Edinburgh Research Explorer Continuous positive airway pressure in older people with obstructive sleep apnoea syndrome (PREDICT)

Summary Background The therapeutic and economic beneﬁ ts of continuous positive airway pressure (CPAP) for moderate to severe obstructive sleep apnoea (OSA) syndrome have been established in middle-aged people; however, the beneﬁ ts in older people are unknown. This trial was designed to address this evidence gap. Methods This 12-month, multicentre, randomised trial enrolled patients across 14 National Health Service sleep centres in the UK. Consecutive patients aged 65 years or older with newly diagnosed OSA syndrome were eligible to join the trial. Patients were randomly assigned (1:1) into parallel groups to receive either CPAP with best supportive care (BSC) or BSC alone for 12 months. Randomisation was done by the Medical Research Council Clinical Trials Unit with computer-generated randomisation. The main investigator at each centre was masked to the trial randomisation. Coprimary endpoints were Epworth sleepiness score (ESS) at 3 months and cost-eﬀ ectiveness over the 12-month trial period. Secondary outcomes were subjective sleepiness at 12 months, plus objective sleepiness, quality of life, mood, functionality, nocturia, mobility, accidents, cognitive function, and cardiovascular risk factors and events at 3 months and 12 months. The analysis was by intention to treat. This study is registered as an International Standard Randomised Controlled Trial, number ISRCTN90464927 . Findings Between Feb 24, 2010, and May 30, 2012, 278 patients were randomly assigned to the trial, of whom 231 (83%) completed the trial. 140 patients were allocated to and received CPAP plus BSC and 138 were allocated to and received BSC only. CPAP reduced ESS by 2·1 points (95% CI –3·0 to –1·3; p<0·0001) at 3 months for 124 (89%) of 140 patients compared with 124 (90%) of 138 patients given BSC, and by 2·0 points (–2·8 to –1·2; p<0·0001) at 12 months for 116 patients compared with 122 patients given BSC. The eﬀ ect was greater in patients with higher CPAP usage or higher baseline ESS. Quality-adjusted life-years were similar between the groups (treatment eﬀ ect 0·01 (95% CI –0·03 to 0·04; p=0·787) and health-care costs were marginally reduced with CPAP (–£35, –390 to 321; p=0·847). CPAP improved objective sleepiness (p=0·024), mobility (p=0·029), total cholesterol (p=0·048), and LDL cholesterol (p=0·042) at 3 months, but these were not sustained at 12 months. Measures of mood, functionality, nocturia, accidents, cognitive function, and cardiovascular events remained unchanged. Systolic blood pressure fell in the BSC group. 37 serious adverse events occurred in the CPAP group, and 22 in BSC group; all were independently classiﬁ ed as being unrelated to the trial and no signiﬁ cant harm was attributed to CPAP use . Interpretation In older people with OSA syndrome, CPAP reduces sleepiness and is marginally more cost eﬀ ective over 12 months than is BSC alone. On the basis of these results, we recommend that CPAP treatment should be oﬀ ered routinely to older patients with OSA syndrome.


Introduction
Obstructive sleep apnoea (OSA) is one of the commonest respiratory disorders aff ecting up to 20% of the general population. 1 When OSA leads to sleep disruption resulting in excessive daytime sleepiness it is known as OSA syndrome. OSA syndrome is thought to aff ect up to 4% of the middle-aged population (30-60 years); 1,2 however, with the current obesity epidemic, the proportion of aff ected patients might require upward revision 3 because obesity is the strongest risk factor for OSA. The long-term consequences of OSA syndrome include reduced social functioning and quality of life; neurocognitive impairment; increased risk of road traffi c accidents; cerebrovascular, cardiovascular, and metabolic morbidity; and higher mortality. 4 Continuous positive airway pressure (CPAP) therapy is established as an effi cacious and cost-eff ective treatment for middleaged patients with moderate to severe OSA syndrome. 5 In older people (older than 60 years), the prevalence of OSA syndrome is increased and the symptoms can be confl ated with the functional impairments of ageing. 6 Thus, reversing the OSA syndrome component of any functional impairment could increase independence and reduce health-care costs in ageing populations. However, the magnitude of the treatment and economic benefi ts of CPAP shown in middle-aged people cannot be extrapolated to older populations. The daytime sleepiness associated with OSA could be less debilitating in older people who have more fl exible schedules allowing for extra sleep opportunities. Conversely, both the prevalence of morbidities leading to polypharmacy and the detrimental eff ect on sleep of the illnesses themselves increase with age. 7 Furthermore, diff erent perceptions of quality of life, higher overall health-care use, and shorter life expectancy could modify the economic benefi ts of CPAP in this population. This trial was designed to assess the effi cacy of CPAP to reduce daytime sleepiness in older people with OSA syndrome and to determine its cost-eff ectiveness.

