Effect of orthopedic and functional orthodontic treatment in children with obstructive sleep apnea: A systematic review and meta-analysis

Orthodontic treatment is suggested in growing individuals to correct transverse maxillary de ﬁ ciency and mandibular retrusion. Since, as a secondary effect, these orthodontic procedures may improve pediatric obstructive sleep apnea (OSA), this systematic review assessed their effects on apnea-hypopnea index (AHI) and oxygen saturation (SaO2). Twenty-ﬁ ve (25) manuscripts were included for qualitative synthesis, 19 were selected for quantitative synthesis. Five interventions were analyzed: rapid maxillary expansion (RME, 15 studies), mandibular advancement (MAA, ﬁ ve studies), myofunctional therapy (MT, four studies)


Introduction
Obstructive sleep apnea (OSA) is a common disease belonging to the sleep-disordered breathing (SDB), and characterized by repetitive episodes of complete and/or incomplete obstruction of the upper airways which occur during sleep [1,2]. Due to the hypoxemia associated with obstruction episodes, untreated OSA is a potentially life-threatening disorder [3]; furthermore, as for in adults, OSA in children presents several metabolic, and cardiovascular consequences, detrimental behavioral effects, neurocognitive impairments and academic underperformance [4]. Common symptoms associated with pediatric OSA are fragmented sleep, mouth breathing, snoring, nocturnal enuresis, headaches, and systemic inflammation [5e8]. The estimated prevalence of OSA among children ranges between 1% and 4%, depending on the different diagnostic method adopted, which could include nocturnal sleep laboratoryebased polysomnography (PSG) and inhome sleep study, and different cut-offs [9]. Furthermore, numerous previous researches have adopted different patientreported and parent-reported questionnaires to determine the presence of OSA in children [9], but studies have shown that questionnaires and clinical history alone are not adequate in clinical practice to distinguish primary snoring from OSA in children [10] and therefore nocturnal sleep laboratoryebased PSG should be used as gold standard for the diagnosis of OSA through the assessment of apnea-hypopnea index (AHI) [11]. However, the high economical costs, the limited availability of sleep centers, the complexity of the equipment, and the need of specialized expertise for diagnosing children, limit the use of PSG in children for routine purposes [12]. Furthermore, the unfamiliar laboratory setting and the placement of sensors and electrodes by a stranger can represent stressful events for the young patient. In turn, this emotional status can affect the compliance and the quality of sleep [13]. Portable and home sleep tests have been suggested as an alternative option to PSG for the screening of OSA in children, especially whenever PSG is not feasible (for instance, in low income countries or rural areas) [14].
The pathophysiology of OSA in children is complex and presents a multifactorial etiology. Major risk factors are adenotonsillar hypertrophy, obesity, neuromuscular disorders and craniofacial anomalies [15]. Since enlarged tonsils and adenoids remain the main anatomical condition that reduces the caliber of the upper airways, it has been shown that the peak of OSA in children can be observed around 10 years of age when the lymphatic system shows its maximum development [6]. Craniofacial features that have been linked with higher risk of OSA symptoms in children include macroglossia and incorrect tongue posture, retrognathic mandible, transversal maxillary deficiency, high palatal vault, increased total and lower anterior facial heights, and a more anterior and inferior position of the hyoid bone [15e19]. Interestingly, a recent epidemiological study pointed out doubled SDB risk prevalence in the pediatric orthodontic population compared with a healthy pediatric population (10.8% vs. 5%, respectively), supporting the idea that orthodontic practitioners should routinely screen SDB in their clinical practice [20,21]. On the other hand, inconsistent findings are reported in the literature regarding the prevalence of malocclusion in children with OSA. In particular, authors reported higher prevalence of posterior crossbite, and abnormal overjet and overbite in children with OSA compared to a control group [22], while other researches pointed out that prevalence of malocclusion in children with suspected sleep-disordered breathing was not greater than what has been reported for the general population [23].
The treatment strategy of OSA in children depends on several factors, including the severity of the syndrome and the etiology of the obstruction. Adenotonsillectomy (AT) is recognized as first-line treatment in case of adenotonsillar hypertrophy, which is supported by evidences of reduced symptoms and improved PSG outcomes after surgery [24]. However, it has been reported that in approximately half of the cases, surgical approach does not completely treat pediatric OSA, and up to 68% of patients can show post-surgical relapse [25], especially when underlying skeletal disharmonies, increased soft tissue crowding of the oropharynx or increased body mass index are present [26e28]. Functional and orthopedic orthodontic treatment (such as mandibular advancement, MAA, or rapid maxillary expansion, RME) have recently been proposed as a part of comprehensive treatment for pediatric OSA patients with craniofacial alterations. A recent systematic review (SR) reported evidence that AT and orthodontic treatment are more effective together rather than separately to cure OSA in pediatric patients [29]. A previous SR with metaanalysis examined the effects MAA on pediatric OSA, pointing out significant reduction in the mean AHI in the treatment group, compared to untreated controls [30]. However, the diagnosis of OSA was not based on PSG in all primary studies included. Other recent SRs investigated the efficacy of single orthodontic approaches (MAA or RME) in the treatment of pediatric OSA [30,31]. Favorable results were observed in both reviews, but the literature searches are currently out of date (search limit 2018 and 2017 respectively), and numerous studies could have been published in the following years considering the growing interest in the topic. Finally, in 2019 authors provided an overview on the effects and a comparison of the outcomes of various treatments in the management of pediatric OSA with adenotonsillar hypertrophy [32]. The authors concluded that, irrespective of the intervention used, complete resolution of OSA was not achieved in most included trials. However, the results of this review are limited to OSA patients presenting with specific adenotonsillary characteristics and cannot be extended to the general population. Furthermore, different type of surgical, pharmacological, and functional/orthodontic treatments were included.
The current study aimed to summaries the evidence on the effects orthodontic treatments on respiratory outcomes in OSA pediatric patients. The goal was to provide an overview of all types of orthodontic/orthopedic approaches available in the literature for the management of OSA pediatric patients. The question to be answered was as follows: 'Is there any improvement in the polysomnographic parameters of OSA growing patients after orthopedic/functional orthodontic treatment?'

