This review was conducted in accordance with the guidelines of the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) (11). The protocol was documented and registered in the PROSPERO database (CRD42020206533).

The inclusion criteria for this systematic review were as follows: (i) DS; (ii) age 0-18 years; and (iii) LT-NIV use, defined as respiratory support delivered via an interface outside the airway for at least 3 months in a non-acute care setting. Studies reporting on a broader range of conditions were included only if they provided separate data on subjects with DS. If articles included subjects over 18 years of age, the mean age at NIV initiation needed to be under 18 years for inclusion. There were no restrictions on study design, outcome eligibility, or language at the title/abstract stage; language was limited to English, French, Spanish, Catalan, and Portuguese at the full text stage. To be included, studies had to report on a minimum of three children with DS.

This systematic review is an extension of a scoping review on LT-NIV in children (10). The search strategy for the scoping review, developed for MEDLINE (Ovid) and later translated to additional databases, used Medical Subject Headings (MeSH) and free-text terms for “child” and “non-invasive ventilation” (Supplement Table 1). Since the first study of LT-NIV use in children was published in 1992, studies of humans published from 1990 onwards were searched in the following databases: Ovid MEDLINE, Embase (Ovid), CINAHL (Ebsco), Cochrane Library (Wiley; Inception to Present), and PubMed. The original search was conducted between November 17-28, 2014 and updated most recently on March 25, 2021.

The titles and abstracts of articles identified by the literature search were reviewed for eligibility for full text retrieval by two reviewers (MCC, TA, JEM). Studies in English, French, Spanish, and Portuguese were considered eligible for full-text retrieval and were reviewed by two reviewers (MCC, TA, JEM). The final list of studies eligible for inclusion from the scoping review were then full-text reviewed by two reviewers (SH and JEM) to identify studies that met the inclusion criteria for this systematic review. Any disagreement at the screening, eligibility, or inclusion levels were discussed by the reviewers until a consensus was established.

Data were collected and entered into a pre-established form in Microsoft Excel (version 16.38, Microsoft Corporation, 2020). These data items included author's name, year of publication, country of publication, journal, study design, study duration, sample size, age at NIV initiation, proportion of females, NIV type, and outcomes. One reviewer (SH) extracted the data and a second reviewer (JEM) verified the data extraction.

Independent assessment for risk of bias and quality assessment of each article was conducted by two reviewers (SH, JEM) using the Cochrane Risk of Bias in Non-Randomized Studies of Interventions (ROBINS-I) tool (12) and the Grading of Recommendations Assessment, Development and Evaluation tool (GRADE) (13). Disagreements were resolved through discussion and consensus.

Studies were grouped by DS subject type; studies that included children with DS as part of a broader cohort of children using LT-NIV and studies that exclusively included children with DS using LT-NIV. Numeric data [median, 95% confidence interval (CI)] was summarized where available and examined for meta-analysis with the data summarized narratively otherwise. Where additional analysis to compare children with and without DS within an article was possible, p < 0.05 denoted significant differences between these groups.

Following removal of duplicates and inclusion of additional records, the scoping review search strategy identified 17608 unique records for review (Figure 1). After screening titles and abstracts, 1389 records were eligible for full-text review and 473 articles met inclusion criteria for the scoping review on LT-NIV use in children. Final full-text review identified 28 articles meeting the inclusion criteria for this systematic review. These studies included 543 children with DS who used LT-NIV. Publication dates ranged from 1995 to 2021. Articles originated from nine countries with the majority of studies stemming from North America (16/28, 57%) and Europe (9/28, 32%).

Most studies were retrospective (21/28, 75%), and single-center (25/28, 89%), and all were quantitative. The majority of studies were observational (26/28, 93%) and described as cohort (7/28, 25%), or cross-sectional (19/28, 68%) studies. Two studies (7%) were randomized trials. Across the 28 studies, 20 included children with DS in a broader group of children (Table 1) while eight were exclusively on children with DS (Table 2). Data for quantitative synthesis was limited to age at NIV initiation and proportion of female subjects with the remainder of the data summarized using narrative synthesis.

In the 20 studies that included children with and without DS using LT-NIV, children with DS made up 18% (357/2005) of the population (Table 1). The proportion of children with DS using LT-NIV did vary widely across studies (7–44%). The median age of subjects in studies including children with and without DS, which included children with OSA who received treatments other than NIV, overlapped with those in studies focused on DS [9.0 (95% CI 6.0–11.3) vs. 8.5 (95% CI 6.2–10.9) years]. The proportion of female subjects in the broader studies is also similar to those studies focused on DS [39% (95% CI 35–44) vs. 44% (33–47%)]. Twelve of these studies included only the number of children with DS using LT-NIV so were not analyzed further (30–41).

