	ICER (£)	P£30k (%)	INB/£30k (95% CI)	Inc. cost (£)	Inc. QALYs	ICER (£)	P£30k (%)	INB/£30k (95% CI)
NHS	108	109	956	869	1.601	1.548	88	0.053	1643	81.60	983 (–1429 to 3449)	111	0.015	7164	57.76	–378 (1673 to 1896)	113	0.009	11919	53.16	154 (–2443 to 2893)
Societal	72	70	1157	952	1.610	1.566	205	0.044	4668	68.60	680 (–2308 to 3547)	213	0.015	13,769	54.34	168 (327 to 1369)	214	0.009	22,589	51.56	142 (–2567 to 2825)

At a typical willingness to pay of £30,000 per QALY gained, mean (95% CI) INB at 2 months is £14 (–£161 to £194) or £9 (–£155 to £191) from the NHS and societal perspectives, and £154 (–£2443 to £2893) and £142 (–£2567 to £2825) at 2 years (Tables 42 and 43, and Figure 16).

FIGURE 16.

Step 3: INB, QALYs. INB and 95% CI are shown for (a) 2-month NHS; (b) 2-month societal; (c) 2-year NHS; and (d) 2-year societal perspectives. Results based on imputed data adjusted for baseline utility.

At the £30,000 threshold, there is between a 51.6% and 54.6% probability of leukotriene antagonist being cost-effective compared with long-acting β₂-agonist in patients at step 3 of the national asthma guidelines (dependent on time horizon and perspective – Tables 42 and 43, and Figure 17).

FIGURE 17.

Step 3: cost–utility analysis: CEACs. Based on imputed, adjusted results of cost–utility analysis, societal/NHS indicates that results were from a societal or NHS cost perspective, respectively, and these outcomes took place at 2 months or 2 years. This figure indicates that in all cases the probability of leukotriene antagonist being cost-effective compared with long-acting β₂-agonist is marginally above 50%, given a typical willingness to pay of £30,000 per QALY gained. At a willingness to pay of £20,000, the probability is between 48.5% and 54.0%, depending upon time horizon and perspective.

Summary of cost-effectiveness analyses

The results of the cost-effectiveness analyses are somewhat equivocal: mean INB is positive when considering MiniAQLQ, as long as the willingness to pay for a point improvement in MiniAQLQ is > £6267 (2-year time horizon, societal analysis). However, when considering ACQ as the outcome measure, mean INB is always negative, yet the point estimate of the ICER is £22,589 (which would yield a positive mean INB at a ‘typical’ £30,000 threshold). The results are therefore contradictory, depending on the outcome measure.

There is at least a 50% probability of leukotriene antagonist being cost-effective, so long as the willingness to pay for a point improvement in MiniAQLQ is, at worst, £6200, and, at best, £47 (depending upon perspective and time horizon), and for a QALY, at least £6180–£23,600.

Whether an intervention is deemed cost-effective is dependent upon this willingness-to-pay threshold. A ‘reasonable’ threshold for QALYs is thought to be in the order of £30,000. The threshold for a point improvement in MiniAQLQ or ACQ is less well established. However, the results of the cost–utility analysis suggest a 51.6–54.6% probability of cost-effectiveness at a willingness to pay of £30,000 per QALY gained.

Adverse events

Interpretation

The results of this study provide no evidence of superiority of inhaled steroid or long-acting β₂-agonist over leukotriene antagonist, nor vice versa, in patients at either step 2 or step 3, in terms of the primary end point of asthma-specific QOL (MiniAQLQ score). For initial controller therapy at step 2, the results were equivalent at 2 months. After adjusting for baseline characteristics at 2 years, the 95% CIs did include the threshold for equivalence of 0.3 in favour of inhaled steroid. This was true of the one-sided 95% CI, i.e. inferiority of leukotriene antagonist could not be excluded. Conversely, the per-protocol population (fixed treatment regime and no changes within randomised therapy class at any stage) results also included the threshold for equivalence of 0.3, but this time in favour of leukotriene antagonist. Therefore, the results were inconsistent between the intention to treat and per-protocol analyses and equivalence could not be excluded. However, any possible advantage of one over the other looks clinically insignificant.

