Surgical vs Conventional Therapy for Weight Loss Treatment of Obstructive Sleep Apnea:  A Randomized Controlled Trial FREE

Obesity is a major risk factor for obstructive sleep apnea (OSA). The estimated prevalence of OSA in obese adults varies from 42% to 48% in men and 8% to 38% in women.¹ Obstructive sleep apnea is strongly related to obesity and associated conditions such as type 2 diabetes and hypertension.² Individuals with OSA are at greater risk of stroke, cardiac disease, psychosocial morbidity, cognitive dysfunction, and impaired quality of life.^3- 5

Therapy for symptomatic OSA focuses on methods to reduce upper airway resistance and prevent obstruction, such as continuous positive airway pressure (CPAP),⁶ oral appliances,⁷ and upper airway surgery. However, in severely obese patients, none of these options treats the accompanying obesity. Observational studies of major weight loss following bariatric surgery suggest substantial remission of OSA symptoms in up to 60% to 80% of patients.⁸ However, in the studies where repeat polysomnography was available, remission of OSA was unusual.⁹ There have not been any randomized controlled trials comparing medically supervised weight loss with bariatric surgery as a treatment for OSA.

Laparoscopic adjustable gastric banding (LAGB) is a safe and effective weight loss treatment, with 2-year weight loss of approximately 20% of initial body weight achieved.¹⁰ We previously demonstrated that randomized controlled trials comparing surgical weight loss with conventional therapy are feasible, and LAGB is effective, safe, and cost-effective.^11- 12 We conducted a 2-year randomized controlled trial involving 60 severely obese patients comparing LAGB with conventional weight loss therapy for the management of moderate to severe OSA.

Patients with obesity and OSA were recruited from 7 Melbourne, Australia, sleep clinics between September 2006 and March 2009 and all data were available for analysis starting in September 2011. All patients provided written informed consent for participation in the trial and additional written informed consent was obtained prior to surgical procedures. The study was approved by the human ethics committees of the Alfred and Avenue hospitals and Monash University in accordance with the guidelines of the National Health and Medical Research Council of Australia and the Helsinki Declaration, as revised in 2000.

The inclusion criteria were patients aged 18 to 60 years, body mass index (calculated as weight in kilograms divided by height in meters squared) of 35 to 55, apnea-hypopnea index (AHI) of 20 events/hour or more diagnosed within the previous 6 months with recommendation to commence CPAP therapy,¹³ and at least 3 prior significant weight loss attempts. The exclusion criteria were previous bariatric surgery, obesity hypoventilation syndrome requiring bilevel positive airway pressure, and contraindications to bariatric surgery including cognitive impairment, drug or alcohol addiction, and significant cardiopulmonary, neurological, vascular, gastrointestinal, or neoplastic disease.

Potential participants were screened initially at sleep clinics and referred to trial physicians (J.B.D. and L.M.S.), who explained the study in detail and assessed eligibility. All were provided with autotitrating CPAP equipment (Autoset S8, ResMed). Baseline anthropometric, functional testing, biochemistry, and questionnaire data were collected immediately prior to computer-derived randomization.

This program delivered the best-available medical practice for the treatment, education, and follow-up of severely obese patients with moderate to severe OSA. Dietary, physical activity, and behavioral programs were individualized. The advice regarding physical activity encouraged walking and 200 minutes/week of structured activity, including moderate-intensity aerobic activity and resistance exercise. Dietary advice was based on the Dietary Guidelines for Australian Adults and the Australian Guide to Healthy Eating and included a planned daily deficit of 500 kcal from estimated energy requirements.^14- 16 All participants were offered an initial intensive very low-energy diet (VLED) (Optifast, Nestle-Australia) program, with the meal replacements provided. The VLED meal replacements continued to be available for further intensive, intermittent, or occasional use throughout the study.

Participants underwent 2 weeks of VLED to reduce liver size prior to placement of an LAGB (LAP-BAND System, Allergan Health) via the pars flaccida pathway by 1 of 3 experienced surgeons (P.E.O. and W.B. and Stewart Skinner, MBBS[Hons], PhD, Department of Surgery, Monash University), within 1 month of randomization.¹⁷ Adjustments to band volume were made using standard clinical criteria.¹⁸

Patients in both programs had open access to a bariatric physician, sleep physician, and dietitian, and had their progress reviewed every 4 to 6 weeks throughout the 2 years. The management of OSA, the intensity, and nature of the lifestyle program were common to both groups.

