Obstruction level associated with outcome in hypoglossal nerve stimulation

About 20% of women and 50% of men suffer from moderate to severe sleep disordered breathing (SDB) according to a recent population-based study [1]. Most patients with SDB suffer from obstructive sleep apnea (OSA) [2]. Patients complain of excessive daytime sleepiness and cognitive deficits [3]. OSA also has a high societal relevance since approximately 20% of car accidents are related to sleep deprivation of which OSA is a main cause [4]. In addition, OSA is associated with secondary diseases especially of the cardiovascular system such as hypertension, coronary artery disease, and cardiac arrhythmias [5]. Several treatment options exist for patients with OSA ranging from conservative methods such as positive airway pressure (PAP) therapy and mandibular advancement devices to surgical interventions [6, 7]. For patients with poor compliance to conservative therapies and moderate to severe OSA, selective hypoglossal nerve stimulation (sHNS) may constitute an effective surgical alternative [8].

In sHNS, the main pharyngeal airway dilatory muscle and tongue protrudor is activated to prevent airway collapse during sleep [9]. Different stimulation techniques have been developed ranging from the activation of proximal sectors to distal fibers of the nerve [10]. One system frequently implanted stimulates the branches of the hypoglossal nerve which are required for tongue protrusion and is termed selective hypoglossal nerve stimulation (sHNS) [11]. The stimulation is synchronized with the breathing cycle in this system [11]. In addition to a significant AHI reduction, sleep architecture was also improved in patients with OSA by using this kind of neurostimulation [12]. Established criteria for selecting patients for sHNS are apnea–hypopnea index (AHI) between 15 and 65/h, BMI ≤ 35 kg/m² and the absence of a complete concentric collapse at the palate level during drug-induced sleep endoscopy (DISE). DISE is the most established method to detect the level and pattern of pharyngeal obstructions in the preoperative assessment for surgical OSA treatment [13]. DISE, however, requires sedation and is operator-dependent [14]. Obstructions can alternatively be monitored during whole night of natural sleep with a multisensory manometry system [15]. Sleep architecture is not influenced by the system [16]. Obstructions measured with this system can be divided into velar (soft palate), infravelar (below soft palate), and multilevel obstructions. Manometry can provide valuable information compared to DISE, since REM sleep does not occur in propofol sedation [17] and infravelar obstructions increase in REM sleep [18].

The aim of this study was to determine if the treatment response of sHNS is associated with the obstruction level detected with manometry.

The characteristics of patients are depicted in Tables 1 and 2. Apnea–hypopnea index (AHI) of patients ranged from 18.0 to 70.6/h.

In this study, the obstruction level was analyzed via manometry throughout an entire night of natural sleep in patients with OSA either before (n = 14) or after treatment (n = 17) with sHNS to determine if treatment effectivenss depends on the anatomic location of obstructions.

Velar (upper level) obstructions were significantly higher in the preoperative measurements in non-responders (AHI > 50% and AHI > 15/h) compared to responders. Inversely, the percentage of infravelar (lower) obstructions was also significantly higher in responders. Furthermore, the preoperative percentage of apneas from all obstructive events at that level was analyzed with regard to treatment response. The relative proportion of apneas to overall obstructive events at the different level was not associated with treatment response. These results suggest that sHNS seems to be more effective on infavelar obstructions and that patients with a high percentage of velar obstructions may potentially be suboptimal cases for sHNS. In the evaluation of these results, it needs to be taken into consideration that a complete concentric collapse (CCC) was excluded prior to implantation with drug-induced sleep endoscopy (DISE) in all patients of our study cohort. A complete concentric collapse (CCC) is an important predictor for treatment success in sHNS [25] and occurs in 20–30% of patients with OSA [26]. Since a correlation between a complete concentric collapse and manometry findings has not yet been demonstrated, both DISE and manometry need to be performed before implantation.

To date, the level of obstruction has not been tested with manometry in patients with sHNS. Nonetheless, sHNS has been demonstrated in numerous studies to effectively reduce AHI [23]. This high degree of effectiveness requires a reduction of obstructions at all level. The resolution of velar obstructions in sHNS is believed to work via palatoglossal coupling and has been validated using drug-induced sleep endoscopy (DISE) [27]. Notwithstanding, Mulholland et al. validated with DISE that patients with worse baseline severity of collapse at the level of the lateral walls were accompanied by worse sHNS outcomes [28]. Contradictory findings remain concerning the relevance of the airway opening at the upper pharynx for sHNS effectiveness. The degree of opening of the retropalatal space with activated sHNS was associated with improved treatment outcome in one study [29]. In contrast, patients with significant airway improvement in the upper pharynx with mandibular advancement during DISE appeared less likely to succeed with sHNS [28]. These parameters are ideally assessed with sleep endoscopy, since the pattern and detailed anatomic location of obstruction are detected. However, DISE is an artificial examination for a limited time frame and not comparable to natural sleep. The alleviation of obstructions at the different level through sHNS measured during a whole night of natural sleep could therefore differ from DISE examination and may provide important additional information. Differing results between DISE and manometry have already been reported especially with regard to the detection of infravelar obstructions [30]. A potential reason for these observed differences could be the change in obstruction level occurring in REM sleep [18].