Study design and participants
The PREDICT trial was a multicentre, investigatormasked, randomised, parallel controlled trial done in 14 National Health Service (NHS) sleep centres in the UK.
Consecutive patients aged 65 years or older with newly diagnosed OSA syndrome were invited to join the trial. A diagnosis of OSA syndrome was based on the clinical practice in each centre; the criteria that needed to be fulfi lled for a diagnosis were 4% or greater oxygen desaturation index (ODI) with more than 7·5 events per h, and Epworth sleepiness score (ESS) of 9 or greater. ODI and ESS were assessed again at enrolment and the (standardised) enrolment measures were used in the primary analysis (as opposed to the measures made during clinical diagnosis). Exclusion criteria were previous exposure to CPAP, awake oxygen saturation (SpO 2 ) less than 90% on air, ratio of forced expiratory volume in 1 s to forced vital capacity of less than 60%, being a professional driver, reporting sleepiness while driving, shift work, or any severe symptom of OSA syndrome for which the referring physician felt CPAP was mandatory. The number of patients assessed for eligibility was documented in screening logs. This trial was approved by a central research ethics committee (REC 09/H0708/33) and all patients gave written informed consent. Further details of the trial management and any changes to the protocol are provided in the appendix pp 4, 5.

Randomisation and masking
Patients were randomly assigned (1:1) centrally by the Medical Research Council Clinical Trials Unit (MRC CTU) by computer-generated randomisation, with minimisation by ESS at enrolment (>13 vs ≤13), functionality (Townsend Disability Scale [TDS] >1 vs ≤1), and recruitment centre. The allocation group was revealed by telephone to the (unmasked) investigator initiating the intervention, once baseline data collection was complete.
Treatment allocation was concealed from the individual completing follow-up assessments. Patients were discouraged from discussing their treatment allocation with the masked research staff and the importance of maintaining blinding was emphasised in the patient information sheets. The case report forms were designed to collect blinded and unblinded data separately. Masking of all trial staff was not possible, although the assessments were done blind wherever possible. The trial manager and trial support staff at the coordinating centres in Oxford and London did not have contact with the patients. The trial statisticians analysed the results using an analysis plan that had been fi nalised before the database was locked and before the blinded data were analysed.

Procedures
All recruitment centres had expertise in the treatment of OSA syndrome and were provided with auto-titrating CPAP devices (S9 Autoset, ResMed [UK] Ltd, Oxfordshire, UK), humidifi ers, and a range of interfaces. CPAP treatment (auto-titrating with default minimum and maximum pressure settings at 4-20 cm H 2 O) was initiated using the standard practice in each centre, by appropriately qualifi ed staff not involved in trial outcomes. Humidifi cation and choice of interface were made on an individual patient basis. At every follow-up visit (3 months and 12 months), the hours of use of CPAP were downloaded from the CPAP machine.
BSC was comprised of advice on minimising daytime sleepiness through sleep hygiene, naps, caff eine, and weight loss, as appropriate to each patient. A standard information booklet was given to all patients. Both groups had identical visit schedules, and were asked to continue with their usual drugs and medical care during the trial. Further details of the visit schedules are outlined in the appendix p 4.
