How does blood sugar level affect wound infection rates and device revision and removal rates after hypoglossal nerve stimulation in obstructive sleep apnea patients? A large database study

Abstract

Purpose

This study aims to evaluate the impact of perioperative hyperglycemia on wound infection, device revision, reimplantation and removal rates after hypoglossal nerve stimulation, utilizing a large-scale and federated, real-world database.

Methods

A retrospective cohort study was conducted using the TriNetX Network on September 26th, 2025. Patients with OSA who underwent HGNS implantation were classified into the hyperglycemic cohort if their laboratory glucose values exceeded the predefined thresholds (126 mg/dL and 154 mg/dL) during the six months before or after implantation. Otherwise, they were in non-glycemic cohort. Postoperative wound infection and device reoperation rates were assessed at one-year and three-year follow-up. Outcomes were analyzed using incidence rates, odds ratios (ORs), and Log-Rank tests, with propensity score matching (PSM) employed to control for confounding variables.

Results

Patients with serum glucose levels ≥ 126 mg/dL and ≥ 154 mg/dL exhibited significantly higher incidence rates of surgery-related wound infections compared to their respective counterparts (one-year OR: 1.56 and 1.82; three-year OR: 1.55 and 1.74; all p ≤ 0.01). These findings remained statistically significant after PSM (one-year OR: 1.51 and 2.08; three-year OR: 1.40 and 1.63; all p < 0.05). However, there were no significant differences, except 3 year OR after PSM, in device revision, replacement or removal rates at both one-year and three-year follow-ups (p > 0.05).

Conclusions

Hyperglycemia is associated with significantly increased risk of wound infection, including long term infection following HGNS implantation, while device reoperation rates remain unaffected. These findings underscore the importance of stringent perioperative glycemic control to mitigate postoperative infection risk in HGNS recipients.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11325-026-03587-1.

Text

Introduction

Obstructive sleep apnea (OSA) is a chronic breathing disorder characterized by repeated collapse of the upper airway during sleep. Untreated OSA has been proven to be associated with systemic comorbidities ranging from cardiovascular diseases, metabolic disorders, respiratory comorbidities, neuropsychiatric impairments, and even malignancies due to chronic intermittent hypoxia, sleep fragmentation, and increased sympathetic activity during sleep [1, 2].

After approval by the United States Food and Drug Administration in 2014, hypoglossal nerve stimulation (HGNS) has been proven as an effective and safe surgical intervention in mitigating OSA burden for select patients. Over the past 5 years, rates of hypoglossal nerve stimulation therapy have increased within the US, from 2,813 implants in 2019, to 22,859 implants in 2023, with the accompanying rise of adverse event reports from 92 in 2019 to 414 in 2023 [3]. Similar to cardiac implantable electronic devices (CIEDs), HGNS systems, as long-term implanted foreign bodies in the chest, carry an intrinsic risk of infection. While the risk factors for infections associated with CIEDs have been extensively explored, large-scale, population-based studies evaluating the infection risk factors unique to HGNS remain scarce, highlighting the need for further investigation. Notably, some complications may not manifest immediately after implantation, making detection and reporting challenging. A surgical wound infection was reported to have occurred 55 months after HGNS implantation [4]. A systemic review of clinical trials showed 7 explantation events at 60-month follow-up [5]. Consequently, it is imperative to emphasize stringent patient selection criteria and implement structured, long-term follow-up protocols.

Hyperglycemic status is widely recognized as a risk factor of wound infection through multiple physiological mechanisms such as immunity impairment, decreased angiogenesis and fibroblast proliferation, and disrupted keratinocyte migration, which hinder normal tissue repair [6]. Although there have been no clinical guidelines on glucose control in HGNS recipients, a certain period of strict perioperative glucose control is deemed imperative. However, OSA is a well-established cause of metabolic disarray including insulin resistance, and the clinical reality is that many of these patients undergo surgery with sub-optimal control. HNGS has been shown to improve metabolic outcomes, including glycemic control [7]. For patients who failed to comply with or see improvements from OSA treatments other than HGNS, strict glycemic control is sometimes challenging and may delay HGNS as proper treatment [8].

