Weight reduction and the impact on apnea-hypopnea index: A systematic meta-analysis

Abstract

Obstructive sleep apnea (OSA) is strongly associated with obesity. While the relationship between weight reduction and apnea-hypopnea index improvement has been documented, to our knowledge, it has not been quantified adequately. Therefore, this study aimed to quantify the relationship between weight reduction and AHI change.

Methods:

A systematic literature search was performed using meta-analyses (PRISMA) guidelines for studies reporting AHI and weight loss in people with obesity/overweight and OSA between 2000 and 2023. A linear and quadratic model (weighted by treatment arm sample size) predicted percent change from baseline AHI against mean percent change from baseline weight. The quadratic term was statistically significant (P < 0.05), so the quadratic model (with 95 % prediction interval) was used.

Results:

The literature search identified 27 studies/32 treatment arms: 15 using bariatric surgery and lifestyle intervention each and 2 using pharmacological interventions. Included studies were ≥3 months with weight intervention and participants had AHI ≥15/h. Weight reduction in people with OSA and obesity was associated with improvements in the severity of OSA. BMI reduction of 20 % was associated with AHI reduction of 57 %, while further weight reduction beyond 20 % in BMI was associated with a smaller effect on AHI. As the prediction intervals are relatively wide, a precise relationship could not be conclusively established.

Conclusion:

The degree of AHI index improvement was associated with the magnitude of weight reduction. The model suggests that with progress in weight reduction beyond 20 %, the incremental decrease in BMI appeared to translate to a smaller additional effect on AHI.

1. Introduction

The obesity pandemic has occurred globally, with its rate initially increasing in affluent countries [1,2] and more recently in developing countries [3]. Projections support that the increase in obesity may negatively impact life expectancy gains from the last century [4,5]. Obesity contributes to the development or worsening of over 200 comorbidities [6], including diabetes, nonalcoholic fatty liver disease, cardiovascular disease, cancer, mental health disease, and obstructive sleep apnea (OSA). Worldwide increases in OSA rate have paralleled the obesity pandemic because of the direct pathophysiological effect of obesity on OSA. Concerningly, OSA is associated with major neurocognitive and cardiometabolic complications [7].

There are a range of therapeutic options for treating OSA. Nasal positive airway pressure (PAP) therapy, either automatic PAP or continuous PAP (CPAP), is the most common therapeutic option for OSA, and it can result in transformative improvements in sleep-disordered breathing, symptoms, blood pressure, and risk of motor vehicle accidents [8–10]. However, PAP is not always well tolerated, leading some health care providers to offer other treatment options. Oral appliances and upper airway surgery have a role in OSA therapy, but the efficacy of these treatments can be inadequate and the ability to predict responders is variable [11,12]. Weight reduction as an intervention can be efficacious for some patients, but it can be difficult to achieve in clinical practice [13,14].

Although obesity plays an important role in OSA pathophysiology, chronic weight management plays only a supportive role in OSA treatment [15]. There is limited research on the effect of different magnitudes of weight reduction on the apnea-hypopnea index (AHI) in people with OSA and obesity.

Weight reduction can be facilitated by lifestyle intervention, although success rates with this approach are often disappointing [16]. Bariatric surgery and antiobesity medications are being increasingly used, but the risks and costs associated with these treatment options need to be weighed against the potential benefits [17]. There are different views on whether weight reduction has a dose-response effect (i.e, for every kg lost there is some improvement in AHI) or a threshold effect (ie, achieving a weight below a certain level leads to resolution of OSA) [18]. Clinically, a dose-response relationship seems likely, although individual variability is clearly present.

Based on this conceptual framework, the objective of this analysis was to explore the effect of different magnitudes of weight reduction on AHI in people with OSA and obesity or overweight based on published literature. These data would help inform interventional studies, given recent developments in obesity pharmacotherapy: there is a growing pipeline of research on incretins for the management of obesity, which includes studies of longer-term use to modify cardiovascular risk and other complications of excess adiposity, including OSA [19].

2. Methods

This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Briefly, searches for peer-reviewed publications were conducted in Cochrane Central Register of Controlled Trials (CENTRAL) and EMBASE and in the MEDLINE^® and MEDLINE^® In-process databases on the OVID^® platform for the period between the year 2000 and February 2023. To complement the peer-reviewed publications, conference abstracts and the ClinicalTrials.gov and WHO International Clinical Trials Registry Platform Search Portal registries were also searched. Reference lists of the included studies were also searched to retrieve additional articles.

2.1. Search strategy

The search strategy employed for the systematic review was based on the patient, intervention, comparator, outcomes, study design (PICOS) framework. The PICOS framework was also used to inform the search strings and the eligibility criteria of the studies. Briefly, terms such as “sleep disordered breathing,” “Obstructive Sleep Apnea,” “Sleep Apnea Syndrome,” “bariatric surgery,” “body weight loss,” “lifestyle,“. “Diet,” and “exercise” and other terms related to pharmacological interventions for weight loss were included in the search strategy. The detailed search strategy included a combination of free text, keywords, subject headings, and limiters and is presented in Table 1 in Supplement.