For add-on therapy at step 3, at 2 months results were equivalent, and at 2 years the intention-to-treat analysis resulted in near equivalence of leukotriene antagonist and long-acting β₂-agonist (the lower bounds marginally missing the equivalence value) with the possibility of a minor advantage for long-acting β₂-agonist. The per-protocol results were consistent with this, but, again, any advantage of long-acting β₂-agonist appears clinically insignificant.

We chose a conservative approach in selecting 0.3 as the threshold for equivalence or non-inferiority, because the minimum clinically important difference for MiniAQLQ is 0.5. 51 Therefore, although a difference from 0.3 to 0.5 does not meet our study definition of equivalence, the outer bounds of the CIs for the differences are less than the minimum that is clinically important difference for this parameter.

There were no significant differences between treatment groups in results for most of the secondary end points, including, most importantly, the two markers of asthma control, ACQ score and exacerbation rate. The 95% CIs for differences in ACQ score were well within the 0.5 minimum clinically important difference for both step 2 and 3 trials at both time points and in both intention-to-treat and per-protocol analyses. In addition, there were no significant differences in secondary health-related QOL measures, asthma symptoms as measured by diary card, the RCP3 questions and respiratory tract infections. Hospitalisations were infrequent in both step 2 and 3 trials.

At step 2, no significant differences were found in short-acting β₂-agonist prescriptions between arms. At step 3, short-acting β₂-agonist prescriptions over 2 years were significantly greater in the leukotriene antagonist arm than the long-acting β₂-agonist arm of the intention-to-treat population but not the per-protocol population. Calculations of short-acting β₂-agonist prescriptions in this study were based on the numbers of inhalers prescribed and do not reliably represent actual short-acting β₂-agonist use, because prescribed inhalers may not have been dispensed or used. Instead, ACQ scores, which incorporate a question on actual short-acting β₂-agonist use (puffs/day) during the prior 7 days, were not significantly different between leukotriene antagonist and long-acting β₂-agonist groups. Composite measures, such as the ACQ, are recommended for evaluating asthma control by the current European Respiratory Society (ERS)/American Thoracic Society (ATS) task force, which notes the challenges in measuring and comparing short-acting β₂-agonist use in clinical trials, suggesting that this outcome be derived from a diary or visit-based questionnaire. 78 Statistically significant differences in diary-recorded lung function between treatment groups in the step 3 trial were small and unlikely to be of clinical significance.

The number of participants included in the per-protocol analysis for both study arms was small [in the step 2 study 65/145 (45%) and 82/155 (53%) of the LTRA and ICS arms respectively, and 60/169 (36%) and 80/181 (44%) of the LTRA and LABA arms respectively in the step 3 study] because of substantial use of self-management plans with prescription instructions permitting adjustments to dosage, prescriptions which were technically changes (often within randomised class) but which were the result of a change of practice prescription policy, or prescriptions from an encounter with a non-study aware provider who made a change as allowed as per normal asthma management, often only for a single prescription issue, and which thus did not truly represent a substantial change in therapy. Nonetheless, these changes resulted in participants being excluded from the per-protocol group. While large and clinically as well as statistically comparable improvements from baseline were seen in all treatment groups investigated in this study, final mean outcomes are still not optimal and suggest that further intervention is required for many of the patients studied.

From a health economic perspective, mean results indicate that in a primary care setting, at step 2, inhaled steroids are more cost-effective than leukotriene antagonists due, principally, to the greater acquisition costs of leukotriene antagonists than inhaled steroids. At step 3 of the guidelines, results are somewhat inconsistent. Whilst leukotriene antagonists were slightly more expensive than long-acting β₂-agonists, patients receiving leukotriene antagonists had marginally better overall health-related QOL (as measured in QALYs), and hence we estimate a mean incremental cost per QALY gained of £22,589 (societal perspective, 2-year time horizon), with a probability of between 51.6% and 54.6% of leukotriene antagonists being cost-effective compared with long-acting β₂-agonist at a willingness to pay of £30,000 per QALY gained. There was no statistically significant difference in INB at any threshold willingness to pay for a point improvement in MiniAQLQ or ACQ, or at a threshold of £30,000 per QALY gained (the exception being at low thresholds for QALYs in step 2 patients, for which the upper bound for the 95% CI for INB is negative – see Figures 6, 8, 10, 13, 15 and 17). Mean results based on ACQ also suggested that long-acting β₂-agonist dominated leukotriene antagonist.