Structured assessments were done at baseline, 3 months, and 12 months. Additionally, all patients received a telephone call at 1 week, 1 month, and 6 months to record symptoms and side-eff ects and to optimise CPAP adherence. Patients also completed monthly diaries recording symptoms, side-eff ects, use of health care, change in medication, functionality, and quality-of-life questionnaires. All enrolled patients completed a domiciliary overnight respiratory polygraphy study (Embletta Gold, Embla, Amsterdam, Netherlands) before initiation of CPAP, the results of which were scored centrally. Domiciliary overnight pulse-oximetry (Konica-Minolta, Osaka, Japan) was done at 3 months and 12 months.

Outcomes
The coprimary outcomes were subjective sleepiness at 3 months measured by the ESS 8 (mean ESS of months 3 and 4) and cost-eff ectiveness of provision of See Online for appendix CPAP over the 12 months measured by quality-adjusted life-years (QALYs) calculated with the European Quality of Life-5 Dimensions (EQ-5D) and use of health-care resources, which was collected monthly from patient diaries. 9 Secondary outcomes were subjective sleepiness at 12 months (mean ESS of months 10, 11, and 12) 16 Trail Making Test B, 17 Digit Symbol Substitution Test, 18 simple and four-choice reaction time 19 ), cardiovascular risk factors (systolic and diastolic blood pressures, fasting blood profi le), and cardiovascular events at 3 months and 12 months. SF-36 was also used to derive short-form six dimensions (SF-6D). Treatment compliance was a tertiary outcome measured objectively by download of the CPAP smart card at every visit.

Statistical analysis
The study was designed to detect a minimally clinically important change of one point in the ESS, with 90% power at the two-sided 5% signifi cance level. Assuming a 10% loss to follow-up and a standard deviation for the change in ESS in each group of 2·4, 5 270 patients were required. Data were held in a central database at MRC CTU. All analyses were prespecifi ed in the analysis plans.
Analysis was by intention-to-treat with adjustment for treatment allocation, minimisation factors, and the corresponding baseline variable of the outcome. Multivariable linear regression models were used for continuous outcomes and logistic regression used for binary outcomes. Data obtained outside the prespecifi ed time period of 2-5 months (primary ESS outcome) and patients with missing data were excluded from the relevant analysis. Multiple imputation with chained equations was used to assess the eff ect of missing data on the primary outcomes, 20 initially under the missing at random assumption. This was followed with several sensitivity analyses in which missing outcomes were assumed to be better or worse on average than those noted in the trial. Exploratory analyses were done to investigate the eff ect of age, ODI, body-mass index, and baseline ESS on the primary treatment eff ect with fractional polynomials. 21 The eff ect of use of CPAP on the primary ESS outcome was also explored. Additionally, a sensitivity analysis of the primary effi cacy outcome was done excluding the patients who switched from BSC to CPAP before 3 months.
The cost-eff ectiveness of CPAP was analysed from the perspective of the UK NHS. Health outcomes were expressed as QALYs with EQ-5D (primary) and SF-6D (secondary). Costs were evaluated in UK pounds sterling (2012 price base) and included the costs associated with general practitioner and nurse visits, phone calls to the general practitioner and NHS Direct, ambulance use, visits to accident and emergency, outpatient appointments, hospital overnight admissions, emergency admissions, and total number of nights in hospital, as recorded in the monthly diaries, plus the cost of CPAP treatment. The unit costs applied to every item to calculate the total cost per patient are given in appendix p 11. The base-case analysis was done in the intention-to-treat groups after multiple imputation with chained equations of missing data and adjustment for imbalances in baseline EQ-5D (or SF-6D) and in the Figure 1: Trial profi le 1614 patients were assessed for eligibility, of these 541 (34%) were eligible. The number of patients included in the ESS and health economic analysis included data obtained from monthly diaries (when available) for those who did not attend the 3 month and/or 12 month visits, but remained in the trial. Information about recruitment is also given in the appendix p 3. BSC=best supportive care. CPAP=continuous positive airway pressure. ESS=Epworth sleepiness score. *The 1073 ineligible patients were grouped into the following categories: 442 (41%) did not meet inclusion oxygen desaturation index or ESS criteria; 79 (7%) had previous exposure to CPAP; 171 (16%) had awake oxygen saturations less than 90% on air or forced expiratory volume in 1 s/forced vital capacity ratio less than 60%; 216 (20%) for being a professional driver, reporting sleepiness whilst driving, shift work, or any severe symptom of OSA syndrome for which the referring physician felt CPAP was mandatory; and 165 (15%) for no information or incomplete data.
