Introduction

Obstructive sleep apnea (OSA) is very common in middle aged adults. Using a polysomnography (PSG) derived definition, defined as an apnea hypopnea index (AHI) of >5, it occurs in 20–30 % of males and 10–15 % of females in North America [1]. OSA patients are at increased risk of a number of adverse clinical outcomes. They experience attention deficits, which translates into a two to threefold higher rate of motor vehicular accidents in those with severe disease [2]. There is an independent association between OSA, insulin resistance and type 2 diabetes [3]. Patients with sleep apnea have a higher incidence of hypertension, coronary artery disease and stroke compared to matched subjects without sleep apnea. Patients with moderate to severe untreated OSA have a two to three fold increased risk of all-cause mortality compared to patients without OSA independent of other risk factors, such as obesity and hypertension [4–6].

Continuous positive airway pressure (CPAP) is the main stay of treatment for OSA. In a meta-analysis of 22 randomized trials, CPAP was effective in reducing sleepiness and improving quality of life measures in patients with moderate to severe OSA [7]. Although lacking validation from a randomized trial, multiple observational studies have reported an association between CPAP use and decreased cardiovascular disease and mortality [8–10]. Unfortunately, CPAP is a cumbersome treatment, and many patients are unable to use it. Approximately 20 to 40 % of patients with OSA are unable to use CPAP at all. Of the remainder many do not use CPAP optimally [11–16]. Thus, there is a clear need for alternative therapies in those that are CPAP intolerant. Surgical options are potentially very attractive since continued nightly usage is not required. However, many surgical options have high failure rates and their long-term efficacy has been poorly characterized.

Temperature controlled radiofrequency tissue ablation (TCRFTA) utilizes a probe to deliver temperature controlled energy to the upper airway tissue while sparing adjacent tissues. It results in less pain than more conventional methods thus potentially speeding patient recovery from the procedure. TCRFTA can be applied at the level of the soft palate, base of tongue or both levels depending on the predominant site of obstruction. It can be performed in an office setting, can be repeated as needed, and is associated with less morbidity than an operative procedure.

Comprehensive studies evaluating TCRFTA are lacking. A meta-analysis was published in 2008 evaluating this procedure. Sixteen studies were included in this analysis, of which 3 were randomized and 13 were non-randomized. The most recent paper included in this analysis was published in 2006. In the prior study, studies that involved TCRFTA at the level of the soft palate, base of tongue, or multilevel were mixed together. Their analysis did show an overall reduction in Epworth Sleepiness Scores (ESS) and Respiratory Disturbance Index (RDI) suggesting TCRFTA might be a useful treatment option in patients with sleep apnea. Since the time of the last meta-analysis, additional studies [17, 18] have been published, which allow a site specific analysis. The aim of our systematic review is to provide a more precise estimate of the effectiveness of TCRFTA performed at the level of the soft palate, base of tongue or multilevel for the treatment of sleep apnea.

Methods

Eligibility criteria

We included observational and randomized controlled trials that satisfied the following inclusion criteria: (1) subjects with OSA confirmed by clinical symptoms and a PSG demonstrated RDI ≥ 5; (2) treatment with TCRFTA of the soft palate, base of the tongue or both as a stand-alone procedure; and (3) an English version of the study was available. We excluded citations where: (1) TCRFTA was used with other surgical interventions such as uvulopalatopharyngoplasty and (2) TCRFTA was used for treatment of socially disruptive snoring or upper airway resistance syndrome. If an article reported data for subjects with OSA and socially disruptive snoring and/or upper airway resistance syndrome separately, only the data for subjects with OSA was abstracted.

Search strategy

A comprehensive computer-based search of the published medical literature was performed using MEDLINE (1950 onwards access via Ovid), PubMed, EMBASE (all years access via Ovid), and Evidence Based Medicine Reviews. We used search terms as outlined in Appendix 1.