Methods
In accordance with the preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement [33], the protocol for the SR was registered (PROSPERO ID: CRD42020180164).

Selection criteria
According to the PICO approach, the inclusion criteria were the following: Population: children and adolescents (less than 18 years old in age), without craniofacial syndromes and with a polygraphic diagnosis of OSA Intervention: all kind of orthopedic/functional orthodontic interventions Comparison: no treatment, waiting list, placebo or other therapies Outcome: primary outcome was AHI; secondary outcome was oxygen saturation level (SaO2; minimum and mean).
The exclusion criteria were: multi-bracket orthodontic therapy and extractive orthodontic therapy; systematic reviews, reviews, opinion articles, letters to the editor, case series, case report, animal studies; studies including; sample size 10 patients studies including syndromic patients and/or patients affected by cleft lip and palate; dual publications; studies in which OSA diagnosis was performed by means of questionnaires, self-report, symptoms, clinical examination, and/or pulseoxymetry.

Literature search
Four electronic databases were investigated from their inception, up to January 2020: Medline (via PubMed), Scientific Electronic Library Online (SciELO), Cochrane Central Register of Controlled Trials, and Scopus. The key words obstructive sleep apnea AND orthodontic* were adopted, and search strings were adapted to each database, according to the appropriate database-specific indexing terms and syntax (Table S1). An update of the search was conducted on January 2021. In addition, Google Scholar was explored for grey literature search, and a hand-search of the reference lists of the included studies was carried out. No restrictions on language or publication year were applied for the electronic searches. Full-texts in Chinese and Japanese languages were excluded.