In the step 2 trial, we observed that, by 2 months, patients receiving leukotriene antagonist had received approximately twice as many NHS contacts as their inhaled steroid counterparts, and that the majority of the contacts were in primary care. Whether this was related to the issuing of prescriptions for leukotriene antagonist and a greater perceived need to follow the patient, given that leukotriene antagonists are rarely used in the UK as first-line anti-inflammatory therapy, is unknown. However, given that leukotriene antagonists are available only in 28-day packs, compared with a 200-dose inhaler that could sustain a patient for up to 3 months, the likelihood of greater follow-up in the leukotriene antagonist group to reissue prescriptions is to be expected due to repeat prescriptions being used as a trigger for review in many practices. This explanation is further supported by the lack of difference in ACQ score and exacerbations. If lack of familiarity with this class of therapy contributed to consultations this would be likely to be a smaller factor in the future and reduce the cost of treatment to the NHS of using a leukotriene antagonist.

Indirect costs are costs attributable to lost productivity and/or time off work due to ill health. Our results indicate that over the longer time horizon mean indirect costs are lower for leukotriene antagonist versus inhaled steroid patients at step 2 but higher for leukotriene antagonist versus long-acting β₂-agonist at step 3. Although we cannot exclude chance in explaining these findings (and response rate to time off work questions was poor), it should be noted that in step 2 patients at 2 years the hours of work lost per patient randomised to inhaled steroid were lower than those lost per patient randomised to leukotriene antagonist, yet the cost is slightly higher. This apparent contradiction is due to the differential timing of costs and the effect of discounting: a greater proportion of the time off work was reported for the inhaled steroid group in the first year.

A key driver of the cost-effectiveness of one drug compared with another is usually the acquisition cost. At current prices, we estimate a mean ICER of £22,589 in step 3 patients. If the price of leukotriene antagonists falls following patent expiry, the ICER will also fall and the probability of their being cost-effective compared with long-acting β₂-agonists in these patients (step 3) will rise.

Study strengths and limitations

Conducting a pragmatic randomised control trial in a primary care setting has advantages and disadvantages over conducting a randomised controlled trial as a clinical and explanatory clinical trial.

Smoking and response to asthma therapies

Active smoking has been shown to reduce the anti-inflammatory efficacy of inhaled steroids. 82,83 As the anti-inflammatory action of long-acting β₂-agonist reported in vitro has not been confirmed in vivo, leukotriene antagonists may provide an alternative anti-inflammatory treatment for asthma in smokers. Indeed Lazarus and colleagues84 recently reported that mild asthmatics who were also active smokers had a significantly lower response to inhaled steroid than non-smoking mild asthmatics. However, the active smokers had a significantly greater response to the leukotriene antagonist montelukast than the non-smoking patients, suggesting that in contrast with inhaled steroid, smoking does not significantly affect the action of the leukotriene antagonist montelukast. Whether this difference reflects changes in the mediation of airway inflammation in smokers or not, is not known. Furthermore, as Lazarus and colleagues84 conducted their study under clinical trial conditions, whether comparable results would be found under ‘real-life’ primary care conditions is unknown.

Comparison with prior studies

To our knowledge, there have been no other published pragmatic, head-to-head studies comparing leukotriene antagonist and inhaled steroid at step 2 or leukotriene antagonist and long-acting β₂-agonist as add-on to inhaled steroid at step 3 for patients with asthma in primary care. Specifically, prior studies have enrolled selective patient populations, requiring evidence of airway reversibility, substantially impaired lung function and/or excluding patients with concomitant conditions or who smoke. They have also suffered from relatively high dropout rates, which may influence intention-to-treat approaches to analysis. Nonetheless, results of longer-term studies could be of relevance in comparison with the present study.