Selection process

Article titles and abstracts were reviewed by two reviewers (RYB and VRM) to determine article eligibility. If an article was thought to be potentially eligible by either reviewer the full-text article was reviewed. The reference lists of relevant review citations and references of all included citations were checked for additional potentially eligible studies. Disagreements between reviewers were resolved either by discussion or by a third reviewer (MJM).

Data abstraction

One reviewer abstracted data from each eligible study using a self-developed standardized form. A second reviewer verified data abstraction, and conflicts, if any, were resolved mutually or by the third reviewer. We collected basic patient population demographic data as well as information about the surgical intervention such as site, number of session, etc. Outcomes were divided into:

  • Objective outcomes: RDI and lowest oxygen saturation (LSAT, %)

  • Subjective outcomes: included bed partner’s visual analogue scale evaluation of snoring (VAS, 0–10) and ESS

Data analysis

Statistical analysis was performed using RevMan Version 5.2 (Review Manager, Cochrane Collaboration 2012). We used a relative measure of effect i.e., ratio of means (RoM) by calculating the ratio of final mean value post-intervention to the mean value prior to intervention, such that ratio of means <1 favors reduction in outcome. Standard error was calculated as described previously [19], and RoM across studies and its associated confidence interval (CI) were pooled using standard equations for inverse variance weighting using a random effects model [20]. Heterogeneity was expressed by I 2, with I 2 values of 25, 50, and 75 % interpreted to indicate low, moderate, and high heterogeneity, respectively [21]. Each surgical site, i.e., base of tongue, soft palate, and multi-level intervention, was analyzed separately. Since only a small number of studies were analyzed in each group, we considered a funnel plot unreliable in the determination of presence or absence of publication bias [22]. The overall quality of evidence was graded using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach [23] (Table 1).

Table 1 Methodological quality
Full size table

Results

The search process is outlined in Fig. 1. Twenty-nine studies described temperature controlled radiofrequency ablation [24–52] as a stand-alone procedure. Three studies were excluded since only the abstracts were available in the English language and there was not enough information available in the abstract to obtain the relevant outcomes [35, 37, 41]. Attempts to contact the first authors for additional information were unsuccessful. Six studies were excluded from the meta-analysis but not the systematic review because patients with OSA could not be extracted from the group data that included snorers (one study [48]), the mean data was not provided (one study [25]), data was not presented in a way to extract relevant data (one study [24]), and three studies contained duplicate patients, the centers provided their further experience with this technique at a later time [43, 45, 50]. The study from these centers that contained the most extractable data on the most patients was chosen for inclusion. In total, 21 of the 26 studies had full in-laboratory polysomnography at baseline and at follow up [24, 26, 28, 29, 31–34, 36, 38–40, 42, 43, 45–47, 49, 51, 52, 50]. Among the other five studies, three had home or unattended polysomnography data ([25, 30, 44], one had level I-III polysomnography based on the clinical probability of severity [27] and one had full polysomnography only at baseline but not at follow up [48].

Fig. 1
figure 1

Search process

Full size image

Based on the site of application of TCRFTA, studies were divided into three groups- base of tongue (BoT), soft palate (SP) and multi-level approach. Tables 2, 3, and 4 contain the methodological qualities and results reported by the 20 studies included in our meta-analysis, and 6 studies (bolded text in table) included in our systematic review but excluded from the meta-analysis as outlined above. Studies were generally of low methodological quality (Table 1).

Table 2 TCRFTA—base of tongue
Full size table
Table 3 TCRFTA—soft palate
Full size table
Table 4 TCRFTA—multilevel
Full size table

TCRFTA: base of tongue

Our literature review included three non-randomized, parallel group comparative trials comparing application of TCRFTA at the BoT with either CPAP [51], submucosal minimally invasive lingual excision (SMILE) [34] technique or lingual suspension [32], and seven prospective case series [39, 45, 36, 46, 40, 26, 31].

Respiratory disturbance index (RDI)

The overall mean RDI showed a significant 40 % reduction after treatment (Fig. 2; RoM = 0.60, 95 % CI 0.47–0.76, p < 0.0001, I 2 = 51 %). Five of the seven studies found significant reductions in RDI with TCRFTA at the BoT [39, 51, 40, 26, 32], whereas the remaining two studies [46, 31] showed a non-significant reduction after TCRFTA treatment. Heterogeneity was moderate.

