Outcomes of Hypoglossal Nerve Stimulator Implant in Obstructive Sleep Apnea Patients With Various Health Statuses

ABSTRACT

Objectives

Hypoglossal nerve stimulation (HGNS) is increasingly used in the management of moderate‐to‐severe obstructive sleep apnea (OSA). In this study, we investigated whether common medical and psychiatric comorbidities are associated with apnea‐hypopnea index (AHI) reduction and treatment success after HGNS implantation.

Methods

Charts of patients who underwent HGNS implantation with pre‐operative AHI of 65 or below were reviewed. ANOVA, linear mixed‐effects model (LMM), and multivariate logistic regression analyses on AHI and demographics and medical comorbidities were performed on RStudio. Outcomes for logistic analysis pertained to whether they met the Sher20 criteria for treatment success.

Results

A total of 204 patients were included, of whom 174 were male and 30 were female. At the nominal 0.05 threshold, obesity was associated with higher AHI across time in the ANOVA analysis (F = 7.539, p = 0.0066) and with lower odds of Sher20 treatment success at 6 months (OR = 0.31, 95% CI: 0.10–0.90). In the small CAD subgroup (n = 16), CAD was associated with a lower longitudinal percent reduction in AHI in the linear mixed‐effects model (β = −31.88, p = 0.033). No other modeled covariates met the nominal significance threshold.

Conclusion

Obesity and CAD emerged as potential, rather than established, predictors of less favorable HGNS response. These findings are exploratory and require validation, particularly the CAD association, which was based on only 16 patients and arose in the setting of multiple unadjusted comparisons.

Level of Evidence

III.

This is a retrospective cohort study evaluating whether specific comorbidities affect response to HGNS. Our analysis revealed that patients with obesity may have less AHI reduction, and patients with coronary artery disease may be less likely to achieve therapeutic response.

1. Introduction

Obstructive sleep apnea (OSA) affects 34% of males and 17% of females in the United States [1]. Severity is classified using the apnea‐hypopnea index (AHI), measured in events per hour, with 5–15 events/h classified as mild, 15–30 as moderate, and ≥ 30 as severe [1, 2]. OSA is significantly associated with type 2 diabetes mellitus (T2DM), metabolic syndrome, obesity, hypertension, cardiovascular disease, and depression, and is classified as an independent risk factor for hypertension and coronary heart disease [3, 4, 5].

Several surgical and non‐surgical interventions have been developed to manage OSA. Treatment selection depends on guideline criteria, OSA severity, patient anatomy, comorbidity burden, and response to prior therapy. Non‐surgical methods include lifestyle modification, medications, positive airway pressure (PAP), oral appliances, and transcutaneous electrical nerve stimulators [6, 7] Hypoglossal nerve stimulation (HGNS) represents an upper airway stimulation therapy for selected patients with moderate‐to‐severe OSA who are unable to tolerate, adhere to, or achieve effective treatment with PAP.

HGNS therapy has demonstrated high efficacy, with substantial reductions in AHI from baseline and stable long‐term adherence above 80% [8, 9]. In the landmark STAR trial, upper airway stimulation produced significant improvements in AHI, oxygen desaturation index, sleepiness, and quality‐of‐life measures among selected patients with moderate‐to‐severe OSA [10]. However, early HGNS studies used strict inclusion criteria, excluding patients with certain comorbidities and body mass indexes greater than 32 kg/m². In 2023, the FDA broadened HGNS criteria to include patients with AHI up to 100 events/h and BMI up to 40 kg/m². Since this change, identifying patient‐specific predictors of HGNS response has become increasingly important, particularly among patients who were underrepresented in early studies.

Obesity is particularly relevant to HGNS outcomes because it may affect OSA through both anatomic and physiologic mechanisms. Increased BMI is associated with greater upper airway collapsibility, altered respiratory mechanics, and increased soft tissue burden around the pharynx and tongue [11]. Because HGNS improves airway patency by stimulating upper airway dilator musculature, obesity‐related changes in airway anatomy may influence the degree of AHI reduction after implantation. Prior studies evaluating BMI and HGNS outcomes have demonstrated mixed results, with some suggesting preserved treatment benefit in patients outside original STAR trial criteria and others showing less favorable outcomes at higher BMI [12, 13, 14].

