The effect of headgear treatment on the development of obstructive sleep apnoea. A systematic review

239 © Australian Society of Orthodontists Inc. 2018 Aim: To evaluate the effect of the cervical headgear on the development of obstructive sleep apnoea and subsequent alterations of oropharyngeal dimensions. Materials and method: An electronic database search of published and unpublished literature was performed (MEDLINE, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, Clinical Trials.gov and National Research Register). Search terms included obstructive sleep apnoea, sleep disorders, pharyngeal dimensions and headgear. A risk of bias assessment was conducted using the ACROBAT-NRSI tool for non-randomised studies. Results: Of the 51 articles initially retrieved, only three were eligible for inclusion, while the remainder were retrospective cohort studies presenting serious risk of bias primarily due to undetected confounding factors or selection bias. No quantitative synthesis was possible. One study assessed the potential effect of isolated headgear treatment on apnoeic indices, while two studies described pharyngeal airway dimensions after the use of headgear alone or in combination with an activator appliance. Overall, increased apnoeic indices and the oxygen desaturation index were detected for headgear users. Dimensional changes in the posterior airway space were comparable after headgear or activator use, while combined headgear-activator treatment led to an increase in posterior pharyngeal area when compared with isolated fixed appliance therapy. Conclusions: Due to methodological inconsistencies and apparent risk of bias of the existing studies, no robust conclusions can be drawn. Prospective controlled or randomised controlled trials are deemed necessary to provide evidence on the effect of headgear treatment on sleep apnoea or pharyngeal airway dimensions. (Aust Orthod J 2018; 34: 239-249)


Introduction
Obstructive sleep apnoea syndrome (OSAS) is common and aligns within the sleep disordered breathing (SDB) spectrum. It is characterised by repetitive episodes of complete respiratory upper airway obstruction during sleep, which is associated with a reduction in blood oxygen saturation, loud snoring, sleep arousal or awakenings, a cessation of breathing and, in severe cases, cyanosis. 1 Upon awakening, patients typically feel wearied and may describe feelings of disorientation, grogginess, mental dullness and incoordination. The prevalence of OSAS has been estimated to be 4% for men, 2% for women and 1-5.7% in the paediatric population. [2][3]4,5,6 A predominance of men suffering from OSAS coupled with an increased risk in obese patients has been reported. 6 Predisposing factors in children are nasopharyngeal abnormalities, as well as hypertrophied tonsils and adenoids that narrow the upper airway. Allergies, asthma and an excessive volume of soft tissues in obese children have been recorded as risk factors. 7-,8,9 Vera Studer: vera-studer@bluewin.ch; Despina Koletsi: d.koletsi@gmail.com; Anna Iliadi: annaeliades@gmail.com; Theodore Eliades: theodore.eliades@zzm.uzh.ch STUDER, KOLETSI, ILIADI AND ELIADES Current task force initiatives on the diagnosis and management of OSAS in 2-to-18-year-old children suggest primary identification and elimination of potential pre-disposing abnormalities, stepwise reevaluation to detect residual disease and, finally, an evaluation of the need for additional treatment. 10 Skeletal jaw relationships and anatomical variations within the maxillomandibular complex may reflect a particular pattern of the oropharyngeal apparatus. It has been claimed that individuals with OSAS may demonstrate maxillary and mandibular retrognathism, larger craniocervical angles, reduced upper airway space, a longer and thicker palate and a low hyoid bone position. [11][12]13,14 However, no causal relationship between these craniofacial characteristics and obstruction of pharyngeal airway has yet been established. 15 A number of orthodontic appliances have been used to correct a skeletal jaw relationship, of which cervical headgear has been a common option. 16 Major indications for headgear use are maxillary space deficiencies, a Class II molar and skeletal relationship and to augment anchorage. Cervical headgear use may have a profound effect on the antero-posterior growth of the maxilla, while indirectly affecting mandibular growth. 17 Whether widening of the maxilla has a favourable effect on sagittal mandibular development is less clear. 18 Some studies suggest an upward and forward rotation of the mandible, 19,20 which could lead to an increase in upper airway space. 21,22 Alternative studies indicate that a forward growth restriction of the maxilla through cervical headgear use 17,23 will not result in an improvement of upper oropharyngeal airway dimensions or, worse, may result in an aggravation of breathing problems. 24 While much of the existing literature has paid special attention to the effect of cervical headgear on Class II correction, [25][26]27,28,29 there is limited evidence with regard to the potential side effects of headgear use, especially related to sleep disorders and associations with upper airway dimensions. Therefore, the aim of the present systematic review was to evaluate the effect of cervical headgear on the potential development of obstructive sleep apnoea and subsequent alterations in the oropharyngeal dimensions in young patients.