Fig. 2
figure 2

RDI: TCRFTA at base of tongue

Full size image

Lowest oxygen saturation (LSAT %)

Short-term (<12 months) post-TCRFTA LSAT improved significantly (RoM = 1.05, 95 % CI 1.01–1.10, p = 0.02, I 2 = 37 %; Fig. 3). Three studies showed a significant improvement, whereas one study showed a non-significant increase with TCRFTA [51]. Heterogeneity was low to moderate.

Fig. 3
figure 3

LSAT: TCRFTA at base of tongue

Full size image

Epworth Sleepiness Score (ESS) and bed partner’s snoring Visual Analogue Scale (VAS)

Short-term (<12 months) reduction in the mean ESS (RoM = 0.59, 95 % CI 0.51–0.67, p < 0.00001, I 2 = 0) and VAS snoring (RoM = 0.48, 95 % CI 0.37–0.62, p < 0.00001, I 2 = 0) was significant (Figs. 4 and 5, respectively). Heterogeneity was low.

Fig. 4
figure 4

ESS: TCRFTA at base of tongue

Full size image
Fig. 5
figure 5

VAS snoring: TCRFTA at base of tongue

Full size image

Other short-term outcomes

Other reported short-term effects were VAS speech, VAS swallowing, and generic quality-of-life questionnaire (SF-36). Only three studies [39, 46, 40] reported VAS speech and swallowing and statistical analysis showed no significant improvement (VAS speech—RoM 0.92, 95 % CI 0.56–1.49, p = 0.72, I 2 = 94 %; VAS swallowing—RoM 0.68, 95 % CI 0.31–1.47), p = 0.33, I 2 = 61 %) post-TCRFTA with significant heterogeneity between studies. Two studies [39, 51] reported the SF-36 health survey, with a homogenous improvement in four of the eight reported domains; role physical—RoM 1.16, 95 % CI 1.02–1.32, p = 0.02, I 2 = 0; bodily pain—RoM 1.14, 95 % CI 1.06–1.24, p = 0.001, I 2 = 0; vitality—RoM 1.24, 95 % CI 1.07–1.44, p = 0.004, I 2 = 0; and social functioning—RoM 1.11, 95 % CI 1.02–1.20, p = 0.01, I 2 = 0.

Long-term outcomes

Long-term (>12 months) follow up of RDI and ESS was reported by two studies [36, 32]. The RDI showed a significant 36 % reduction (RoM 0.65, 95 % CI 0.53–0.80, p < 0.00001, I 2 = 0), whereas improvement in ESS was not maintained after 2 years (RoM 0.62, 95 % CI 0.33–1.16, p = 0.13, I 2 = 83 %). When compared to short-term (<12 months) results, both RDI and ESS demonstrated a trend towards worsening at long-term follow up (>12 months). One study reported a significant improvement in the change scores between post-treatment and long-term follow up for VAS snoring (p = 0.01), VAS speech (p = 0.02), VAS swallowing (p = 0.09), and two of the eight SF-36 domains (vitality, p = 0.05; social functioning, p = 0.03); and also the mental component score, p = 0.03 [36].

Adverse events

Adverse events following TCRFTA were reported by most studies (Tables 2, 3, and 4). Side effects of treatment included ulceration of the tongue base mucosa, odynophagia, pharyngodynia, mild-to-severe tongue edema, ecchymosis, hematoma, transient neuralgia, transient tongue deviation, and hypoglossal nerve injury. Eight cases of infection and two cases of tongue base abscess were reported by studies that did not use perioperative antibiotic prophylaxis [39, 51, 46]. Oral thrush and postoperative vasovagal reaction were relatively rare complications.