2.2. Study selection criteria

Randomized controlled trials and uncontrolled interventional studies were eligible for inclusion if they met the study selection criteria summarized in Table 1.

Table 1.

Study selection criteria.

Inclusion Criteria	Exclusion Criteria
Age ≥18 years, BMI≥30 kgm−2	Underlying pathophysiology modifying the effect of weight reduction on AHI
AHI ≥15/h; participants without OSA not included	Study intervention modifying the effect of weight reduction on AHI
AHI measured by PSG or HSAT; studies with RDI or REI included	OSA treatment with mechanical devices other than PAPa
Weight loss intervention	Duration of intervention <3 months/12 weeksb
Data on weight/BMI and AHI before and after intervention
Analysis of all arms involving rigorous weight loss intervention

Abbreviations: AHI, apnea-hypopnea index; APAP, automatic positive airway pressure; BiPAP, bilevel positive airway pressure; BMI, body mass index; CPAP, continuous positive airway pressure; HSAT, home sleep apnea test; OSA, obstructive sleep apnea; PAP, positive airway pressure; PSG, polysomnography; RDI, respiratory disturbance index; REI, respiratory event index.

^aCohorts with the use of any PAP-like treatments (CPAP, BiPAP, APAP) were not excluded.

^bCohorts with outcome assessments in less than 3 months/12 weeks from the beginning of intervention were excluded.

2.3. Study selection

Study selection was performed in 2 stages. In the first stage, 2 independent reviewers screened articles by their titles and abstracts based on the predefined eligibility criteria. Conflicts around articles for which there may have been uncertainty in decision making by the 2 independent reviewers were resolved through “arbitration” by a third independent reviewer. Publications included at the end of this stage were retained for the second stage of screening—full-text review. The full text of articles was independently screened by the 2 reviewers against the eligibility criteria, and the approach for resolving uncertainty regarding inclusion of publications was the same as that in the first stage of the screening process. Studies (N = 2) were excluded from the analysis as the data were expressed as median instead of mean.

All publications included after completion of full-text review were retained for data extraction and synthesis. Details of the eligible studies are presented in Table 2.

Table 2 –

Clinical studies and treatment arms included in the analysis.

	First author	Study design	N	Gender (% females)	BMI (mean, kg/m2)	AHI (events/h)	Weight Intervention	Duration
1	Al-Jumaily AM [37]	C	10	100	48.5	38.1	BS	12 months
2	Bae EK [38]	C	10	50	39.9	51	BS	12 months
3	Dixon JB [39]	R	30	43	46.3	65	BS	2 years
4	Dixon JB [40]	C	25	28	52.7	61.6	BS	1–2 years
5	Yilmaz Kara B [41]	C	31	55	49.8	36.1a	BS	12 months
6	Leentjens M [42]	C	31	81	43.7	28.7	BS	6–9 months
7	Magne F [43]	C	44	80	46.1	52.8b	BS	avg. 7.9 months
8	Poitou C [44]	C	35	83	51.3	24.5a	BS	1 year
9	Priyadarshini P [45]	C	27	78	48.4	31.8a	BS	3–6 months
10	Ravesloot MJ [46]	C	110	61	45.4	39.5a	BS	7.7 months
11	Del Genio G [47]	C	36	67	51.3	32.8a	BS	5 years
12	Lettieri CJ [48]	C	24	75	51.0	47.9a	BS	1 year
13	Sutherland K [49]	C	18	89	44.1	23.6a	BS	6 months
14	Haines K [50]	C	101	92	56.0	51a,c	BS	11 months
15	Nasir I [51]	C	12	67	45.5	24.8a	BS	3 months
16	Barnes M [52]	C	12	75	36.1	24.6a	LCD	16 weeks
17	Blackman A [53]	R	180	28	38.9	49	LCD	32 weeks
18	Dixon JB [39]	R	30	40	43.8	57.2	LCD	3 years
19	Foster GD [13]	R	125	62	36.8	22.9a	LCD	1 year
20	Hoyos CM [54]	R	34	0	36.6	33.2a	LCD	18 weeks
21	Johansson K [55]	C	63	0	34.8	36	LCD	52 weeks
22	Nerfeldt P [56]	C	23	27	40	41a	LCD	24 months
23	Ng S [57]	R	61	21	30.2	43.4	LCD	12 months
24	Papandreou C [58]	R	10	19	37.1	45.8	LCD	6 months
25	Papandreou C [58]	R	11	19	36.2	46.6	LCD	6 months
26	Winslow DH [59]	R	23	35	35.3	45.2	LCD	28 weeks
27	Georgoulis M [60]	R	59	29	34.9	60	LCD	6 months
28	Georgoulis M [60]	R	59	23	35.8	62	LCD	6 months
29	Lopez-Padros C [61]	R	18	17	34.5	69.8b	LCD	12 months
30	Truby H [62]	R	47	25	34.2	43.6	LCD	12 months
31	Blackman A [53]	R	179	28	39.4	49.3	P	32 weeks
32	Winslow DH [59]	R	22	59	36.0	44.2	P	28 weeks

