Beta-Agonist Lung injury TrIal-2 (BALTI-2): a multicentre, randomised, double-blind, placebo-controlled trial and economic evaluation of intravenous infusion of salbutamol versus placebo in patients with acute respiratory distress syndrome

Article history

This issue of Health Technology Assessment contains a project originally commissioned by the MRC but managed by the Efficacy and Mechanism Evaluation Programme. The EME programme was created as part of the National Institute for Health Research (NIHR) and the Medical Research Council (MRC) coordinated strategy for clinical trials. The EME programme is funded by the MRC and NIHR, with contributions from the CSO in Scotland and NISCHR in Wales and the HSC R&D, Public Health Agency in Northern Ireland. It is managed by the NIHR Evaluation, Trials and Studies Coordinating Centre (NETSCC) based at the University of Southampton.

Declared competing interests of authors

FGS and GDP have received an investigator-led research grant from GlaxoSmithKline. GDP and DFM have consulted for, sat on advisory boards for, and received lecture fees from GlaxoSmithKline, and have received lecture fees from AstraZeneca for educational meetings (all unrelated to β-agonists). All other authors declare that they have no competing interests.

Copyright statement

© Queen's Printer and Controller of HMSO 2013. This work was produced by Gates et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) 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: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Description of acute respiratory distress syndrome

Acute respiratory distress syndrome (ARDS) is a condition characterised by a failure of pulmonary oxygen exchange due to increased alveolar–capillary permeability and resultant pulmonary oedema. It can be caused by primary lung conditions such as aspiration pneumonitis, or can arise as a complication of non-pulmonary conditions such as severe sepsis. The syndrome was first described by Ashbaugh and colleagues in 19671 in a group of 12 patients with acute onset of dyspnoea, tachypnoea, refractory hypoxaemia, reduced pulmonary compliance and diffuse alveolar shadowing on their chest radiographs. All patients required positive-pressure mechanical ventilation with positive end-expiratory pressure to maintain arterial oxygenation. The term ‘adult respiratory distress syndrome’ was initially used to describe the condition,2 but it was subsequently renamed as ARDS because it may also occur in children. 3 The current definition arose from the American–European Consensus Conference in 19944 which recognised two grades of disease, based on the degree of hypoxaemia. ARDS was reserved for the more severe grade, with acute lung injury (ALI) being used to describe the less severe form. The definition of ALI/ARDS requires:

1. acute onset

2. bilateral infiltrates on chest radiographs

3. pulmonary artery occlusion pressure < 18 mmHg (if measured), or absence of clinical signs of left atrial hypertension

4. ratio of partial pressure of oxygen in arterial blood (PaO₂) to the fraction of inspired oxygen (FiO₂) < 200 mmHg (26.7 kPa) for ARDS, or PaO₂–FiO₂ ratio < 300 mmHg (40 kPa) for ALI.

Incidence and burden of disease

The population incidence of ARDS in Europe and Scandinavia has been estimated by several studies at between 7.8 and 28 cases per 100,000 population per year. 5–9 This translates as up to 16,800 cases per year in the UK, similar to all new cases of lymphomas and leukaemias combined. ARDS is a common condition among intensive care unit (ICU) patients, affecting 6–8% of all ICU admissions, and patients with ARDS have a very high risk of death. Recent multicentre cohort studies from Europe, the USA and Australia have given mortality estimates of between 34% and 61% (measured over different timescales). 5,7,8,10,11 Two studies conducted in the UK and Europe, both in 1999, found the highest of these mortality figures, with rates of death in hospital of 61% and 58%. 5,7 There has been a trend to reduced mortality in epidemiological and clinical trials in recent years. 9 Nevertheless, there may be 7000 deaths per year in the UK in patients with ARDS. As well as a high mortality, ARDS causes long-term health problems and reduction in quality of life (particularly physical activity) for survivors. 12 A recent study in survivors of ARDS found evidence of exercise limitation, physical and psychological sequelae and decreased physical quality of life up to 5 years later. 13 In addition, ARDS has significant resource implications as it prolongs ICU and hospital stay and requires convalescence in hospital and subsequent rehabilitation in the community. 14

Existing evidence

A systematic review published in 2004, based on an electronic search of seven electronic databases and a hand search of reference lists from review articles and relevant papers,15 found three clinical studies (one randomised, controlled, cross-over trial16 and two non-randomised studies17,18) using β₂-agonists in patients with ARDS. These studies examined the effects of nebulised16,17 or intravenous (i.v.)18 β₂-agonists on the respiratory mechanics of artificially ventilated patients with ARDS and found that β₂-agonists reduced airway resistance and peak and plateau airway pressures. There were no clinical studies addressing the effects of β₂-agonists on alveolar fluid clearance or on outcome.

A single-centre Phase II trial [Beta-Agonist Lung injury TrIal (BALTI)-1] investigating the efficacy of i.v. salbutamol on in vivo fluid clearance through serial measurement of extravascular lung water in 40 patients with ARDS was conducted between 2001 and 2003. 19 An initial dose-ranging study determined that the maximum infusion rate for salbutamol that did not cause tachydysrhythmias in patients with ARDS was 15 μg/kg ideal body weight (IBW)/hour. This is the maximal recommended dose for the treatment of airflow obstruction in acutely ill patients. The trial showed that an infusion of salbutamol over 7 days significantly reduced lung water [day 7 lung water mean (standard deviation; SD), 9.2 (6) vs. 13.2 (3) ml/kg; p = 0.038] and plateau airway pressures [day 7 plateau airway pressures mean (SD) 23.9 (3.8) vs. 29.5 (7.2) cmH₂O; p = 0.049], providing proof of concept that treatment with i.v. β₂-agonists can influence alveolar fluid clearance, but it was not designed to address important clinical outcomes and a subsequent larger trial was needed to evaluate this therapy.

As part of the BALTI-2 funding application, the investigators updated the literature search (unpublished) using the same keywords combined with terms to identify randomised controlled trials (RCTs). No studies using i.v. salbutamol infusion were identified. The only relevant additional publication was a retrospective case review of 86 patients with ALI suggesting that high-dose nebulised salbutamol may be superior to a low dose. 20

At the time that BALTI-2 was initiated, the only treatment of proven effectiveness for ARDS was use of a lung-protective (pressure and volume limited) strategy of mechanical ventilation. 21 There were no additional treatments known to improve outcome. A Cochrane review of pharmacological treatments that included 22 studies of 14 different drugs concluded that ‘Effective pharmacotherapy for ALI and ARDS is extremely limited, with insufficient evidence to support any specific intervention’. 22 Recently, a single trial has been published showing a reduction in mortality in ARDS patients treated with neuromuscular blocking agents. 23 This is the only trial that has demonstrated an effective pharmacological treatment for ARDS.

Rationale for beta-agonists in acute respiratory distress syndrome

There is good evidence from in vivo and in vitro human and animal studies that β₂-agonists may have a range of important beneficial effects in ARDS patients. 15 First, they can affect epithelial and endothelial function to reduce alveolar–capillary permeability, accelerate alveolar fluid clearance and increase surfactant secretion, all of which may help to reduce pulmonary oedema. Second, they modulate the inflammatory cascades and regulate neutrophil recruitment, activation and apoptosis. 24 This may improve outcomes, as high titres and persistence of inflammatory cytokines are associated with poor outcome and stimulating neutrophil apoptosis may lead to reduced lung injury and improved survival. Third, they enhance epithelial wound repair and promote alveolar–capillary healing. 25,26

Salbutamol is a low-cost treatment and is readily available from generic drug manufacturers. A 7-day infusion is cheap compared with the cost of ICU care (the NHS reference cost for a day in ICU care is £1390). 27

Nebulised compared with intravenous salbutamol

The optimal route for delivering β₂-agonists in patients with ARDS with a goal of increasing alveolar fluid clearance has not been determined. Nebulising drugs into the breathing circuits of mechanically ventilated patients appears attractive as it results in high lung concentrations but low blood concentrations and so may reduce the incidence of systemic side effects compared with parenteral treatment. 16 However, nebulised drugs might not reach the alveolar space in the consolidated and poorly ventilated lungs found in patients with ARDS. A trial [ALbuterol for the Treatment of ALI (ALTA)] conducted in the US concurrently with BALTI-2 evaluated the effects of nebulised salbutamol (albuterol) in patients with ALI. Recruitment was terminated early by the Data Monitoring and Ethics Committee (DMEC) on the grounds of ‘futility’, with no clear differences between the albuterol and placebo groups. 28

Trial summary

Beta-agonist Lung Injury Trial-2 was a multicentre, pragmatic, randomised, double-blind, placebo-controlled clinical trial. Patients fulfilling the American–European Consensus Conference definition of ARDS were randomised in a 1 : 1 ratio to receive an i.v. infusion of either salbutamol (15 µg/kg IBW/hour) or placebo (0.9% sodium chloride solution), for a maximum of 7 days. Allocation to randomised groups used minimisation to ensure balance with respect to hospital of recruitment, age group and PaO₂–FiO₂ ratio. The trial was fully blinded and all drugs were packaged identically, so that patients, clinicians and investigators did not know which patients were in each arm. The primary outcome was mortality at 28 days after randomisation, with follow-up for mortality and quality of life to 12 months. The target sample size was 1334 patients, to be recruited from about 50 ICUs in the UK. The trial protocol has been published. 29

Pilot and main study

The trial was structured into a pilot phase and a main trial. The pilot phase was conducted at five hospitals in the West Midlands, and was conducted while substantive funding was obtained. There were no changes to the trial protocol between the pilot and main trial phases, except that, for resource reasons, the pilot phase did not include long-term follow-up at 6 and 12 months. However, patients recruited later in the pilot phase, whose follow-up points fell within the main trial period, were followed up. As there were no substantial differences in the trial between the pilot and main periods, we included all patients in the final analysis.

Objectives

The primary objective of the trial was to assess whether or not an i.v. infusion of salbutamol given at 15 μg/kg IBW/hour for up to 7 days reduces 28-day all-cause mortality in patients with ARDS compared with a placebo (0.9% sodium chloride) infusion.

The secondary objectives were:

1. to evaluate the effects of i.v. salbutamol on mortality in ICU, mortality in hospital, ventilator-free days (VFDs), organ failure-free days, length of ICU and hospital stay, mortality up to 12 months after randomisation and health-related quality of life at 6 and 12 months after randomisation

2. to evaluate the safety of i.v. salbutamol for ARDS patients

3. to evaluate the cost-effectiveness of i.v. salbutamol for patients with ARDS

4. to explore whether or not the effects of salbutamol vary between patients of different age, initial disease severity, mortality risk at ICU admission and ARDS aetiology.