TCRFTA: soft palate

Five articles describing outcomes following TCRFTA of the SP were included [28, 27, 48, 24, 25] (Table 3) in the review

Respiratory Disturbance Index (RDI)

After excluding Terris et al. [48], a study that lacked follow up PSG data, the remaining two observational studies [28, 27] were included in the analysis. A non-significant trend towards improvement of short-term mean RDI (RoM 0.67, 95 % CI 0.43–1.03, p = 0.07, I 2 = 83 %) was observed with significant heterogeneity between the two studies.

Epworth Sleepiness Score (ESS) & bed partner’s snoring Visual Analogue Scale (VAS)

One study [24] did not report ESS, all five studies reported VAS scores. ESS did not improve significantly with TCRFTA treatment (RoM 0.85, 95 % CI 0.63–1.15, p = 0.29, I 2 = 0 %), However, reduction in mean VAS snoring was significant (RoM 0.39, 95 % CI 0.30–0.50, p < 0.00001, I 2 = 0 %).

TCRFTA: multilevel

Eleven articles were identified by literature search and reviewed in detail (Table 4). Nine of the 11 studies met inclusion criteria for the meta-analysis [33, 52, 42, 47, 44, 29, 38, 49, 30]. Only one study was a randomized controlled prospective trial [52], two were non-randomized comparative parallel group trials [42, 38], and the remaining six were either prospective [33, 47, 44, 29, 49] or retrospective [30] case series. Although most studies restricted application of radiofrequency to the BoT and SP only, three studies [33, 29, 30] included additional sites such as nasal turbinate and tonsils.

Respiratory Disturbance Index (RDI)

Short-term improvement in RDI was calculated using one randomized, and seven observational studies (Fig. 6). The overall 41 % reduction in mean RDI was significant (RoM 0.59, 95 % CI 0.45–0.79, p = 0.0003, I 2 = 93 %), but the results were significantly heterogeneous (p < 0.00001) between the included studies. Sub-group analysis identified two studies that reported significantly better results as compared to the other included articles [29, 38]. Neruntarat and Chantapant [38] and Ceylan et al. [29] reported a 65 and 46 % reduction in RDI with TCRFTA respectively. The patients in these two studies were somewhat less obese on average than those from the remaining studies and this might be responsible for the different success rates. Indeed, in one of these studies (36), BMI was a univariate predictor of surgical success. We reanalyzed the data excluding these studies. The remaining six studies continued to show a significant reduction in RDI (RoM 0.70, 95 % CI 0.63–0.77, p < 0.00001, I 2 = 0 %) with a high-level of agreement between the pooled studies (Fig. 7).

Fig. 6
figure 6

RDI: multilevel TCRFTA (3.1.1 and 3.1.2)

Full size image
Fig. 7
figure 7

RDI: multilevel TCRFTA (3.3.1 and 3.3.2)

Full size image

Lowest oxygen saturation (LSAT %)

Pre- and post-TCRFTA LSAT were reported by one randomized and three observation studies [52, 42, 29, 38]. There was a non-significant trend towards a short-term increase in LSAT with treatment (RoM 1.03, 95 % CI 1.00–1.06, p = 0.08, I 2 = 73 %), but significant heterogeneity was noted (p = 0.01) (Fig. 8).

Fig. 8
figure 8

LSAT: multilevel TCRFTA

Full size image

Epworth Sleepiness Score (ESS) & bed partner’s snoring Visual Analogue Scale (VAS)

Seven of the included nine studies reported short-term post-treatment ESS, including one randomized prospective placebo controlled trial [52] and six observational studies [33, 42, 47, 29, 38, 49]. Four observational studies reported VAS snoring [33, 47, 38, 49]. Initial analysis showed significant heterogeneity between the results of these trials. After exclusion of the two studies that showed unusually favorable results [29, 38], a significant 22 % reduction in short-term mean ESS (RoM 0.78, 95 % CI 0.69–0.89, p = 0.0001, I 2 = 0 %) and 46 % reduction in mean VAS snoring (RoM 0.54, 95 % CI 0.41–0.70, p < 0.00001, I 2 = 59 %) were observed (Fig. 9).