Study selection
Two reviewers (BZ and RB) independently screened the articles by title and abstract, using the software Rayyan (http://rayyan.qcri. org) [34]. In case of disagreements, a third reviewer (VD) was contacted. In case of uncertainty, all potentially eligible studies were retrieved in full texts for further evaluation.

Data extraction
The following data were collected from each included study: author and year of publication, country, sample characteristics (sample size, gender, age), drop out, study design, description of treatment and control group, interventions, orthodontic diagnosis, OSA diagnosis, follow-up periods, success endpoint, outcome data (limited to the respiratory function) and general conclusion. Whenever relevant information was not available, the authors were privately contacted by email. Data extraction was performed independently by two reviewers (BZ and RB) using a preformed, standardized spreadsheet that was developed and agreed upon by the review team. A third reviewer (VD) was contacted to solve residual disagreements.

Risk of bias (quality assessment) of the studies
A methodological quality assessment of each study was performed by BZ and SIP independently, and any discrepancies were resolved through discussions with RB.
To evaluate the risk of bias of randomized controlled trials (RCT), the Cochrane Collaboration "risk of bias" (RoB-2) tool was used [35]. Similarly, to evaluate the risk of bias of non-randomized studies, the Cochrane Collaboration "risk of bias in nonrandomized studies e of interventions" (ROBINS-I) tool was applied [36]. In particular, reporting of previous AT, measurement body mass index, and reporting of ear, nose and throat assessment were considered as confounders. Also, age ranges and orthodontic diagnosis at the baseline were evaluated in the selection of participants.

Certainty of the evidence
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach was used to judge the quality of the evidence, using GRADEpro software [37]. Each outcome could obtain a high, moderate, low, or very low evidence value depending on the study design, risk of bias, inconsistency, indirectness, imprecision, and publication bias [38]. The downgrading of evidence was based on the following criteria: (1) for study limitation if the majority of studies (>50%) were rated as high risk of bias; (2) for inconsistency if heterogeneity was considerable (I2>75%) [39]; (3) for indirectness if the baseline orthodontic diagnosis or the device used for treatment were extremely diverse among studies or not clearly described; and (4) for imprecision if meta-analysis had a small sample size (n<300) or large CI.

Data synthesis
Statistical heterogeneity was explored using a the I2 statistics, and the level of significance was set at P < 0.05. To analyze changes in AHI and SaO2, a random-effects model was chosen, and a moderator analysis was performed. Data collected for the moderator analysis were the following: study design of primary studies, risk of bias, % of males, age, type of appliance, activation protocol, previous AT, OSA diagnosis (PSG or polygraph), and baseline OSA severity.
Quantitative data were computed as post-treatment minus pretreatment and different meta-analysis were performed according to different follow-up times. The pooled estimate of the standardized mean difference (SMD) for each outcome and respective 95% Confidence Intervals (CIs) were computed. The statistical significance of the hypothesis test was set at P < 0.05 (two-tailed Z tests). Studies with incomplete statistical reporting (e.g. absence of standard deviation values, different number of patients between time points) or non-comparable assessments were excluded from the meta-analysis. The meta-analysis was performed with ProMeta software (Internovi, Cesena, Italy). Publication bias was evaluated through a visual inspection of funnel plots, in case more than 10 studies per outcome could be retrieved.

Study selection
A total number of 764 records were identified through electronic search, and 36 additional records were identified through hand search. One-hundred and eight (108) full-text were read, and 84 articles were excluded (references and reasons for exclusion are listed in Table S2). Finally, 25 manuscripts were included in the review for qualitative synthesis [40e64]. Nineteen (19) articles were included in the quantitative synthesis. PRISMA flow diagram of study identification, screening, eligibility and inclusion phases is shown in Fig. 1.