Abbreviations: AHI, apnea-hypopnea index; BS, bariatric surgery; BMI, body mass index; C, prospective cohort study; LCD, low-calorie diet; OSA, obstructive sleep apnea; N, population size; P, pharmacological intervention; R, prospective randomized study. All cohorts include patients with AHI ≥15/h.

^aInclude also patients with ≤AHI<15/h.

^bInclude only patients with AHI ≥30/h.

^cRespiratory distress index (RDI) was used in this study.

2.4. Risk of bias assessment

The validity of the results of an individual study depends on the robustness of its overall design and execution and its relevance to the decision problem. Each study that met the inclusion criteria for inclusion was therefore critically appraised as per National Institute for Health and Clinical Excellence guidelines.

2.5. Data synthesis

The primary outcomes of interest were AHI and body mass index (BMI). To ensure that the meta-analysis included relevant data, we included a study that used respiratory disturbance index instead of AHI, as the 2 measures are very highly correlated (depending on exact definitions used). If a study tracked weight instead of BMI change, we included it as well because we believe that weight represents a valid BMI substitute (height is constant).

2.6. Meta-analysis procedure

A literature search was conducted to identify studies that examined the relationship between reduction in BMI and changes in AHI based on the predefined eligibility criteria. Linear and quadratic models were developed to predict population percent change from baseline (defined as the mean change in the variable divided by the baseline mean of the variable for the population) AHI against population mean change from baseline weight. The linear and quadratic models were weighted proportional to treatment arm sample size. The quadratic term was statistically significant (P < 0.05), so the quadratic model (with 95 % prediction interval) was used for inference. All analyses were conducted using RStudio, Posit PBC (R version 4.1.2) [20].

3. Results

3.1. Search results and risk of bias

The literature search identified 9798 potential articles that met the aforementioned criteria. Identified articles were screened and 70 clinical studies were selected for detailed review. The studies were read in detail and assessed versus inclusion/exclusion criteria. Studies (N = 27) and treatment arms (N = 32) were identified for inclusion in the analysis (Fig. 1).

Fig. 1.

Identification of studies for inclusion.

Overall, the risk of bias assessment indicated that a majority of the studies did not have a high risk of bias. Among the randomized studies, a source of bias was inadequate blinding. Additionally, the search criteria for this meta-analysis included a mix of randomized clinical trials and non-randomized observational trials, and the combination of these sources of data and the non-randomized nature of the observational trials introduce additional potential bias. Table 2 presents information on the studies and treatment arms included.

3.2. Population description

Population information is presented in Table 2. The population analyzed in this meta-analysis included a wide range of cohort sizes, study durations/follow-up periods, and OSA severity. The 27 studies with AHI and weight measurements before and after weight intervention in people living with obesity or overweight and OSA included the following cohorts: 15 from bariatric surgery, 15 from diet and exercise intervention, and 2 from antiobesity medication studies. Seventeen of these cohorts were from prospective cohort studies and 15 were from randomized controlled trials.

Cohort size varied significantly, ranging from 10 to 180 participants. Participants in all cohorts had overweight or obesity (participants with overweight were present in 5 cohorts from lifestyle intervention studies, and 15 cohorts with bariatric surgery intervention had participants with BMI ≥40 kg/m² or BMI ≥35 kg/m² with obesity-related complications) and average AHI spanned from 22.9 to 69.8 events per hour. Males and females were equally represented, with 15 cohorts including predominantly females, 16 including predominately males, and 1 including both males and females in equal number. The severity of OSA in all participants was assessed by polysomnography (no cohort with home sleep apnea test (HSAT) met all the inclusion/exclusion criteria) before intervention and after the follow-up period. Follow-up periods of the studies varied from 18 weeks to 5 years.

3.3. Relationship between weight reduction and AHI improvement

The quadratic meta-regression model (Fig. 2) revealed that weight reduction was associated with reduced AHI in patients living with obesity and OSA. The model suggests that with progress in weight reduction, the incremental decrease in BMI appeared to translate to a smaller additional effect on sleep-disordered breathing: 10 % change in BMI was associated with a 36 % reduction in AHI, a 20 % lowering in BMI was associated with a 57 % reduction in AHI, and a 30 % decrease in BMI corresponded to a 69 % decrease in AHI. As the prediction intervals are relatively wide, a precise relationship could not be dependably established. As a sensitivity check, a linear regression model was also used and showed a significant relation between change in BMI and change in AHI with substantial AHI lowering associated with 20 % reduction in BMI.