Fig. 9
figure 9

ESS and VAS: multilevel TCRFTA

Full size image

Discussion

Untreated OSA has a number of adverse clinical implications. CPAP is a well-validated first line treatment option for OSA; however, a substantial number of patients are unable to use CPAP. Among surgical options, UPPP is the best studied; however, limited effectiveness, potential complications, and a prolonged recovery [53] make it less attractive.

TCRFTA appears to be a safe alternative that can be used in conjunction with surgery or as a stand-alone procedure. It can be performed safely, potentially as an outpatient procedure under local anesthesia. The procedure can be repeated in several sessions to achieve the desired result. Radiofrequency ablation can be performed at the base of the tongue, soft palate or as a multi-level procedure, while safely sparing adjacent tissue.

Due to significant heterogeneity between studies, it is hard to draw any strong conclusions on the ideal patient population who will benefit from TCRFTA. However we can state the following:

  1. 1.

    Flexible nasopharynoscopy with Mueller’s maneuver (to verify site of obstruction) and cephalometric radiography (to rule out concomitant skeletal abnormalities) are adjunct tests to physical examination that may aid in patients selection. Other tests to be considered included volumetric tongue MRI and drug induced sleep endoscopy (DISE). DISE is emerging as the tool of choice to plan surgical therapy for sleep apnea. A recent study suggested that DISE provided more information on the site of obstruction in the upper airway particularly regarding the hypopharynx than awake screening tools [54]. However, consensus on an accepted scoring and classification system has yet to be achieved [55]. Nevertheless, future trials on TCRFTA will likely use DISE for patient selection. .

  2. 2.

    Most studies evaluated this therapy in patients less than 65 years of age with a BMI < 35.

  3. 3.

    TCRFTA has been trialed mostly in mild to moderate OSA. Data is sparse in the severe OSA group.

  4. 4.

    Majority of the studies included patients who had failed other modalities such as CPAP or mandibular advancement devices.

  5. 5.

    Some trials included patients with prior pharyngeal surgery; however, most excluded these patients.

Most trials excluded patients with severe underlying co morbidities, high surgical risk profiles or history of pacemakers.

Complications can occur but appear to be less frequent than in surgical upper airway procedures. Our review primarily aimed at examining the short-term efficacy of TCRFTA performed at different anatomical sites. The primary outcomes variables available for analysis were polysomnography based (RDI and LSAT), a measure of sleepiness and reduction in snoring.

TCRFTA at base of the tongue

We found significant reductions in RDI, LSAT, ESS and snoring. This suggests that this procedure has efficacy at least in the short term. There was considerable heterogeneity between subjects within studies with some subjects not showing any improvement or occasionally even worsening after the procedure. In addition, it would be useful to know how many subjects had a reduction in AHI to less than 15/hour or 10/hour after the procedure, but this information was rarely reported. Subjects whose AHI remains above 15/hour or above 5/hour with residual sleepiness would be considered to require additional therapy. There are now more options for treatment for sleep apnea than there was in the past. Mandibular advancement device efficacy has been well characterized and they have the advantage that if ineffective the patients just needs to stop using them. In contrast, if TCRFTA is ineffective, the patient has gone through a surgical procedure with possible side effects without benefit. Nevertheless, there are patients who would be willing to accept this risk if it meant that they might not have to use a nightly therapy. Our meta-analysis suggests promising short-term results. Longer-term results were quite limited but suggested that some of the improvements might wane with time. A recent meta-analysis [17], included primarily patients who were treated for snoring suggested that a reduction in snoring was maintained for over a year but truly long-term follow up was lacking and is essential before the true benefits of this therapy can be appreciated.

TCRFTA at soft palate

There was a paucity of studies examining TCRFTA of the soft palate alone. Other than an improvement in snoring, no significant short-term improvements could be found. Thus, based on the limited available data, TCRFTA of the soft palate alone appears to have limited efficacy. This result is not unexpected, as any improvement in obstruction at the level of the soft palate only may not be sufficient to significantly improve OSA. It is for this reason that ENT surgeons have moved towards multi-level TCRFTA. However, TCRFTA at the level of the soft palate may have some benefits in terms of reduction in snoring index and Epworth scores in habitual snorers without OSA [56]. However, a recent study examining long-term follow up for TCRFTA of the soft palate [18] found that snoring returned in most patients over time albeit its intensity was lower.