Material and methods
The Guidelines for Meta-Analyses and Systematic Reviews of Observational Studies (MOOSE) 30 were followed for the reporting of this systematic review.

Eligibility criteria
The following selection criteria were applied for this review: • Study design: Randomised Controlled Trials (RCTs), Controlled Clinical Trials (CCTs) and observational studies including a comparison group (cohort-type, case-control) were considered.
• Participants: Children or adolescent patients undergoing orthodontic treatment using headgear.
• Interventions: Any type of headgear appliance alone or in combination with other fixed or removable appliances.
• Comparators: Appliances other than headgear, or untreated control groups.
• Outcome measures: Changes in dimensions related to the pharyngeal airway or apnoea indicators. Both conventional cephalometric measurements and cone beam computed tomography (CBCT) radiography were considered, where relevant. These included but were not confined to: the distance between the soft palate and posterior pharyngeal wall, hyoid bone measurements, apnoeic index, oxygen desaturation, and the number and duration of apnoeic episodes.
• Exclusion criteria: Studies involving patients with systematic or other diseases undergoing orthodontic treatment, studies involving adult patients (>18 years of age), and cohort studies without a comparison group.

Search strategy
An electronic search within the following databases was undertaken in July, 2016  An eligibility assessment, data extraction and Risk of Bias (RoB) assessment was implemented independently and, in duplicate, by two reviewers (VS and DK). Disagreements were resolved through discussion and after consultation with a third author (AI).

Data extraction
Data extraction was performed on standardised prepiloted forms by two independently-working reviewers (VS and DK) who were not blinded to author identity nor study origin. The titles and abstracts were examined first, followed by full text screening for the possibility of including articles. Information was obtained from each included study on its design, observation period and methods, participants, interventions, comparators and outcomes.

Risk of bias within studies
The risk of bias in individual studies was assessed according to the ACROBAT-NRSI 31 tool as described by the Cochrane Collaboration. In particular, the following domains were considered: (1) bias due to confounding, (2) bias in the selection of participants in the study, (3) bias in the measurement of interventions, (4) bias due to departures from intended interventions, (5) bias due to missing data, (6) bias in the measurement of outcomes, (7) bias in the selection of the reported result. An overall assessment of the risk of bias was made for each included study (critical, serious, moderate, low, or no information). Studies receiving an assessment indicating a critical risk of bias in several domains were considered a critical risk of bias, while those with at least one item were designated to be at a serious risk of bias. Reports with a moderate risk of bias for one or more key domains were considered to be at moderate risk of bias, while those with low risk of bias in all domains were rated a low risk. No information corresponded to domains in which there was no information on which to base a bias judgement.

Summary measures and data synthesis
The clinical heterogeneity of the included studies was assessed through the examination of individual trial settings, eligibility criteria, appliances used and data collection methods. Statistical heterogeneity was to be examined through visual inspection of the confidence intervals (CIs) for the estimated treatment effects on forest plots. Also, a chi-square test was to be applied to assess heterogeneity; a p-value below the level of 10% (p < 0.1) was considered indicative of significant heterogeneity. An I 2 test for homogeneity was also to be undertaken to quantify the extent of heterogeneity.
Only studies at a moderate or low risk of overall bias were intended to be included in meta-analyses. Random effects meta-analyses were conducted, as they were considered more appropriate to better approximate expected variations in trial settings. Treatment effects were calculated through pooled standardised mean differences (SMD) in cephalometric/CBCT measurement changes along with associated 95% Confidence Intervals (95% CIs) and Prediction Intervals where applicable (at least three trials needed). For binary outcomes, Odds Ratios (OR) were considered.