To further examine the effectiveness of sHNS on different levels of obstructions, sHNS patients were examined with manometry under stimulation. The rate of residual multilevel obstructions per hour with active stimulation was significantly higher in non-responders compared to that in responders. Also, relevant residual velar obstructions occurred in responders as well as non-responders. This could be in line with sHNS being less effective on velar and multilevel obstructions. Another explanation could be a higher sHNS efficiency on infravelar and multilevel obstructions in the responder group and an impaired efficiency at all obstruction levels in the non-responder group. In addition, the rates of obstructive events per hour in the different levels were tested preoperatively and postoperatively with active stimulation in five patients, and the highest reduction was seen in multilevel obstructions. However, the validity of this comparison is limited since there were three non-responders in this small group.

Because a bilateral protrusion of tongue base in DISE could be correlated with a better opening of the soft palate in a previous study [27], we analyzed the relationship between the percentage of velar obstructions and tongue protrusion (which leads to the protrusion of tongue base). In this study, 24% of patients in the postoperatively measured cohort showed a right tongue protrusion and 76% a bilateral tongue protrusion. The tongue motion was not associated with the percentage of residual velar obstructions with active stimulation. A possible explanation is that tongue motion observed in awake patients is not correlated with the bilateral protrusion of the tongue base visualized in DISE. Also, as outlined above, the relative share of velar obstructions in manometry during a whole night of sleep can differ from the opening of the soft palate in DISE. In addition, the baseline soft palate anatomy could be a confounding factor, especially since sHNS is a non-anatomically modifying surgery. Schwab et al. evaluated the soft palate volume on computed tomography on baseline and demonstrated that smaller soft palate volumes were associated with a favorable response [31].

In patients with insufficient treatment results with sHNS, several options exist to improve the therapy. For example, it has been suggested that palatal surgery could improve the outcome in non-responder with obstructions at velum level or oropharynx proven by DISE [32]. In reality, probably not all non-responder will profit from palatal surgery, and DISE seems not to discriminate the non-responder group well enough; as in the study by Steffen et al., almost 90% of suboptimal responders had a complete collapse at velum level [32]. Potentially, manometry could assist stratifying patients with suboptimal response profiting from palatal surgery and warrants further exploration. The preoperative localization of obstructions with manometry to select patients for uvulopalatopharyngoplasty (UPPP) was beneficial in one study [33]. However, a smaller study showed no association between the level of obstruction and patients’ UPPP outcome [34].

Despite thorough selection of patients, about 20–30% of patients are suboptimal responders according to the Sher criteria [23, 35]. It would be highly relevant to identify this patient group of suboptimal responders preoperatively. Several studies have been published on objective factors for the prediction of treatment success in HNS. Patients with lower therapeutic level of PAP were found to be more likely to have a treatment success in sHNS compared to patients with higher pressure requirements [36]. In the ADHERE registry, a lower BMI, higher age, and female gender were significant predictors of therapy response in a multivariate model [35, 37]. These are important findings but the difficulty remains how to stratify the “problematic” patient with high PAP pressure and BMI — potentially, manometry could be of use. Based on the findings in this study, it seems possible that the inadequate treatment response is at least partially caused by a high percentage of velar obstructions. Manometry could therefore possibly enrich the framework of objective factors for optimal HNS candidate selection. DISE would only be performed in preselected patients.

There are several limitations to our study. Most importantly the small sample size (due to the difficult recruitment of patients) restricts the detection of changes in subgroups and the results therefore need to be confirmed in a larger multi-center study. The percentage of non-responders in the study cohort is higher than reported in the literature [23, 38]. The high percentage in the postoperatively measured patients can be explained by more frequent visits of non-responders in our sleep laboratory, e.g., advanced titration. Also, all patients were preselected with DISE to rule out a complete concentric collapse. The overlap between patients with a complete concentric collapse at velum and a high percentage of velar obstructions can therefore not be discerned. Furthermore, the obstruction level could be subject to night to night variance since obstruction levels change with sleep stages. In subsequent projects, the night to night variance should be analyzed to verify our findings.

In summary, this study emphasizes the soft palate area as critical for HNS success since a high percentage of velar obstruction was associated with treatment response. Manometry may be a complementary diagnostic procedure for the selection of patients for HNS.