Of the remaining eight studies that included children as part of a broader group of children using LT-NIV (Table 1) (14–21), only one provided direct comparisons between children with DS and non-DS children (21). In this study, children with DS made up 41% (44/106) of children using LT-NIV with overall similar age and sex distribution between the DS and non-DS groups; in the sub-group of children 0-5 years of age, the DS group was younger than the non-DS group [0.42 (interquartile range, IQR, 1.04) vs. 2.1 (IQR 3.4) years]. The apnea-hypopnea index (AHI), as a measure of OSA severity, was higher in the DS group (DS: 16.8 ± 1.85 vs. non-DS: 11.73 ± 1.5 events/h). While LT-NIV adherence was greater in the DS group (DS: 79% vs. non-DS: 72% of days used) there was no difference in adequate usage (>4 h) on the nights used. Data from machine downloads showed that AHI (5.71 ± 0.47 vs. 4.42 ± 0.27 events/h), system leak, and percentage time in excess leak from device downloads were higher in the DS compared to non-DS group though there were significant improvements in AHI in both groups with LT-NIV use.

The other seven studies reporting on children with DS as part of a broader group of children using LT-NIV included 83 children with DS (Table 1) (14–20). The indications for LT-NIV was reported for six of these studies; while OSA was an indication for all six, two studies included hypoventilation without OSA (16) one study limited indications to residual OSA post-adenotonsillectomy or medical management (18), and one limited indications to OSA post-adenotonsillectomy or preference for non-surgical management/poor surgical candidate (15). Four of these studies included only CPAP treatment (14, 15, 18, 19) while two included CPAP/BPAP (16, 17), and one included CPAP/BPAP as well as high flow nasal cannula (20).

Age of LT-NIV initiation for children with DS was reported in one study and was 10 y (IQR 11.2) compared to 6.6 (IQR 12.95) for the full group (16). One study reported the proportion of children started on CPAP without prior adenotonsillectomy or adenoidectomy; 32% (7/22) were children with DS compared to 53% (30/57) of non-DS children (Chi-square 2.8, p = ns) (15). A single study reported on location of NIV initiation and showed that children with DS were started during a hospitalization (17%, 1/6) or after polygraphy/polysomnography (83%, 5/6) while non-DS children started in the PICU (21%, 15/70), during a hospitalization (24%, 17/70), and after polygraphy/polysomnography (54%, 38/70) (17). Interface type for children with DS was reported in two studies. One reported the use of full face mask in 50% (11/22) of children with DS in contrast to 7% (4/57) of non-DS children (Chi-square 19.1, p < 0.001) (15). The second included five children with DS and reported that all children used a nasal mask or prongs (16). This study also reported on NIV mode where 60% (3/5) of children with DS were started on CPAP and 40% (2/5) on BPAP compared to 78% (49/63) started on CPAP and 22% (14/63) on BPAP for non-DS children (Chi-square 0.84, p = ns).

Data on adherence in DS and non-DS children were reported in two studies. One study that included 22 children with DS (19% of subjects) reported that adherence in the full group did not differ between those with and without impaired cognition, with and without adenotonsillectomy prior to NIV, full face vs. nasal mask, and with and without psychological support for NIV initiation (15). This same study reported that 27% (6/22) of children with DS were referred for psychological assistance compared to 14% (8/57) of non-DS children (Chi-square 1.91, p = ns). Finally, while 20% of children took longer than 90 days to establish CPAP use, 60% (6/14) of this group were children with DS (Chi-square 1.91, p = ns). In the second study which included 31 children, four never initiated CPAP use ≥4 h; the rate of never initiating CPAP use ≥4 h was 43% (3/7) for children with DS compared to 4% (1/24) for non-DS children (Chi-square 6.9, p < 0.01) (18).

The eight studies focused on DS included 186 children who were using LT-NIV (Table 2). Four studies included data on body composition with 55% of all subjects being obese in one study (25) and three studies reporting BMI z-scores of 1.4 (IQR 1.4) in subjects with OSA in one cohort, 1.7 ± 4.0 for the full cohort in the second (some of whom were treated with surgery rather than NIV), and 0.89 (IQR 1.85) in the third (26–28). The indications for use were similar to the studies reporting data on children with DS in larger cohorts of children using LT-NIV; OSA, OSA post-adenotonsillectomy, OSA with preference for non-surgical management, and hypoventilation or raised CO₂ overnight. One study reported that 55% of children with DS and OSA required LT-NIV after adenotonsillectomy (23) where another reported that 33% of their DS cohort required LT-NIV after baseline sleep study (27).