138 were allocated to and received BSC 12 were lost to follow-up 10 withdrew consent 1 was unable to be contacted 1 withdrew because of other medical condition 2 discontinued BSC requesting CPAP 124 attended 2 did not attend (but remained in the trial) 124 included in the ESS analysis 6 were lost to follow-up 2 withdrew consent 3 were unable to be contacted 1 died 6 discontinued BSC and started CPAP 2 requested CPAP 4 CPAP prescribed by physician 117 attended 3 did not attend 118 included in the health economic analysis 140 were allocated to and received CPAP plus BSC 15 were lost to follow-up 11 withdrew consent 2 were unable to be contacted 2 withdrew because of other medical condition 20 discontinued CPAP 121 attended 4 did not attend (but remained in the trial) 124 included in the ESS analysis 10 were lost to follow-up 4 withdrew consent 5 were unable to be contacted 1 died 21 discontinued CPAP 114 attended 1 did not attend 113 included in the health economic analysis health-care costs in the month before enrolment. 22 The probability of CPAP being cost eff ective under the conventional UK NHS thresholds was calculated with semi-parametric bootstrapping. 23 The sensitivity analysis tested fi ve scenarios: (1) the assumption that the CPAP device was replaced every 3 years, the masks replaced every 3 months, and the fi lters replaced monthly; (2) the assumption that the CPAP device was used for 1 year and discarded (no annuitisation of costs); (3) complete case analysis in which patients with missing data were removed from the analysis; (4) missing data imputed with mean interpolation between follow-up points; and (5) the assumption that individuals with missing data had 25% greater costs or had 25% lower health-related quality of life than what otherwise would be predicted from the multiple imputation model. All analyses were done with Stata version 12.0 (StataCorp LP, College Station, TX, USA).
The trial is registered with International Standard Randomised Controlled Trial number ISRCTN90464927.

Role of the funding source
The National Institute of Health Research Health Technology Assessment Programme funded this trial (08/56/02) and the views and opinions expressed therein are those of the authors and do not necessarily refl ect those of the Health Technology Assessment Programme, National Institute of Health Research, National Health Service, or the Department of Health. Neither the funder nor the company (RedMed UK) had any role in the trial design, data collection, data analysis, data interpretation, or writing of the report. The raw data were accessible by DJB, AJN, and the Independent Data Monitoring Committee only during the trial, and on completion to AM, RF, SG, JRS, RLR, and MJM. The corresponding author had full access to all of the data on completion of the trial and the fi nal responsibility to submit for publication.

Results
Between Feb 24, 2010, and May 30, 2012, 278 patients were randomised. 138 patients were allocated to and received BSC and 140 were allocated to and received CPAP plus BSC. Follow-up visits were done for 245 (88%) of 278 patients at 3 months and 231 (83%) of 278 patients at 12 months (fi gure 1). The main reason for loss to follow-up was withdrawal of consent; 21 (8%) patients at 3 months and 27 (10%) patients at 12 months. The table and appendix (pp 12, 13) show the baseline characteristics and overnight domiciliary respiratory polygraphy measurements of the enrolled patients. The mean age across both treatment groups was 71·1 (SD 4·6) years, ODI 28·7 (19·1) events per h, and ESS 11·6 (3·7). The baseline characteristics were broadly similar between the two groups, although by chance the BSC group seemed to have a slightly higher incidence of comorbidities than did the CPAP group.