Other common comorbidities may also influence HGNS outcomes through different mechanisms. Cardiometabolic conditions such as T2DM and coronary artery disease (CAD) are highly prevalent among patients with OSA and may reflect greater cumulative disease burden, autonomic dysfunction, metabolic dysregulation, and long‐term exposure to intermittent hypoxemia. These conditions may not directly alter neurostimulation itself, but may identify patients with more complex or less reversible sleep‐disordered breathing. Prior studies have also suggested that HGNS may be associated with improvements in secondary health outcomes, including mean arterial pressure, antihypertensive requirements, and A1C, further supporting the importance of evaluating cardiometabolic comorbidity in this population [15].

Psychiatric comorbidities may influence HGNS outcomes through separate pathways. Depression and anxiety are frequently seen in patients with OSA and may affect sleep quality, symptom perception, treatment engagement, and tolerance of nightly stimulation [16, 17]. Although these conditions are common in OSA, their relationship to objective AHI reduction after HGNS remains incompletely characterized. Gastroesophageal reflux disease was also evaluated as a common sleep‐disrupting comorbidity that may contribute to symptom burden in patients with OSA.

While OSA risk factors and other therapies are well studied, the literature on patient‐ and disease‐related predictors of HGNS outcomes remains limited. The comorbidities analyzed in this study were selected because they are common among patients with OSA, clinically relevant to HGNS candidacy and treatment response, and reliably documented in the electronic health record. In this study, we aimed to evaluate how patient‐specific factors and disease characteristics affect outcomes after HGNS therapy. Specifically, we investigated whether obesity, CAD, T2DM, anxiety, depression, gastroesophageal reflux disease, and patient sex were associated with baseline AHI, longitudinal AHI reduction, and treatment success after HGNS implantation.

2. Methods

2.1. Study Design

A retrospective chart review of patients who had been implanted with a hypoglossal nerve stimulation device at a single center from October 1, 2015 to August 31, 2024, was conducted. Two hundred and four patients with a diagnosis of moderate to severe OSA, and baseline AHI ≤ 65 events/h, and implanted with a HGNS were identified. Ethical approval was acquired through the institution's review board. A large proportion of patients were enrolled in the ADHERE registry, allowing for multiple postoperative polysomnography (PSG) reports across the follow‐up period. Patients' data was collected from Epic electronic medical records (Epic Hyperspace, Epic Systems Corporation, Madision, WI), and from Inspire SleepSync (Inspire Medical Systems Inc., Golden Valley, MN).

The patients' information from both sources was cross‐matched and then selected in accordance with the inclusion criteria. Only patients with baseline measurements (i.e., pre‐operative AHI, health, and demographic status collected prior to their surgery) and at least one postoperative follow‐up (collected ideally every 3, 6, and 12 months) were included.

2.2. Data Collection

Data was collected and stored using REDCap. Patient demographic and comorbidity data were extracted from the EHR. Comorbidities were defined via the entries in patient problem lists, and the active medications documented at time of surgery. The comorbidities included obesity, type 2 diabetes mellitus, anxiety, depression, gastroesophageal reflux disease (GERD), and CAD, given their established epidemiologic co‐occurrence with OSA. SleepSync data were collected from within the timeframe. Data gathered from SleepSync included information about baseline (preoperative) AHI, and post‐implantation AHI values at follow ups of 3, 6, and 12 months to profile treatment response over time. Data from both sources was cross‐matched and compiled in Redcap for analysis.

All patients underwent at least one sleep study prior to implantation, either in‐laboratory PSG or a home sleep test (HST), with hypopneas scored using a 4% oxygen desaturation threshold (pAHI). Preoperative baseline AHI was derived from PSG in 77% of patients and from HST in 23%. PSG values were preferred when both studies were available. For all PSG studies, whether pre or postoperative, the AHI used was the full‐night value; AHI obtained from split‐night or device‐titration was not used. The unattended AHI (uAHI) reported on HST likewise reflected whole‐night recording, allowing for consistent standard with both modalities.