Study characteristics
Summaries of study characteristics and results of the included studies, divided according to the different orthodontic therapy, are shown in Table 2 to 4 In particular, five interventions were collected: RME (15 studies, Table 1), MAA (five studies, Table 2), myofunctional therapy (MT) (four studies, Table 3), and RME in combination with MAA (one study, Table 4). Furthermore, different appliances and protocol were adopted. RME was performed with fixed tooth-anchored appliances, supported either by two-or fourbands, or by dental acrylic segments. MAA was performed with Twin-Block, modified monobloc, modified Planas appliance, or acrylic resin personalized oral appliance. MT included both active and passive approaches: passive approaches involve the use of custom-made adjustable oral devices, while active approaches rely solely on isometric and isotonic exercises which involves tongue, soft palate, and lateral pharyngeal wall designed in order to improve suction, swallowing, chewing, breathing, and speech functions. The largest sample included 110 patients [47], while the smallest included 11 patients [40]. In the vast majority of studies (20) the OSA diagnosis was performed with laboratory PSG; three studies adopted portable home cardiorespiratory monitoring type 3 devices [46,49,53], one study employed a portable polygraph [57], and another study adopted an ambulatory polygraph [40]. Participants' age ranged from 5.03 ± 2.03 years [61] to 12.27 ± 1.93 [46]. The timing of the post-intervention assessments varied widely (from three week to six years follow-up). Of the 25 included records, six articles were RCTs while 19 were non-randomized studies (five retrospective and 14 prospective). Most of the prospective studies were uncontrolled before-after studies (ten studies). Among the four remaining prospective nonrandomized studies with a control group, two compared different therapies (AT vs. RME [53,61]), one adopted a control group of non-OSA untreated patients [42], and one compared full-birth and premature birth children [41]. With regards to the RCTs, three compared different therapies or combination of therapies (AT þ RME vs. RME þ AT [43], active MT vs. passive MT [47], and MT þ nasal wash vs. nasal wash only [63]), while three adopted an untreated control group [46,48,58].
Some of the data collected for the meta-analsysis (activation protocol, and previous AT) were not included among moderators since information were not consistently reported across studies (frequently non reported or not clearly described). In addition, "type of appliance" was not included as a moderator in the RME meta-analysis since all the devices analyzed in the included primary studies were fixed expanders. Finally, publication bias was not explored since none of the assessed outcomes included more than 10 studies.

AHI
Eight studies evaluated AHI changes within six months from the end of active RME, of which six reported a significant reduction of the AHI compared to baseline values, while the remaining two also reported a reduction of the AHI, but no statistical analysis was performed. The only study with an untreated control group [46] reported reduction of the AHI also among controls, and nonsignificant differences between treated and untreated individuals in the short-term were observed.
Between six months and one-year follow up, 11 studies reported significant reduction of the AHI compared to baseline values, and 1 study pointed stable results (non-significant changes) compared to the immediate post-expansion findings. In addition, another study reported reduction of the AHI after one year compared to baseline values, but no statistical analysis was performed. The only study reporting non-significant changes in the AHI one year after RME [43] supported the need of treatment combination with AT, since significant AHI reduction was observed following RME þ AT, despite treatment sequence.
Only one study [44] reported a very long-term follow-up assessment after RME, pointing out recurrence of the pathology in some individuals during adolescence.
Immediately after RME, the meta-analysis showed significant reduction in the AHI values, with considerable heterogeneity (SMD: *Consider, if feasible to do so, reporting the number of records identified from each database or register searched (rather than the total number across all databases/registers). **If automation tools were used, indicate how many records were excluded by a human and how many were excluded by automation tools.       Figs. 2 and 3), and the level of evidence supporting this finding was very low (Table S3). The risk of bias the primary studies was a significant moderator, pointing out that higher effects were reported by studies presenting higher risk of bias ("serious risk" SMD: -6.54 vs "moderate risk" SMD: -3.32).
In the short (within six months) and in the medium term (within 12 months) follow up, significant reduction of AHI following RME was found in the meta-analysis (SMD: -2.97, CI: -4.35, -1.59, P<0.001, and SMD: -1.08, CI: -1.56, -0.61, P<0.001, respectively, Figs. 2 and 3). Concerning both time-points, high rate of heterogeneity was observed (I 2 : 95.03%, and I 2 : 80.04%, respectively) and the quality of the body evidence was very low (Table S3). At six months follow-up, risk of bias, OSA diagnosis and baseline AHI were significant moderators. In particular, the two RCTs with some concern of risk of bias and the non-randomized study with moderate risk of bias showed higher effects, as compared to the two non-randomized studies with serious risk of bias ("some concern of risk of bias" SMD: -5.54; "moderate risk of bias" SMD: -5.22; "serious risk of bias" SMD: -0.35). In addition, five studies adopting PSG as diagnostic method present a pooled SMD (SMD: -3.85) higher than those using polygraphic recording (SMD: -0.65). Finally, studies including sample presenting severe AHI index (AHI >10) at the baseline, showed the highest findings (SMD: -6.65).
At 12 months of follow-up, the only significant moderator was the baseline AHI index, supporting highest effect in individuals presenting with severe AHI before treatment (SMD: -4.10).