At 3 months, ESS was signifi cantly reduced in patients receiving CPAP treatment (-3·8, SD 0·4) compared with those given BSC (-1·6, 0·3) with a treatment eff ect of -2·1 (fi gure 2). Adjustment for age, sex, body-mass index, and baseline ODI made no diff erence to the result, neither did the sensitivity analysis, which excluded two patients who swapped from BSC to CPAP before the 3-month visit. Imputation analyses showed the primary result was robust to missing data (appendix p 6). The treatment eff ect was signifi cantly greater in patients with higher baseline ESS (appendix p 7) or higher CPAP use (appendix p 14).
The average QALYs obtained with the EQ-5D were 0·68 (95% CI 0·64 to 0·72) for CPAP (n=113) and 0·67 (0·63 to 0·71) for BSC (n=118) with the adjusted increase in QALYs for CPAP being non-signifi cant at 0·01 (-0·03 to 0·04; p=0·787). The EQ-5D scores did not diff er signifi cantly between groups at each month (appendix p 8) and resource use per costs were similar across groups (appendix p 15).  The annual cost of CPAP treatment per patient was estimated at £201. This amount included the annual equivalent costs of the CPAP device at £70 and the humidifi er at £27 (the humidifi er was only given to 82 [59%] of the 140 patients), totalling £86 on average per patient, plus the cost of masks at £114 (£104, assuming that 10% of patients received two masks each) and the cost of two fi lters per patient per year at £1·13 (average cost of fi lters was £0·58). Overall the average cost per patient allocated to CPAP was £1363 (95% CI 1121-1606) and for BSC was £1389 (1116-1662). After adjustment for costs incurred to the month previous to enrolment, the CPAP group accrued on average -£35 (95% CI -390 to 321) health-care costs. The results were not sensitive to diff erent assumptions regarding missing data. However, when alternative assumptions were made for the frequency of replacing equipment, the cost per QALY increased (appendix p 16). Additionally, the cost-eff ectiveness of CPAP was more robust in patients with higher baseline ESS (appendix p 17). The probability that the CPAP was cost eff ective at the thresholds conventionally used in the NHS (£20 000 per QALY gained) was 0·61.
The improvement in the ESS on CPAP compared with BSC was maintained at 12 months, with a treatment eff ect of -2·0 (fi gure 2; appendix p 14).
Of 140 patients randomly assigned to CPAP, 120 (86%) at 3 months and 99 (71%) at 12 months self-reported they were still using CPAP. Usage data for CPAP were obtained for 117 patients at 3 months with a median usage of 1 h 52 min (IQR 19 min to 5 h 12 min) per night and for 102 patients at 12 months with a median usage of 2 h 22 min (10 min to 5 h 9 min) per night. Assuming zero usage in those patients who stopped treatment during follow-up or had missing data gave a more conservative estimate of median usage of CPAP of 1 h 33 min (IQR 13 min to 5 h) per night at 3 months and 1 h 26 min (4 min to 4 h 45 min) per night at 12 months. Additional data for CPAP usage are shown in appendix p 24. 37 serious adverse events were recorded during the trial; 15 (in 12 patients, including one death) in the CPAP group and 22 (in 13 patients, including one death) in the BSC group; all adverse events were independently classifi ed as being unrelated to the trial. CPAP was associated with common self-reported side-eff ects (appendix p 25). No clinically important harm from use of CPAP was noted.