The primary outcome measure was the change in AHI, measured by diagnostic laboratory polysomnography from baseline to 2 years. Polysomnography was performed using standard electroencephalogram, electrooculogram, electromyogram, nasal pressure cannulae, oronasal thermistor, respiratory inductance plethysmography, finger oximetry, electrocardiography, and video monitoring for body position. Diagnostic polysomnography at years 1 and 2 was performed after a 48-hour CPAP washout at the same institution as the initial test, scored by staff blinded to randomization group, and using the same precise AHI scoring criteria for each study.¹⁹

One of the secondary measures was CPAP adherence, which was monitored with 3 monthly computerized downloads recording mean CPAP adherence (in hours) and pressure (in cm H₂O). Other diagnostic polysomnography measures included arousal index, percentages of rapid eye movement and slow wave sleep, and the mean and minimum oxygen saturation. We also assessed measures of cardiometabolic risk (office blood pressure, resting heart rate, 6-minute walk test, levels of fasting plasma glucose, insulin, triglycerides, total and high-density lipoprotein cholesterol, and calculated low-density lipoprotein cholesterol); measured weight, waist, hip, and neck circumference; and reported metabolic syndrome status.²⁰ Other secondary outcomes included the following functional status measures: Epworth Sleepiness Scale score,²¹ which was scored from 0 to 24 with 10 or more considered abnormal, Short Form-36 Health Survey,²² and Beck Depression Inventory.²³ Adverse events were recorded throughout the study.

The sample size was selected to provide a statistical power of at least 80% to detect a between-group difference of 50% in AHI (mean of 20 events/hour for conventional program vs 40 events/hour for surgical program with an SD of 20 events/hour) at 2 years (P < .05), with modeling based on our observational data and expectations of other devices.^24- 26 Fifty patients were needed to provide a 95% confidence interval of this difference with a power of 0.80 or more (2-tailed). Recruitment size was therefore set at 60 to allow for a dropout rate of up to 20% or for failure to follow the randomly assigned allocation.

Univariate statistical analysis was performed using SPSS statistical software version 18 (SPSS Inc), with baseline comparisons made using χ² tests for equal proportion, t tests for normally distributed outcomes, or Mann-Whitney tests. A multivariate longitudinal analysis was performed using the PROC Mixed procedure in SAS version 9.2 (SAS Institute Inc), with each patient treated as a random effect.

Longitudinal models were fitted using main effects for group, time, and an interaction between group and time to ascertain if the groups behave differently over time. Covariance structures were chosen to minimize the corrected Akaike and Bayesian information criteria. Multiple imputation was generated for missing data using SAS version 9.2 and was based on the assumption that data were from a multivariate normal distribution and were missing at random. The regression method was used for imputation with 5 imputed data sets used for each variable. The post hoc relationship between changes in weight and AHI were explored by dividing weight loss into quintiles and using comparisons of Pearson correlation coefficients and 95% confidence intervals performed using the Fisher r-to-z transformation. A locally weighted smoothing (loess) curve was used to indicate nonlinear trends in the scatter plots.

All data were analyzed using the intention-to-treat principle. Continuous variables were expressed as means and standard deviations with differences expressed as means and 95% confidence intervals. Sensitivity analyses examining assigned only completer and per-protocol completer were performed. A 2-sided P value of .05 was considered to be statistically significant.

Study participant flow and reasons for exclusion are shown in Figure 1. There were no statistically significant between-group differences in the baseline characteristics (Table 1). In the surgical group, 4 participants did not consent to surgery but continued in the study and received conventional therapy; complete 2-year follow-up data were available in 28 (93%) of the 30 participants. Of the 30 participants in the conventional group, the 2-year program was completed by 28 (93%), and all data were collected for 26 (87%). One conventional group participant elected to undergo LAGB surgery soon after randomization. Another conventional therapy participant was unable to undergo a repeat diagnostic polysomnography because he or she was unable to sleep without the CPAP device.

All bands were placed laparoscopically with a mean (SD) procedure time of 70 (14) minutes (range: 40-105 minutes). The lengths of hospital admission were 1 day for 24 participants (92%) and 2 days for 2 participants (8%). In the conventional group, 17 participants chose to use an intensive VLED program for 4 to 6 weeks, another 5 used a more conservative initial program; 11 of these 22 elected to use a partial or intermittent intensive VLED meal replacement throughout the study.