Interface type was specified in two studies with 75% (33/44) of children using nasal masks and 25% (11/44) using nasobuccal or full face masks (27, 28). Subjects in four studies used exclusively CPAP (22, 24, 26, 29) while the remaining four studies reported on CPAP and BPAP or PAP usage (23, 25, 27, 28). Two studies reported the proportion of subjects using CPAP and BPAP; overall, 75% (33/44) used CPAP and 25% (11/44) used BPAP (27, 28). Two studies reported on NIV pressure settings and adherence (27, 28) with only one reporting pressure settings at follow-up (28). CPAP pressures ranged from 6–10 cmH₂O with BPAP median inspiratory positive airway pressure of 12 (range 12–17) and 14 (IQR 6) cmH₂O, and median expiratory positive airway pressure of 8 (range 5–8) and 7 (IQR 4) cmH₂O (27, 28). The study that included pressure settings at follow-up showed no change in CPAP or BPAP pressures after a median follow-up of 1.9 y (28). With respect to adherence, one study reported an average use of 8 h 46 ± 6 h 59 per night and 82% of children showing use >4 h/night (27) while another reported 56% regular use at four months and 46% at 1.9 y with a median use/night of 4 h (IQR 6) for CPAP and 8 h (IQR 3) for BPAP at 1.9 y (28), and a third reported average nightly use of 116 min (IQR 69) (26). Adherence was described as high in the first study, good with no difference between CPAP and BPAP in the second, and low though typical of other studies in the third. In a study of 37 children initiated on CPAP, 44% were rated by parents as “always” or “nearly always adherent” while 24% were “never” adherent; children who were adherent were more likely to report perceived benefit (29).

Three studies reported on outcomes of NIV treatment in children with DS. One study compared treatment outcomes between LT-NIV and upper airway surgery (adenoidectomy ± tonsillectomy ± turbinectomy) and reported improvements, compared to baseline, in oxygen parameters on sleep studies for children using LT-NIV; while the same pattern of changes was seen after surgery, these differences did not differ from baseline measures (27). Of note, the criteria for starting LT-NIV differed from the criteria for surgery (AHI>10 vs. >5 events/h) so the groups were not comparable at baseline (27). In a randomized trial of CPAP vs. sham CPAP for OSA in children with DS, where none of the children had pulmonary hypertension at baseline based on echocardiography, cardiovascular outcomes did not differ between groups at 4-months follow-up though there was an improvement in LV diastolic function with CPAP use (26). In a study of sleep, sleep apnea, and neuropsychological function in children with DS, those successfully treated for OSA, either with adenotonsillectomy or CPAP, showed greater improvement in attention though results were not presented separately by treatment type (24). Finally, LT-NIV was ceased because of improvements in 8–50% of children with DS across three studies (22, 27, 28).

Overall risk of bias across studies was serious for all studies (Table 3). This was predominantly attributable to study design as most were observational. The included data was heterogenous in nature with limited comparative evidence; quality assessment tools could not be applied.

The most prominent limitation of this systematic review is the serious risk of bias of the individual studies, which were mainly retrospective and observational. Risk of bias is serious overall due to baseline confounding variables, differences in type of support included in the definition of NIV (e.g., BPAP alone vs. CPAP/BPAP modes), limited consideration of adherence to NIV and subsequent impact on outcomes, loss of participants to follow-up, and selective reporting of results. As such, while this review provides a summary of current evidence, the serious risk of bias limits the ability to provide strong recommendations. In addition to limited outcomes data and comparisons of LT-NIV in children with and without DS, gaps in knowledge relevant to LT-NIV use in DS include the impact of craniofacial differences in DS on NIV use, the impact of LT-NIV use on facial growth, and an understanding of how LT-NIV use impacts quality of life for both children with DS and their caregivers.

The preceding scoping review did not include articles from all possible languages; this led to exclusion of 8% of articles which may have potentially included articles on children with DS. Furthermore, the search strategy was limited to children and may have excluded relevant studies on young adults where the mean age of subjects exceeded the set maximum of 18 years for this review. OSA risk in DS does not stop in adulthood (56, 57), therefore, conclusions of this review may not be applicable to young adults with DS. The heterogeneity of the subject groups and limited outcome data precluded meta-analysis.