Combination Tonsillectomy and Hypoglossal Nerve Stimulation for Sleep Apnea Patients With Oropharyngeal Lateral Wall Collapse

Abstract

Objective.

The efficacy of hypoglossal nerve stimulation (HGNS) therapy is limited by obstruction of the oropharyngeal lateral walls (OLWs). Our objective was to investigate the effect of palatine tonsillectomy on HGNS efficacy in obstructive sleep apnea (OSA) patients with OLW collapse.

Study Design.

Case-control study of patients with moderate-to-severe OSA, complete-or-partial OLW collapse, and small tonsils (1 – 2+). Concomitant palatine tonsillectomy and HGNS (HGNS+T) were compared against a control group of patients who underwent HGNS alone.

Setting.

Single academic institution.

Methods.

Study outcomes were measures of HGNS efficacy defined as a %reduction in apnea-hypopnea index (AHI) (primary) and successful treatment response (50% AHI reduction to <15/h, logistic regression), respectively. Regression analyses quantified the additional effect of tonsillectomy (HGNS+T vs HGNS alone, independent variable) on HGNS efficacy. Analyses were adjusted for OLW collapse severity (complete vs partial), tonsil size, age, sex, body mass index, and baseline AHI.

Results.

Nineteen patients underwent HGNS+T and had follow-up sleep testing for the current analysis. The control group (HGNS alone) consisted of 78 patients. Baseline demographics and OSA severity were similar between the groups, except HGNS+T group had increased prevalence of complete OLW collapse. Linear regression demonstrated that adding tonsillectomy resulted in an additional 22.9% [7.5, 35.2] reduction in AHI [95% confidence interval, CI] (P = .006), and 8.6 [1.7,43.4] (P =.010) greater odds [95% CI] of a successful treatment response with HGNS.

Conclusion.

Compared to historically poorer outcomes of HGNS in patients with OLW collapse, these early results suggest combining tonsillectomy with HGNS may represent a promising strategy to improve success rates.

Obstructive sleep apnea (OSA) is a common chronic condition with high rates of nonacceptance or nonadherence to positive airway pressure therapy (PAP).^1-3 Hypoglossal nerve stimulation (HGNS) has emerged as a popular alternative to PAP therapy that benefits from higher rates of adherence compared to PAP.^4-6 However, the efficacy of HGNS therapy is highly dependent on a multitude of anatomic and nonanatomic pathophysiologic factors, one of which is the pattern of oropharyngeal lateral wall (OLW) collapse,^7-10 as determined by drug-induced sleep endoscopy (DISE).¹¹ In particular, complete concentric collapse of the soft palate is a contraindication for HGNS implantation.¹² Patients with OLW, while still eligible for HGNS implantation, exhibit a comparably reduced efficacy with HGNS therapy.^8,13-16

Optimal surgical options for patients with OSA that are ineligible for, or unlikely to respond to, HGNS tend to be more invasive procedures such as maxillomandibular advancement (MMA) or expansion pharyngoplasty.^17,18 Success rates for traditional OSA surgery are also reduced when OLW collapse is present, with the exception of MMA.^18-21 When hypopharyngeal obstruction is coexistent, palate surgery typically must be combined with other procedures, further adding morbidity that may not be acceptable for all patients. As a potential solution, we considered concurrent tonsillectomy (HGNS+T) to reduce the OLW obstruction that hampers HGNS efficacy. This strategy pairs a relatively low morbidity procedure (HGNS) with a moderately morbid one (tonsillectomy), which may be a more appealing approach to many patients, especially those with more significant comorbidities or advanced age. Given the high rates of OLW collapse,²² this combination of procedures may benefit a significant number of patients who are interested in HGNS but have a historically lower likelihood of a successful response. We hypothesized that combining tonsillectomy with HGNS is safe, feasible and increases treatment efficacy in patients with OLW collapse, compared to HGNS alone.