Outcome measures

1. Primary outcome.

   1. All-cause mortality 28 days after randomisation.

2. Secondary outcomes.

   1. Mortality at (first) discharge from ICU.

   2. Mortality at (first) discharge from hospital.

   3. Number of VFDs.

   4. Number of organ failure-free days.

   5. Mortality at 12 months post randomisation.

   6. Side effects (tachycardia/new arrhythmia/lactic acidosis) sufficient to stop treatment with trial drug.

   7. Serious adverse events (SAEs) and suspected unexpected serious adverse reactions (SUSARs).

   8. Health-related quality of life: European Quality of Life-5 Dimensions (EQ-5D) and Short Form questionnaire-12 items (SF-12) at 6 and 12 months after randomisation.

   9. Length of stay in critical care unit.

   10. Length of stay in hospital.

3. Economic outcomes.

   1. Health service contacts up to 12 months after randomisation.

   2. Patient out-of-pocket expenditure and time away from work.

Ventilator-free days were defined as the number of calendar days after initiating unassisted breathing to day 28 after randomisation, assuming a patient survived for at least 48 consecutive hours after initiating unassisted breathing. 30 For example, if a patient initiated unassisted breathing on day 16 and survived to day 28, he/she was assigned a value of 12 VFDs. If a similar patient began unassisted breathing on day 16 but died on day 25, the number of VFDs was 9. If a patient survived for > 48 consecutive hours of unassisted breathing but required assisted breathing (for any reason) before day 28, the number of VFDs was the number of days of unassisted breathing before day 28. Patients who died without initiating unassisted breathing or before 48 consecutive hours of unassisted breathing were assigned a value of zero VFDs. Patients transferred to another hospital or other health-care facility prior to day 28 (intermediate care, nursing home, etc.) while still on positive pressure ventilation were followed to assess this outcome.

In the assessment of VFDs, unassisted breathing was defined as:

1. extubated with face mask, nasal prong oxygen, or room air; or

2. T-tube breathing; or

3. tracheostomy mask breathing; or

4. continuous positive airway pressure = 5 cmH₂O without pressure support or invasive mechanical ventilation assistance. 30

Organ failure-free days was defined as the number of days in the first 28 days after randomisation that the patient had no cardiovascular support, renal support, hepatic support or neurological support, according to Critical Care Minimum Data Set definitions. 8

Inclusion/exclusion criteria

Consent

In the majority of cases patients were unable to consent for themselves, and consent was initially sought from the patient's ‘personal legal representative’, who was a relative, partner or close friend. The representative was informed about the trial by the responsible clinician and provided with a copy of the patient information sheet and additional information for personal legal representatives. If the representative decided that the patient would have no objection to participating in the trial they were asked to sign three copies of the consent form which were then counter-signed by the responsible clinician. The representative retained one copy of the signed consent form, one copy was placed in the patient's medical records and one copy was retained in the trial site file. If no personal legal representative was available, a doctor who was not connected with the conduct of the trial acted as a professional legal representative.

Patients for whom consent was given by a personal legal representative or professional legal representative were informed of their participation in the trial by the responsible clinician once they regained capacity to understand the details of the trial. The patient was asked for consent to continue participation in the trial and to sign the consent to continue form. Patients were specifically asked whether or not they were happy for data collection to continue, to receive follow-up questionnaires and for data already collected to be used.

If a patient or their representative requested termination of infusion of the trial drug during the treatment period, the drug infusion was stopped but the patient continued to be followed up. If a patient or their representative withdrew consent for trial participation during trial treatment, the trial drug was stopped but permission was sought to access medical records for data related to the trial. If a patient or their representative withdrew from the trial after completion of the trial treatment, permission to access medical records for trial data was sought.

Randomisation

Patients were randomised using a 24-hour telephone randomisation service located at the University of Aberdeen.

Randomisation was minimised by centre, PaO₂–FiO₂ ratio (≤ 6.7, 6.8–13.2, ≥ 13.3 kPa) and age (≤ 64, 65–84, ≥ 85 years). The minimisation criteria were used to ensure balance within centres, and within strata that were expected to differ in mortality risk, according to published reviews. 5,10

The randomisation service used a computer-generated random number sequence, and allocated a numbered treatment pack to each patient. Each pack contained all of the drugs necessary for giving a complete course of trial treatment to one patient.

Each patient was allocated a unique six-digit patient trial number that was used throughout the trial as their unique identifier.

Trial treatments

The trial drug boxes contained 50 × 5-ml ampoules containing either salbutamol (Ventolin^TM solution, GlaxoSmithKline) for i.v. infusion 1 mg/ml, or placebo (sodium chloride injection BP, Hameln Pharmaceuticals Ltd) 0.9% weight/volume.

Drug pack preparation and supply

All trial drugs were packaged identically and identified only by number. Patient drug packs were prepared by Bilcare Global Clinical Supplies (Europe) Ltd (Elvicta Business Park, Crickhowell, Powys, UK). Each ampoule of salbutamol or placebo had a black out label applied, and 50 ampoules of either salbutamol or placebo were packaged in a white cardboard box in 10 trays containing five ampoules each. Each box contained sufficient material for the treatment of one patient for 7 days. The outside of the boxes were labelled only with the drug box number and Medicines and Healthcare Products Regulatory Agency (MHRA)-approved labelling identifying the contents as BALTI-2 study drugs. The drug packs were stored by Bilcare and dispatched by them to participating hospital pharmacies, as required. All clinical and trial personnel were blind to study treatment, and all assessment of outcomes was done without knowledge of treatment allocations.

Hospital pharmacies dispensed the trial drugs to their ICU. Because patients could be recruited outside normal pharmacy opening hours, two or more patient drug packs (at least one each of salbutamol and placebo) were kept available on each ICU at all times. When a patient was recruited, the randomisation service informed the recruiting clinician of the drug pack number to be allocated to the patient and the number of another drug pack, of the same treatment allocation, to be obtained from pharmacy. This ensured that there was always at least one drug pack of each allocation available in the ICU.

Administration of trial drug

Prior to infusion, two ampoules of trial drug were diluted with 40 ml of saline in a 50-ml syringe. Infusion syringes were made up immediately prior to use.

Salbutamol and placebo infusions were administered through a dedicated i.v. line at a rate of 0.075 ml/kg IBW/hour (equivalent to 15 µg salbutamol/kg IBW/hour). IBW was calculated from the patient's height; the patient was measured from heel to vertex using a soft tape measure and the IBW and infusion rate were obtained from the conversion table (Table 1). Trial drug infusions were started immediately after randomisation.

Alteration of infusion rate

Sinus tachycardia or arrhythmias are known side effects of i.v. salbutamol administration. If a patient receiving a trial drug infusion had tachycardia [heart rate (HR) > 140 beats/minute] or any new arrhythmia, the drug infusion rate was adjusted according to a prespecified protocol (Figure 1). Standard antiarrhythmic therapy was given if indicated in addition to alteration of infusion rate.

FIGURE 1.

Protocol for adjustment of study drug infusion rate. ECG, electrocardiogram.

Infusion termination criteria

The trial drug infusion was terminated before 168 hours in the following circumstances:

• death

• HR > 140 beats/minute despite two adjustments in infusion rate

• new arrhythmias despite adjustment in infusion rate

• development of a significant lactic acidosis, which in the opinion of the treating clinician was attributable to infusion of the trial drug

• 24 hours after discontinuation of mechanical ventilation (of any sort)

• discharge from critical care environment

• discontinuation of active treatment

• request to withdraw from personal legal representative or patient

• decision by the attending clinician that the infusion should be discontinued on safety grounds.

Otherwise, the infusion was terminated 7 days (168 hours) after randomisation. Reasons for early termination of the infusion were recorded in the CRF.

Clinical management of patients in the trial

Patients involved in BALTI-2 were managed according to best practice established locally on each unit. The only specific requirement was that patients were not routinely administered nebulised β₂-agonists or other i.v. β₂-agonists such as isoprenaline. The uncontrolled use of nebulised bronchodilators in the control group would limit the ability of the trial to detect a difference in outcomes and their use in the treatment group would expose patients to a risk of toxicity.

Serious adverse events and suspected unexpected serious adverse reactions

A SAE is defined as an adverse event that fulfils one or more of the following criteria:

• results in death

• is immediately life-threatening

• requires hospitalisation or prolongation of existing hospitalisation

• results in persistent or significant disability or incapacity

• results in congenital abnormality or birth defect

• requires medical intervention to prevent one of the above, or is otherwise considered medically significant.

Suspected unexpected serious adverse reactions are SAEs that are also unexpected, i.e. their nature or severity is not consistent with the Summary of Product Characteristics, and are considered to be caused by the study drug.

As BALTI-2 recruited a population that was already in a life-threatening situation, many of the participants were expected to experience SAEs. Events that were expected in this population and those that were collected as outcomes of the trial were not reported as SAEs. This included death and organ failure. SUSARs and side effects of salbutamol sufficiently severe to be fatal or immediately life-threatening were reported, using a specific SAE reporting form.

Data collection

Statistical methods

Statistical analysis

All analyses were conducted as far as possible by intention to treat, i.e. all patients were analysed in their randomised group regardless of the treatment actually received, and we sought to include all randomised patients in the analyses. We did not impute values for missing data.

Ethics and regulatory approvals

Ethics approval was given for the study by the Oxfordshire A Research Ethics Committee in September 2006. Local Research Ethics Committee (LREC) approval and permission from the research and development (R&D) department of each participating NHS Trust were required until April 2009, when the system was changed. LREC approval was no longer required after this date.

Funding and registration

The pilot phase of the trial was funded by the Intensive Care Society. The main phase of the trial was funded by the Medical Research Council (grant number 84730). Authorisation was given by the MHRA in 2006 (24698/0004/001). The trial was registered with the EudraCT (European Union Drug Regulatory Authorities Clinical Trials) database (2006-002647-86) and with the International Standardised Randomised Controlled Trial Number database (ISRCTN38366450).

Overview of recruitment

Patients were recruited between November 2006 and March 2010. Initially five hospitals in the West Midlands were recruited to the pilot phase, prior to acquisition of substantive funding, which allowed further centres across the UK to be opened. The main trial commenced recruiting in August 2008. Recruitment was terminated following the second interim analysis, in March 2010, when the DMEC reviewed the results for 273 patients. Owing to a significant adverse effect of salbutamol on 28-day mortality, the DMEC recommended closing recruitment to BALTI-2. The Trial Steering Committee endorsed the DMEC recommendation and closed recruitment on 23 March 2010. All patients receiving study drug at that time had their infusion discontinued (one salbutamol, two placebo).

Recruitment of centres

The pilot study was conducted at five centres in the West Midlands, which opened to recruitment between September and December 2006. The first patient was recruited in November 2006. The pilot study continued until July 2007, when it was terminated at four sites but continued at a single site. Following award of funding for the main trial in September 2007, a new supply of study drugs was ordered and further sites were opened. Recruitment to the main phase of the trial began in September 2008. Further sites were opened throughout recruitment and in total 46 ICUs in the UK recruited to the trial. A further 25 ICUs obtained approvals to start the trial but were unable to do so before recruitment was stopped.