At step 2, open-label treatment for 36 weeks with fluticasone or montelukast gave comparable results for some patients with mild asthma, whereas fluticasone improved asthma control more than montelukast for patients who had decreased lung function and greater albuterol use at baseline. 87 In another study, the effectiveness of montelukast and inhaled beclometasone was similar over 2+ years of open-label treatment, and the authors speculated that the initially greater mean effect of beclometasone on lung function was offset over time by better adherence to orally administered montelukast. 48 We found that adherence to treatment with leukotriene antagonist was substantially better than that to inhaled therapy by patients in this study who had at least 6 months of unchanged therapy. At step 2, median adherence was 61% and 41% to leukotriene antagonist and inhaled steroid, respectively. At step 3, median adherence was substantially higher in the leukotriene antagonist arm than in the long-acting β₂-agonist arm, both to inhaled steroid (82% versus 65%) and to add-on therapy (90% with montelukast versus 49% with long-acting β₂-agonist).

With regard to step 3, a recent systematic review looking only at studies of ≥ 12 weeks’ duration comparing montelukast with salmeterol as add-on to inhaled steroid found that while salmeterol may be more effective with regard to most clinical outcomes in the medium term, over 48 weeks, the proportions of patients with ≥ 1 exacerbation were similar, as were hospitalisation and emergency treatment rates. 36 The rate of serious adverse events over 48 weeks was significantly higher with add-on salmeterol; thus, montelukast may have a better long-term safety profile.

Prior reviews of health economic studies in asthma have pointed out the need for longer-term studies using a pragmatic trial design and outcome measures that reflect asthma control and are clinically meaningful and relevant to patients. 88,89 We believe that the design of this study addresses that need and provides results meaningful for decision-makers.

Statistical issues

Missing data are a limitation in any clinical trial. This is a particular issue in economic evaluations, where, typically, not only are multidimensional QOL and other outcome measures collected, but complex resource use questionnaires may also be required. Indeed, in this study, we collected data on the costs associated with prescription and over-the-counter medications, NHS resources and time off work. The number of observations for which both outcome and cost data were available is therefore less than the number of observations for either outcomes or cost data alone.

We used Rubin’s multiple imputation technique76 to handle missing data. Imputation of missing values is feasible in this study and desirable as while there were a large number of observations with incomplete data, the actual number of data elements missing from each individual observation was small [218/683 patients (32%) had complete data, yet 514/683 (75%) had only four or fewer of the 13 data items missing]. The complete case analysis therefore excludes a lot of valid data. The particular strength of Rubin’s multiple imputation approach is that it acknowledges that missing values are uncertain, and therefore estimates several possible values for each (five imputations are usually considered sufficient76). It therefore provides a better characterisation of uncertainty than single-imputation techniques.

Nevertheless, the use of multiple imputation rests on a number of assumptions. Firstly, that the data are at least missing at random (that is, the probability that an observation is missing can depend on the observed variables, but not on the missing variable itself), and, secondly, that the data follow a multivariate normal distribution.

If data are not missing at random then, in principle, the missingness process could be modelled. However, this adds to the (already substantial) complexity of the model and studies have shown that even when ‘missing at random’ is violated, multiple imputation may still be superior to other approaches. 90 Furthermore, multiple imputation may provide a sufficient approximation of missing data with appropriately transformed variables, and it is suggested that even binary variables may be approximated by estimating under a normal assumption and rounding the continuous values. 91

We presented results of the economic evaluation in terms of the ICER. Whether or not an intervention is deemed cost-effective depends on whether or not the ICER is below some threshold of willingness to pay for that unit of outcome. Where the unit of outcome is QALYs gained, a threshold of approximately £30,000 is considered a de facto standard. However, such a threshold for point improvement in MiniAQLQ and ACQ is currently undefined. This has implications for representing uncertainty in the estimate of cost-effectiveness because a standard 95% CI is not necessarily defined for the ICER due to its ratio properties. We therefore calculated CEACs as well as the INB and its associated CI. However, to estimate a meaningful value of the INB requires knowledge of the threshold, which for MiniAQLQ and ACQ is unknown. Therefore we presented INB based on MiniAQLQ and ACQ plotted for an arbitrarily wide range of values (see Figures 6, 8, 13 and 15), rather than stating a point estimate (with CI) in Tables 14–17 and 37–40.