Initial device voltage settings were determined postoperatively by the Inspire clinical team in accordance with Inspire protocol. Subsequent voltage titration at follow‐up visits, along with determining whether a patient had reached an adequate therapeutic outcome, was managed by the Inspire team based on patient‐reported feedback.

2.3. Statistical Methods

Statistical analysis was performed using RStudio (Posit team, 2025) and Microsoft Excel (Microsoft Office 365, version 2411). Descriptive analyses summarize demographic and disease‐related characteristics. Categorical variables were reported as proportions and percentages. Continuous variables such as age and BMI were presented as medians with interquartile ranges (IQR). Mean baseline AHI was compared across groupings of sex and individual comorbidities using independent samples t‐tests.

Repeated‐measures ANOVA was performed to assess whether comorbidities influenced AHI over time for obesity, anxiety, and depression. Other covariates were excluded from this analysis due to insufficient data at multiple postoperative timepoints. A linear mixed‐effects model (LMM) was used to evaluate the longitudinal trend of AHI from baseline through 12 months postoperatively, with adjusted effects for time, sex, and comorbidity status, and a random intercept for each patient.

Treatment success was defined using the Sher20 criteria, namely a ≥ 50% reduction in AHI from baseline in conjunction to a postoperative AHI < 20 events/h [18] Multivariable logistic regression was performed to assess predictors of responder status at each postoperative timepoint, with respect to the covariates. Odds ratios are reported with 95% confidence intervals.

3. Results

3.1. Patient Characteristics

In this study, 204 patients who underwent HGNS implantation were identified, and their EHRs were obtained. Within this cohort, mean BMI was 29.3 kg/m², mean age was 58.7 years, N = 174 patients were male, and N = 30 patients were female (Table 1). All EHRs had baseline AHI data collected pre‐operatively, N = 110 had AHI data collected 3 months postoperatively, N = 76 had AHI data collected 6 months postoperatively, and N = 54 had AHI data collected 12 months postoperatively.

TABLE 1.

Patient demographics.

	Male	Female
Sex	85.3% (174)	14.7% (30)
	Average
Age, years, median (IQR)	58.7 years (15.25)
BMI, mean (SD)	29.3 kg/m2 (3.53)
Comorbidities	Percent (N)
Obesity	41.2% (84)
Type 2 diabetes	15.2% (31)
Anxiety	10.8% (22)
Depression	19.1% (39)
Gastroesophageal reflux	28.9% (59)
Coronary artery disease	7.84% (16)
Smoking (tobacco)
Current	9.3% (19)
Former	26.5% (54)
Never	64.2% (131)
Alcohol
Current	65.2% (133)
Former	8.3% (17)
Never	26.5% (54)
AHI assessment modality	PSG	HST
Baseline	168 (77.1%)	50 (22.9%)
3 months post‐op	60 (54.5%)	50 (45.5%)
6 months post‐op	30 (40.5%)	44 (59.5%)
Sleep study at 12 months	18 (34.0%)	35 (66.0%)

At baseline, 77.1% of patients underwent PSG. For postoperative AHI assessment, PSG was performed in 54.5%, 40.5%, and 34.0% of patients at 3, 6, and 12 months, respectively, while HST was used in 45.5%, 59.5%, and 66.0% at the corresponding time points.

3.2. Impact of Sex and Co‐Morbidity Status on Baseline AHI

Baseline AHI was obtained for each patient in our cohort and analyzed with respect to gender and comorbidity status (Table 2). At the nominal, unadjusted 0.05 threshold, baseline AHI differed according to obesity status (p = 0.0045) and did not vary significantly with comorbidity status for sex, hypertension, type 2 diabetes, anxiety, depression, gastroesophageal reflux disease (GERD), or coronary artery disease (CAD).

TABLE 2.

Baseline AHI by demographics and co‐morbidities.