To our knowledge, the role of hyperglycemia in post-HGNS complications has yet to be investigated. We herein hypothesize higher glucose levels will raise post-HGNS infections in both the short and long term. Accordingly, HGNS device reoperation rates will also be investigated. This study aims to analyze how different blood glucose levels affect the perioperative morbidities of HGNS and provide a clinical reference for establishing practicable perioperative glucose control protocols.

Materials and methods

This retrospective cohort study was designed to analyze the risk of post-HGNS wound infection and device reoperations (revision, replacement, and removal) in obstructive sleep apnea patients using a global federated medical network, TriNetX. Data were retrieved using a structured query language to extract patients’ demographic characteristics, diagnoses, laboratory results, and medical procedures. When the outcome events turn out to be single digit, they were automatically rounded as “ ≤ 10” to protect patient confidentiality, and the system will be unable to display detailed analytic data accordingly.

The query criteria were set to include the following: First, age at index event range from 18 to 90 years between 2014 and 2025. Secondly, diagnoses related to obstructive sleep apnea (G47.30 and G47.33). And finally, had received implantation of a hypoglossal nerve neurostimulator as our index event, by the correlated Current Procedural Terminology (CPT) and International Classification of Diseases 10th Revision, Procedure Coding System (ICD-10-PCS) codes (Table 1). Patients who took immunosuppressants were excluded due to the potential confounding effects on infection. Patients with epilepsy and recurrent seizures (G40) in the subgroup 01HY0MZ and 0JH60BZ were also excluded to avoid the unwanted involvement of vagus nerve neurostimulator implantation as a refractory or drug-resistant treatment. We chose glucose level from serum, plasma or blood (shortened as “blood glucose”) as our independent variables. We ran four analyses for the same study population with 2 laboratory thresholds (≥ 126 or ≥ 154 mg/dL of glucose level) and 2 outcome endpoints (post-implantation 1 and 3 years). Patients were categorized into hyperglycemic arm if they had at least one laboratory data above the designated threshold 6 months before and after the index event (defined as the surveillance window), and vice versa. Outcome variables were defined by the CPT and ICD-10-PCS codes to include procedure-related wound infection events, device removal/replacement or revision events, and device removal events only, within 1 and 3 years after the HGNS procedure (Table 1). To enhance the robustness of evaluating hyperglycemic effects, an additional analysis was performed in the same setting, incorporating HbA1c ≥ 7% as a predefined variable. Patients with outcomes before the HGNS procedure were excluded. Age, sex, overweight and obesity, dermatitis and eczema, and nicotine dependence were considered potential confounding factors with the outcomes of interest.

Table 1.

ICD-10-CM, ICD-10-PCS and CPT codes applied in query

	ICD-10-CM, ICD-10-PCS, or CPT codes		Time window
Inclusion criteria
Obstructive sleep apnea	G47.30 G47.33		Anytime
Hypoglossal nerve stimulation therapy	64,568 AND 0466 T 64,582
	01HY0MZ 0JH60BZ AND cannot have G40, to avoid unwanted inclusion of vagus nerve stimulation system
Exclusion criteria
Immune suppressants	L04	Anytime
Outcomes
Postoperative wound infection	T81.3 T81.30XA T81.31XD T81.4 T81.49XA T85.732 T85.734 T85.738 T85.79XA L08.89 L08.9		1 or 3 years after HGNS therapy
Device revision/replacement/ removal	61,886 61,888 64,569 64,570 64,583 64,584 0467 T 0468 T
Device removal only	64,570 64,584 0468 T

Data analysis

After the total numbers and patients with outcomes of each cohort were collected, TriNetX autogenerated a report of risk difference, risk ratio and odds ratio (OR) along with bar charts. Censored survival analyses and Log-rank test were also conducted. To control confounding factors, propensity score matching (PSM) was also performed. The level of significance was set to p-value < 0.05.

Results

Since the database was updated continuously, and the responding healthcare organizations (HCOs) varied with time, the total number fluctuated minorly with different timepoints. On September 26th, 2025, a maximum of 9,977 eligible patients were collected among the cohorts (Fig. 1). The racial distribution was as follows: 83.3% white, 6.8% unknown, 4.7% Black or African American, 1.5% Asian, 0.4% American Indian, Alaska Native or Pacific islanders, and 3.3% other races (Fig. 2). The overall infection rates in 1- and 3- years were 2.1% and 3.2%; overall device revision, replacement or removal rates in 1- and 3- years were 2.0% and 3.2%. The glucose < 126 mg/dL cohort included 7,567 patients (mean age: 60.6 ± 12.2 years, 63.0% male), while the glucose ≥ 126 mg/dL cohort had 2,410 patients (mean age: 64.2 ± 10.6 years, 64.7% male). There were 29 patients with unknown sex (Table 2). Similar demographic characteristics distributions were also shown when categorizing by serum glucose at ≥ 154 mg/dL. From the characteristics, significant differences in age and several comorbidities could be observed between these cohort groups.