Discussion
This 12-month randomised, controlled trial has unequivocally shown that CPAP reduced subjective sleepiness in older people with OSA syndrome at 3 months, despite low overall CPAP usage. The benefi cial eff ects were maintained at 12 months, and the magnitude of the improvement is similar to that seen in middle-aged patients with similar levels of disease severity treated with CPAP. 5 The reduction in subjective sleepiness was corroborated by a signifi cant improvement in objective sleepiness, measured by the OSLER test at 3 months. CPAP also produced signifi cantly better quality-of-life outcomes, as measured with the SAQLI and SF-36. The relative increase in QALYs was not signifi cant in the primary cost-eff ectiveness analysis; this could have been because the EQ-5D is a less sensitive measure of health status attributable to sleepiness because it contains no relevant dimension for this symptom. Overall, the marginal economic benefi t of CPAP was linked to a reduction in health-care use, off setting the cost of the CPAP equipment; this was more robust if using SF-6D to measure health benefi ts and in patients with higher ESS. Secondary outcomes related to cognitive function did not show any diff erence between the two groups, despite improvements in sleepiness in the CPAP group. Additionally, mood, which might also aff ect cognitive function, was not signifi cantly diff erent between the two groups. Although, patients in this trial had a low prevalence of depression compared with those in a recent study, 24 and the baseline cognitive scores were often within the age-adjusted normative range, which could have resulted in a ceiling eff ect. The low overall use of CPAP in our trial might also have been a factor, or older people prepared to participate in a year-long trial might be constitutionally diff erent to those whose cognitive function leads to clinically signifi cant compromise. Cognitive defi cits have been reported in middle-aged patients with moderate to severe OSA syndrome; 25 however, the eff ect of OSA syndrome on cognitive defi cits in mildly symptomatic patients is questionable. 26 In terms of the cardiovascular outcomes, signifi cant improvement was noted for total cholesterol at 3 months in the CPAP group, but this was not sustained at 12 months. The fall in cholesterol was similar to fi ndings in a group of patients with more severe OSA syndrome after a 1-month CPAP trial. 27 CPAP produced no improvement in blood pressure. At fi rst sight, this might be surprising because CPAP has been shown in other randomised controlled trials to reduce blood pressure by roughly 2-10 mm Hg in patients with OSA syndrome. 28,29 However, our fi ndings are consistent with a recent metaanalysis in patients with minimally symptomatic OSA, which showed that CPAP does not have a benefi cial eff ect on blood pressure. 29 In the BSC group, systolic blood pressure fell, a fi nding also reported in another recent study of minimally symptomatic patients with OSA. 30 We speculate that this could be because the control group followed the BSC advice more closely than did the CPAP group. This suggestion cannot be verifi ed, although the lack of weight loss in both groups could imply similar adherence with the BSC information (appendix p 26). Further research is needed to clarify the cardiovascular eff ect of CPAP treatment in older people with OSA syndrome. Other secondary outcomes, including nocturia and home and driving accidents (appendix p 19), also did not improve with CPAP, which could be because of their multifactorial causes.
The mean usage of CPAP was low at 3 months and 12 months and this might have diluted the signifi cant treatment eff ect we noted between the groups. We adopted a standard NHS clinical (real-world) approach to initiate and manage CPAP treatment across the 14 sleep centres in the UK that participated in the trial. The centres varied in size and experience; however, any eff ect of the diff erences between centres was accounted for by adjustment for centre in a random eff ects model in every analysis.
The real-world clinical management adopted in PREDICT could have resulted in the lower CPAP usage compared with that in a more intensive trial approach 26 or shorter duration studies. 27 However, such adoption ensured that the outcomes of PREDICT refl ect clinical practice, which in turn strengthens the validity and applicability of our results. Furthermore, the mean usage of CPAP was similar to a 6-month, randomised, controlled trial of CPAP in minimally symptomatic patients with OSA, 30 and we noted a dose-response relation between the treatment eff ect and CPAP usage, consistent with previous trials in middle-aged populations. 5 Adherence to auto-titrating and fi xed CPAP has not been shown to diff er, although other factors, such as reduced social support, could have contributed to lower CPAP adherence, because 50% of the patients in our trial reported sleeping alone. The ESS was selected as the primary outcome measure for sleepiness in PREDICT because it is the most widely used subjective scale of sleepiness severity in clinical and research practice; it is also the measure from which the UK National Institute for Health and Care Excellence guidelines are drawn. 5 The eff ect of CPAP on ESS in middle-aged patients with mild OSA has been reported as -1·07 (SD 2·4). 5 We therefore powered PREDICT for a one point change in ESS; however, we did not know if this would translate to functionality changes in older patients. The improvements in ESS and quality of life at 3 months and 12 months might go some way to support the notion that changes of this magnitude in ESS are clinically meaningful in this age group. However, some of the secondary endpoints might have lacked power to detect a clinically meaningful eff ect, and we also noted that many of the measures started at normal levels (eg, depression), thus there was minimum room for improvement.