	Male	Female	p
Sex, mean AHI (SD)	39.1 (14.0)	33.4 (12.5)	0.0576
	Condition present	Condition not present	p
Obesity, mean AHI (SD)	42.1 (14.9)	35.3 (13.2)	0.0045
Type 2 diabetes, mean AHI (SD)	36.7 (14.3)	38.7 (13.6)	0.572
Anxiety, mean AHI (SD)	36.2 (13.7)	38.3 (13.7)	0.513
Depression, mean AHI (SD)	35.9 (14.1)	34.5 (12.3)	0.570
Coronary artery disease, mean AHI (SD)	33.6 (13.8)	35.8 (13.8)	0.54
Gastroesophageal reflux disease, mean AHI (SD)	34.1 (14.1)	36.9 (13.5)	0.19

3.3. Impact of Co‐Morbidity Status and Time on Treatment Response

Overall, the entire cohort of patients saw a 54.9% success rate according to Sher20 criteria. Male patients had a 58.6% rate, while female patients saw a 58.6% success rate. Success rates for the obesity, T2DM, CAD, anxiety, depression, and GERD patient subgroups had respective success rates of 49.4%, 48.1%, 42.9%, 50.0%, 54.1%, and 51.2% (Table 3). ANOVA was performed to evaluate whether patient sex or co‐morbidity status could influence treatment response over time (Table 4). Although obesity status significantly impacted treatment response (F = 7.539, p = 0.0066), there was no significant interaction between obesity status and time (F = 0.499). Patient sex, anxiety, and depression had no significant effect on treatment response over time according to ANOVA.

TABLE 3.

Sher20 treatment success and mean AHI percent reduction by demographic and comorbidity subgroup. Comorbidity rows reflect patients with the condition present.

	Overall latest	3 months	6 months	12 months
Sher20 success, n/N (%)
Full cohort	95/173 (54.9%)	68/110 (61.8%)	31/76 (40.8%)	32/55 (58.2%)
Male	78/144 (54.2%)	60/96 (62.5%)	24/60 (40.0%)	27/45 (60.0%)
Female	17/29 (58.6%)	8/14 (57.1%)	7/16 (43.8%)	5/10 (50.0%)
Obesity	41/83 (49.4%)	30/54 (55.6%)	10/33 (30.3%)	15/26 (57.7%)
T2DM	13/27 (48.1%)	10/17 (58.8%)	4/10 (40.0%)	6/9 (66.7%)
CAD	6/14 (42.9%)	5/9 (55.6%)	1/6 (16.7%)	1/3 (33.3%)
Anxiety	10/20 (50.0%)	10/17 (58.8%)	1/4 (25.0%)	3/6 (50.0%)
Depression	20/37 (54.1%)	14/25 (56.0%)	7/16 (43.8%)	8/12 (66.7%)
GERD	22/43 (51.2%)	18/25 (72.0%)	8/25 (32.0%)	8/13 (61.5%)
Mean AHI percent reduction
Male	49.2%	55.2%	38.2%	46.6%
Female	46.2%	48.3%	39.5%	7.6%
Obesity	46.8%	52.2%	33.6%	39.3%
T2DM	37.6%	48.9%	24.1%	43.2%
CAD	28.6%	17.2%	−5.7%	30.2%
Anxiety	53.9%	59.3%	21.3%	48.6%
Depression	52.1%	56.1%	41.0%	56.6%
GERD	42.7%	64.9%	28.4%	34.8%

Abbreviations: CAD, coronary artery disease; GERD, gastroesophageal reflux disease; T2DM, type 2 diabetes mellitus.

TABLE 4.

ANOVA of AHI over time by demographic and co‐morbidity.

Effect	F	p
Gender
Time	1.037	0.356
Gender	0.130	0.718
Time × gender	0.603	0.548
Obesity
Time	1.073	0.344
Obesity	7.539	0.0066
Time × obesity	0.499	0.608
Anxiety
Time	1.044	0.354
Anxiety	1.251	0.265
Time × anxiety	0.787	0.457
Depression
Time	1.053	0.351
Depression	0.481	0.489
Time × depression	2.074	0.128