Further study

There are several validated assessment tools that are currently used for different aspects of asthma compliance and control. In addition to the Juniper ACQ,62,64 MiniAQLQ,50 EQ-5D71 and RCP3 questions68,69 that we used here, additional measures that have been used to evaluate asthma compliance and control include the Beliefs about Medicines Questionnaire,92 the Illness Perception Questionnaire,93 the Medication Adherence Report Scale (R. Horne, University of London, 2004, personal communication), and the Satisfaction with Information about Medicines Scale. 94 However, a limitation of these tools is that no one single measure allows the evaluation of patient perceptions about their illness and therapy, and current symptoms, adherence, side effects and control. In addition, the RCP3 questions is a set of questions that has not been validated. The development of a unified and yet easily performed test that would aid in the assessment of all of these areas would provide a valuable tool for both research use as well as clinical management of patients. A Minimal Asthma Assessment Tool was developed alongside these trials, partially utilising data collected from the patient-reported outcome measures. 95

Twenty-eight in-depth interviews have also been conducted and are currently being analysed to understand patients’ perceptions of preventative therapies. This substudy has been described in a PhD thesis96 and is being analysed for potential further publication.

Implications for health care

The evidence suggests that while any advantage of one treatment over the other appears to be clinically unimportant, leukotriene receptor antagonists are unlikely to be a cost-effective alternative to inhaled corticosteroids, at 2005 prices, as initial asthma controller therapy at step 2. In addition, the evidence suggests that leukotriene antagonists may be clinically equivalent to long-acting β₂-agonists as add-on to inhaled steroids in terms of QOL as well as secondary measures, and, furthermore, suggests that leukotriene antagonists could be repositioned as an equal alternative to long-acting β₂-agonists at step 3 of the BTS guidelines. When generic leukotriene antagonist formulations become available in the next few years their cost-effectiveness as an alternative to ICS may justify further evaluation, particularly in the subgroup of patients with limited impairment of lung function, those newly diagnosed with asthma to minimise inhaler education and those with fears about inhalers or inhaled steroids.

Recommendations for research

• Establish in primary care, whether leukotriene antagonists will be more or less beneficial than inhaled steroids alone or as an add-on to inhaled steroids in treating patients with asthma who are also active smokers.

• Determine why the ACQ correlates more poorly with economic outcomes of asthma than the MiniAQLQ and EQ-5D.

• Understand further the reasons why patients were switched from study medication when there was no real clinical indication to do so, and examine ways to minimise this happening in future pragmatic primary care-based clinical trials.