At the start of the study, approval was required from a LREC and the R&D department of each participating NHS Trust before recruitment could commence. The requirement for LREC approval was removed in April 2009, so sites initiated after this date needed only R&D approval. As well as the approvals process, each site needed to sign a site agreement with the sponsor of the trial (University of Warwick) and the co-ordinating centre needed to organise training of the relevant clinicians and arrange drug supplies with the hospital pharmacy and Bilcare.

The process of setting up sites was extremely time consuming (Table 3). The average time for LREC and R&D approval was 92 days. In most cases the R&D approval was the rate-limiting step; submissions to LRECs and R&D were normally done concurrently and LREC took an average of only 34 days. It took an average of 123.5 days from the date of approval to the recruitment of the first patient. This was partly owing to time taken to get the site ready to recruit (training, site agreement and drug delivery) and partly due to time taken to identify, approach and consent an eligible patient once the site was open. The average time from submission of LREC and R&D approvals to recruitment of the first patient was about 218 days, with a range from 84 to 452 days.

Participants

A total of 326 patients were enrolled from December 2006 to March 2010. Of the 326 patients, 162 were randomly assigned to salbutamol and 164 to placebo. During the pilot phase, up to September 2008, 63 patients were recruited, and 263 were recruited to the main phase of the trial. One patient in each arm withdrew consent before assessment of the primary outcome and hence no outcome data were available for these patients. The statistical analysis of the primary outcome and other short-term outcomes was therefore based on 161 patients in the salbutamol arm and 163 patients in the placebo arm (Figure 2). One additional patient in the salbutamol arm withdrew from the follow-up. Survival status at 12 months could not be determined for 13 patients and in the survival analysis these patients were censored at the last time they were known to be alive (discharge from hospital or 6-month follow-up). A relatively small number of patients were followed up for quality-of-life outcomes at 6 and 12 months to evaluate quality of life and record economic data. This was for several reasons: a high proportion of patients died, some patients recruited to the pilot phase before the start of the main trial could not be followed up for resource reasons, some were not contacted because it was considered inappropriate by clinicians or carers and the response rate of those who were contacted was poor (Figure 3).

FIGURE 2.

Flow chart for 12-month mortality.

FIGURE 3.

Flow chart for 6- and 12-month follow-up for quality-of-life outcomes.

One centre (Birmingham Heartlands Hospital) recruited about 25% of the trial population, far more than any other centre (most of which recruited < 10 patients over the course of the trial) (Figure 4). This was achieved by the hospital recruiting for a longer period of time than any other centre and also as a result of a high monthly recruitment rate.

FIGURE 4.

Number of recruits for each participating centre.

The recruitment rate projected in the study protocol was one patient/centre/month. We therefore estimated that 37 centres would be needed to complete recruitment of 1334 patients in 36 months. It was planned to recruit in up to 50 centres, to give extra capacity in case of poor recruitment and to allow for the fact that some smaller units would be unlikely to recruit one patient/month. It became clear during recruitment that in addition to substantial delays in starting recruitment, few centres were achieving the target recruitment. Only six centres achieved a recruitment rate (after recruiting their first patient) above one patient/month (Figure 5). Additional centres were added to the trial to make up for the shortfall in recruitment and 72 centres were eventually involved.

FIGURE 5.

Number of recruits per month for each participating centre. The recruitment rate was calculated over the period from the centre's first recruit to the end of recruitment. Centres that recruited for < 30 days were omitted from Figure 5 (n = 2).

There appear to be several contributory reasons for poor recruitment. First, in some sites there were considerable delays from receiving approvals to recruitment of the first patient (see Table 3). These were caused by various factors, including organising drugs supplies with the pharmacy and supplier, signing site agreements with the sponsor, dissemination of information about the trial and training of staff, and developing systems for identifying eligible patients. Second, it is likely that in most sites a high proportion of eligible patients were missed. At one centre, Birmingham Heartlands Hospital, there was sustained careful screening of all patients against the eligibility criteria and it is unlikely that more than a few eligible patients were missed. This unit recruited at a rate of more than two patients/month. This suggests that most other sites would not have been limited by a shortage of patients and probably failed to recruit to target because many patients who were eligible were not identified.

Baseline characteristics

Baseline characteristics were, as expected, similar between the randomised groups (Table 4). There were very few patients in the oldest age stratum (aged ≥ 85 years), or the most severe stratum of PaO₂–FiO₂ ratio. Collection of end-expiratory tidal volume was only started about halfway through recruitment, hence the large number of participants with missing data for this variable. As the variable was collected continuously after its introduction, there is no concern that any selection bias could affect the comparison between the groups.

Acute Physiology and Chronic Health Evaluation II mortality risks could not be calculated for about 20% of participants. This was for several reasons, including failure of centres to provide data for calculation of the APACHE II score, unknown primary reason for admission, patient could not be identified in the ICNARC database and reason for admission was excluded from calculation of mortality risk.

Treatment with study drug

The study drug infusion was not given to two patients in the salbutamol arm. One of these required a β-blocker between randomisation and starting the study drug and the other patient's next of kin refused to have a separate i.v. line inserted for the study drug infusion, after initially giving consent.

Patients in the salbutamol group were more likely to have their infusion terminated early than patients in the placebo arm. This was mainly due to death (14/161 vs. 8/163), or the development of significant side effects (47/161 vs. 13/163). The duration of infusion was on average 24.5 hours shorter in the salbutamol group [mean (SD), 114.1 (62.7) hours vs. placebo group 138.6 (47.9) hours; p < 0.01]. The risks of developing a tachycardia, new arrhythmia or lactic acidosis severe enough to warrant stopping the study drug were substantially higher in the salbutamol group (Table 5).

A small number of patients had infusions that were given in error for more than 168 hours. Most of these were terminated soon after 168 hours when the error was noticed, but a few infusions were continued for considerably longer. Infusions of > 168 hours were more common in the placebo group (10 vs. 5 patients), presumably because more patients in the salbutamol group had already had their infusion terminated before reaching 168 hours and therefore the chance of accidentally exceeding 168 hours was lower.

Patients who withdrew from the study were specifically asked for permission to continue data collection, including the 6- and 12-month follow-ups. Two patients (one in each arm) withdrew and did not allow further data collection; hence, no outcome data were available for these patients. One patient withdrew from treatment only and was happy for data collection to continue, a further patient withdrew from the follow-up only and is therefore included in the analysis of short-term outcomes and long-term mortality.

Outcomes

Primary outcome

The risk ratio for death at 28 days in the salbutamol group compared with the placebo group was 1.47 (95% CI 1.03 to 2.08; p = 0.03) (Table 6). Salbutamol resulted in a 10.9% (95% CI 1.0% to 20.4%) absolute increase (34.2% vs. 23.3%) in 28-day mortality (see Table 6). There was one additional death for every 9.2 (95% CI 4.9 to 100.9) ARDS patients treated with salbutamol (number needed to treat for harm). Survival analysis for the primary outcome (Figure 6) showed a hazard ratio of 1.56 (95% CI 1.03 to 2.36).

TABLE 6 - Short-term outcomes

IQR, interquartile range; RR, risk ratio.

Data collection for high-frequency oscillatory ventilation was introduced part way through recruitment.

	Salbutamol (%) (n = 162)	Placebo (%) (n = 164)	Statistics (95% CI)
Primary outcome; mortality at 28 days post randomisation	55 (34.0)	38 (23.2)	RR = 1.47
Missing	1	1	(1.03 to 2.08)
Death before discharge from ICU	58 (35.8)	45 (27.4)	RR = 1.31
Missing	1	1	(0.95 to 1.80)
Death before discharge from hospital	62 (38.3)	53 (32.3)	RR = 1.18
Missing	1	1	(0.88 to 1.59)
Tachycardia sufficient to stop treatment with study drug	23 (14.2)	2 (1.2)	RR = 11.71
Missing	1	0	(2.81 to 48.88)
New arrhythmia sufficient to stop treatment with study drug	14 (8.6)	3 (1.8)	RR = 4.75
Missing	1	0	(1.39 to 16.23)
Other side effects sufficient to stop treatment with study drug	10 (6.2)	1 (0.6)	RR = 10.19
Missing	1	0	(1.32 to 78.66)
VFDs
Mean (SD)	8.46 (8.83)	11.14 (9.32)	Difference –2.68
Median (IQR)	6 (0–16)	13 (0–20)	(–4.67 to –0.70)
Range	0–26	0–27
Missing	1	1
Organ failure-free days
Mean (SD)	16.2 (10.7)	18.5 (9.8)	Difference –2.30
Median (IQR)	19 (5–26)	23 (12–26)	(–4.54 to –0.06)
Range	0–28	0–28
Missing	1	1
High-frequency oscillatory ventilationa	10	7	RR = 1.53
Missing	81	77	(0.61 to 3.84)
Duration of ICU stay (days)
Mean (SD)	17.6 (14.3)	17.1 (14.0)	Difference +0.5
Median (IQR)	15 (8–23)	13 (7.5–21.5)	(–2.6 to 3.6)
Range	0–85	0–91
Missing	1	1
Duration of hospital stay (days)
Mean (SD)	32.5 (35.9)	34.9 (36.3)	Difference –2.4
Median (IQR)	23 (12–39)	22 (14–42)	(–10.3 to 5.5)
Range	0–191	0–243
Missing	1	1
Duration of ICU stay excluding deaths (days)	n = 103	n = 118
Mean (SD)	20.5 (15.3)	17.1 (12.6)	Difference +3.4
Median (IQR)	17 (11–23.5)	13 (8–22)	(–0.3 to 7.1)
Range	3–85	1–82
Duration of hospital stay excluding deaths (days)	n = 99	n = 110
Mean (SD)	42.4 (37.8)	40.7 (38.9)	Difference +1.7
Median (IQR)	32 (20–49)	26 (17–49)	(–8.8 to 12.2)
Range	4–277	7–243
Missing	1	1
Days level 3 care
Mean (SD)	14.1 (8.0)	13.2 (8.0)	Difference +0.9
Median (IQR)	13 (7–20)	11 (7.5–19)	(–0.85 to 2.65)
Range	1–28	1–28
Missing	1	1

FIGURE 6.

Kaplan–Meier plot for survival over the first 28 days.

The result for the primary outcome at the second interim analysis, when the DMEC took the decision to recommend stopping recruitment based on data from 273 patients, was a risk ratio of 1.55 (95% CI 1.07 to 2.24) and the 99.8% CI excluded a benefit for salbutamol of the size anticipated in the protocol.

Short-term secondary outcomes

Consistent with 28-day mortality, the risk ratio for death within ICU was 1.31 (95% CI 0.95 to 1.80; p = 0.10) and for death within hospital was 1.18 (95% CI 0.88 to 1.59; p = 0.26) (see Table 6). There was an 8.4% absolute increase (95% CI –1.7% to 18.3%) in ICU mortality and 6.0% increase in hospital mortality (95% CI –4.4% to 16.2%) in the salbutamol group.

The differences between the groups in ICU and (especially) hospital mortality were smaller than the difference in 28-day mortality. This was because the deaths in the salbutamol group occurred earlier, although the final proportions that eventually died in hospital were fairly similar (Figure 7). Deaths in hospital continued to occur up to nearly 200 days after randomisation.