Materials and Methods

This is a prospective observational study investigating the effect of concomitant tonsillectomy and unilateral HGNS (Inspire Medical Systems) in patients with OLW collapse at a single academic medical center by a single surgeon. This clinical procedure has been performed routinely at this center since November of 2020, and remains ongoing. For the current study, we included the first 30 patients that underwent HGNS+T from November 2020 to October 2023. Among this group, 19 patients underwent follow-up sleep testing within the window for outcomes collection (February 2024). The remaining 11 are still being followed up clinically but have either had scheduling delayed (n = 4), declined follow-up sleep testing (n = 6), or were denied by insurance. No subjects were lost to follow-up. All patients were part of a larger prospective observational cohort study aimed at investigating outcomes of HGNS and predictors of efficacy (Mass Eye and Ear Sleep Surgery [MEESS] Patient Registry). Institutional review board (IRB) approval was granted by the Mass General Brigham IRB and all patients signed informed consent.

Patients eligible for HGNS+T had partial or complete OLW collapse in the setting of small tonsils (1+ or 2+) and otherwise met all criteria for HGNS (AHI >15 events/h, failure of PAP therapy, <25% mixed/central AHI, body mass index <32 or 35 kg/m²). The presence of OLW collapse was determined by DISE, performed as part of routine clinical care to identify sites of anatomic collapse.²³ The DISE exams were scored by the treating surgeon using the VOTE classification.²⁴ Decision-making was shared between the patient and treating surgeon to determine whether to proceed with HGNS+T or not. The control group is HGNS patients from the MEESS Patient Registry who had complete or partial OLW collapse, 1+ or 2+ tonsils, and no other surgical treatment between HGNS implantation and had already completed follow-up sleep testing (Figure 1).

Figure 1.

Participant enrollment flow chart. HGNS, hypoglossal nerve stimulation; HGNS+T, concomitant palatine tonsillectomy and HGNS; OLW, oropharyngeal lateral wall.

All demographic and sleep study data were maintained prospectively in the MEESS Patient Registry. Preoperative sleep studies were a mixture of in-lab polysomnography (PSG, n = 15) or home sleep apnea tests, were scored using American Academy of Sleep Medicine guidelines²⁵ and converted to 3% oxygen desaturation scoring for this analysis (see Supplemental Materials, available online, for details).

All patients were implanted with the HGNS in the standard manner with or without concurrent coblation tonsillectomy. This was followed by device activation 1 month later and a titration PSG (tPSG) no sooner than 3 months post-activation (see Supplemental Materials, available online, for details on surgical technique, activation, and tPSG protocols). Entire night AHI computes AHI over the entire night and thus includes both sub- and supratherapeutic amplitudes, as well as periods with therapy off. Optimal setting AHI computes AHI over the period that stimulation amplitude is at the optimal setting.

Safety and feasibility of HGNS+T was assessed by evaluating operating room (OR) time, recovery room (ie, postanesthesia care unit, PACU) time, and postoperative complications. OR time was defined as the amount of time the patient spent in the OR, not operative time. PACU time was defined as the amount of time from completion of OR time to discharge. Patients were excluded if they underwent additional procedures, such as septoplasty. All patients were discharged to home following the procedure. Postoperative complications, including bleeding and infection, were assessed in the medical record.

Statistical Analysis

In primary analysis, multivariable linear regression evaluated the effect of adding tonsillectomy to HGNS (binary independent variable) on percent reduction in entire night AHI, adjusting for degree of OLW collapse, tonsil size, age, sex, body mass index (BMI), and baseline AHI. Percent reduction in AHI was transformed to provide a normal distribution before analysis $(x_{\text{transformed}}=x∕(2-x))$ and 95% confidence interval (CI) is presented based on back-transformed results. P <.05 was used to indicate the statistical significance of the primary outcome test. In the secondary analysis, we repeated the above analysis using successful treatment response as the outcome (binary) and logistic regression instead of linear regression. We defined successful response as a 50% entire night AHI reduction to <15/h (modified Sher criteria),²⁶ consistent with our previous publication evaluating the effect of site of collapse on HGNS outcomes.⁸

To demonstrate the consistency of the effect of tonsillectomy on treatment response, we also explored the effect of tonsillectomy on other commonly used definitions of response: 50% AHI reduction to <10/h^27,28 and 70% AHI reduction.²⁹ As exploratory analyses, we repeated the above analyses except for outcomes of percent reduction AHI and successful treatment response were defined based on the optimal setting AHI. Finally, to investigate feasibility and safety, we used linear regression to evaluate the effect of adding tonsillectomy to the HGNS (binary independent variable) on differences in OR and PACU time, adjusting for time since first surgery to adjust for any effect of surgeon experience over time.