A possible limitation of the PREDICT trial was that sham CPAP was not used as a comparator, although any placebo eff ect there might have been in the CPAP group could reasonably be expected to have disappeared by 12 months. Additionally, the objective OSLER test and the dose-response relation between the treatment eff ect and CPAP usage support a real eff ect. Sham devices have been validated as a placebo for CPAP, but there is no consensus on the ideal comparator, and trials with BSC produce results essentially identical to those from trials with subtherapeutic or sham CPAP. 5 BSC was chosen as the trial comparator for PREDICT because it was more appropriate for a pragmatic multicentre design with a 12-month follow-up.
The strength of the PREDICT trial was that patients presenting to our sleep clinics, requiring investigation and treatment for OSA syndrome were drawn from geographically diverse areas (appendix p 27) and were treated in a real-world clinical setting. Additionally, to our knowledge, PREDICT is one of the fi rst trials specifi cally aimed at older people (≥65 years). The mean age of 70 years, with no participants younger than 65 years, diff ers signifi cantly from the mean age of 58 years (SD 7) in another recent UK trial, 30 the MOSAIC study (appendix p 10). The PREDICT trial has also been one of the longest, randomised, trials of CPAP treatment for OSA syndrome, directly measuring both therapeutic and economic benefi ts.
The PREDICT trial was designed to be done in sleep clinics in the UK where polysomnography is not routinely done for the diagnosis of OSA syndrome. Use of polysomnography would have been fi nancially and practically prohibitive. Respiratory polygraphy could have reduced the ODI if the older patients slept poorly, because the ODI is calculated as apnoeas and hypopnoeas divided by time asleep, and using the polyrgraphy meant that the time asleep could only be estimated from total time in bed (which might be an overestimation of the total time asleep). However, the randomised, controlled trial design meant that any underestimation of the change in ODI would be evenly distributed across groups. Likewise, the Panel: Research in context Systematic review Several systematic reviews have assessed the effi cacy of continuous positive airway pressure (CPAP) treatment for patients with obstructive sleep apnoea (OSA) syndrome. One of the most recent and comprehensive concluded that CPAP was an eff ective and cost-effi cient treatment for moderate to severe OSA syndrome in well defi ned middle-aged populations. 5 However, the study emphasised evidence gaps with a need for trials in other patient groups, one such group being older people. The authors concluded that ''clinical trials to defi ne treatment eff ects at the extremes of age particularly in the elderly where cardiovascular co-morbidity complicates assessment would be benefi cial''. Therefore, despite the high prevalence of obstructive sleep apnoea in older people there is a paucity of evidence on the relative clinical benefi ts or risks of CPAP treatment in older people. We then updated the scientifi c literature search done by McDaid and colleagues 5 March 19, 2012, without language restrictions, for full articles reporting randomised controlled trials assessing the effi cacy of CPAP treatment in OSA syndrome, with participants aged on average 60 years or older with the capacity to give informed consent, and identifi ed only three studies. [31][32][33] None of these studies assessed subjective sleepiness as a primary endpoint or obtained generic measures of health utility, and two of the studies recruited only patients with chronic heart failure.