FIGURE 7.

Cumulative numbers of deaths in hospital with time from randomisation.

Ventilator-free days and organ failure-free days during the first 28 days after randomisation were both reduced in the salbutamol group (see Table 6). No clear differences were detected between the groups in lengths of ICU and hospital stays, or in level 3-care bed days (see Table 6). ARDS survivors in the salbutamol group required on average 3.4 more days in ICU than those in the placebo group (95% CI of difference –0.3 to 7.1 days).

Long-term outcomes

Mortality at 12 months was higher in the salbutamol group than in the placebo group (42.0% vs. 38.5%), but the 95% CI included 1, and the data are compatible with a small reduction in 12-month mortality or an increase (Table 7). The survival analysis of death up to 12 months after randomisation gave a hazard ratio of 1.18 (95% CI 0.84 to 1.68); the Kaplan–Meier plot showed that although the final difference in the proportions that died at 12 months was moderate, the majority of deaths occurred earlier in the salbutamol group (Figure 8).

FIGURE 8.

Kaplan–Meier plot for 12-month survival.

The proportion followed up at 6 and 12 months was disappointingly low (see Figure 3). This was for a combination of reasons. First, owing to a lack of resources the pilot study did not include the follow-up; the intention was to begin following up patients when the main study started and resources allowed. Hence, 33 patients were not included in the follow-up. Second, 22 patients at 6 months and 16 patients at 12 months were not contacted for a range of other reasons (see Figure 3). Third, the proportion of questionnaires returned was low (52.7% and 55.0% of those contacted at 6 and 12 months, respectively).

No clear differences were found in the SF-12 and EQ-5D scores at 6 months, though all were lower in the salbutamol group. A similar pattern was found at 12 months; all of the scores were lower in the salbutamol group and for the SF-12 physical component score the 95% CI did not include zero. These results suggest that, consistent with the short-term outcomes, salbutamol may be associated with a lower quality of life 6 and 12 months after randomisation, but these results should be interpreted with considerable caution because of the high number of missing data.

TABLE 7 - Long-term outcomes

RR, risk ratio.

	Salbutamol (%) (n = 162)	Placebo (%) (n = 164)	Statistics (95% CI)
Death in 12 months after randomisation	66 (40.7)	60 (36.6)	RR = 1.09
Missing (status unknown)	5	8	(0.83 to 1.43)
Health-related quality of life at 6 months
	n = 42	n = 32
SF-12 physical component score; mean (SD)	34.9 (11.0)	38.9 (11.3)	Difference −3.61 (−8.85 to 1.63)
SF-12 mental component score; mean (SD)	42.1 (13.4)	44.4 (12.5)	Difference −2.25 (−8.30 to 3.79)
	n = 43	n = 33
EQ-5D	0.52 (0.39)	0.60 (0.33)	Difference −0.09 (−0.25 to 0.08)
Health-related quality of life at 12 months
	n = 37	n = 37
SF-12 physical component score; mean (SD)	37.8 (12.3)	43.6 (12.6)	Difference −5.78 (−11.56 to −0.01)
SF-12 mental component score; mean (SD)	45.0 (12.4)	50.4 (11.7)	Difference −5.39 (−10.97 to 0.19)
	n = 39	n = 38
EQ-5D	0.59 (0.37)	0.76 (0.22)	Difference −0.14 (−0.28 to 0.01)

Sensitivity analyses

We performed sensitivity analyses to explore adjustment of the analysis of the primary outcome for baseline variables (age, gender, PaO₂–FiO₂ ratio and aetiology). No adjustment for any of the baseline factors alone or in combination made a substantial difference to the estimate of the treatment effect of salbutamol or altered the conclusions. The unadjusted odds ratio (OR) was 1.71 (95% CI 1.05 to 2.78) and the adjusted ORs were between 1.70 and 1.76, and all had a lower 95% confidence limit > 1.

Subgroup analyses

We investigated modification of the treatment effect by four prespecified factors: aetiology, age, severity of hypoxaemia and APACHE II mortality risk. Subgroup analyses did not suggest that the effects of salbutamol were modified by any of the variables investigated. For aetiology (categorical subgrouping variable) the ratio of risk ratios was 0.96 (95% CI 0.46 to 2.01) (Table 8). The analysis by age suggested weak evidence of a possible interaction effect, whereby salbutamol is superior to placebo in the oldest patients [ratio of ORs 0.97 (95% CI 0.93 to 1.00); p = 0.07 (Table 9)]. However, as the effect is small and there were very few patients in the oldest age stratum (n = 4 aged > 85 years) this is likely to be a chance finding. For the other continuous variables, there was no evidence of an interaction (Tables 10 and 11 and Figure 9). The ratios of ORs for each variable investigated were: severity of hypoxaemia 1.02 (95% CI 0.92 to 1.14); p = 0.66 and mortality risk 1.29 (95% CI 0.08 to 22.04); p = 0.86.

FIGURE 9.

Subgroup analyses: variation of estimated odds of primary outcome with (a) age; (b) severity of hypoxaemia; and (c) APACHE II mortality risk.

Serious adverse events

Serious adverse event reports were received for 14 patients, 10 in the salbutamol group and 4 in the placebo group (see Appendix 1).

Causes of death

Data on cause of death were returned for 91/93 patients who died by day 28 (97.8%): 55/55 in the salbutamol group and 36/38 in the placebo group.

Owing to the diversity of individual diagnoses, causes of death results were grouped according to organ system. Respiratory system diagnoses were the most common primary cause of death in both groups (salbutamol 50.9% vs. placebo 52.6%), followed by multiorgan failure (salbutamol 21.8% vs. placebo 36.8%). ARDS was recorded on the death certificate for 11/55 patients in the salbutamol group and 8/38 patients in the placebo group.

Respiratory system diagnoses most commonly accounted for the primary (1a) cause of death in both groups, 50% of the salbutamol group and 52.6% of the placebo group (Table 12). Pneumonia/lower respiratory tract infection was the single most frequent cause of death recorded in 1a in the salbutamol group (28.6%), while multiorgan failure was most commonly recorded in the placebo group (36.8%). Respiratory system diagnoses predominated in cause of death 1b in both groups (30.4% vs. 29%; Table 13). The single most frequent diagnosis in the salbutamol group was pneumonia/lower respiratory tract infection (19.6%) compared with sepsis/septicaemia in the placebo group (23.7%).

In the majority of patients in both groups no entry was recorded under cause of death 1c. Respiratory system diagnoses continued to predominate in both groups where a diagnosis was recorded in this section of the certificate; 10.7% in the salbutamol group and 13.2% in the placebo group. Pneumonia/lower respiratory tract infection was the commonest single diagnosis in both groups (8.9% and 10.5%).

The number of patients with comorbid conditions recorded under cause of death 2 on the certificate was significantly greater overall in the salbutamol group (62.5%) compared with the placebo group (36.8%). Cardiovascular system diseases were most frequently recorded in the salbutamol group (23.2%) but rarely in the placebo group (5.3%). There were no other major differences between individual or system diagnoses between the two groups.

Acute respiratory distress syndrome was recorded on the death certificate for 11/56 patients in the salbutamol group and 8/38 patients in the placebo group. In the salbutamol group, the death certificate indicated a direct aetiology for ARDS in seven patients and an indirect aetiology in four patients. In the placebo group the death certificate indicated a direct aetiology for ARDS in six patients and an indirect aetiology in two patients.

Aim and perspective

Initially the aim of the BALTI-2 economic analysis was to assess the within-trial cost-effectiveness of a 7-day continuous infusion of salbutamol at 12 months and to construct a cost-effectiveness model with a lifetime horizon. However, owing to the limitations of available data (particularly follow-up data post hospital discharge) resulting from the trial being stopped early, the economic analyses were changed from those described in the original protocol. In this chapter we present the following:

• analysis 1: health-care costs at 28 days

• analysis 2: cost-effectiveness analysis at 28 days based on life-years gained

• analysis 3: cost–utility analyses at 6 and 12 months.

Although analyses 1 and 2 allow a larger sample to be included in the evaluation, giving more robust results, they do not provide any information on longer-term cost-effectiveness or quality of life. Analysis 3, including follow-up data at 6 and 12 months, gives a more comprehensive picture of cost-effectiveness; however, there is greater uncertainty in the estimates because of the considerably smaller sample size.

Description of methods of analyses

Time frame

Costs were calculated for the duration of the initial hospital stay up to 28 days post randomisation for analyses 1 and 2 and for a 6- and 12-month time frame for analysis 3.

Sample size

The sample sizes that were used for the analyses are provided in Table 14. For the cost analysis and cost-effectiveness analyses at 28 days (analyses 1 and 2) the total sample included 162 patients who had been randomised to the salbutamol arm of the trial and 164 patients who were randomised to placebo. For the cost–utility analysis (analysis 3) a sample of 75 patients had completed the health-related quality of life questionnaire at 6 months. Only 54 patients completed both the 6- and 12-month questionnaires. As highlighted earlier, in order to maximise the sample size we imputed missing values using the LOCF method for all those patients who were alive at 12 months and had completed the questionnaire at 6 months but had not completed the questionnaire at 12 months.

Resource use

Resource use relating to participants' initial hospital episode was collected prospectively from patients' hospital records. The mean values and the SDs of the individual items in the prospectively collected hospital data are reported in Table 15. These data are reported for the full sample and for a subsample of those with follow-up data at 6 months. Non-parametric tests found no significant differences in resource use between the two arms of the trial.

Post-discharge use of health and social care was collected retrospectively by way of postal patient self-completed questionnaires at 6 and 12 months. Table 16 shows the resource use from discharge to 6 months and Table 17 shows the resource use for 6–12 months.

Unit costs

The analyses assume price year of 2010. The cost of salbutamol was derived from the British National Formulary. Within the trial patients received an i.v. infusion of salbutamol given at 15 μg/kg IBW/hour for up to 7 days compared with a placebo (0.9% sodium chloride) infusion. The economic analysis assumes an average body weight of 70 kg. Given a net price of £2.48 for a 5-ml ampoule [Ventolin^® (A&H) solution for i.v. infusion, salbutamol (as sulphate) 1 mg/ml] this equates to a cost per hourly infusion of £0.62 as shown below:

It is assumed that each 5-ml ampoule will allow for 4 hours of treatment with some wastage.

Cost per hour = £2.48/4 = £0.62 (2012 price) (deflated using Consumer Price Index to 2010 price = £0.58).

The unit costs for hospital resources collected prospectively through patient records during participants' initial hospital episodes are provided in Table 18. The unit costs used in the study for health- and social-care services collected retrospectively by way of patient-completed questionnaire inputs are provided in Table 19.

Results

Analysis 2: cost-effectiveness analysis at 28 days

The data on life-years gained at 28 days for both arms of the trial are shown in Table 22. An individual who was alive after 28 days was allocated 0.0767 (28/365) of a life-year gained. In the case of those patients who died within 28 days, the values are based on number of days the patient survived post randomisation.