Fig.1.

Flow Diagram of Patient Inclusion (with glucose threshold setting at 126 mg/dL)

Fig.2.

Racial distribution of the study participants

Table 2.

Cohort 1 (blood glucose level < 126 mg/dL) and cohort 2 (blood glucose level ≥ 126 mg/dL) characteristics before propensity score matching

Demographics
Cohort			Mean ± SD		Patients	% of Cohort	P-Value			Std diff
1 2	AI	Age at Index	60.6 ± 12.2 64.2 ± 10.6		7,567 2,410	100% 100%	< 0.001	0.314
1 2	F	Female			2,715 851	35.8% 34.4%	0.203	0.030
1 2	M	Male			4,780 1,602	63.0% 64.7%	0.126	0.036
Diagnosis
1 2	E08-E13	Diabetes mellitus			881 1,577	11.6% 63.7%	< 0.001		1.274
1 2	G89.2	Chronic pain, not elsewhere classified			1,610 832	21.2% 33.6%	< 0.001		0.280
1 2	G62.9	Polyneuropathy, unspecified			400 285	5.3% 11.5%	< 0.001		0.226
1 2	F17	Nicotine dependence			915 427	12.1% 17.2%	< 0.001		0.147
1 2	R11	Nausea and vomiting			1023 553	13.5% 22.3%	< 0.001		0.232
1 2	E66	Overweight and obesity			2,611 1,249	34.4% 50.4%	< 0.001		0.329
1 2	L20-L30	Dermatitis and eczema			1020 475	13.4% 19.2%	< 0.001		0.156
1 2	D10-D36	Benign neoplasms, except benign neuroendocrine tumors			1,747 790	23.0% 31.9%	< 0.001		0.200
1 2	D60-D64	Aplastic and other anemias and other bone marrow failure syndromes			825 549	10.9% 22.2%	< 0.001		0.308

Before matching, the median follow-up time in 1- and 3-years outcomes from all cohorts were 365 and 608–643 days, respectively. Patients with at least one serum glucose levels ≥ 126 mg/dL during the surveillance window exhibited significantly higher incidence rates of surgery-related wound infections compared to their respective counterparts (one-year OR: 1.56, 95%CI 1.16–2.09, p = 0.004; three-year OR: 1.55, 95%CI 1.21–1.98, p < 0.001). Similarly, patients with at least one serum glucose level ≥ 154 mg/dL during the surveillance window revealed significantly higher incidence rates of surgery-related wound infections (one-year OR: 1.82, 95%CI 1.30–2.55, p < 0.001; three-year OR: 1.74, 95%CI 1.31–2.31, p < 0.001). Additionally, we performed a survival Log-Rank test, and all cohorts were statistically significant in both 1- and 3-years outcomes (all p < 0.05). However, there was only one significant difference in device revision, replacement, or removal rates between hyperglycemic and non-hyperglycemic cohorts at threshold glucose 154 mg/dL, 3-years follow-up (p = 0.038). As we analyzed only device removal, no significant differences were presented. The cohorts with outcome less than 10 disabled further statistical analysis. The details were summarized in Table 3 and Fig. 3.

Table 3.

Risk analyses between the hyperglycemic and non-hyperglycemic cohorts before propensity score matching

