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 Centre for Reviews and Dissemination/Centre for Health Economics (CRD/CHE) Technology Assessment Group, University of York. (See the "Availability of Companion Documents" field).
Clinical Effectiveness
Data Extraction
The authors of a recent systematic review provided the extracted data from their review to avoid duplication of work. This also included some unpublished data. These data had been independently extracted by two reviewers. Data from the new studies, as well as any additional data required from the studies previously extracted were extracted by one reviewer and checked by another. Discrepancies were resolved by discussion and, if necessary, a third reviewer was consulted. Where there were multiple publications from the same study, the main publication for each study was identified and data were extracted from that paper. Where additional relevant outcomes were available in a related paper these were also extracted.
Data were extracted into Revman and into a standard form in Word. Data extracted included patient characteristics (age, sex, severity of obstructive sleep apnoea/hypopnoea syndrome [OSAHS], body mass index), details of the intervention (fixed or autotitrating continuous positive airway pressure [CPAP], use of humidifier), comparator (details of placebo, conservative management or dental device), adherence (usually reported as the average number of hours the machine was running at night), length of follow-up, outcomes and study quality.
Predominantly endpoint data were reported in the trials, except for blood pressure where a mixture of change and endpoint data were reported. Where both endpoint and change data were reported, preference was given to endpoint data for all outcomes except blood pressure where change data were used (provided the variance for the change score was reported). Where only change data were reported, the variance was imputed if necessary.
Paired data were extracted from crossover trials where available. If the standard deviation (SD) or standard error (SE) from a paired analysis was not reported, the standard error was imputed from the t-statistic, the p-value or the confidence interval from a paired analysis. For one crossover study it was necessary to impute the standard error for blood pressure: a within-person correlation of 0.5 was used and a within-person correlation of 0.1 and 0.9 for a sensitivity analysis.
Due to time limitations and the quantity of cognitive data from crossover trials it was not feasible to impute data for a paired analysis, where these were not reported, for all the cognitive outcomes. Where three or more studies were available for potential pooling, the SE was estimated where data were available as above. For the other cognitive outcome measures the mean end value at follow-up and the SD for the intervention and control group with the associated p-value were extracted. Where available the SD or SE from a paired analysis were extracted.
Quality Assessment
Study quality was assessed based on criteria from CRD Report No 4 and additional criteria were used to assess crossover trials (see Section 5.2.1.2 of the Assessment Report [see the "Availability of Companion Documents" field]). The criteria assessed were broad in anticipation that a narrative synthesis may have been necessary. Quality was assessed by one reviewer and checked by another. Discrepancies were resolved by discussion and, if necessary, a third reviewer was consulted.
Data Analysis
Where sufficient data were available, they were pooled in quantitative syntheses using a random effects model. Studies comparing CPAP to placebo or best supportive/usual care were pooled separately from studies comparing CPAP to dental devices. Where data sets included both study designs, parallel and crossover trials were pooled together. The generic inverse variance method in Revman was used to pool data sets which included both parallel and crossover designs, or only crossover trials. When only parallel trials were being pooled the weighted mean difference method in Revman was used. To transform the parallel data for entry into the generic inverse variance facility the standard error for the mean difference was calculated from the 95% confidence interval (CI). This was calculated using the formula SE = (upper CI- lower CI)/3.92. This method assumes a sample size of at least 30, however, given the number of outcomes and studies included in the review it was not considered feasible in the time available to use the t-statistic.
Statistical heterogeneity between trials was assessed using the I2 statistic. Five sources of potential clinical and methodological heterogeneity were identified a priori as being of priority: baseline disease severity, baseline daytime sleepiness, study design, type of placebo, and study quality. The Assessment Group planned to investigate these for the primary outcomes using sub-group analysis, since clinically important variations in the magnitude of treatment effects are likely in different severity groups. The sub-groups specified in advance were as follows.
- Population sub-groups:
- Baseline disease severity, as classified using the apnoea/hypopnoea index (AHI) or the desaturation rate using the mean baseline score for each study: mild (AHI 5-14/hr or oxygen desaturation rate 5-10/hr), moderate (AHI 15-30/hr or oxygen desaturation rate 10-30/hr) and severe (AHI >30/h or oxygen desaturation rate >30/hr)
- Baseline symptom severity, as classified using the mean baseline Epworth Sleepiness Scale (ESS) score for each study: mild (0 to 9 points), moderate (10 to 15 points) and severe (16 to 24 points).
- Comparator sub-groups:
- Sham CPAP, oral placebo, and best supportive care.
- Study design sub-groups:
- Parallel and crossover.
- Endpoint data and change from baseline data.
The Assessment Group planned to investigate the influence of study quality on the treatment effect by pooling studies with adequate concealment of allocation separately from those with inadequate or unclear adequacy of concealment. This analysis was limited due to the small number of studies that reported an adequate method of concealing treatment allocation.
The pooling of the primary outcomes and blood pressure were rerun using a fixed effect model to test the impact of the model of analysis used. The robustness of the findings for these outcomes was also investigated by assessing the impact on the treatment effect of removing each study singly.
Refer to Sections 5.1.3 -- 5.1.5 and 5.2 of the Assessment Group Report (see the "Availability of Companion Documents" field) for more information.
Cost-Effectiveness
Cost-Effectiveness Review Methods
Data were extracted using a data extraction form that was developed for use in previous Technology Assessment Reviews. The quality of the cost-effectiveness studies was assessed based on a checklist developed by Drummond et al (2005) and which reflects the criteria for economic evaluation detailed in the methodological guidance developed by NICE http://www.nice.org.uk/. (Refer to Appendix 11.6 of the Assessment Report [see the "Availability of Companion Documents" field] for economic evaluation data extraction table and Table 6.27 of the Assessment Report for economic evaluation quality assessment table).
In an attempt to make full use of all of the available evidence on therapies for the treatment of OSAHS and in order to overcome some of the limitations of the published and manufacturer's cost-effectiveness studies, a new cost-effectiveness model was developed by the Assessment Group.
See Section 6.2 of the Assessment Report (see the "Availability of Companion Documents" field) for information about the York economic model.