TABLE 22 - Life-years gained (28 days)

Life-years gained, mean (SD)
Salbutamol	Placebo
0.061 (0.024)	0.066 (0.021)

Costs to the NHS for the 28-day period were calculated based on patient resource use (days in hospital by level of care and drug use) (see Table 18). The mean costs are shown in Table 23.

TABLE 23 - Mean cost per patient (28 days)

Total cost (£), mean (SD)
Salbutamol	Placebo
23,083.97 (11,248.64)	22,462.07 (10,953.14)

The ICER shown in Table 24, calculated using deterministic values and the results from bootstrapping, shows salbutamol to be dominated by the placebo as it is associated with higher resource costs and is less effective (results in fewer life-years gained).

TABLE 24 - Expected ICER (28 days)

	Sample mean		Bootstrapped estimated mean
	Salbutamol	Placebo	Salbutamol	Placebo
Sample size	162	164	10,000
Expected incremental life-years (effectiveness)	−0.0047		−0.005
Expected incremental cost (£)	623.13		701.00
Expected ICER (£)	Dominated (−131,385.83)		Dominated (−142,297.73)

In order to explore the levels of uncertainty surrounding this result, Figure 10 presents the cost-effectiveness plane. This shows a large degree of uncertainty surrounding the results of the ICER calculation. The CEAC (Figure 11) shows for each threshold of willingness to pay the probability that salbutamol would be cost-effective.

FIGURE 10.

Incremental cost-effectiveness plane (28 days).

FIGURE 11.

Cost-effectiveness acceptability curve (28 days).

Incremental cost-effectiveness ratios

The ICERs showing the ratios of the incremental cost and incremental benefits (QALYs) between the salbutamol and placebo arms over 6 and 12 months are given in Table 28. The results indicate that the salbutamol arm of the trial is more costly and less effective than the placebo arm of the trial over both 6 and 12 months (i.e. dominated by the placebo).

TABLE 28 - Expected ICERs

	6-month data (baseline = −0.402)		12-month data (baseline = −0.402), using 6-month data for missing data		6 months (baseline = −0.402)	12 months (baseline = −0.402), including 6 months data
	Salbutamol	Placebo	Salbutamol	Placebo	Simulation	Simulation
Based on sample size	43	33	43	32	10,000	10,000
Difference in QALYs	−0.021		−0.081		−0.021	−0.079
Difference in costs (£)	1873.83		1798.80		1488.51	1854.00
ICER (£)	−88,878.63		−22,176.99		−72,062.85	−23,478.09

In order to explore the levels of uncertainty surrounding the result, Figures 12–14 present CEACs and cost-effectiveness planes. The data indicate, using the mean values for both the 6- and 12-month analyses, that the placebo is more cost-effective. Over a 12-month period (see Figure 12), while there are some points on the right-hand side of the quadrants, indicating that there is a possibility in the longer term that salbutamol could be more cost-effective, the majority of points are still on the left-hand side of the quadrants. Similarly, Figure 13 shows that there are few examples at a 6-month level where salbutamol is more effective than the placebo, with the majority of the simulated points in the left two quadrants.

FIGURE 12.

Cost-effectiveness plane for salbutamol vs. placebo (bootstrapping output) 12-month time frame with imputed values.

FIGURE 13.

Cost-effectiveness plane – 6-month time data.

FIGURE 14.

Cost-effectiveness acceptability curves – 12-month data (with imputed values).

Summary

There were several limitations in this economic evaluation. The analysis at 28 days (analysis 1) indicates that mean cost was higher for patients in the salbutamol arm of the trial than in the placebo arm, both for those who survived 28 days from randomisation and those who died within 28 days. Analysis 2 also shows that the mean costs to the NHS were higher and the life-years gained were lower for the salbutamol arm of the trial. Both these analyses are based on data from the patients' initial hospital episode and as such do not include health- and social-care use between discharge and 28 days (for those participants who survived and were discharged within the 28-day period). Also, no additional costs were included for administering the infusion. As such, the costs are likely to be underestimated. It was not originally intended to undertake analysis at 28 days and while data were collected on health- and social-care use post discharge, and have been included in analysis 3, it was not possible to disaggregate those relating to this period for analyses 1 and 2.

We made assumptions regarding the baseline utility values. Baseline utility value was not collected within the study because of the nature of the intervention, with patients residing in ICU following admittance to the hospital. Within analysis 3 we have assumed a baseline utility value of –0.402 representing ‘unconscious’, which is in line with previous studies. 37

In order to facilitate analysis of the longer-term cost-effectiveness it was necessary to impute missing data given a large proportion of patients not completing the follow-up questionnaires (full details of the follow-up quality-of-life data have been given in the results section of the main trial). Zero utility scores were assigned to participants who had died and we used the method of LOCF for the remaining missing values. The advantage to this approach is that it minimises the number of subjects who are eliminated from the analysis, and it allows the analysis to examine the trends over time, rather than focusing simply on the end point. Other studies have shown a gradual improvement in quality-of-life scores over 12 months following admission to ICU50 so LOCF is likely to be a conservative way of dealing with missing utility data.

As highlighted earlier, in addition to the prospective resource use data used in analyses 1 and 2, analysis 3 also uses retrospective resource use data collected from patients by postal questionnaire. No resource use data were available post discharge for those patients who survived 28 days but died within 6 months. For this group we have assumed costs equivalent to those in analysis 1. This is likely to be an underestimate as no cost is applied for use of health or social care post discharge. If data at 12 months were missing but the individual was still alive, we have assumed that their costs between 6 and 12 months are equivalent to the average costs for this time frame for the arm of the trial that they were allocated to. However, this also has the potential to bias the results based on the small sample size.

The analysis 3 results indicate that patients who were randomised to the salbutamol arm of the trial had worse health outcomes at 6 months (lower utility values) and incurred a higher mean cost to the health- and social-care sector; however, given the restrictions owing to the small sample size and limited follow-up data, the results should be treated with caution.

Overview of the trial findings

The Beta-Agonist Lung injury TrIal-2 found that i.v. administration of salbutamol to patients with ARDS increased 28-day mortality and reduced VFDs and organ failure-free days compared with treatment with placebo. 51 The findings were unexpected, but the trial provided a definitive answer to the question of whether or not i.v. infusion of β₂-agonists should be used in patients with ARDS.

Internal validity and methodological limitations

This evaluation used a randomised, placebo-controlled design, which should ensure that the result is as reliable as possible. Very small numbers of patients did not receive the planned interventions (2/326; 0.6%) or had missing data for the primary outcome (2/326; 0.6%); these are very unlikely to have affected the estimated treatment effect.

The most notable methodological issue is that the trial was terminated at a smaller sample size than originally intended, on the recommendation of the DMEC, because of excess mortality in the salbutamol group found at an interim analysis. This has a number of potential consequences. First, the precision of treatment effect estimates is inevitably lower than anticipated, because of the smaller sample size. This is likely to make the effects of the intervention on some outcomes less clear than would be the case if the trial had reached its planned sample size. A larger sample size and narrower CIs would probably clarify salbutamol's effects on some outcomes such as mortality in ICU and in hospital. Second, there is a small possibility that the increased risk of death in the salbutamol group was a chance finding, arising because the interim analysis happened to coincide with a time when more deaths had occurred in the salbutamol group. The trial was stopped when < 25% of the planned sample size had been recruited, and the treatment effect of salbutamol could have changed substantially if the trial had continued to recruit up to its original target. The excess of deaths in the salbutamol group might then have been small or non-existent, and it is even possible, though unlikely, that the treatment effect could have been reversed into a benefit for salbutamol. Thus, there is a remote possibility that the early termination of the trial may have led to a beneficial treatment effect being missed.

A substantial number of earlier studies, including one clinical trial, indicated potential benefit for salbutamol, and there was sufficient clinical interest in β₂-agonists to initiate trials in the UK and North America. 28 However, the final result of BALTI-2 was in the opposite direction. This may have occurred partly because of the different outcomes measured by different studies. The early studies were concerned primarily with physiological outcomes, and it is possible that these are simply poor surrogates for substantive clinical outcomes and were inaccurate predictors of salbutamol's effects on mortality. Alternatively, the results on physiological parameters may have been misleading because of chance and bias. There are numerous examples in many fields of medicine where a treatment that seemed promising based on preliminary studies was found by a subsequent RCT to be ineffective or harmful. There is currently a huge investment of time and resources in large-scale RCTs that fail to show any treatment benefit, which suggests that current ways of selecting interventions for evaluation in large-scale RCTs may not be optimal. More rigorous selection of promising interventions could increase the number of large trials that provide evidence of benefit, and hence speed up improvements in health care.

We were successful in following up a high percentage of the population for 12-month mortality. The survival status at 12 months could not be ascertained for only 13 patients (4%) and most of these provided some data to the survival analysis, censored at the last point they were known to be alive. Follow-up for mortality used routinely collected data systems and did not require a response from the individual participants. Eleven patients could not be traced for mortality; one was known to have emigrated, another was from overseas and did not have a permanent address in the UK, and the remaining nine had insufficient identifying details for tracing because the recruiting hospitals were unable to supply them, despite efforts by the study team to retrieve this information. For these nine patients, the only information provided was initials, gender and date of birth, whereas address and preferably NHS number are also needed for successful and accurate tracing. Inability to trace participants was therefore mainly due to problems in data collection rather than the tracing system. A recommendation for future trials is to set up mechanisms to collect identifying details, especially NHS numbers, to facilitate long-term follow-up. A possible weakness in the data collection system of BALTI-2 was that it relied on hospital staff to provide contact information; if this could be collected directly from participants or their relatives, it would probably be more complete and would allow collection of alternative addresses and modes of contact (such as mobile telephone or e-mail). Achieving this in a multicentre trial is, however, challenging.

Previous intensive care trials have found it difficult to achieve high rates of long-term data collection52 and the experience of BALTI-2 confirms that this is a difficult population to follow-up. Only about 40% of survivors were successfully followed up at 6 and 12 months for quality-of-life outcomes. This was due to a high level of participants who were not contacted, plus a very poor return rate for postal questionnaires, despite the use of evidence-based strategies to improve the response rate. The protocol for chasing questionnaires that were not returned has been used successfully in previous trials,53 but was not noticeably successful in BALTI-2. In part its usefulness was limited by lack of contact information; telephone numbers were only supplied by hospitals for a limited number of participants. Telephone data collection was therefore not possible in many cases. A combination of factors was probably responsible for the low response rate; many of the population were elderly or disabled, and their motivation to return questionnaires may have been low because they were recruited to the trial while unconscious and did not feel any sense of personal investment. The follow-up questionnaires incorporated both clinical and economic data collection, and may have appeared long and time-consuming, discouraging completion. A possible strategy in future trials would be to separate the economic and clinical parts of the questionnaire. This would increase the administrative burden of follow-up and increase the number of questionnaires that participants were expected to complete, but might promote higher return rates of a shorter clinical questionnaire, although it could have a negative impact on return of the economic (or both) questionnaires. So far this idea has not been tested.