	Glucose level (mg/dL)	< 126	126 ~ 500	< 154	154 ~ 500
Wound infection events	1 year events/total patients (risk ratio)	138/7242 (0.019)	66/2250 (0.029)	160/8203 (0.020)	45/1289 (0.035)
	odds ratio (95% CI) p-value	1.56 (1.16, 2.09) 0.004		1.82 (1.30, 2.55)  < 0.001
	Log-Rank test Hazard ratio (95% CI)	p = 0.005 1.51 (1.13, 2.03)		p < 0.001 1.78 (1.28, 2.48)
	3 year events/total patients (risk ratio)	203/7242 (0.028)	96/2250 (0.043)	236/8203 (0.029)	63/1289 (0.049)
	odds ratio (95% CI) p-value	1.55 (1.21, 1.98)  < 0.001		1.74 (1.31, 2.31)  < 0.001
	Log-Rank test Hazard ratio (95% CI)	p = 0.001 1.49 (1.17, 1.90)		p < 0.001 1.67 (1.27, 2.21)
Device revision/ replacement/ removal events	1 year events/total patients (risk ratio)	144/7516 (0.019)	52/2439 (0.021)	164/8536 (0.019)	30/1416 (0.021)
	odds ratio (95% CI) p-value	1.12 (0.81, 1.54) 0.506		1.11 (0.75, 1.64) 0.619
	3 year events/total patients (risk ratio)	233/7516 (0.031)	87/2439 (0.036)	260/8536 (0.030)	58/1416 (0.041)
	odds ratio (95% CI) p-value	1.16 (0.90, 1.49) 0.257		1.36 (1.02, 1.82) 0.038
Device removal only	1 year events/total patients (risk ratio)	30/7588 (0.004)	12/2474 (0.005)	*
	3 year events/total patients (risk ratio)	66/7588 (0.009)	24/2474 (0.010)	72/8627 (0.008)	18/1433 (0.013)
	odds ratio (95% CI) p-value	1.12 (0.70- 1.79) 0.647		1.51 (0.90- 2.54) 0.119

^*the patient count is too small to display detailed results

Fig.3.

One-year infection-free survival curves for wound infection in cohorts stratified by glucose thresholds of 126 mg/dL and 154 mg/dL before propensity score matching

To minimize the effect from cofounding factors and retain reasonable patient numbers, we selectively used age, sex, overweight and obesity, dermatitis and eczema, and nicotine dependence for PSM. The after-matching demographics and diagnoses between hyperglycemia and non-hyperglycemia groups are shown in Table 4. After matching, the median follow-up time in 1- and 3-years outcomes from all cohorts were 365 and 629–643 days, respectively. Significantly higher 1- and 3-years incidence rates of surgery-related wound infections were still present in cohorts with glucose level above 126 (one-year OR: 1.51, 95%CI 1.03–2.21, p = 0.035; three-year OR: 1.40, 95%CI 1.03–1.92, p = 0.033), and 154 mg/dL (one-year OR: 2.08, 95%CI 1.25–3.45, p = 0.005; three-year OR: 1.63, 95%CI 1.09–2.44, p = 0.017). The Log-Rank test showed consistent results (Table 5, Supplementary Fig. 1). Regarding device revision, replacement, and removal, only the cohort with glucose level exceeding 154 mg/dL exhibited a statistically significant OR 1.69 (95%CI 1.10–2.58, p = 0.016) in 3 years follow-up. Similar to the before matching results, there is no significant difference in device removal only, with most of the outcome events being too small to be analyzable (Table 5).

Table 4.

Cohort 1 (blood glucose level < 126 mg/dL) and cohort 2 (blood glucose level ≥ 126 mg/dL) characteristics after propensity score matching

Demographics
Cohort			Mean ± SD	Patients	% of Cohort	P-Value	Std diff
1 2	AI	Age at Index	64.5 ± 10.3 64.2 ± 10.5	2,462 2,462	100% 100%	0.440	0.022
1 2	F	Female		818 847	33.2% 34.4%	0.382	0.025
1 2	M	Male		1,624 1,591	66.0% 64.6%	0.323	0.028
Diagnosis
1 2	E08-E13	Diabetes mellitus		367 1,565	14.9% 63.6%	< 0.001	1.149
1 2	G89.2	Chronic pain, not elsewhere classified		639 822	26.0% 33.4%	< 0.001	0.163
1 2	G62.9	Polyneuropathy, unspecified		182 282	7.4% 11.5%	< 0.001	0.139
1 2	F17	Nicotine dependence		385 414	15.6% 16.8%	0.262	0.032
1 2	R11	Nausea and vomiting		358 547	14.5% 22.2%	< 0.001	0.199
1 2	E66	Overweight and obesity		1,222 1,234	49.6% 50.1%	0.732	0.010
1 2	L20-L30	Dermatitis and eczema		453 463	18.4% 18.8%	0.714	0.010
1 2	D10-D36	Benign neoplasms, except benign neuroendocrine tumors		697 780	28.3% 31.7%	0.010	0.074
1 2	D60-D64	Aplastic and other anemias and other bone marrow failure syndromes		345 541	14.0% 22.0%	< 0.001	0.208

Table 5.