Fig. 2.

Apnea-hypopnea index versus weight reduction inference. Abbreviations: AHI, apnea-hypopnea index; N, population size.

4. Discussion

The degree of AHI improvement was associated with the magnitude of weight reduction. Based on the meta-regression, the first 10 % reduction in BMI was associated with a greater than 20 % reduction in AHI. With a BMI reduction beyond 20 %, the decrease was associated with a proportionately smaller effect on sleep-disordered breathing. A further 10 % reduction in BMI was associated with a less than 15 % decrease in AHI. We have observed a dose-response relationship in which weight reduction was associated with clinically relevant improvements in sleep apnea as assessed by AHI. These findings may have clinical implications since they could motivate some patients and clinicians to pursue weight reduction as a therapeutic strategy.

Based on the data included in this study, it seems the AHI improvements were consistent regardless of the intervention used. However, one study which was a Phase 2 pharmacological intervention trial of phentermin/tompiramate [59] showed larger effect of the intervention. This drug combination was not tested in a Phase 3 study, and therefore, there are no data investigating consistency of this effect. Future studies with anti-obesity medications may be useful in investigating the effect of different pharmacological interventions.

Although the most commonly prescribed treatment for OSA is PAP therapy, its effect on body weight can be variable: several studies in a meta-analyses, have documented minor weight gain with PAP therapy [21]. The mechanism underlying this finding is unclear, but several possibilities have been proposed. First, fluid accumulation has been observed with PAP therapy in some studies, suggesting that the observed weight gain with PAP may not be related to adipose tissue [22]. Second, in clinical practice, some patients on PAP therapy will resume social activities that promote weight gain. Third, respiratory work in untreated sleep apnea may be substantial such that PAP may actually reduce nocturnal caloric expenditure [23]. Thus, treatment options for OSA directed at weight reduction may be an important strategy for OSA patients either as a primary approach or in combination with PAP therapy.

The mechanism underlying weight reduction–induced apnea improvement is also unclear [24,25]. OSA is now known to be a heterogeneous condition with multiple underlying pathophysiological mechanisms or endotypes [26,27]. Weight reduction likely improves pharyngeal mechanics based on measurements of critical closing pressure as a surrogate for airway collapsibility [28]. Weight reduction may yield improvements in control of breathing that are also thought to be important in OSA pathogenesis [29–32]. Moreover, body weight, particularly abdominal fat, can reduce the functional residual capacity (FRC) of the lung, and thus weight reduction may ameliorate the obesity-associated reduction in FRC [33]. Improved FRC may reduce desaturations but can also improve pharyngeal patency through caudal traction forces [34,35]. Regardless of the mechanism, weight loss is likely beneficial for people with OSA.

This meta-analysis is, based on our knowledge, the first to quantify the effect of weight reduction on sleep disordered breathing (measured by change of AHI) across full spectrum of weight interventions. Despite our study’s strengths, we acknowledge several limitations. First, we conducted a meta-analysis and thus are limited by the published literature. There were major variations between studies, such as different treatment protocols, co-interventions, confounders, outcome measures and follow-up duration. In general, studies are sparse and lacking mechanism, and thus we are supportive of further rigorous research in this area. Second, the magnitude of AHI reduction may be influenced by other factors that were not examined in the analysis. For example, it may be reasonable to expect lower AHI reduction with BMI reductions <30 kg/m², but because of the low number of cohorts achieving such weight reduction in selected studies we could not examine the relationship. Third, the published studies often lacked information on secondary outcomes such as hypoxic burden and/or patient reported outcomes. Thus, further work is clearly needed to fully quantify the benefits of weight reduction on OSA symptoms and the associated cardiometabolic risk. , Fourth some agents, including glucagon-like peptide-1 receptor agonists (GLP-1 RA’s) and GLP-1 RA/GIP agonists are relatively new [36]. Thus, published literature regarding these antiobesity medications in the context of OSA is lacking.

5. Conclusions

Weight reduction in people with obesity and OSA was associated with improvements in OSA severity as assessed by AHI. The degree of AHI improvement was associated with the magnitude of weight reduction. Based on the meta-regression, the effect of greater decrease in BMI was associated with smaller incremental benefit on sleep disordered breathing as measured by percent change in AHI. It is challenging to establish a precise relationship between percent change in AHI and percent weight reduction given the sparsity of data and the relatively few randomized clinical trials. Ultimately, weight management should be a standard part of care for patients with OSA and obesity or overweight. Optimal treatment of OSA and the associated complications may involve PAP therapy combined with effective weight loss strategies for patients with OSA and obesity or overweight.