Obviously, the large number of missing data means that conclusions can only be very tentative because of the risk of bias, but the data suggested that quality of life may be lower at 12 months in the salbutamol group. This was in the same direction as the results for mortality.

A possible limitation is that we did not measure tidal volume after randomisation. It is possible that the use of salbutamol may have made low tidal volume ventilation more difficult, and this may have influenced the outcomes. This could occur through stimulation of metabolic activity and increased production of carbon dioxide, although there was no evidence of such an effect in the earlier BALTI-1.

Salbutamol and increased mortality

The finding of excess mortality in the salbutamol group mirrors that of a study of early goal-directed therapy in which critically ill patients were treated with i.v. dobutamine, which has mixed beta 1 and dopaminergic receptor actions. 54 The mechanisms underlying the increase in mortality because of salbutamol are unclear. BALTI-2 was designed as a pragmatic trial and hence did not collect data that might help to explain the increase in mortality, such as cardiovascular comorbidity and causes of death. It may be significant that the survival curves for salbutamol and placebo appear to continue to diverge after the end of the study drug infusion (see Figures 6 and 8). This suggests that the mechanisms may be complex, and involve indirect effects. Several possible causes of increased mortality can be suggested. First, salbutamol is known to cause arrhythmia and tachycardia, and as many patients with critical illness have comorbid cardiovascular disease, it may cause adverse cardiovascular events, leading to greater mortality in these patients. Second, known side effects of salbutamol include hypokalaemia, hypomagnesaemia and lactic acidosis, which may adversely affect outcomes in some patients. Third, the dose of salbutamol used in BALTI-2 (15 µg/kg IBW/hour) was selected after an early dose-ranging study identified it to be the maximum dose that critically ill patients could receive without an increase in ventricular or atrial tachycardia or ectopy. The dose is at the higher end of the manufacturer's recommended dosing regimen, and it is possible that a lower dose might have been better tolerated and caused fewer adverse outcomes.

External validity and generalisability

The intention was for BALTI-2 to be a pragmatic trial, embedded as far as possible in usual UK practice for caring for patients with ARDS. Therefore, apart from the trial intervention, patients were treated according to their hospital's standard care. The inclusion criteria were not tightly controlled, but allowed scope for slightly different practice in different centres in, for example, definition of the time of onset of ARDS and interpretation of chest radiographs. This is appropriate, because it should ensure that the population recruited to the trial is representative of ARDS patients treated in UK hospitals. The trial should have involved minimal change to the way patients were identified, although it is possible that in some centres screening of patients for the trial may have identified ARDS patients more quickly than would have been the case without the trial.

Two other randomised trials investigating the effects of β₂-agonists in patients with ALI/ARDS have been published. The first BALTI19 used the same intervention as BALTI-2, but was designed as a Phase II trial and was not powered to address clinical outcomes. It suggested an advantage to salbutamol based on physiological outcome; however, mortality was considerably higher than we found in BALTI-2 (58% of the salbutamol group and 66% of the placebo group died). The patient characteristics were similar to those recruited to BALTI-2, so the reasons for higher mortality in the earlier trial are unclear.

ALbuterol for the Treatment of ALI was a multicentre RCT of nebulised salbutamol in patients with ALI, conducted in the USA. 28 Patients were randomised to receive either salbutamol 5 mg every 4 hours or saline placebo, for up to 10 days. The primary outcome was VFDs. Recruitment started in August 2007 with a target sample size of 1000 patients, but was terminated by the Data and Safety Monitoring Board after 282 patients had been enrolled, on the grounds of futility. There was no clear difference in VFDs between the salbutamol and placebo arms (14.4 vs. 16.6 days; 95% CI –4.7 to 0.3 days) or hospital mortality (salbutamol 23.0% vs. placebo 17.7%; 95% CI –4.0% to 14.7%). Although the intervention was delivered by a different route in ALTA, and the early termination of recruitment means that the CIs are wide, the results are similar to those of BALTI-2 (there was an increase in mortality and decrease in VFDs in the salbutamol group).

The mortality in the placebo group (23.7%) was considerably lower than assumed in the original sample size calculation. This is likely to have been, at least in part, due to a reduction in the mortality caused by ARDS in recent years due to improvements in treatments. 55 We used a lower estimate of mortality than was found in earlier published studies, because it was likely that mortality had improved since these data were collected in 1999. However, mortality was also considerably lower than was suggested by ICNARC data from 2005 (23.7% vs. 41.2%). It is possible that the unexpectedly low mortality in the BALTI-2 population may be partly owing to selection effects; it is very likely that at most centres, only a fraction of the eligible patients were recruited to the trial, and it is possible that patients with the highest risk of mortality tended not to be recruited. There is little direct evidence of this, but it could have implications for the generalisability of the results.

Recruitment of centres and patients

Recruitment to BALTI-2 was more difficult than anticipated, for two main reasons: (1) the process of obtaining approvals and setting up sites was time-consuming; and (2) once recruitment had started, few centres were able to recruit as many patients as was anticipated.

Many trials experience significant delays in obtaining NHS approvals. We found that the main delay was with NHS Trust R&D approvals, a step that is supposed to be a simple assessment of whether or not there are particular local reasons that would prevent that hospital from running the research project. We also found substantial delays from granting of NHS R&D approval to recruiting the first patient. In part, this was a consequence of the delays in R&D approval; the trial co-ordinating centre was not willing to commit resources to centres before R&D approval had been given because of the uncertainty about the amount of time this would take. Time spent in training staff could easily be wasted effort if recruitment was not able to begin until several months later.

Only a small number of centres achieved the target of one recruit/month once they had commenced recruitment (and fewer would achieve the target if delays in starting recruitment were included). The centres that did manage to recruit to target were generally not centres with particularly large or high-risk populations, and we therefore speculate that the main reason for slow recruitment was that eligible patients were not recruited, rather than that there was a lack of eligible patients. There may be several reasons why patients were not recruited: they may not have been recognised as eligible, they may have been identified as eligible but not approached or they may have been approached but not given consent. A further possibility is that the eligibility criteria were applied in different ways in different centres. Although they were as objective as possible, there remains some subjectivity within the definition of ARDS, and it is possible that some patients would have been regarded as eligible in some centres but not others. Recruiting centres were asked to complete screening logs of potentially eligible patients, but these were not well completed, so we do not have enough information to perform any analysis of the reasons for non-recruitment. It is likely to be true that greater resources for participating centres might allow them to screen patients more effectively, to ensure that relatives of eligible patients are always contacted about randomisation, and that initial contacts are followed up. It would also be helpful for trial co-ordination to ensure that detailed logs of eligible patients are kept by participating centres, which should record the number of patients fulfilling the eligibility criteria, the number excluded and the number actually recruited, with reasons for non-recruitment. This would enable the trial co-ordination team to see where problems in recruitment were occurring. However, most centres found that screening was time consuming and with limited resources they were unable to do it consistently. Obtaining detailed screening and eligibility information is therefore likely to be expensive.

Implications for health care

The results of this trial showed that i.v. salbutamol at a dose of 15 µg/kg IBW/hour, early in the course of ARDS, was poorly tolerated, is unlikely to be beneficial and could worsen outcomes. The evidence from this and other trials suggests that salbutamol is unlikely to be a beneficial treatment for ARDS.

Implications for research

Further trials of β-agonists in patients with ARDS are unlikely to be conducted. Some questions remain, such as whether or not there may be benefit at a different dose or in specific populations, but any studies investigating these would require a very strong rationale.

Contribution of authors

Simon Gates (Principal Research Fellow): design of study, funding application, supervision of project, statistical analysis and writing study report.

Gavin D Perkins (Professor of Critical Care): concept and design of study, funding application, supervision of project and writing study report.

Sarah E Lamb (Professor of Rehabilitation, Director of Clinical Trials Unit): design of study, funding application, supervision of study and writing study report.

Charlotte Kelly (Senior Research Fellow, Health Economics): health economic analysis and writing study report.

David R Thickett (Wellcome Intermediate Clinical Fellow): supervision of study and clinical expertise.

Duncan Young (Consultant in Anaesthetics and Critical Care): concept and design of study, funding application and writing study report.

Daniel F McAuley (Professor of Critical Care): recruitment, interpretation of results and writing study report.

Catherine Snaith (Clinical Fellow): data collection and analysis for causes of death and writing study report.

Christopher McCabe (Professor of Health Economics): design of study and supervision of health economic analysis.

Claire T Hulme (Senior Lecturer, Health Economics): supervision of health economic analysis and writing study report.

Fang Gao Smith (Professor of Anaesthesia, Critical Care and Pain): concept and design of study, funding application, supervision of project and writing study report.

BALTI-2 Study Group

Trial Management Group: F Gao (chief principal investigator), G Perkins, S Gates (statistician), S Lamb, D Young, V Barber, C McCabe (health economist), S Duggan and T Melody.

Trial Steering Committee: S Baudouin (chairperson), B Cuthbertson, K Rowan, B Williams, F Gao, S Gates, S Lamb and D Young.

DMEC: K Wheatley, J Bion and G Bellingan.

Statistics: S Gates and JL Smith.

Health economics: C McCabe, CE Hulme and C Kelly.

Trial management: S Duggan and T Melody.

Trial co-ordinators: T Latif, J Bell, R Ezra and B Hoddell.

Administration and data entry: H Johnson, I Ruders and C Snaith.

Recruitment facilitators: S Dale, H Reay, A McLaughlin and V Gordon.

Clinical Collaborators

Birmingham City Hospital: Z Khan.

Birmingham Heartlands Hospital: F Gao.

Queen Elizabeth Hospital Birmingham: W Tunnicliffe.

Selly Oak Hospital: W Tunnicliffe.

Royal Blackburn Hospital: A Krige.

Good Hope Hospital: J Hull.

Ipswich Hospital: R Howard-Griffin.

Royal Berkshire Hospital: C Danbury.

Hereford County Hospital: R Harding.

Broomfield Hospital: E Makings.

Watford General Hospital: S Afolami.

University Hospital of Wales: S Shah.

Royal Preston Hospital: S Laha.

Royal Victoria Hospital: D McAuley.

Ulster Hospital: J Trinder.

Papworth Hospital: A Vuylsteke.

University Hospital of Coventry and Warwickshire: D Watson.

Dumfries and Galloway Royal Infirmary: P Jefferson.

Whiston Hospital: J Wood.

Arrowe Park Hospital: H Black.

University Hospital Aintree: G Dempsey.

Solihull Hospital: F Gao.

Cheltenham General Hospital: R Orme.

Sandwell General Hospital: A Arora.

Hammersmith Hospital: S Brett.

Charing Cross Hospital: S Brett.

Russells Hall Hospital: J Sonksen.

Barnet General Hospital: R Schoub.

Bedford General Hospital: S Snape.

Southend University Hospital: D Higgins.

Norfolk and Norwich Hospital: J Nortje.

James Paget Hospital: K Blenk.

Nevill Hall Hospital: M Martin.

Worcester Royal Hospital: S Graystone.