SaO2
Few studies evaluated the changes in the mean and minimum SaO2 within six months following RME, and the results pointed mainly toward lack of changes. On the other hand, the majority of the studies reported results considering one-year of follow-up pointing out significant increase in both minimum and mean SaO2, compared to baseline values. As for AHI, the only study reporting long-term data [44] showed recurrence of the pathology during adolescence also with regards to SaO2 values.
Immediately after RME, the meta-analysis pointed out significant increase of the minimum SaO2, with moderate heterogeneity (SMD: 2.03, CI: 1.06, 3.00, P<0.001, I 2 ¼ 65.17 %, Figs. 2 and 4). This finding was provided with low quality of evidence (Table S3) and none of the studied moderators was significant. Similarly, also in the short (within six months) and in the medium term (within 12 months), significant increase of minimum SaO2 was found after RME (SMD: 4.54, CI: 2.12, 6.96, P<0.001, and SMD: 2.60, CI: 0.41, 4.79, P<0.001, Figs. 2 and 4). High rate of heterogeneity was observed both in the short term (I 2 : 95.20%), and in the medium term (I 2 : 91.74%). These findings present with low and very low quality of evidence, respectively (Table S3). At six months followup, OSA diagnosis was a significant moderator, showing increased effects from the three studies using the PSG (SMD: 7.17), as compared to the only study adopting polygraphic recordings (SMD: 0.43). At 12 months, baseline AHI severity was found as significant moderator, showing enhanced effects in individuals presenting with severe AHI before treatment (SMD: 3.78).  In the pre-post treatment assessment, both at six and at 12 months from the beginning of the treatment, all included studies assessing this outcome (four studies) reported significant reduction of AHI. Also, these results were significantly different from those obtained from the untreated groups that showed non-significant changes or even increase of the AHI [48,58].

AHI
Significant changes in the AHI values (SMD: -1.33, CI: -2.18, -0.47, P<0.005, Figs. 2 and 5) with considerable heterogeneity (I 2 ¼ 84.18%) were observed in the meta-analysis considering the data within six months after treatment. This finding was supported by very low quality of evidence (Table S4). The type of appliance was found as significant moderator, with the best results showed by the modified monobloc (SMD: -2.40).

SaO2
Three studies evaluated SaO2 following MMA: two of them reported non-significant changes in the mean SaO2, while one study supported a significant increase of minimum value only and one study reported significant increase of both mean and minimum values following treatment. The meta-analysis showed nonsignificant changes in the minimum SaO2 within six months of treatment (SMD:0.13, CI: -1.46, 1.73, P ¼ 0.869, Figs. 2 and 6) with high heterogeneity (I 2 ¼ 94.63%). This result was supported by very low evidence (Table S4). The moderator analysis showed significant findings with regards to the type of appliance and OSA diagnosis. In particular, the device made by two thermoplastic bite splints showed some positive effects (SMD: 0.95), while the modified monobloc showed negative findings (SMD: -0.67). Similarly, the study using PSG for OSA diagnosis presented favorable effects, while the study adopting polygraphic recordings showed unfavorable results.