Risk analyses between the hyperglycemic and non-hyperglycemic cohorts after propensity score matching

	Glucose level (mg/dL)	< 126	126 ~ 500	< 154	154 ~ 500
Wound infection events	1 year events/total patients (risk ratio)	46/2327 (0.020)	52/2238 (0.029)	23/1342 (0.017)	45/1288 (0.035)
	odds ratio (95% CI) p-value	1.51/(1.03, 2.21) 0.035		2.08/(1.25, 3.45) 0.005
	Log-Rank test Hazard ratio (95% CI)	p = 0.037 1.49 (1.02, 2.17)		p = 0.004 2.05 (1.24, 3.39)
	3 year events/total patients (risk ratio)	72/2327 (0.031)	96/2238 (0.043)	41/1342 (0.030)	63/1288 (0.049)
	odds ratio (95% CI) p-value	1.40 (1.03, 1.92) 0.033		1.63 (1.09, 2.44) 0.017
	Log-Rank test Hazard ratio (95% CI)	p = 0.672 1.37 (1.01, 1.87)		p = 0.025 1.56 (1.05, 2.32)
Device revision/ replacement/ removal events	1 year events/total patients (risk ratio)	39/2433 (0.016)	52/2424 (0.021)	22/1417 (0.016)	30/1415 (0.021)
	odds ratio (95% CI) p-value	1.35 (0.89, 2.05) 0.165		1.37 (0.79, 2.39) 0.2633
	3 year events/total patients (risk ratio)	65/2433 (0.027)	87/2424 (0.036)	35/1417 (0.025)	58/1415 (0.041)
	odds ratio (95% CI) p-value	1.36 (0.98, 1.88) 0.067		1.69 (1.10, 2.58) 0.016
Device removal only	1 year events/total patients (risk ratio)	*		*
	3 year events/total patients (risk ratio)	19/2461 (0.008)	24/2459 (0.010)	*
	odds ratio (95% CI) p-value	1.27 (0.69, 2.32) 0.443		*

^*the patient count is too small to display detailed results

In the additional analysis using HbA1c level as a variable, eligible outcome events for both wound infection and device revision/removal were obtainable only when the threshold was set at 7% with a 3-year follow-up. Before propensity score matching, this analysis yielded an OR of 1.7 (95% CI: 1.08–2.69, p = 0.023); however, no statistically significant differences were observed after matching (supplementary Table 1).

Discussion

By leveraging the large, real-time, federated health record network, this study demonstrates that analyzable data on this uncommon complication can still be obtained. From an eligible population of nearly 10,000 patients, we found that those with hyperglycemic status within 6 months before and after hypoglossal nerve stimulation surgery had significantly higher wound infection rates. However, despite this increase, there were no significant differences in device revision, replacement or removal. Admittedly, the small number of device removal events limited the strength of our conclusions. These suggest that many wound infections can be managed non-surgically and that reoperations were often attributable to factors other than wound infections.

The characteristics of our study population showed a roughly 1.8 to 1 male-to-female ratio, a mean age of around 61.6 years, and primarily Caucasian patients (83.3%). This was similar to an observational study conducted in 2024 that analyzed longitudinal data from an Optum database which revealed consistent results in their HGNS population, with 67.1% male, a mean age of 62.0 years, and a significant HGNS percentage of Caucasian patients (81.2%) [9]. Consistency from two different databases implies good representations of real-world data in the United States highlighting current inequities in access to OSA treatment.