Yeovil District Hospital: R Daum.

Glan Clwyd Hospital: R Pugh.

Hull Royal Infirmary: I Smith.

Bristol Royal Infirmary: J Bewley.

Addenbrookes Hospital: A Johnston.

Edinburgh Royal Infirmary: M Beatty.

Great Western Hospital: M Watters.

Southampton General Hospital: M Grocott.

North Middlesex Hospital: JM Cuesta.

Royal Derby Hospital: D Rogerson.

James Cook University Hospital: J Park.

Frenchay Hospital: J Soar.

George Eliot Hospital: M Ranganathan.

Frimley Park Hospital: S Tote.

Warwick Hospital: T Long.

Wythenshawe Hospital: P Alexander.

Walsall Manor Hospital: M Khalil.

Warrington Hospital: J Little.

Royal Hallamshire Hospital: G Mills.

Macclesfield Hospital: J Hunter.

Northern General Hospital: G Mills.

Salford Royal Hospital: M Ghrew.

Kings Mill Hospital: L Milligan.

Antrim Area Hospital: R Ballie.

Royal United Hospital Bath: A Padkin.

Royal Victoria Infirmary: S Wright.

Freeman Hospital: S Wright.

St Johns Hospital: M Beatty.

Newcastle General Hospital: S Wright.

Chase Farm Hospital: R Schoub.

Scunthorpe Hospital: R Sharawi.

Southmead Hospital: J Soar.

Royal Lancaster Infirmary: D Highley.

Hairmyres Hospital: D Allen.

Aberdeen Royal Infirmary: D Noble.

Monklands Hospital: J Ruddy.

West Middlesex Hospital: M Popescu.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, the MRC, NETSCC, the HTA programme, the EME programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme, the EME programme or the Department of Health.

BALTI-2 PROTOCOL 15 JULY 2008

A MULTICENTRE RANDOMISED, DOUBLE-BLIND, PLACEBO-CONTROLLED TRIAL TO EVALUATE THE EFFECT OF INTRAVENOUS INFUSIONS OF SALBUTAMOL VERSUS PLACEBO ON 28-DAY MORTALITY IN PATIENTS WITH ACUTE RESPIRATORY DISTRESS SYNDROME

1. Background

1.2. Rationale for Beta Agonists in ARDS

1.2.1 A High Burden of Disease and Lack of Effective Therapies

ARDS is common, 13.3% of patients who require mechanical ventilation have ARDS, which is up to 40 times as high as previous studies have indicated. 15 ARDS is frequently fatal; in-Intensive Care Unit (ICU) mortality is estimated at 41–46%, corresponding to about 2,200 deaths per year in the UK. 1 ARDS is costly in health economics terms: these patients consume significantly more resources than matched patients without ARDS since they require a longer ICU and hospital stay (median 17 vs 8 days and 31 vs 25 days, respectively),17 and convalescence on the ward and subsequent rehabilitation in the community. 18 The quality of life after ARDS is significantly reduced with 35% unable to return to work 24 months after hospital discharge. 19;20 ARDS has no primary treatments proven to improve outcome other than supportive care with a lung-protective ventilator strategy. 21

1.2.2 Basic Science Data Support a Clinical Trial of a β₂ Agonist in ARDS

Experimental studies in animals, as well as in the ex-vivo human lung, have demonstrated that β adrenergic agonists accelerate the rate of alveolar fluid clearance predominantly through stimulation of the β₂ receptor on alveolar type I and II cells. 14 β₂ receptor activation increases intracellular cAMP resulting in increased sodium transport across alveolar cells by up-regulation of the apical sodium and chloride pathways, Na⁺/K⁺ ATPase and probably cystic fibrosis transmembrane conductance regulator. 1 This leads to the development of a micro-osmotic gradient between the alveolar space and interstitium which drives the movement of water and accelerates alveolar fluid reabsorption.

β₂ agonists have been shown to reduce neutrophil sequestration, activation and inflammatory cytokine production in-vitro and in animal models of ARDS. 22 In humans, inhaled salmeterol (long acting β₂ agonist) given prior to lipopolysaccharide inhalation reduces neutrophil influx, degranulation and tumour necrosis factor-α release. 23

In ARDS, β₂ agonists reduce endothelial permeability in animal models and humans24;25 and afford a degree of epithelial cytoprotection from infection related epithelial cell injury. 2 In BALTI 1, we found in-vivo evidence of reduced alveolar capillary permeability26 (fig 1) and in-vitro evidence of enhanced epithelial monolayer wound repair in patients treated with salbutamol27 (fig 2). This effect is protein kinase A dependent and occurs predominantly due to cell migration/spreading. These data suggest that, in addition to enhancing alveolar fluid clearance, β₂ agonists may maintain alveolar-capillary integrity, thereby reducing alveolar flooding and the development of ARDS or promote alveolar capillary repair in those with established ARDS.

FIGURE 1.

FIGURE 2.

1.2.3 Lack of Published Randomised Controlled Trials of β₂-Agonists in Patients with ARDS

In 2004, we conducted an electronic search of the on-line bibliographic databases Medline, PubMed, Current Contents, Clinical Evidence, the Cochrane Library, EBM and bmj.com for all publications in English using key words “acute lung injury” (or “ALI”), “ARDS”, “alveolar epithelium”, “β₂-agonists” and “pharmacotherapy” in the title, abstract, or Medical Subject Headings. A ‘hand search’ of the full reference lists from review articles and individual relevant papers in peer reviewed English language respiratory and critical care journals was also performed in order to cross check the quality of the computer retrieval method. The results were published as a review. 22 Three clinical studies (one randomised controlled cross over trial28 and two non-randomised studies29;30) using β₂-agonists in patients with ARDS were identified. These studies examined the effects of nebulised28;29 or intravenous (IV)30 β₂-agonists on the respiratory mechanics of artificially-ventilated patients with ARDS and found that β₂-agonists reduced airway resistance, peak and plateau airway pressures. There were no clinical studies addressing the effects of β₂-agonists on alveolar fluid clearance, or on outcome.

Recently, we conducted a further literature search (unpublished), using the same keywords combined with terms to identify randomised controlled trials, to identify any recent studies of the treatment of human ALI or ARDS with β₂-agonists. No studies using IV Salbutamol infusion were identified. A retrospective case review of 86 patients with ALI was the only relevant publication. 31 The patients with ALI who also received high dose nebulised Salbutamol (2.5–6.4 mg day^–1) had a significantly more days alive and free of ALI (n = 22, 12.2 (4.4) days) compared with the group receiving ≤ 2.4 mg day^–1 (n = 64, 7.6 (1.9) days) although both groups had similar hospital mortality rates of 48% vs 50%.

1.2.4 A Pilot Study of the β₂ Agonist Salbutamol in ARDS Confirms Laboratory Findings

After an initial dose-ranging study to determine the maximum infusion rate for salbutamol that did not cause tachydysrhythmias in patients with ARDS we undertook a single centre randomised, double blind, placebo-controlled phase II study (BALTI 1) in 40 adult patients with ARDS to determine if an intravenous infusion of salbutamol 15 µg kg ideal body weight^–1 hr^–1 for 7 days would accelerate clearance of alveolar oedema. As shown in the figure below, salbutamol significantly reduced lung water (left) (day 7: mean (SD), 9.2 (6) (•) vs 13.2 (3) (▴) ml kg^–1, P = 0.038) and plateau airway pressures (right) compared with placebo, and trend towards reduced lung injury score. 32

Patients with ARDS who have impaired alveolar fluid clearance have a higher hospital mortality than those with normal clearance (fig 3). 33

This association suggests that the improved clearance of extravascular lung water seen in the salbutamol-treated patients in the BALTI 1 study may lead to a survival benefit. We could not demonstrate this as the study was powered to detect a reduction in extravascular lung water. Therefore, a large-scale definitive trial with a survival endpoint is required.

1.2.5 The Research Proposed is Supported by the Worldwide Critical Care Community

The American-European Consensus Conference on ARDS in 199834 first supported the hypothesis that β₂ agonists could accelerate alveolar fluid clearance in ARDS and called for a clinical trial to investigate if β₂ agonists would alter outcomes in ARDS. The National Heart, Lung and Blood Institute Working Group considered the future research directions in ARDS in 2002 and concluded that clinical trials to investigate strategies targeting alveolar fluid clearance were required. 35 More recently, Professors Matthay and Abraham, two leading American critical care physicians, endorsed the need for a clinical trial with β₂ agonists in ARDS in their editorial which accompanied the BALTI 1 publication. 36

The BALTI 1 trial was funded and heavily supported by the West Midlands Intensive Care Society. At a national critical care research strategy meeting in November 2005, held by the UK Intensive Care Society (ICS), to assess the feasibility of undertaking ICU based multi-centre randomised clinical trials, the BALTI 2 trial was most highly ranked by over 50 active ICU researchers and the expert panel. The ICS funded a feasibility and pilot study on BALTI 2 for one year which has allowed the piloting and refinement of the trial protocol.

1.2.6 Reliable Drug Delivery

The optimal route for delivering β₂ agonists in patients with ARDS with a goal of increasing alveolar fluid clearance has not been determined. Nebulising drugs into the breathing circuits of mechanically ventilated patients appears attractive as it results in high lung concentrations but low blood concentrations and so may reduce the incidence of systemic side effects compared with parenteral treatment. 28 However, nebulised drugs might not reach the alveolar space in the consolidated and poorly ventilated lungs found in patients with ARDS.

Prior to the BALTI 1 study we conducted a dose-ranging study to determine the maximum tolerable dose of intravenous salbutamol that critically-ill patients could receive without an increase in ventricular or atrial ectopy. The maximal tolerable dose was 15 µg kg ideal body weight^–1 hour^–1 which is the maximal recommended dose for the treatment of airflow obstruction in acutely ill patients. This dose was used in the BALTI 1 study. This dose achieved plasma levels of salbutamol (10^–6M) which are associated with 100% increase in basal alveolar fluid clearance in animal models of ARDS. 37

1.2.7 Acceptable Tolerability and Side Effects

The administration of β₂-agonists can lead to important cardiovascular, metabolic and renal complications. Stimulation of cardiac and vascular β₁ and β₂ receptors causes tachycardia, arrhythmias, exacerbation of myocardial ischaemia, pulmonary vasodilatation and loss of hypoxic pulmonary vasoconstriction. 38;39 Metabolic sequelae include hypokalamaemia, hyperinsulinaemia and hyperglycaemia. 40 The use of intravenous β₂ agonists for tocolysis during pregnancy has been associated with the development of maternal pulmonary oedema. 41;42 Studies investigating this phenomenon in vivo in rabbits and humans found that intravenous injection of β₂-agonists caused reduced sodium, potassium and water excretion leading to a reduced haematocrit and intravascular hypervolaemia. 43;44 These adverse effects are usually more marked following intravenous rather than nebulised administration. However, in general, these drugs are well tolerated in the critically ill. These potentially deleterious effects may limit the potential beneficial effects of β₂-agonists described in this review.