AHI
Significant reduction in the AHI was observed at six months follow-up with passive MT. Also, passive MT showed significantly better results with regards to AHI reduction, compared to active MT. In the long-term follow-up (more than three years), significant effect of MT following orthodontic treatment was observed, compared to individuals who did not performed any reeducation.
No meta-analysis was performed regarding this outcome since different therapeutic approaches were found.

SaO2
Two studies supported non-significant changes of the mean SaO2, following both passive and active MT. On the other hand, one study underlined that nasal washes in combination with MT provided significant improvement of mean SaO2. Finally, in the longterm follow up, minimum SaO2 was significantly increased in individuals performing regular MT, compared to those not following any reeducation program.
3.6. RME þ MAA Treatment combination supported significant decrease of AHI after 9 months of therapy.

SaO2
None of the included studies assessed SaO2 changes following combined treatment.

Risk of bias (quality assessment) of the included studies
-RCT Two RCTs on RME were rated with some concern of bias ( Table S5), while of the two RCTs on MAA, one was rated with some concern and one with high risk of bias (Table S5). Finally, both RCTs (two out of two) on the MT were rated with high risk of bias (Table S5).
-Non-randomized studies Eight studies on RME presented serious risk of bias, while the remaining five studies presented moderate risk of bias (Table S6). With regards to MAA, two studies presented serious risk of bias and only one study presented moderate risk of bias (Table S6). Finally, the only study on MAA þ RME (Table S6) and the two studies on MT (Table S6) presented with serious risk of bias. Common reasons for loosing points in the quality assessment were: poor description and evaluation of the sample characteristics, variability in the age range of the sample, absence of the uniform orthodontic diagnosis at baseline, lack of appliance description (type of appliance, design, screw, anchorage teeth), lack of description of activation protocol.

Discussion
The aim of this SR was to assess the effects of different types of orthopedic and functional treatment on the respiratory outcomes in OSA children and adolescents. Overall, within one year from the baseline, favorable respiratory outcomes have been found following different types of orthodontic treatments, suggesting that interceptive orthodontics might play a role in the multidisciplinary therapeutic approach of pediatric OSA in children. However, the level of the body evidence supporting those findings ranged between very low and low for the majority of the outcomes of the studies, thus limiting the applicability of the findings. In addition, in the longer distance, few data and mainly non-significant findings have been found.