Appendix 1
MEDLINE search (via Pubmed) Limits: 'Humans', no language restriction applied Publication date: no restriction Search Builder: 'All Fields' Two consecutive searches combined with "AND" Boolean operator, using "OR" between MeSH terms or keywords:

Risk of bias across studies
If more than 10 studies were included in the metaanalysis, publication bias was to be explored through standard funnel plots. 32

Additional analyses
Sensitivity analyses were predetermined to explore and isolate the effect of studies with a moderate risk of bias on the overall treatment effect if both low and moderate risk of bias studies were included.

Study selection
The initial search yielded 51 studies. A total of six studies 24,33,34,35,36,37 were left for full text evaluation and potential inclusion in the review. Three of these studies were rejected after full text screening, due to an absence of comparator groups or irrelevant outcomes in relation to the formulated question of the present review. Finally, three studies 24,33,34 were included in the qualitative synthesis.

Study characteristics
The three selected studies were controlled, all published in English and classified as retrospective cohort studies.
Godt et al. 33 and Hänggi et al. 34 reported on pharyngeal dimensions after headgear treatment in Class I or Class II patients. Godt et al. 33 also investigated whether pharyngeal narrowness was expected in conjunction with differential vertical growth patterns. Hänggi et al. 34 compared the physiological changes in the pharyngeal area of healthy individuals after the use of a combined type of headgear and activator treatment. The total sample size used in both studies was 372 patients, with an age range of 9.3 to 11.2 years at the beginning of the study. Conventional cephalometric radiographs were assessed and compared at baseline and at the end of the active phase of treatment in both studies to record alterations in the pharyngeal airway.
Pirilä-Parkkinen et al. 24 reported a mixed population of 30 children who presented with apnoea, and who were either healthy or headgear users with an average initial age of 8.2 years. The proportions of breathing abnormalities, apnoea indices and oxygen saturation were recorded under laboratory conditions (Table I).

Risk of bias within studies
Based on the assessment of ACROBAT-NRSI, the three studies acquired a rating 'serious risk of bias' in the overall risk of bias judgement. To achieve a selection of participants unrelated to intervention was difficult, since headgear is normally used in Class II patients and, therefore, children with such malocclusions were selected for these studies. Confounding was another parameter that yielded a serious risk of bias, as all the studies were retrospective. In addition, a number of baseline factors could possibly affect the association between headgear use and airway dimensions or breathing conditions (i.e., cephalometric/CBCT measurements that differed between the study groups), while none were taken into account during statistical analysis. Parameters associated with missing data and selection of reported outcomes/results were less prone to bias, as complete patient records were recoded for all follow-up time-points and all reported outcomes had been pre-specified in the article methodologies (Table II; Table III).

Results of individual studies
In an assessment of nasopharyngeal dimensions, the use of headgear was found to present reduced posterior airway space at all levels from the nasopharynx to the mandibular base at the initial headgear phase of treatment; however, this finding was also reported for other occlusion-modifying, first phase appliances. These changes were not related to different vertical patterns of growth. 33 The combined use of an activator-headgear appliance revealed increases in pharyngeal airway parameters related to area, length and the smallest distance between the tongue and the posterior pharyngeal wall after the active phase of treatment. In the post-treatment period, changes in nasopharyngeal dimensions were established and were of the same amount as in the untreated control group. 34 Related to breathing pattern and apnoea symptoms, the headgear group presented an increased oxygen desaturation index (ODI 10 ) when compared with the sample of apnoea children. In addition, other apnoea related parameters such as apnoea indices, obstructive, central and mixed apnoea periods, and total apnoea time were elevated during sleeping with a headgear appliance. 24 However, headgear patients were those previously reported as having presented at least one apnoea symptom and were selected based on this characteristic.