3.4. Predicted Treatment Response According to Comorbidity Status

To evaluate treatment success, defined based on Sher20 criteria as a ≥ 50% reduction in AHI (responder status) plus a posttreatment AHI < 20, multivariable logistic regression was performed at various time intervals (Table 5). The odds ratios for achieving responder status across the 3‐, 6‐, and 12‐month time points are visually summarized in a forest plot (Figure 1). At 3 months, none of the covariates demonstrated associations with likelihood of surgical success. At 6 months, obesity met the nominal threshold for an association with lower odds of treatment success (OR = 0.31, 95% CI: 0.10–0.90). At 12 months, none of the covariates were significantly associated with surgical success. However, several variables demonstrated large effect estimates with wide confidence intervals, including T2DM (OR = 3.59, 95% CI: 0.45–28.47), CAD (OR = 0.14, 95% CI: 0.007–2.87), anxiety (OR = 0.23, 95% CI: 0.026–2.09), and depression (OR = 5.05, 95% CI: 0.73–34.99).

TABLE 5.

Multivariate logistic regression of responder status at 3, 6, and 12 months postoperatively.

Predictor	OR	95% CI
3 months
Sex (male)	1.06	0.30–3.71
Obesity	0.56	0.24–1.30
T2DM	0.82	0.27–2.49
CAD	0.52	0.12–2.29
Anxiety	0.96	0.26–3.58
Depression	0.70	0.23–2.12
GERD	2.03	0.72–5.67
6 months
Sex (male)	1.88	0.47–7.49
Obesity	0.31	0.10–0.90
T2DM	1.91	0.34–10.82
CAD	0.12	0.01–1.50
Anxiety	0.28	0.22–3.72
Depression	1.78	0.46–6.90
GERD	0.39	0.13–1.19
12 months
Sex (male)	2.89	0.56–14.81
Obesity	0.98	0.30–3.18
T2DM	3.59	0.45–28.47
CAD	0.14	0.007–2.87
Anxiety	0.23	0.02–2.09
Depression	5.05	0.73–34.99
GERD	0.90	0.21–3.77

Abbreviations: 95% CI, 95% confidence intervals; OR, odds ratio.

FIGURE 1.

Forest plot of odds ratios from multivariate logistic regression for ≥ 50% AHI reduction at 3, 6, and 12 months postoperatively. Error bars represent 95% confidence intervals. Dashed line indicates OR = 1 (no effect).

Linear mixed‐effects models were used to evaluate the association between patient covariates and longitudinal changes in percent reduction in AHI over time (Table 6). The beta coefficients and their 95% confidence intervals are visually summarized in a forest plot (Figure 2). In the linear mixed‐effects model, CAD met the nominal 0.05 threshold for an association with a lower longitudinal percent reduction in AHI (β = −31.88, p = 0.033). Only 16 patients in the cohort had CAD. Among the covariates analyzed, coronary artery disease was significantly associated with a lower percent reduction in AHI (β = −31.88, p = 0.029). Patient sex, obesity, type 2 diabetes, anxiety, depression, and GERD were not significantly associated with longitudinal changes in AHI.

TABLE 6.

Linear mixed‐effects model of longitudinal percent changes in AHI.

Predictor	β	p
Sex (male)	11.77	0.26
Obesity	−6.70	0.39
T2DM	0.09	0.99
CAD	−31.88	0.033
Anxiety	1.69	0.90
Depression	3.53	0.73
GERD	−3.47	0.69

Abbreviation: β, beta coefficient.

FIGURE 2.

Forest plot of beta coefficients from a linear mixed model for percent AHI improvement. Error bars represent 95% confidence intervals. Dashed line indicates beta coefficient = 0 (no effect).

4. Discussion

In this retrospective cohort of patients, who underwent hypoglossal nerve stimulator implantation for moderate‐to‐severe obstructive sleep apnea, treatment response varied according to select demographic and co‐morbid illness factors. Although obesity was associated with significantly higher AHI values across all time points, the absence of a significant time‐by‐obesity interaction suggests the trajectory of AHI improvement is comparable to that of non‐obese counterparts. Multivariate logistic regression modeling of treatment response defined using Sher20 criteria (≥ 50% reduction + < 20 posttreatment AHI), revealed that obesity was associated with significantly lower odds of achieving therapeutic response at 6 months postoperatively. Patient sex, type 2 diabetes, depression, anxiety and GERD demonstrated no significant effects on response to treatment. These findings suggest that while HGNS appears effective across all patients in our study, specific patient characteristics and co‐morbidities may influence severity of OSA and impact therapeutic response to HGNS.