Interpretation
Before the PREDICT trial, very little information was available for clinicians and health-care professionals regarding the best way to treat OSA syndrome in older people, and even less information was available about how CPAP treatment aff ected quality of life and cost-eff ectiveness in this population. The results of the PREDICT trial show that CPAP reduces symptoms of excessive daytime sleepiness in older patients with OSA syndrome, as it does in middle-aged populations, and that these clinical benefi ts are associated with some reduction in health-care use. Therefore, CPAP is more likely to be a costeff ective option for older patients with OSA syndrome at the cost-eff ectiveness thresholds used by the UK National Institute for Health and Care Excellence (probability 0·61 with EQ-5D and 0·96 with SF-6D at the threshold of £20 000 per qualityadjusted life-year gained). On the basis of these fi ndings, we recommend that CPAP treatment should be off ered routinely to older patients with OSA syndrome. plus governance of the trial. SG was health economist for PREDICT with specifi c responsibility for the health economic analysis and write-up of the report. RJD was a senior investigator for PREDICT, and a member of the Trial Steering Committee who was involved in all aspects of the study design and development of case report forms; however, he died during the early part of the study. AJN was a statistician for PREDICT, a member of the Trial Management Committee, Trial Steering Committee, and Data Monitoring Committee with specifi c responsibility for the statistical analysis of the trial. JRS was a senior investigator for PREDICT with specifi c responsibility for clinical oversight of the trial. RLR and MJM were coprincipal investigators for PREDICT; MJM and RLR share last authorship. MJM was corresponding author with the fi nal responsibility to submit for publication.

Declaration of interests
JRS reports grants from NIHR and ResMed during the conduct of the study, plus personal fees from ResMed outside the duration of the submitted work. The other authors declare no competing interests. distribution of comorbidities likely to produce central events, such as heart failure, was similar across groups; notably, no central or mixed events were scored, and patients with other sleep disorders were excluded. For these reasons we do not think the use of polygraphy restricted the fi ndings of PREDICT.
To our knowledge, PREDICT is the fi rst trial to measure both SF-36 and EQ-5D over a 12-month period, enabling a full health economic evaluation of the treatment of OSA syndrome in older people (panel). As already noted, the SF-6D has a dimension on vitality, which could render it more sensitive to changes in sleepiness and sleep quality than EQ-5D. Other studies have compared SF-36 with EQ-5D and also noted that SF-36 is more sensitive to the eff ect of CPAP on health-related quality of life. 34,35 Diff erences in data collection processes could also have aff ected the health-related quality of life reported by our patients. The EQ-5D was collected every month through the sleep diary, which was fi lled in by the patient at home. The SF-36, from which SF-6D was derived, was collected in a clinic visit at baseline, 3 months, and 12 months.
With respect to generalisability, the PREDICT trial did not focus on asymptomatic older people with OSA and although it could be argued that the patients studied had a fairly low mean ESS at baseline, they were suffi ciently symptomatic to seek treatment. At the other end of the disease spectrum, exclusion of highly symptomatic patients with OSA syndrome (216 [20%] of 1073 patients; appendix pp 3, 4) in whom CPAP was considered mandatory, is likely to have diminished the treatment eff ect. The exploratory analyses showed that the treatment eff ect was larger in patients with a higher baseline ESS or greater use of CPAP. The marginal improvement in cost-eff ectiveness was also greater in the more symptomatic patients.
In summary, the PREDICT trial addresses the inequality of research in older people with OSA syndrome, and incorporates data for comorbidity that is often lacking in clinical guidelines. 36 The trial shows that in older patients with OSA syndrome, CPAP treatment reduces symptoms of excessive daytime sleepiness, as it does in middle-aged populations. On the basis of these fi ndings, we recommend that CPAP treatment should be off ered routinely to older patients with OSA syndrome, especially those with higher ESS. Future research should focus on how best to optimise CPAP delivery in this age group.

Contributors
All authors (except RJD) were members of the Trial Management Committee and Trial Steering Committee with responsibility for the progress and conduct of the trial. All authors were involved in the study design and development of case report forms, analysis of data, and all aspects of writing the report. AM was Clinical Research Fellow for PREDICT with specifi c responsibility for all data collection and oversight of the trial, and fi rst author of the report. DJB was a statistician for PREDICT, a member of the Trial Management Committee, Trial Steering Committee, and Data Monitoring Committee. RF was a health economist for PREDICT with specifi c responsibility for the health economic analysis and write-up of the report. ML-S was trial manager for PREDICT with specifi c responsibility for all aspects of data collection and recording,