Participating practices

Thorpewood Surgery, Norwich; Bacon Road Medical Centre, Norwich; Drayton and St Faith Medical Practice, Drayton; Botesdale Health Centre, Botesdale; Fakenham Medical Practice, Fakenham; Beccles Medical Centre, Beccles; Siam Surgery, Sudbury; Beach Road Surgery, Lowestoft; Grimston Medical Centre, Grimston; Cutlers Hill Surgery, Halesworth; Mattishall Surgery, Mattishall; Aldborough Surgery, Norwich; Needham Market Country Practice, Needham Market; Costessey Medical Centre, Old Costessey, Norwich; Attleborough Surgeries, Attleborough; Coltishall Surgery, Coltishall; Bungay Medical Practice, Bungay; South Quay Surgery, Great Yarmouth; Millwood Surgery, Bradwell, Great Yarmouth; Lattice Barn Surgery, Ipswich; Leiston Surgery, Leiston; Harleston Medical Centre, Harleston; East Thurrock Medical Centre, Grays; Stowmarket Health Centre, Stowmarket; Orchard Medical Centre, Ipswich; Church Plain Surgery, Loddon; Holt Medical Practice, Holt; Mersea Road Surgery, Colchester; Birchwood Surgery, North Walsham; Leighton Road Surgery, Linslade; Kingswood Medical Centre, Basildon; Parish Fields Practice, Diss; Barrow Hill Surgery, Bury St Edmunds; Little Waltham Surgery, Chelmsford; John Tasker House, Great Dunmow; Yaxley Group Practice, Yaxley, Peterborough; James Fisher Medical Centre, Bournemouth; Neera Medical Centre, Stanford-le-Hope; Hethersett Surgery, Hethersett; The White House Surgery, Weston, Southampton; West Pottergate Health Centre, Norwich; Grays Surgery, Grays; Laindon Health Centre, Basildon; Rigg-Milner Medical Centre, East Tilbury; Sheringham Health Centre, Sheringham; The Barn Surgery, Gillingham; Lynwood Health Centre, Romford; The Burnhams Surgery, King’s Lynn; Angel Hill Surgery, Bury St Edmunds; Framfield House, Woodbridge; Hoveton & Wroxham Medical Centre, Hoveton; The Woottons Surgery, King’s Lynn; Rookery Medical Partnership, Newmarket; Wells Health Centre, Wells Next The Sea; Wood Lane Surgery, Hornchurch; Capital Road Surgery, Higher Openshaw, Manchester; The Witham Health Centre, Witham.

Contribution of authors

David Price (Professor, Primary Care Medicine) was involved in original design and grant submission, management and steering committee activity, data validation, analysis plan production, interpretation of data, study report and final publication. Stanley Musgrave (Research Associate, Norwich Medical School) was involved in study implementation, management and steering committee activity, data validation, analysis plan production, interpretation of data, study report and final publication. Ed Wilson (Lecturer, Health Economics), was involved in data validation, analysis plan production, interpretation of data, study report and final publication. Erika Sims (Research Fellow, Health Economics) was involved in management and steering committee activity, data validation, analysis plan production, interpretation of data, study report and final publication. Lee Shepstone (Professor, Medical Statistics) was involved in original design and grant submission, management and steering committee activity, data validation, analysis plan production, interpretation of data, study report and final publication. Annie Blyth (Research Associate, Norwich Medical School) was involved in study implementation, management and steering committee activity, data validation, study report and final publication. Jamie Murdoch (Research Associate, School of Nursing and Midwifery) was involved in study implementation, management and steering committee activity, data validation, analysis plan production, interpretation of data, study report and final publication. Miranda Mugford (Professor, Health Economics) was involved in original design and grant submission, management and steering committee activity, analysis plan production, interpretation of data, study report and final publication. Elizabeth Juniper (Professor Emeritus, Clinical Epidemiology and Biostatistics) was involved in original design and grant submission, management and steering committee activity, analysis plan production, interpretation of data, study report and final publication. Jon Ayres (Professor, Environmental and Respiratory Medicine) was involved in original design and grant submission, management and steering committee activity, analysis plan production, interpretation of data, study report and final publication. Stephanie Wolfe (Respiratory and Research Nurse) was involved in original design and grant submission, management and steering committee activity, analysis plan production, interpretation of data, study report and final publication. Daryl Freeman (General Practitioner) was involved in study management and steering committee activity, analysis plan production, interpretation of data, study report and final publication. Alistair Lipp (Director, Public Health) was involved in study management and steering committee activity, analysis plan production, interpretation of data, study report and final publication. Richard Gilbert (General Practitioner) was involved in study management and steering committee activity, analysis plan production, interpretation of data, study report and final publication. Ian Harvey (Professor and Dean, Faculty of Health) was involved in original design and grant submission, management and steering committee activity, data validation, analysis plan production, interpretation of data, study report and final publication.

Conflicts of interest

David Price has consultant arrangements with Astra Zeneca, GlaxoSmithKline, Merck, Sharpe and Dohme, Schering Ploughm and Teva; has grants/research support from AstraZeneca, GlaxoSmithKline, Merck, Sharpe and Dohme, Mundipharma, Novartis, Schering Plough and Teva; and is a member of the speakers’ bureau fgor GlaxoSmithKline, Merck, Sharpe and Dohme, Mundipharma, Novartis, Schering Plough and Teva.