In BALTI 1 treatment was generally well tolerated. 7 There was a trend towards higher heart rates in the salbutamol group at day 4 (means (SD), 103(22) vs 88(16) beats min^–1, salbutamol vs placebo, p = 0.06) and day 7 (94(14) vs 86(22), p = 0.264). 19 patients received intravenous salbutamol for a total of 2148 hours. During salbutamol or placebo infusions, seven patients (n = 5 – salbutamol, n = 2 – placebo, p = 0.164) developed new onset of supraventricular tachycardia. These arrhythmias did not cause significant hemodynamic compromise and were short lived. No patients sustained serious ventricular arrhythmias. There were no substantial differences in electrolyte concentrations or acid-base balance between salbutamol and placebo for K⁺, Mg⁺⁺ H⁺ or glucose as shown in Figure 4:

FIGURE 4.

Electrolyte, pH and glucose levels in salbutamol and placebo arms.

1.2.8 Lactic acidosis

Lactic acidosis is reported in the literature as a recognised side effect of intravenous and nebulised β₂ agonists. 45 This effect is probably mediated by β₂-adrenoreceptors and is hypothesised as being due to an increase in skeletal muscle glycogenolysis leading to a rise in peripheral lactate production. Splanchnic glucose production and lactate extraction are also increased, probably secondary to increases in hepatic glycogenolysis and gluconeogenesis. Acidosis does not develop until the bicarbonate buffering system is saturated, and this usually does not occur until lactate concentrations exceed 5 mmol L^–1. 46

There was no significant difference in lactate levels between placebo and treatment arms in the BALTI-1 study and no patients required discontinuation of the trial drug due to lactic acidosis. Two patients (out of 53) recruited to the BALTI-2 pilot study developed a lactic acidosis which the treating clinicians attributed the trial drug and discontinued the infusion. In both cases, the lactic acidosis resolved spontaneously after discontinuation of the trial drug over the next 6 hours.

1.2.9 The Intervention is Simple and Cheap

Salbutamol is a low cost treatment, and is readily available from generic drug manufacturers. A seven day infusion for a 70 kg patient will cost just £98. By comparison the NHS Reference cost for a day ICU care (2004) is £1,328.

2. Trial Design

2.4 Flow Diagram of Trial Design

2.9 Trial Treatments

2.9.1 Test Treatment

Active ingredient: Salbutamol Sulphate.

Trade name: Ventolin^TM Solution for Intravenous Infusion.

Concentration: 1 mg ml^–1.

Excipient: Sodium chloride, sodium hydroxide and Water for Injection.

Container: Clear glass ampoules, 5 ml.

Pharmaceutical Form: Sterile injection.

Manufacturer: GlaxoSmithKline Manufacturing S.p.A.

2.9.2 Control (Placebo) Treatment

Name: Sodium chloride Injection BP 0.9% w/v.

Concentration: 9 mg ml^–1.

Container: Clear glass ampoules, 5 ml.

Pharmaceutical Form: Sterile injection.

Manufacturer: Hameln Pharmaceuticals Ltd.

2.9.3 Diluent

Name: Sodium chloride Injection BP 0.9% w/v.

Concentration: 9 mg ml^–1.

Pharmaceutical Form: Sterile injection.

2.9.4 Drug Pack Preparation and Supply

Patient drug packs will be prepared by Bilcare GCS (Europe) Limited (Elvicta Business Park, Crickhowell, Powys, UK). Salbutamol and sodium chloride ampoules will be supplied to Bilcare. Each ampoule will have a randomised black out label applied and 50 ampoules of either salbutamol or placebo will be packaged in a white cardboard box in ten trays containing five ampoules each. Boxes will be sealed and labelled. Each box will contain sufficient material for the treatment of one patient for seven days. All trial drugs will be packaged identically and identified only by number.

Study: BALTI 2 Drug pack: VV

EUDRACT No. 2006-002647-86

Salbutamol 5 mg per 5 ml / Placebo

FOR IV INFUSION ONLY

FOR CLINICAL TRIAL USE ONLY

Batch: V Expiry: V

Store between 15°C – 25°C

Chief Investigator: Prof Fang Gao

Co-ordinating Centre: Warwick Medical School

Clinical Trials Unit, University of Warwick, Coventry

CV4 7AL. Tel: 02476 575848

AMPBAL

Drug boxes will be stored by Bilcare and dispatched by them to participating Hospital pharmacies.

2.9.5 Dispensing of Drug Packs

Hospital pharmacies will dispense the trial drugs to their ICU. Because patients may be recruited into the trial outside normal pharmacy opening hours, two or more patient drug packs (at least one each of salbutamol and placebo) need to be available on each hospital ICU at all times. When a patient is recruited, the randomisation service will inform the recruiting clinician of the drug pack number to be allocated to the current patient and the number of the next drug pack to be obtained from pharmacy. A retrospective prescription will be completed by the recruiting physician when the drug pack has been allocated to a patient along with a request form for the next pack required.

2.9.6 Calculation of Infusion Rate

Salbutamol and placebo infusions will be administered through a dedicated intravenous line at a rate of 0.075 ml (kg ideal body weight)^–1 hour^–1 (equivalent to 15 µg salbutamol (kg ideal body weight)^–1 hour^–1). Ideal body weight will be calculated from the patient's height. The patient will be measured in centimetres from heel to vertex using a soft tape measure and the ideal body weight and infusion rate obtained from the conversion table on next page.

Height (cm)	Male IBW (kg)	Infusion Rate (ml hr–1)	Female IBW (kg)	Infusion Rate (ml hr–1)	Height (cm)	Male (IBW) (kg)	Infusion Rate (ml hr–1)	Female IBW (kg)	Infusion Rate (ml hr–1)
146	44.2	3.3	39.7	3.0	174	69.7	5.2	65.2	4.9
148	46.0	3.5	41.5	3.1	176	71.5	5.4	67.0	5.0
150	47.8	3.6	43.3	3.2	178	73.3	5.5	68.8	5.2
152	49.6	3.7	45.1	3.4	180	75.1	5.6	70.6	5.3
154	51.5	3.9	47.0	3.5	182	76.9	5.8	72.4	5.4
156	53.3	4.0	48.8	3.7	184	78.8	5.9	74.3	5.6
158	55.1	4.1	50.6	3.8	186	80.6	6.0	76.1	5.7
160	56.9	4.3	52.4	3.9	188	82.4	6.2	77.9	5.8
162	58.7	4.4	54.2	4.1	190	84.2	6.3	79.7	6.0
164	60.6	4.5	56.1	4.2	192	86.0	6.5	81.5	6.1
166	62.4	4.7	57.9	4.3	194	87.9	6.6	83.4	6.3
168	64.2	4.8	59.7	4.5	196	89.7	6.7	85.2	6.4
170	66.0	5.0	61.5	4.6	198	91.5	6.9	87.0	6.5
172	67.8	5.1	63.3	4.7	200	93.3	7.0	88.8	6.7

2.9.7 Treatment Preparation and Administration

Prior to infusion, two ampoules of trial drug will be diluted with 40 ml of saline in a 50 ml syringe. Infusion syringes should be made up immediately prior to use.

Trial drug infusions should be started immediately after randomisation. If at the time of attempting to commence the trial drug the patient's heart rate exceeds 140 beats min^–1, the administration should be delayed until the heart rate is less than 140 beats min^–1 for at least 30 minutes. Every attempt should be made to complete the treatment infusion without interruption for a maximum of seven days (i.e. until 168 hours after randomisation).

2.9.8 Alteration of Infusion Rate

Sinus tachycardia or arrhythmias are known side effects of intravenous salbutamol administration. If a patient receiving a trial drug infusion is noted to have tachycardia (HR > 140 beats min^–1) or any new arrhythmia occurs, the dose rate of drug will be adjusted according to the flow diagram (page 23). Dose adjustments for renal or hepatic failure will be driven by the cardiovascular response to the infusion rather than on the degree of renal or hepatic impairment. Standard anti-arrhythmic therapy will be given if indicated in addition to alteration of infusion rate.

Example 1 – Patient develops sinus tachycardia rate 145 beats min^–1, sustained for greater than 30 minutes without haemodynamic compromise. Infusion running at rate A. Action – sedation and analgesia requirements reviewed; other causes of tachycardia sought but not identified. Infusion reduced to rate B. One hour later HR 118 beats min^–1. Infusion continued at rate B.

Example 2 – Patient with pre-existing atrial fibrillation develops tachycardia (AF) rate 160 beats min^–1. Infusion running at rate A. Action – sedation and analgesia requirements reviewed; other causes of tachycardia sought but not identified. Infusion reduced to rate B. One hour later HR 148 beats min^–1 (AF). Infusion stopped. Clinical decision taken to treat AF with anti-arrhythmic according to unit protocol. Patient reassessed 12 hours later. HR 80 beats min^–1 (AF). Infusion restarted at rate C. If further tachycardia or new arrhythmia infusion should be stopped and remaining drug returned to pharmacy.

Example 3 – Patient develops new onset atrial fibrillation. Infusion running at rate A. Action – Stop infusion. Check electrolytes, record 12 lead ECG. Clinical review and decision taken to treat AF with anti-arrhythmic according to unit protocol. Patient reassessed 12 hours later. HR 80 beats min^–1 (sinus rhythm) Infusion restarted at rate B. If further tachycardia or new arrhythmia infusion should be stopped and remaining drug returned to pharmacy.

2.9.9 Infusion Termination Criteria

Termination of the infusion is defined as discontinuation of the trial drug infusion without intention to restart the infusion at a later time. Patients whose infusion is terminated before 7 days after randomisation are not withdrawn from the trial, but will remain in the trial until twelve months after randomisation or death. Trial drug infusion will be terminated in the following circumstances:

• Death.

• Heart rate > 140 beats min^–1 despite two adjustments in infusion rate.

• New arrhythmias despite adjustment in infusion rate.

• Development of a significant lactic acidosis, which in the opinion of the treating clinician is attributable to infusion of the trial drug.

• 24 hours after discontinuation of mechanical ventilation (of any sort).

• Discharge from Critical Care environment.

• Discontinuation of active treatment.

• Request to withdraw from PerLR or patient.

• Decision by the attending clinician that the infusion should be discontinued on safety grounds.

• 7 days (168 hours) after randomisation.

2.9.10 Treatment Compliance

Treatment will be administered by site personnel with relevant training and experience at the hospital. Trial infusions will be recorded in the Case Report Forms to monitor treatment compliance.

2.9.11 Drug Accountability

Hospital pharmacies will be responsible for recording trial drug packs dispensed to the ICU.

Preparation of all drug infusions will be recorded on the Nursing Staff Drug Accountability Form and drug administration on the patients prescription chart. The trial drug packs will include a sheet on which the fate of all ampoules will be recorded (infused, opened but not infused, discarded, unused). At the end of the treatment period any remaining unused drug will be returned to the hospital pharmacy for recording and will then be destroyed.

3. Data Management

4. Data Analysis

5. Trial Organisation

Reference List