RME
RME is an effective procedure for the correction of maxillary transverse deficiency, with the primary treatment goal to increase the widths of the maxilla through the opening of the mid-palatal and peri-maxillary sutures [65]. In turn, in growing patients, this treatment can significantly affect the dimension of the nasal vault and increase the tridimensional volume of the nasal cavity [66e68] resulting in decreased nasal resistance and increased nasal flow [69e71]. These secondary effects may improve OSA [72] and support the significant improvements observed in the current SR, concerning respiratory outcomes (both AHI and SaO2%) at different time-points. Furthermore, the augmented maxillary width provides increased space available for a more forward and upward positioning of tongue, thus indirectly enhancing the oropharyngeal/retrolingual air space [73]. The results of the meta-analysis on the effects of RME showed substantial heterogeneity in all the studied timepoints. That could be explained by several factors Table 4 Characteristics of the included studies addressing the effects of Rapid Maxillary Expansion   related to patients' characteristics and treatment modalities. Some factors were included in the statistical analysis as moderators, but still several confounders were not addressed due to the lack of transparency in reporting sample characteristics across studies. In particular, despite the multifactorial etiology of the OSA, few studies considered the body mass index of the participants. For instance, in one of the included primary studies published by Quo and colleagues [56], the authors retrospectively observed, albeit not significant, an increase in the AHI in one third of the sample following expansion of the upper and lower jaw. Moreover, those patients who had a positive response to therapy did show residual OSA. These findings could be explained by the lack of control of confounding factors that might have contributed to the worsening of the symptoms during the follow-up period and could also be ascribed to the lack of an adequate diagnosis at baseline. Furthermore, the assessment of adenotonsillar hypertrophy as either inclusion or exclusion criteria was not consistently reported among studies, and presence or absence previous adenotonsillectomy was also extremely diverse among studies. The type of appliance was not considered a potential confounder as all included studies measured the effects of tooth-anchored fixed expanders. However, the devices presented different designs (2-bands, 4-bands, acrylic supported, or not reported) and different anchor teeth (second deciduous molars, first permanent molars, or not reported). In addition, although the majority of the studies reported a screw activation protocol of 2 turns/day (approximately 0.50 mm/ day), other studies applied slower expansion rates [43,46], or adapted the expansion protocol according to participants' age [56]. Moreover, few studies included also participants performing transversal expansion at the lower jaw [43,44,56].
Interestingly, most of the included studies on the effects of RME reported a subjective diagnosis of maxillary constriction, narrow maxilla, cross bite or ogival palate at the baseline, thus supporting an indication for an orthodontic treatment of maxillary expansion, independently from the OSA management. Therefore, the results of the current meta-analysis can be extended only to those patients presenting with transversal discrepancies, but there is still no indication for RME in OSA patients in absence of maxillary constriction, in accordance with the recommendations of the American Association of Orthodontists white paper which support the use of RME only when there is an appropriate underlying skeletal condition [72]. Furthermore, the vast majority of the RME studies were uncontrolled studies (11 studies), or compared different therapies (3 studies); only 1 study included an untreated   control group composed by individuals with the same malocclusion pointing out reduction of the AHI also among controls and nonsignificant differences between treatment arms after 5 months of follow-up [46]. Similarly, a large, randomized, controlled trial of therapy for the pediatric OSA comparing the efficacy of early AT versus watchful waiting with supportive care, showed a significant reduction of the AHI in both groups after 7 months of follow-up [74]. These results could be explained by a trend of natural remission of OSA in children. A longitudinal study of objectively measured OSA in a community-based child cohort examined OSA's natural history from middle childhood to adolescence. A very small percentage (approximately 4%) of children and adolescents presented an objective OSA diagnosis at each time point, thus suggesting that OSA rarely persists from middle childhood through late adolescence [75]. However, both studies were conducted without taking into account the craniofacial morphology of the children included in the study sample, therefore it is still to be elucidated whether in diagnosed OSA patients with a dentofacial deformities (either transversal maxillary deficiency or Class II due to mandibular retrusion) the natural course of OSA presents a favorable prognosis with spontaneous improvement in absence of treatment. Notwithstanding, in order to clearly elucidate the role of growth in the changes of respiratory variables, controlled studies are critically needed, although ethical concerns of not treating children with OSA make it extremely difficult to recruit such control group.

MAA
The retruded position of the mandible is considered, as well as the narrow dental arches, a common feature in pediatric OSA [16]. In fact, anterior jaw-positioning was the second most studied therapy for the treatment of childhood OSA in the current SR (five studies). The primary goal of this treatment modality is to correct the dento-skeletal class II discrepancy. As a secondary effect, the anterior displacement of the mandible and of the hyoid bone leads to anterior repositioning of the tongue, thus widening the available space in the upper airways and consequently potentially improving OSA [76]. In adulthood, mandibular advancement devices are recognized as an effective, non-invasive and safe approach for mild or moderate OSA [77,78]. While for adult patients mandibular advancement devices therapy is to date a symptomatic treatment, in growing patients MAA, unlike adults, could act as a causal treatment, aiming at eliminating the retruded position of the lower jaw. According to the results of the current SR, different appliances have been used to attempt forward positioning the mandible in growing patients, but interestingly only two studied [48,64] adopted devices (Twin Block and Planas appliance) which are specifically designed to correct dentoskeletal malocclusion in patients clearly diagnosed with a Class II and/or mandibular retrusion at the baseline [79]. The remaining studies adopted devices which resemble more the concept of the adult mandibular advancement device, thus providing a more forward position of the mandible with the main purpose of increasing the airway patency, rather than an orthopedic functional treatment. Furthermore, in those studies, the baseline diagnosis of Class II and/or mandibular retrusion was not clearly stated, thus limiting the external validity of the current findings. Interestingly, in the short-term follow up meta-analysis (< six months), the appliances used in the three studies were all monobloc devices. The study by Cozza and coworkers [42] provided the highest results, probably due to the fact that the PSG was performed with the appliance "in-situ". On the other hand, in the meta-analysis performed in the longer term (<12 months) the appliances compared were both composed by two separate acrylic plates, and no difference was observed between the two devices.