The overall infection rates (1- and 3- years: 2.1% and 3.2%) shown in our study appeared to be high among the reported literature, namely, 0.55% by Bentan and Nord, 0.9% by Lorenz and Goyal, 1.4% by Ali et al., and 3.8% by Chieffe and colleagues [3, 10–12]. One potential explanation is the use of certain CPT and ICD-10-PCS codes to collect outcomes that are not HGNS-specific, potentially including wound infection events unrelated to the procedure itself. Additionally, publication bias and underreporting of adverse events contribute to discrepancies in the data, emphasizing the value of real-world data in mitigating such limitations. It is noteworthy that our overall infection rates appeared to be close to those of infectious complications related to CIEDs, which was reported by Polewczyk et al. to occur in 2.3–3.4% of CIED recipients, and 1% at 1-year after CIED procedure based on large, prospective studies [13, 14]. Of note, we observed that overall infection rates increased by 50% at a median follow-up of 1.7 years compared with the first year. This pattern resembles findings from a study on CIED infections, which reported 30.79% of events within 1 year, 29.89% between 13–36 months, and 39.32% beyond 3 years after the last procedure. Notably, that study suggested that infections occurring after 12 months are more often related to lead-dependent factors rather than procedure-related factors [13]. It should be emphasized that any analogy drawn between CIED and HGNS must be interpreted with caution, as they differ in patient characteristics, the depth and complexity of implantation procedures, follow-up protocols, and the diversity of CIED device types. Future studies on HGNS infections should focus on clarifying the incidence rates, underlying mechanisms, and demographic factors associated with different onset timings.

This study identified an overall device reoperation rate of 2.0% at 1 year and 3.2% at 3 years post-implantation. These findings align closely with those reported in a separate TriNetX study by Lorenz and Goyal, which documented a combined rate of 2.6% for device revision or replacement (1.4%) and removal (1.2%) over a mean duration of 1.8 years from implantation to the indexed procedures, and also consistent with a postmarket surveillance data reporting a combined rate of 2.2% for device revision (1.5%) and explantation (0.7%) at first year [10, 15]. Although hyperglycemia was associated with higher infection risk, the rates of device revision, replacement, or removal were not increased. According to the study utilizing the Manufacturer and User Facility Device Experience database reported in 2024, infection accounted for 24.0% of all adverse events, yet only 4.9% of reoperations (repositioning, reinstallation, or explantation) were infection-related. Explantation was attributed to infection in 153 cases versus 185 for other causes. Among 315 infection cases, 77.5% received medical management, and only 59.2% required surgical intervention. These findings suggest a dissociation between HGNS-related infections and device reinterventions, which are often driven by other complications such as skin erosion, device extrusion, stimulation discomfort, lead migration/tethering, Twiddler syndrome, MRI restrictions, or patient request. [3].

Factors influencing infection risk of implantable electric devices can be derived from patient characteristics, surgical techniques, perioperative medical care, and device design. A retrospective cohort study compared the infection rate before and after the introduction of a MRSA decolonization protocol, although there were no significant differences, the authors believed there might be a downward trend as the patient database increases [11]. Given that the risk factors of HGNS infection remain unclear, studies from CIED could be acceptable references due to their similarities as electronic implantable devices designed to reside long-term in the chest. From over 60 years of clinical cases accumulation, the risk factors of CIED were well-documented and there are various risk scoring systems developed for risk stratification and corresponding prevention strategies [16, 17]. A nationwide study from Danish researchers also indicated that young age, male sex, severe renal insufficiency/dialysis, dermatitis, and CIED reinterventions were associated with CIED infections [14]. Before matching, the hyperglycemia group was on average 3.5 years older and had higher rates of comorbidities, including obesity, dermatitis/eczema, nicotine dependence, and polyneuropathy (Tables 2 & 4). Because multivariate regression is not feasible in TriNetX, propensity score matching was applied to minimize these confounders. Nonetheless, the influence of age, sex, and other comorbidities on infection outcomes warrants further investigation.

Among the aforementioned infective risk scoring systems for CIED infections, 4 out of 6 use the presence of diabetes as a risk factor [16]. However, patients diagnosed with diabetes could be in different hyperglycemic status from well, moderately, to poorly controlled. According to the 2017 guidelines for the prevention of surgical site infections, perioperative blood glucose control targeting levels below 200 mg/dL is strongly recommended for patients, irrespective of diabetes status [18]. In this study, we explored the blood glucose thresholds at 126 mg/dL and 154 mg/dL—corresponding to HbA1c values of 6% and 7%, respectively—to distinguish normal from hyperglycemia, which associated with a 1.5- to 2.1-fold increase in post-HGNS 1 year hyperglycemic group [19]. This is consistent with a meta-analysis study, showing a 4.6% of wound infection rates with postoperative glucose <150mg/dL vs 7.2% with glucose >150mg/dL, leading to a risk ratio of 1.6 [20]. Considering the potential for treatment delays arising from stringent glucose control, further investigation is warranted to determine the optimal glucose threshold that balances timely OSA management with surgical safety. With our pilot positive findings of hyperglycemia as a risk factor of post-HGNS wound infection, the OR of higher blood glucose thresholds like 200 mg/dL can be further analyzed when a larger population is obtainable.