Stanley Musgrave owned stock in Merck Sharpe and Dohme (MSD) within 3 years of beginning this work. He disposed of that stock prior to beginning this work and has had no further financial interest in MSD. During the time that he owned MSD stock he had no role in the preparation of this study. He has grants/research support from MSD and AstraZeneca, and Altana Pharma.

Ed Wilson has received a consultancy fee from a manufacturer of a budesonide inhaler (Meda Pharmaceuticals, 2006).

Erika Sims has received funding from MSD and AstraZeneca to attend international conferences. She has previously undertaken research studies sponsored by MSD, although the funding for these trials did not support her salary.

Elizabeth Juniper has consultant arrangements with GlaxoSmithKline and Astra Zeneca. She also has stock/other equity ownership in GlaxoSmithKline and Pfizer; and is a copyright holder of the MiniAQLQ, ACQ and MiniRQLQ. During the last 5 years she has had financial support from Johnson & Johnson, Schering Plough, Medpointe Pharma, CMP Therapeutics, Alterhealth, Pierre Fabre, GlaxoSmithKline, Munipharma, Ception Therapeutics, Sunten Phytotech, UCB Pharma, Activus Pharma, Novartis, Wyeth, Allergopharma, Airsonett, Embria Health Sciences, CV Technologies, Kalobios, Stallergens, AstraZeneca, Alk Abello, Medicinova, Artu Biologicals, Curalogic, IFE Europe, Genentech, Epigenesis, Merck, Cipla, Amgen, Rottapharma, Allergy Therapeutics, NovoNordisk, Inflazyme, Allergy Choices, Abbott Laboratories, Mitsubishi, Boehringher Ingelheim, Cytos Technologies, Wellpoint/Healthcore, Altanapharma, Capnia, Ivax Laboratories, Critical Therapeutics, Kyowa Hakko, Asthmatx, Neolab, Protein Design Laboratories, ClinPhone, Arriva, Schulman Ronca Bucuvals, Laboratories SMB, Vista Health Plan, Hal Allergy BV, Suburban Lung Associates, Galephar MF SA, Mannkind Corp, Prognostix, Formix Biosciences, Oxagen, SC Ellen Fast Comimpex SRL, Medimmune, Quintiles, PDL Biopharma, Bencard, Efficas, Pfizer, Alcon Lab, Aventis, Pahrmaxis, Assist Tech, Hoffmann LaRoche, Aerocrine, Dynavax, Medtap, Pharmaengine, Topigen, ClineDavis & Mann, Teikoku Pharma, NeoLab, KGK Synergise, and AB Science.

Stephanie Wolfe is on the advisory boards of Teva, Novartis, Astra Zeneca and GSK. She also attends speaker meetings for: Pfizer, Boehringer, Astra Zeneca, Altana Pharma (a Nycomed company), GSK and MSD.

Publications

Murdoch J, Musgrave S, Price D, Rust-Andrews S, Juniper E. Profiles of asthma control in 12488 patients in 30 practices in East Anglia. Prim Care Resp J 2003;12:70.

Wilson ECF, Sims EJ, Musgrave SD, Shepstone L, Blyth A, Murdoch J, et al. The cost-effectiveness of leukotriene receptor antagonists versus inhaled corticosteroids for initial asthma controller therapy: a pragmatic trial. Pharmacoeconomics 2010;28:585–95.

Wilson ECF, Price D, Musgrave SD, Sims EJ, Shepstone L, Murdoch J, et al. The cost-effectiveness of leukotriene receptor antagonists versus long acting beta-2 agonists as add-on therapy to inhaled corticosteroids for asthma: a pragmatic trial. Pharmacoeconomics 2010;28:597–608.

Price D, Musgrave SD, Shepstone L, Hillyer EV, Sims EJ, Gilbert RFT, et al. Leukotriene antagonists as first-line or add-on asthma-controller therapy. N Engl J Med 2011;364:1695–707.

Disclaimers

The views expressed in this publication are those of the authors and not necessarily those of the HTA programme or the Department of Health.