Results

Table 1 shows the demographic and OSA severity data in patients who underwent HGNS+T as well as those who had partial or complete OLW collapse and underwent HGNS alone (controls). Age, sex, BMI, and OSA severity were similar between the groups. However, patients that underwent HGNS+T had a higher prevalence of complete OLW collapse and 2+ tonsils. Supplemental Table S1, available online, also shows the VOTE scores for each of the 19 HGNS+T patients. As described in Figure 2, treatment response was higher in patients that underwent HGNS+T based on both %reduction AHI and response rate. Mean time from surgery date to tPSG was 160 days (range: 70-251 days). As hypothesized, adding tonsillectomy to HGNS was associated with a significant increase in early treatment efficacy, compared to HGNS alone. Specifically, compared to HGNS alone, concomitant tonsillectomy and HGNS were associated with an additional 22.9 [7.5, 35.2] percent reduction [95% CI] in entire night AHI (P = .006) and 8.6 [1.7, 43.4] times increased odds [95% CI] of a successful treatment response (P = .010). The effect of tonsillectomy remained, and was in fact stronger, when response rates were defined based on other commonly used criteria: 50% AHI reduction to <10/h (odds ratio [95% CI]: 13.5 [2.5, 74.2]) and 70% AHI reduction (12.0 [2.4, 59.9]).

Table 1.

Patient Characteristics of Those Undergoing HGNS With Tonsillectomy and HGNS Alone

	HGNS+Tonsillectomy	HGNS alone (control)
N	19	78
Age, years	55 [48, 61]	59 [53, 65]
Men, n	15 (89%)	57 (73%)
BMI, kg/m2	31.0 [27.9, 31.8]	29.8 [26.7, 31.3]
AHI, h	26.7 [21.1, 44.2]	30.9 [22.0, 45.4]
% 2 + Tonsils	47	14
% Complete OLW	76	18

Tonsil size (per Brodsky scale) is scored as 0 (absent), 1+, 2+, 3+, 4+ (higher values indicate larger tonsils); all patients here had small tonsils (“1+“ or “2+“). Abbreviations: AHI, apnea-hypopnea index; BMI, body mass index; HGNS, hypoglossal nerve stimulation; OLW, oropharyngeal lateral wall collapse (scored as absent, partial, or complete; all patients here had partial or complete).

Figure 2.

Treatment response in patients that underwent concomitant hypoglossal nerve stimulation and tonsillectomy (HGNS+T) and HGNS alone in terms of percent reduction apnea-hypopnea index (AHI, A), percentage of patients that achieved a successful response to therapy (50% AHI reduction to <15/h) (B), and residual AHI with treatment (C).

HGNS+T was also associated with improved treatment response as defined by optimal setting AHI, compared to HGNS alone. Specifically, adding tonsillectomy to HGNS was associated with an additional 14.3 [5.5, 21.8] percent reduction [95% CI] in optimal AHI and 14.9 [1.2, 188.3] increased odds [95% CI] of a successful response to therapy, compared to HGNS alone (Supplemental Table S4, Available online).

After adjustment for surgical experience (time from the first procedure), OR and recovery room times [95% CI] were both longer with the HGNS+T procedure by 16.6 [8.7, 24.5] and 25.0 [−6.3, 56.2] minutes, respectively, with the latter being non-significant. In the HGNS+T group, there were no cases of infection and 2 (out of 30, 6.7%) posttonsillectomy hemorrhages requiring return to the OR. The control group had no complications. Although post-operative pain and analgesic consumption were not objectively tracked, larger prescriptions for analgesics were written for those who underwent HGNS+T (see Supplemental Materials, available online).