To our knowledge, the present study represents the largest population-based study of post-HGNS infections and device reinterventions with a well-established federated database. A maximum of 9,977 patients diagnosed with OSA who received HGNS devices were carefully included with selective ICD-10-CM and CPT codes with exclusion of vagus nerve stimulation systems and immunosuppressants usage. Our study population demonstrated consistency with other literature. The nationwide data base also avoided systemic bias by single operators or HCOs. Furthermore, we compared two glucose thresholds at different follow-up endpoints to shed light on the dynamic influences of blood glucose level on the monitored outcomes, revealing a greater risk with increased glucose level.

However, there were some limitations related to the retrospective study design. First, the HGNS-specific ICD-10-CM, ICD-10-PCS, and CPT codes are not widely used by medical practitioners, hence the application of codes with broader medical conditions. Limiting surgical site infection to a one-year window reduced the chance of non-HGNS therapy-associated infections [21]. Second, the blood glucose data could be a combination of random, fasting, or even postprandial blood glucose, thus cannot reflect the actual constitution of glucose condition. This could be addressed by the analysis of HbA1c data (supplementary Table 1). Although HbA1c could serve as a stable, long-term index of average blood glucose level of 2–3 months, the examination was not taken routinely. Consequently, there were inadequate outcome events for statistical analyses, especially after PSM. Third, we did not exclude patients receiving antidiabetic agents, corticosteroids, or antibiotics in order to maintain an adequate sample size. In addition, diabetes mellitus was not included in the PSM. These may have introduced residual confounding, and thus the findings should be interpreted with caution. Lastly, the inability to control for medications, perioperative management, and other infection-related risk factors, plus the lack of detailed information on smoking, surgical techniques, perioperative antibiotics, and center-specific effects also represents limitations.

Conclusions

Hypoglossal nerve stimulation therapy-related infections may cause reinterventions and additional hospital treatments. Hyperglycemia status was shown to significantly increase the wound infection rate, while device revision, replacement, and removal rates remain unaffected. These findings underscore the importance of stringent perioperative glycemic control to mitigate postoperative infection risk in HGNS recipients and help establish realistic levels at which it is considered safe to perform HGNS implantation. Further stratification of higher glucose levels could help set up a risk score system regarding HGNS infection. Future studies with more HCOs enrollment, general use of HGNS-specific codes, and concurrent perioperative monitoring of HbA1c are warranted for a better understanding of this topic.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CIEDs

Cardiac implantable electronic devices

CPT

Current procedural terminology

HCOs

Healthcare organizations

HGNS

Hypoglossal nerve stimulation

ICD-10-PCS

International Classification of Diseases 10th Revision, Procedure Coding System

ORs

Odds ratios

OSA

Obstructive sleep apnea

Authors’ contributions

Conceptualization: Kai-Yuan Hsiao, Sandro Marques, Thomas Joseph O'Neil, Thomaz A. Fleury Curado; Methodology: Kai-Yuan Hsiao, Sandro Marques, Capasso Robson, Thomaz A. Fleury Curado; Data acquisition: Sandro Marques, Scott Howard; Data analysis and investigation: Kai-Yuan Hsiao, Thomaz A. Fleury Curado; Writing- original draft preparation: Kai-Yuan Hsiao, Thomas Joseph O'Neil; Writing- review and editing: Sandro Marques, Scott Howard, Capasso Robson, Thomaz A. Fleury Curado.

Funding

Open access funding provided by National Cheng Kung University. No funding was received for conducting this research.

Data availability

The data are available from the corresponding author on reasonable request.

Declarations

All authors have seen and approved this research.

Ethical approval.

This study utilizes deidentified patient records and was exempted by the University Hospitals of Case Western Reserve University Institutional Review Board review (STUDY20250257).

Conflict of interest

All authors certify that they have no affiliations with or involvement in any organization or entity with any financial interest or non-financial interest in the subject matter or materials discussed in this manuscript.