Note from the National Guideline Clearinghouse (NGC): The National Institute for Health and Clinical Excellence (NICE) commissioned an independent academic centre to perform a systematic literature review on the technology considered in this appraisal and prepare an assessment report. The assessment report for this technology appraisal was prepared by the Peninsula Technology Assessment Group (PenTAG), Peninsula Medical School and Southampton Health Technology Assessments Centre (SHTAC), University of Southampton. (See the "Availability of Companion Documents" field.)
Clinical Effectiveness
Data Extraction Strategy
All trials, except those included in the relevant Cochrane reviews, were fully data extracted. Data were entered into a structured template by one reviewer and checked by a second. Any discrepancies between the data extracted and the original trial report were resolved and the data extraction was finalised.
Critical Appraisal Strategy
The methodological quality of the trials supplemental to the Cochrane reviews was assessed according to criteria specified by the Centre for Reviews and Dissemination (CRD). (Refer to Appendix 4 of the Assessment Report [see the "Availability of Companion Documents" field]). Quality was assessed by one reviewer and their judgements were checked by a second. Where there was disagreement a third reviewer was consulted and a final judgement agreed.
Methods of Data Synthesis
Results of the included trials were synthesised narratively with use of meta-analyses where possible and where appropriate. A framework was devised for the analysis and presentation of results, based on the step wise approach recommended in the British Thoracic Society/Scottish Intercollegiate Guidelines Network (BTS/SIGN) Guidelines for the management of asthma.
The review questions were:
- Which inhaled corticosteroid (ICS) is the most-effective at low doses (200–800 micrograms per day beclometasone dipropionate/budesonide [BDP/BUD] equivalent) (Step 2 of the guidelines)
- Which ICS is the most-effective at high doses (800–2000 micrograms per day BDP/BUD equivalent) (Step 4 of the guidelines)
- Which is the more clinically effective approach to introducing a long-acting beta-2 agonist (LABA) into a treatment regimen:
- To increase the dose of ICS alone or to add a LABA to treatment with ICS using a combination inhaler? (Steps 2-3 of the guidelines)
- To continue with an ICS alone or to add a LABA to treatment with a similar dose of ICS using a combination inhaler? (Steps 2-3 of the guidelines)
- Which is the more clinically effective treatment: fluticasone propionate (FP) and salmeterol (SAL) in a combination inhaler or given in separate inhalers? BUD and formoterol fumarate (FF) in a combination inhaler or given in separate inhalers?
- Which is the most-effective—a combination inhaler containing BUD/FF, or a combination inhaler containing FP/SAL? (Step 3 of the guidelines)
Each included trial was coded according to which of the review questions it was relevant to. Some trials were relevant to more than one review question as they tested multiple doses of inhaled steroids.
Each review question was stratified according to a number of pair-wise comparisons of the inhaled steroids and, where relevant, LABAs (where evidence allows). In addition, some trials were included in more than one pair-wise comparison as they evaluated two or more ICS (e.g., a three arm trial comparing FP with BUD and BDP).
Trials were also divided according to whether or not a parallel-group or cross-over design was used. It is generally considered inappropriate to pool these designs together within a meta-analyses. Where necessary trials were then further divided according to the nominal dose ratio employed, following the approach used in the Cochrane review of FP compared to BUD or BDP.
In summary, the framework comprised sets of trials grouped according to which review question, pair-wise comparison, study design, and dose ratio they related to.
Narrative Synthesis
Within each pair-wise comparison all included trials were tabulated for their key characteristics and described in the text (e.g., trial duration, patient profile, outcome measures, methodological quality). In addition, more detailed data on the trials are available in Appendix 4 of the Assessment Report (see the "Availability of Companion Documents" field), for those trials which were supplemental to the Cochrane reviews (and which underwent full data extraction).
Meta-Analysis
The feasibility and appropriateness of meta-analysis was considered once narrative syntheses had been completed. The decision to pool was mediated by the likelihood that the trials were clinically homogenous, and that the necessary data were available. Potential clinical heterogeneity was assumed if there were differences between trials in:
- Dose
- Disease severity
- Treatment duration
If pooling was considered appropriate the data in each trial were examined to ascertain whether or not sufficient details were reported to facilitate meta-analysis. The Cochrane
Airways Group kindly supplied their Review Manager software files containing extracted and analysed data. These files were edited to correspond to the Assessment Group's review questions and framework (i.e., they were assembled into smaller sets of studies based on dose, design, and pair-wise comparisons). Data from trials included in the Cochrane reviews which did not meet the inclusion criteria for this review were removed. Data from trials supplemental to the Cochrane reviews were added, based on the data extracted to our standardised template.
For continuous outcome measures (e.g., lung function, symptoms) mean values and standard deviations were required in order to calculate mean differences. These were entered where available from the trial reports. Where standard deviations were not reported they were converted from standard errors, p values, or confidence intervals provided in the trial reports (where available), using standard formulae within a spreadsheet. Authors were not contacted to supply missing data.
Cross-over trials were only pooled where data were reported to facilitate appropriate analysis. Many cross-over trials report results as if the trial used a parallel-group design and pooling is not advised, as this results in a unit of analysis error. In such cases cross-over trials were described narratively, with appropriate caveats.
Pooled data were expressed separately in terms of change from baseline to end-point, and as end-point values. Trials were pooled within a meta-analysis as either one of these, but not both. The Assessment Group chose not to impute change values where not reported by authors as it requires estimations of the variance around mean differences, which involves assumptions about within-patient differences. Data were not available to allow within-patient differences to be estimated (e.g., from an appropriate correlation co-efficient).
Much of the data were continuous and where it was apparent that the same measurement scale had been used across studies a weighted mean difference (WMD) was used to summarise treatment effects. If it appeared that different measurement scales were employed a standardised mean difference (SMD) was used. Dichotomous data (e.g., rate of adverse events) were pooled using odds ratios. 95% confidence intervals were used for all measures of effect. A fixed-effects model was used, with random-effects model used if statistical heterogeneity was apparent. Statistical heterogeneity was measured using a chi-squared test with p<0.10 as the level of significance. The I2 statistic was also used, whereby a value in excess of 50% indicates substantial heterogeneity.
Refer to sections 5.1.3 to 5.1.5 of the Assessment Report (see the "Availability of Companion Documents" field) for more information.
Cost-Effectiveness
Refer to Section 6.5 in the Assessment Report (see the "Availability of Companion Documents" field) for information about the methods used in the original economic analysis.