The surgical group achieved a significantly greater mean weight loss of 27.8 kg (95% CI, 20.9-34.7 kg) from 134.9 kg to 107 kg or a mean of 20.6% (95% CI, 15.4%-25.7%) of initial body weight at 2 years compared with 5.1 kg (95% CI, 0.8-9.3 kg) from 126.0 kg to 121.8 kg or a mean of 2.9% (95% CI, 0.6%-7.3%) among the conventional group (P < .001; Table 2, Figure 2, and eFigure 1). This represents a reduction of body mass index from 46.3 to 36.6 with surgery vs 43.8 to 42.3 with conventional therapy.

Both groups had a significant decrease in total AHI between baseline and 2 years, with a decrease of 25.5 events/hour (95% CI, 14.2 to 36.7 events/hour) from 65.0 events/hour to 39.5 events/hour or a mean of 31.4% (95% CI, 12.7% to 50.2%) (calculated for individuals) in the surgical group and 14.0 events/hour (95% CI, 3.3 to 24.6 events/hour) from 57.2 events/hour to 43.2 events/hour or a mean of 13.5% (95% CI, −5.8% to 32.9%) in the conventional group (Table 2 and Figure 2). The between-group difference was −11.5 events/hour (95% CI, −28.3 to 5.3 events/hour; P = .18) or percentage change in AHI of −17.9% (95% CI, −44.3% to 8.5%; P = .18). The sensitivity analysis confirmed that the results were not altered by inclusion of only completers when those in their assigned or per-protocol group followed the program. Thus, despite the substantially greater weight loss with surgical therapy, there was not a greater improvement in AHI. An AHI of less than 15 events/hour, which indicates mild OSA, was achieved by 8 participants in the surgical group (27%) and by 2 participants in the conventional group (7%) (P = .04); only 1 participant in the conventional group achieved an AHI of less than 5 events/hour, which indicates OSA remission.

A post hoc analysis showed a significant positive relationship between the change in weight and change in AHI (r = 0.45 [95% CI, 0.22 to 0.63]; P < .001). However, when examined in the individual groups, the relationship was demonstrated only in the conventional group (r = 0.69 [95% CI, 0.44 to 0.84]; P < .001); in the surgical group, the analysis yielded r = 0.33 (95% CI, −0.03 to 0.62; P = .07) (conventional group r of 0.69 vs surgical group r of 0.33; P = .06). Scatterplots (with loess curves) of change in weight and change in AHI, and percentage weight loss and percentage change in AHI (Figure 3) demonstrate an attenuated benefit beyond the first 8% to 10% or 10 kg of weight loss, and great variability of any change with weight loss (eFigure 2). Table 2 details the treatment effects on other polysomnography variables.

Within the surgical group (n = 30), 28 participants had CPAP initiated; however, at 3 months, only 20 participants were adherent (mean pressure: 11.9 cm H₂O used for 4.9 hours/night) and at 2 years, only 14 participants were adherent (mean pressure: 11.6 cm H₂O for 5.2 hours/night). Within the conventionally treated group (n = 30), 25 participants had CPAP initiated; however, at 3 months, only 21 participants were adherent (mean pressure: 11.7 cm H₂O used for 5.9 hours/night) and at 2 years, only 18 participants were adherent (mean pressure: 11.5 cm H₂O used for 5.6 hours/night). There were no significant between-group differences in CPAP adherence or mean pressure delivered.

eTable 1 shows changes in clinical and laboratory measures of health at 2 years. There were no between-group differences in blood pressure, fasting glucose and insulin, or plasma lipid profile. There were no within- or between-group differences in the use of antihypertensive, diabetes, lipid-lowering, or antidepressant medication between baseline and 2 years. Twenty participants had type 2 diabetes at baseline (10 in each group) and a further 2 in the conventionally treated group developed type 2 diabetes during the study. For those with diabetes (n = 22), the mean hemoglobin A_1c levels were 7.2% (95% CI, 6.4% to 7.9%) for those in the surgical group (n = 10) and 7.3% (95% CI, 6.4% to 8.1%) in the conventional group (n = 12) at baseline and 6.0% (95% CI, 5.6% to 6.4%) and 7.0% (95% CI, 6.3% to 7.7%), respectively, at 2 years. There was a significant within-group change of −1.26% (95% CI, −1.97% to −0.55%; P = .003) in those with diabetes in the surgical group compared with −0.3% (95% CI, −1.16% to 0.5%; P = .46) in the conventional group; however, the between-group difference of −0.96% (95% CI, −2.04% to 0.12%; P = .08) was not statistically significant.