Obesity was associated with significantly higher AHI values at baseline, and across follow‐up visits, as well as odds of response to HGNS at specific timepoints. These results suggest obesity may impact overall OSA severity and AHI, as well as response likelihood. Landmark studies in HGNS implantation, such as the STAR trial from 2014, excluded patients with BMIs exceeding 32 kg/m², limiting early evaluation of treatment efficacy in patients with obesity [10]. Subsequent literature has yielded mixed results, with prior studies demonstrating poorer surgical success in patients with higher BMI, and others reporting preserved treatment effectiveness in patients with BMIs > 32 kg/m² [12, 13, 14]. Our findings raise the possibility that obesity is associated with greater residual AHI and lower odds of Sher20 treatment success at 6 months; however, they do not establish obesity as a determinant of HGNS response. The loss of this association at 12 months is worth noting. The cohort size contracted from 76 patients to 54 patients, and the obesity odds ratio shifted from 0.31 (95% CI: 0.10–0.90) at 6 months towards the null (OR 0.98, 95% CI: 0.30–3.18), at 12 months. This pattern is consistent with diminished power and selective attrition, rather than a true change in effect (i.e., 12‐month analysis, neither confirms nor refutes 6‐month finding). If obese non‐responders were preferentially lost to follow‐up, pursued adjunctive weight reduction, or had devices reprogrammed, the subgroup retained would be enriched for responders, biasing the estimate towards null. The rising use of HST at later timepoints, from 59.5% to 66% of studies, may also contribute to this, as HST underestimates AHI (relative to PSG), and a greater reliance on HST would disproportionately shift higher‐AHI obese patients below the Sher20 threshold. Obesity also met nominal significance thresholds in the baseline and ANOVA analyses. Although the consistency of these observations is suggestive, the analyses were not adjusted for multiplicity and remain susceptible to type I error. The 6‐month finding should therefore be interpreted as hypothesis‐generating rather than as evidence of a stable or causal effect of obesity on HGNS response. Among the 16 patients with CAD, the linear mixed‐effects model yielded an association between CAD and a lower longitudinal percent reduction in AHI following HGNS implantation (β = −31.88; nominal p = 0.033). This finding should be considered exploratory. The CAD subgroup was small, the time‐specific logistic regression models did not meet the nominal 0.05 significance threshold for CAD, and multiple comorbidities, outcomes, and postoperative time points were examined without formal adjustment for multiplicity. Consequently, the overall risk of type I error is increased, and the observed CAD association may represent a chance finding. CAD should therefore be regarded only as a potential predictor of HGNS response pending confirmation in an independent, adequately powered cohort.

If this association is replicated, one possible explanation is that CAD serves as a marker of broader cardiopulmonary or nonanatomic disease burden rather than directly modifying the efficacy of HGNS. The established relationship between OSA and CAD, including intermittent hypoxemia and sympathetic activation, provides biologic plausibility for this hypothesis [5, 18]. However, this mechanistic explanation remains speculative. The present findings do not justify using CAD to determine HGNS candidacy or to provide individualized prognostic counseling. Larger prospective studies with prespecified analyses and adequate representation of patients with CAD are needed to determine whether this exploratory association is reproducible. In our multivariate logistic regression model, male sex was not associated with higher differential odds of achieving treatment success, suggesting that sex‐related differences may not have clinically meaningful influences on HGNS response likelihood. This finding is inconsistent with prior literature demonstrating female sex was associated with significantly higher odds of favorable treatment response [19]. Furthermore, prior literature has demonstrated differing patterns in airway collapsibility, alongside symptom presentation, possibly leading to improved response to HGNS implantation in comparison to male patients [20, 21]. However, interpreting this data warrants caution, considering our cohort's disproportionately male predominance, and relatively small number of female patients. Further investigation of these differences in larger, more gender‐balanced cohorts is justified to substantiate this finding.