MT
Even though RME and MAA in children might be useful in achieving normal upper airway size, these treatments do not ensure correct tongue posture and function and normal orofacial and pharyngeal muscle tonicity, which are crucial factors in the maintenances of upper airway patency [45]. In this context, MT may implement the correct function of oral muscles in order to avoid OSA residues and recurrence of symptoms [44,45]. One previous SR on the role of MT, pointed out a reduction in the AHI by approximately 50% in adults and 62% in children affected by OSA [80]. Extremely diverse MT protocols have been found in the current SR, including active and passive therapies, and combination of both. Hence, it was not possible to perform a meta-analysis due to the observed heterogeneity in outcomes, protocols and timing. Notwithstanding, the analyzed MT protocols supported positive effects on the PSG outcomes, with slightly more favorable results observed with passive MT as compared to active MT. It has to be mentioned that the efficacy of MT is strongly influenced by patients' compliance and adherence to the exercise protocol, which seem to be lower in younger children, and largely related to parental cooperation [41]. Furthermore, reported compliance for active MT seem to be lower than passive MT [81].
As pointed out by several authors, the primary focus of the OSA therapy should be to identify the cause of airways obstruction: the pathophysiology of pediatric OSA is complex and the respiratory disorders could be induced by a combination of factors [82]. The major confounding factors emerged from the current scientific literature on pediatric OSA patients are the lack of accuracy in the baseline ENT assessment, and the unclear exclusion/inclusion of subjects with adenotonsillar hypertrophy, or those who have previously undergone AT surgery, although it is widely recognized that tonsils dimensions and nasal flow are critical element for the prognosis of OSA patients [83].
Following the PRISMA statements, the present SR was based on strict inclusion and exclusion criteria. Assuming PSG findings as primary outcomes dramatically restricted the search results, as numerous studies evaluated treatment efficacy based only on anatomical changes of nasal cavity volume and pharyngeal space, measured from radiographic images and not on respiratory outcomes.
Although differences in the anatomy of the upper airways can be encountered [84], the radiographic assessment does not allow to directly relate changes in the respiratory function and the reliability of the radiographic assessment in OSA is questionable since it is a static measurement performed in wakefulness and in upright position, to evaluate a pathology that is dynamically expressed during sleep and in clinostatism.
Moreover, instrumental diagnosis of OSA at the baseline by means of PSG diagnosis as an inclusion criterion also limited the retrieved results; numerous studies include diagnoses based on self-reported or parental-reported questionnaires, and/or anamnestic findings.

Conclusion
RME shows significant improvement of the analyzed PSG parameters within one-year from the beginning of the therapy, but further studies are needed to determine whether these effects are stable in the longer term. MAA provides positive effects on AHI after six months of therapy, but studies lack of a clear definition of the study population. MT might be a valid adjunct to other OSA treatments, and combination of treatments should be investigated.
However, due to the overall low-very low quality of the body evidence, supported mainly by uncontrolled clinical trials, the current scientific literature does not support orthodontic interceptive treatment as the elective treatment for OSA growing patients.

Declaration of competing interest
The authors have no conflict of interest to declare.