Baseline health-related quality-of-life and summary scores on the Short Form-36 scale were not statistically different between the 2 groups. The surgical group had greater improvement in scores between baseline and 2 years for physical role, general health, vitality, and the physical component summary (mean, 9.3 [95% CI, 0.5 to 18.0]; P = .04 for physical component summary score; eTable 2). Both groups experienced a reduction in daytime sleepiness as measured with the Epworth Sleepiness Scale, reductions in symptoms of depression, and improvement in the 6-minute walk test, but there were no significant between-group differences.

There was no mortality in either group. One patient experienced an acute proximal gastric pouch dilatation causing obstructive symptoms and requiring elective laparoscopic replacement of the LAGB 1 month later. This patient's hospital stay was less than 1 day and there were no complications (eTable 3).

Comparing surgical and conventional weight loss therapy for the management of OSA demonstrated greater weight loss in the surgically treated group, however, this did not translate into significantly greater improvements in OSA. Our study did confirm that weight loss is associated with an improvement in AHI, but we found great variability in the individual effect. The pattern of improvement in AHI suggests that much of the benefit is associated with mild to moderate weight loss, with limited additional benefit with further weight loss.

Our findings support the 2 independent meta-analyses by Greenburg et al⁹ and Chang et al²⁷ of before and after case series trials, in which they concluded that only a minority of obese patients with OSA are free of OSA after bariatric surgery. This contrasts with a meta-analysis of bariatric surgical treatment groups that described resolution of OSA symptoms in 85.7% of patients. However, only 4 of 24 studies (17%) reported polysomnography or AHI outcomes.⁸

Data from the present study (and that from the systematic reviews) do not support routine cessation of CPAP or other therapies for OSA following bariatric surgery.^9,27 Formal assessments, including polysomnography, should be considered before informing a patient that he or she may cease CPAP therapy. The clinical picture can be confounded by patients self-reporting improved quality of life, sleepiness, sleep quality, and exercise tolerance after bariatric surgery.

Our findings support a complex, rather than pure mechanical load, pathogenesis of OSA in obese individuals. The nonlinear relationship may indicate the importance of the state of weight loss rather than the extent of weight loss in generating an altered neurohumoral and metabolic-inflammatory milieu influencing defects in neuromuscular responses to mechanical load.^28- 29 Population studies indicate the influence of weight on OSA varies with age and sex, and additional factors such as boney structures are also contributors to risk.^30- 31 Thus the relationship of the extent of weight loss with improvement of OSA needs further investigation. However, the benefits of modest weight loss have several other major public health implications^32- 33 that should not be diminished by the outcomes of this study.

Several limitations of this study need to be recognized. First, the study may have been underpowered given the extraordinary variance in the effect of weight loss on AHI. We relied on data from our observational study that may have been biased because it relied on patients volunteering for a follow-up polysomnography study.²⁵ However, the clinical relevance of a much larger study, even if a significant effect of weight loss on AHI were demonstrated, would be limited because the effect of weight loss is highly variable and incomplete. Second, we used LAGB surgery, which is associated with a slower weight loss and a lower mean weight loss than gastric bypass and biliopancreatic diversion at 2 years following surgery.³⁴ However our data do not provide a signal that greater weight loss would achieve a better response. Third, we restricted the study to those with a body mass index of 35 to 55 with moderate to severe OSA. Results in those in a lower weight range or with less severe OSA may be different.³³ In addition, this study was not powered for safety, the wide range of secondary outcome measures, or to detect differences in hard end points such as mortality or cardiovascular events.

In conclusion, greater weight loss in the surgical group did not translate into better effect as a therapy for OSA compared with conventional weight loss therapy. While there is a clear overall response to weight loss, the individual effects are variable, incomplete, and appear to be related to modest weight loss. This study suggests caution be used in counseling patients about the expected benefits of weight loss for OSA, and that patients losing weight be assessed appropriately before ceasing effective therapy.