Discussion
Cervical headgear is a common orthodontic appliance used for managing space problems, the correction of Class II skeletal and molar relationship and anchorage reinforcement. [16][17][18][19][20][21][22][23][24][25][26][27][28] The effect of cervical headgear on the maxilla and/or the mandible has been reviewed by previous studies. Nevertheless, only a small number of publications have evaluated the potential side effects of cervical headgear on sleep disorders and pharyngeal airway space. Therefore, this systematic review was designed to provide clear evidence related to the development of obstructive sleep apnoea and subsequent alterations in the naso-and oropharyngeal dimensions in young patients during the use of headgear as part of orthodontic treatment.
Only three studies 24,33,34 were related to the research question and were included in this systematic review, which is indicative of the scarcity of publications   1,3,9 Therefore, the identification and isolation of potential triggering factors would be a significant step towards prevention of the disease.
Alterations in pharyngeal dimensions have been described after the use of growth modification appliances, including headgear treatment in children, 33,34 as an additional and potential triggering factor in the development or aggravation of obstructive sleep apnoea. Alternatively, when evaluating children with confirmed OSA during early ages, the natural history of the disease should be considered, as a number of cases may resolve during adolescence. 38 Common practice when evaluating treatment outcomes in relation to oropharyngeal dimensions has been the assessment and interpretation of treatment effect though within group comparisons. Invariably, this involves a comparison against a baseline and over time, separately for each experimental group. However, the interpretation of the findings following such a comparison should be treated with caution as there have been associations with flawed inferences, increased likelihood of false positive errors or confounding of the outcome due to natural improvement over time. 38 The apparently contradictory findings with regard to naso-and oro-pharyngeal airway dimensions in two of the included studies, 33,34 notwithstanding the potential risk of bias and methodological shortcomings of each, are most likely related to the difference in the biological mechanisms activated by the different treatment procedures followed. The use of a combination of a headgear-activator appliance in one study 34 might have been an influential factor inducing a more anterior repositioning of the lower jaw, resulting in an increase or a non-decrease of the pharyngeal dimensions. This might potentially counteract the initial effect of a headgear-only use and a reduced posterior airway space.
The ACROBAT-NRSI tool 31 was used in the present review to identify the risk of bias involved in the included studies. Sterne et al., in 2014 31 indicated that this tool used a number of signalling questions specifically designed to assess potential sources of bias that may arise in non-randomised studies, mainly represented by: issues on the selection of participants, confounding, bias in the measurement of the outcomes or selection of the reported results. Improvements have followed in the newest ROBINS-I tool 41 designed and validated in 2016 and practically based on the same signalling questions. The Cochrane Collaboration has proposed the use of these tools over previous scales or scoring systems when evaluating non-randomised studies. 42 The orthodontic literature is still new in the use of these tools, while the vast majority of systematic reviews have addressed the methodological quality of observational studies based on customised non-validated scales or tools. Modifications of the Newcastle-Ottawa scale 43 have been described; however, these may consist of questions/items nonspecific to the design of individual studies or lack relevant guidance on how to be used, resulting in varying interpretations considered by different investigators. 44 The review is not free of limitations, which are mainly dependent on the structure/design of the available studies. Only three studies, retrospective in nature, were deemed eligible for inclusion and all presented a serious risk of bias. Study design, confounding factors as well as the recruitment of participants and methods of outcome measurements/analyses were the primary determinants of bias detection in the included studies.
No quantitative synthesis was possible in view of the different settings, populations and outcomes assessed and subsequently evaluation of publication bias was not possible, although pre-specified.

Conclusion
Based on the appraised literature, the evidence is not sufficiently solid to determine the effect of the cervical headgear appliance on the development of obstructive sleep apnoea and subsequent alterations of the nasoand oro-pharyngeal dimensions. Further prospective studies or randomised controlled trials are necessary to fill knowledge gaps and eliminate biases within the available evidence. The elimination of confounding factors and appropriate sample matching at the design level and at the analysis stage are important prerequisites when considering evidence from observational research. Only then can clear recommendations for future practice be identified to enable informed and optimal clinical decision making.