Discussion

HGNS is a well-accepted treatment for PAP-intolerant patients with moderate-to-severe OSA that has numerous strengths of low morbidity, multilevel effect, adjustability, and reversibility.³⁰ Unfortunately, laterally oriented obstruction (OLW)—seen in a substantial number of patients (20%-48%²²)—hampers a successful therapeutic response.^13,31 The current study is the first to present a promising adjunctive therapy to improve early success rates with HGNS in this challenging subgroup of patients. Our case-control analysis demonstrated that the combination of HGNS+T provided clinically significant improvements in early treatment efficacy compared to the HGNS alone- an additional 23% reduction in AHI, a 9-fold greater odds of attaining a successful treatment response, and a residual treatment AHI was nearly halved (Figure 2). Notably, the additional intervention added only 17 minutes of OR time and 25 minutes of recovery time (nonsignificant) and complications were uncommon (6.7%) with no lasting sequelae. Overall, combining tonsillectomy with HGNS could offer considerable benefit to a significant portion of patients in whom HGNS therapy is most commonly unsuccessful.

The high prevalence of OLW collapse in OSA patients and the well-established poorer outcomes with existing non-PAP treatment options in this subgroup emphasizes the critical need to prioritize the development of new treatment strategies.^19,20 Previous literature suggests that among OSA patients that seek surgical treatment and ultimately undergo the DISE procedure, ~20% have partial or complete OLW collapse. Data from our own database shows an even higher prevalence of OLW collapse at 48% (13% complete OLW collapse). Among those with OLW collapse, 61% had small tonsils (1+ or 2+) and, therefore, nearly one-third of our patients could be eligible for HGNS+T and associated improvement in outcomes.

The challenges in treating OLW collapse with HGNS, despite their eligibility for the therapy, was recently documented in a retrospective multicenter study of 343 OSA patients who underwent HGNS implantation. The authors found that the response rate (50% AHI reduction <15/h using optimal setting AHI) in patients with complete OLW collapse was significantly lower (58%) than in patients without OLW collapse (74%).³¹ Data from our study registry of HGNS patients (n = 297), shows similar differences in response rates with and without complete OLW collapse (61% vs 86%, see Supplemental Table S2, available online). Matching the definition of response, the response rate in the patients receiving HGNS+T was 95% in the whole group (n = 19) and 93% in patients with complete OLW collapse (n = 14). In other words, removing the tonsils in OLW collapse patients results in a treatment response to HGNS that closely resembles—and perhaps exceeds—that of non-OLW collapse patients.

Tonsillectomy is a well-established treatment for OSA when OLW collapse and tonsillar hypertrophy (large tonsils, size 3+ or 4+) is present.^32,33 The effect of removing small tonsils on moderate-to-severe OSA in adults has not been formally studied to our knowledge. However, among the studies available that report results by tonsil size, it is likely that removing small tonsils alone has low efficacy in treating OSA. Therefore, it may not be the case that improved treatment efficacy of HGNS+T versus HGNS alone is simply due to the added treatment benefit of removing small tonsils and debulking of lymphoid tissue. Rather, we speculate there may be some synergistic mechanisms by which removing small tonsils makes HGNS more effective, such as decreased OLW compliance through stiffening (scar formation) and perhaps glossopharyngeal coupling where scar from the inferior tonsillar pole may extend onto the tongue base. To better understand the mechanism, future studies should determine the effect of tonsillectomy alone and evaluate OLW collapse severity on DISE before and after tonsillectomy with longer term follow-up.

One of the realities of treating moderate-to-severe OSA in PAP failure is that multimodality therapy, both surgical and nonsurgical, is often needed. A major advantage to HGNS+T as a multimodal approach is the relative simplicity of adjunctive tonsillectomy and its widespread availability. Many surgeons offering HGNS may not be experienced in expansion pharyngoplasty, MMA or other hypopharyngeal procedures. We suspect that a palatine tonsillectomy is in the skillset of most, if not all, HGNS providers which would expand access to patients who otherwise might go untreated or undertreated. Combining HGNS+T when there is a high likelihood of medical necessity improves the efficiency of care as well as convenience to the patient with 1 recovery period, 1 exposure to general anesthesia, and 1 insurance copayment, albeit with increased postoperative pain. Indeed, we note that the addition of tonsillectomy only modestly increased OR and recovery room times (17 and 25 minutes, respectively) when compared to HGNS alone, figures that likely do not negatively impact OR workflow and scheduling and are outweighed by the substantial improvements in OSA observed. We also highlight that to date there have been no cases of infection, which may be of concern when combining the sterile and clean-contaminated portions of the procedure.