Other factors demonstrated possible influences on the likelihood of therapeutic response to HGNS. Large odds ratios were observed among patients with depression and anxiety. While not statistically significant, depression was associated with higher odds of achieving treatment success, while anxiety was conversely associated with a lower likelihood of treatment success. Although these findings did not reach thresholds for significance, they raise the possibility that psychiatric comorbidities may differentially influence treatment outcomes. Few studies have examined the impacts of mood disorders on treatment success; however, higher GAD‐7 and PHQ‐9 scores have been demonstrably higher in cohorts with low adherence to HGNS [22] Thus, treatment engagement could perhaps explain any observed differences in AHI reduction; however, adherence was not measured; thus, this remains speculative. Furthermore, given the non‐significant results from this cohort, these results may be interpreted as hypothesis‐generating rather than definitive. Larger prospective studies may be warranted to elucidate whether a relationship between mood disorders and treatment success indeed exists.

Similarly, a large odds ratio was observed for type 2 diabetes in multivariate logistic regression; however, it did not reach statistical significance. The magnitude of the estimated effect likely relates to our study's limited sample size and potential statistical instability within subgroup estimates. Associations have been found between metabolic dysfunction and increased airway collapsibility, neuropathy and abnormal neural upper airway control, thereby perhaps underlying the observed differences in our cohort [23, 24]. However, given the wide confidence intervals, these findings are, at best, to be viewed as exploratory. Therefore, studies with greater representation of patients with diabetes are necessary to examine whether metabolic dysfunction meaningfully impacts HGNS success.

This study has several limitations. As with any retrospective chart review, the analyses relied on the validity and completeness of patient charts in EMR, and this may result in misclassification of patients and their comorbidities. Additionally, the relatively small cohort size limited statistical power, particularly for subgroup analyses. Because multiple comorbidities were evaluated across several analytic models, outcomes, and postoperative time points without formal adjustment for multiplicity, all p values are nominal and the overall risk of one or more type I errors is increased. The obesity and CAD findings should therefore be considered exploratory and hypothesis‐generating. This concern is particularly important for CAD, which was present in only 16 patients, making the estimate potentially imprecise and unstable. Furthermore, the patient cohort was overwhelmingly male, restricting generalizability, and disproportionately non‐obese in the context of obstructive sleep apnea. Follow‐up AHI data were not uniformly available across all time points, potentially introducing attrition bias. Additionally, a progressive shift towards HST at later time points—from 23% at baseline to 66% at 12 months—may have led to underestimation of AHI in postoperative follow‐up, given HST's tendency to underestimate AHI relative to in‐laboratory PSG. Finally, device adherence data were not systematically captured. Although usage metrics are recorded by the SleepSync platform, these were not uniformly retrievable across the cohort, as data from older‐generation devices were stored locally on individual programmers rather than synchronized with the cloud. This prevented consistent extraction across the study period, and consequently, could not reliably disentangle whether differences in AHI reduction reflect true treatment non‐response or suboptimal device use. Ultimately, future multicenter studies with larger and more diverse populations are necessary to validate these findings and refine patient selection criteria for HGNS therapy.

5. Conclusion

In this retrospective cohort, obesity and CAD emerged as potential predictors of less favorable HGNS outcomes. These associations should be considered exploratory as multiple comparisons were performed without formal adjustment; follow‐up data were incomplete; and only 16 patients had CAD. Larger prospective studies with prespecified hypotheses are needed before either factor is used to guide HGNS candidate selection or patient counseling.

Funding

The authors have nothing to report.

Disclosure

AI assistance was used to improve grammar and sentence structure. No content was generated using AI.

Conflicts of Interest

A.K. previously was involved in Inspire clinical trials, has interests in Regeneron pharmaceuticals, and is an owner of Valam. The other authors declare no financial disclosures or conflicts of interest.

Saeed A., Biskaduros A., Wilmerding N., and Kacker A., “Outcomes of Hypoglossal Nerve Stimulator Implant in Obstructive Sleep Apnea Patients With Various Health Statuses,” Laryngoscope Investigative Otolaryngology 11, no. 4 (2026): e70541, 10.1002/lio2.70541.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.