TCRFTA multi-level

There was a significant reduction in RDI after multilevel TCRFTA. However, significant heterogeneity was noted between studies. Two studies appeared to have better results than the remaining studies. The patients in these two studies were somewhat less obese than in the other studies and this might be responsible for the differing results. Racial differences seem an unlikely explanation as the two studies were from Thailand and Turkey, respectively. Thai and Turkish ancestry are very different, and this difference exceeds the difference between Turkish ancestry and the North American and European patients that make up the other studies, we reanalyzed the data excluding these studies and were reassured that the significant reduction in RDI remained. Oxygenation parameters were not significantly improved after TCRFTA. Sleepiness as measured by the ESS and VAS snoring both improved significantly post-TCRFTA, but heterogeneity between included studies was high. After exclusion of the aforementioned two studies causing heterogeneity in our results, a significant 22 % reduction in short-term mean ESS (RoM 0.78, 95 % CI 0.69–0.89, p = 0.0001) and 46 % reduction in mean VAS snoring (RoM 0.54, 95 % CI 0.41–0.70, p < 0.00001) was observed. Thus, multilevel TCRFTA resulted in both objective and subjective improvement in the short term. Again, long-term results are largely lacking. The sparse long-term data available suggests some diminution of response over time particularly if weight gain occurs (36). Similar to TCRFTA at base of tongue, there is considerable inter-subject variability with some subjects showing no improvement or even worsening after the procedure.

Study limitations

The studies on which this meta-analysis is based were not of high methodological quality which lends caution to our conclusions. In particular, only 2 studies [52, 25] had a placebo control group. ESS and VAS snoring were the most frequently measured subjective measures. Disease specific quality of life questionnaires or other measures of sleepiness were not available for analysis. Generic quality of life measurements were available only in a few studies. Long-term outcome measurements were sorely lacking. Stratification of subjects by severity of sleep apnea and BMI could not be performed as only one study stratified responses based on the severity of sleep apnea [31] and no study stratified based on the BMI. Evaluating response rates based on an AHI below a threshold value (15, 10, or 5/hour) could not be performed as this information was provided in only 1 study [30]. In this study, none of the subjects (0/36) had an AHI less than 5/hour after TCRFTA. However, this may be too strict a criteria for surgical success. It would also have been helpful to study the role of PAP in patients who failed TCRFTA to see if prior TCRFTA adversely affected subsequent CPAP acceptance; however, this data was rarely provided. While analyzing data for Multi-level TCRFTA we failed to conclusively explain the heterogeneity generated by two trials and hence analyzed the data with and without them.

Clinical implications

Temperature controlled radiofrequency tissue ablation appears to be an effective treatment option for OSA based on short-term follow up. Ablation at the base of the tongue or at multiple levels appears to be more efficacious than ablation at the level of the soft palate only. Complication rates were infrequent and rarely serious in nature. Patients with sleep apnea develop obstruction because of a small and more importantly a more collapsible upper airway. TCRFTA can increase upper airway size and the fibrosis produced may increase pharyngeal stiffness and reduce pharyngeal collapsibility. Pharyngeal collapse can occur at multiple sites in the upper airway so treatment at one site may not be sufficient to control sleep apnea. Patients are often reluctant to undergo staged multi-level surgeries to treat sleep apnea. TCRFTA can allow a multi-level approach at a single sitting and the procedure can be easily repeated to achieve the desired result. This makes TCRFTA potentially quite attractive as a stand-alone therapy for OSA in patients intolerant to CPAP. Long-term follow-up studies and better data on short-term cure rates (percentage of patients who have an AHI below a defined threshold (10 or 15/hour) are needed before this procedure can be confidently added to the treatment armentarium for sleep apnea.