While there are somewhat limited treatment options that specifically target OSA caused by OLW collapse, other surgeries exist that may be considered in place of the HGNS+T combination. For instance, expansion sphincter pharyngoplasty (ESP) would be an appropriate consideration.¹⁷ However, when compared to HGNS,³⁴ this procedure has increased morbidity associated with the surgery such as a longer recovery time and more severe pain. While there is no literature comparing the 2, it is the first author's experience that the severity and duration of pain after ESP are also considerably worse than palatine tonsillectomy. Indeed, in evaluating cases over the past 12 months, opioid prescribing for ESP was more than double that of isolated tonsillectomy and HGNS+T patients—the details of which are summarized in Supplemental Materials, available online. Many patients therefore will not or cannot pursue ESP due to comorbidities, age, and/or preference. Furthermore, if a multilevel surgical approach is employed for co-existent tongue or epiglottic collapse, further morbidity is added. MMA is another alternative that has high efficacy including for lateral wall collapse^18,21 but is a major surgery with long recovery times compared to HGNS+T, such that many patients, especially those with older age, are unlikely to select this intervention.

There are limitations of this study. First, patients were not randomized to receive tonsillectomy versus no tonsillectomy so there may be uncontrolled differences between the HGNS and HGNS+T populations. To reduce bias to the greatest extent possible, our comparison group consists of OLW patients with tonsils present who could have potentially been (and oftentimes were) offered HGNS+T. The only major difference between the HGNS and HGNS+T groups was a higher proportion of patients with complete OLW collapse in the HGNS+T group, a factor that is expected to have made the treatment group inherently more challenging to treat, thus biasing toward finding no effect.³⁵

Second, in this study, we did not assess the change in OSA after tonsillectomy alone, that is, with HGNS off, so it is possible that tonsillectomy, and not HGNS, is driving the treatment efficacy. This is somewhat unlikely given that all but 2 patients had multilevel collapse (Supplemental Table S1, available online), such that collapse at other pharyngeal sites would remain with tonsillectomy alone. We also point out that outcomes were assessed over the medium term (mean time between surgery and tPSG was 160 days). It is possible that efficacy may wane as scar tissue remodels over time, although there did not appear to be a difference in outcomes in those with earlier versus later tPSG. This long-term evolution of scar tissue may help understand the variable impact of prior pharyngeal surgery on HGNS outcomes reported in the literature.^36-40 Given the promising magnitude of the current initial findings, future mechanistic and longer-term research is warranted.

Third, a strength of our study was that DISE videos were scored by a single surgeon, therefore, removing a potential source of bias. Given the subjectivity of DISE, it remains unproven whether our findings will apply to patients at other centers, although interrater reliability of OLW scoring was found to be moderate to substantial, so it is likely that most other surgeons view OLW collapse similarly.^8,39,41

To homogenize scoring criteria to allow for the comparison of absolute AHI changes between patients, we employed a method that converted AHI values to a single scoring criterion (see details in Supplemental Materials, available online). While model fits were good, this approach likely introduces a small amount of noise into the absolute AHI values that may have a marginal effect on interpreting differences in absolute AHI values between groups. We emphasize that findings were not meaningfully different without this additional methodological rigor and would not affect percent change in AHI within a given patient (see Supplemental Table S4, available online).

Finally, primary outcome analysis employed the entire night AHI from the titration PSG rather than the optimal setting AHI. While the majority of the HGNS literature uses optimal setting AHI as the outcome measure defining success, there is a growing consensus that it may be misleading by overestimating treatment response. Entire night AHI includes both sub- and supratherapeutic amplitudes, as well as periods with therapy off, which likely produces a slightly elevated treatment AHI and is therefore a more conservative measure of treatment response. Regardless, all analyses were repeated with optimal setting AHI (see Supplemental Tables S3 and S4, available online) and similarly show significant benefit in HGNS+T.

In conclusion, the study is the first to provide evidence that combining HGNS and palatine tonsillectomy is a promising and efficacious strategy to improve early outcomes with HGNS in patients with lateral wall collapse. A larger, randomized trial with longer term follow-up is needed to confirm the increased effectiveness of this therapeutic combination prior to widespread adoption of this technique.

Supplementary Material

Additional supporting information is available in the online version of the article.