A computerized systematic search was conducted on PubMed, MEDLINE, Scopus, and Embase from January 1986 to January 2021 using the following search terms: (((((idiopathic [Title/Abstract]) AND (scoliosis [Title/Abstract])) AND (brace [Title/Abstract])) OR (bracing [Title/Abstract])) AND (nighttime [Title/Abstract])) OR (part-time [Title/Abstract]).

In addition, reference lists of all included articles were screened by two reviewers to identify other potential studies relevant to this systematic review. Only articles written in English were included. The review was registered in PROSPERO (ID: CRD42020179767).

Title and abstract screening was independently performed by the two reviewers, and potentially relevant articles were selected. The two reviewers then independently performed a full text review based on inclusion/exclusion criteria. Any discrepancy between the reviewers was resolved by a consensus meeting or by consulting a third reviewer.

The inclusion criteria were as follows: individuals diagnosed with AIS, age 10–18 years, Risser sign of 0–2, Cobb angle of 20°–45° at initiation of bracing, treatment with overcorrection nighttime bracing, no other previous treatment, and at least one predictive factor of treatment outcome (failure and/or success) assessed. Articles were included if bracing continued until the end of skeletal maturity or spinal fusion. Studies that evaluated the efficacy of full-time bracing or included AIS/juvenile idiopathic scoliosis and other types of scoliosis other than idiopathic were excluded. Review articles, case reports or case study designs, and conference abstracts were also excluded.

The following data were extracted from the included studies: study design, number of patients, age at initiation of bracing, initial Risser sign, initial curve magnitude, curve type, type of brace, IBC, brace treatment length, follow-up duration, and definition for treatment failure and success (Table 1).

Two reviewers independently evaluated the studies using the method developed by Hayden et al. [26] and suggested by Lievense et al. [27] and Scholten-Peeters et al. [28]. This method consists of 13 items categorized into five domains (Table 2). A criteria list was prepared to assess the methodological quality of prediction studies. The score of each item was 0 (missing or insufficient data) or 1 (sufficient data reported). Average scores for the quality assessment tool ranged from 0 (lowest) to 13 (highest). Score cut-offs of ≥9 and <9 indicated high and low-quality, respectively (Table 3). A similar method of quality assessment was used to systematically review observational studies on predictive factors for full-time brace treatment outcome in AIS [23], prognostic factors for curve progression in degenerative lumbar scoliosis [29], prognostic factors of progression of knee and hip osteoarthritis [27,30], and predischarge prognostic factors of physical function in older adults with hip fracture surgery [31]. Scoring disagreement was resolved by discussion or by consulting a third reviewer.

To determine the predictive factors for the effectiveness of nighttime bracing in AIS, the level of evidence of each factor was rated using best-evidence synthesis [32]. The evidence was considered strong in the case of consistent findings in multiple (≥2) high-quality studies, moderate for findings reported in one high-quality study and multiple (≥2) low-quality studies, limited for findings in one high-quality study or consistent findings in ≥3 low-quality studies, inconclusive for findings in ≤3 low-quality studies, and conflicting for consistent findings in <75% of the studies.

Table 1 presents the characteristics of the nine eligible studies. Overcorrection nighttime bracing was defined as brace wearing for 8–10 hours during the night. The sample size of included studies was 34–166 patients, with a total sample of 675. The brace wearing time was 12–40 months, and the follow-up period varied between 12 and 130 months. In six studies, the device was a Providence brace, whereas in three studies, it was a Charleston bending brace. Curve progression of >5° and ≤5° was defined as a treatment failure and success, respectively.

The nine studies were scored by the two reviewers. Of the nine studies, 5 (55.5%) were of high-quality [8,33–36] and 4 (44.5%) of low-quality [37–40]. Table 3 shows the results of their quality assessment.

The studies considered various factors as potential predictors for brace treatment outcome, classified as radiographic or clinical characteristics. Of the nine studies, most (six studies) did not report the results of regression analysis. As a result of heterogeneity of studies, pooling was not possible (Table 4).

Two high-quality studies suggested a strong evidence of an association between curve flexibility [8,35] and brace treatment success. Three high-quality studies suggested strong evidence of an association between the Risser sign [8,33,34] and brace treatment success. The curve progression among patients with Risser signs of 0 and 1 was higher than those with a Risser sign of 2. The evidence for curve type was inconclusive.

Davis et al. [33] divided participants into two groups based on the Risser sign (0 and ≥1). A curve progression rate of <5° in the group with Risser sign ≥1 was 78.6% compared with 40% for the group with Risser sign=0. Results suggested that when combining the Risser sign of ≥1, initial Cobb angle of <35°, and curves with an apex at T10 or lower, the success of overcorrection nighttime bracing is 100% for AIS patients if all these parameters exist, 64% for two of three parameters, 25% if one parameter exists, and 0% when none of the parameters exists. D’Amato et al. [8] found that the success rate of treatment for patients with Risser sign=0 is significantly lower (67%) than for those with Risser sign=1 or 2 (86% or 83%, respectively). However, Lee et al. [38], in a low-quality study, found no significant differences between the brace success rate among patients with a Risser sign of 0–2. The results of three high-quality studies [8,33,34] and one low-quality study [39] revealed that the degree of IBC is associated with the success of overcorrection nighttime bracing. In contrast, one high-quality [33] and one low-quality [41] study found no association between IBC and the success of overcorrection nighttime bracing.

Katz et al. [34] reported that when the IBC is >80%, the probability of successful brace treatment is significantly greater compared to an IBC of <80% (p=0.006). Ohrt-Nissen et al. [35] found that the IBC for patients with successful and failed Providence brace treatment is 66% and 53%, respectively. Overall, the evidence for IBC was inconclusive.

In two high-quality studies [33,34], strong evidence indicates a negative association between the Risser sign and overcorrection nighttime bracing failure. In one high-quality study [35], moderate evidence indicates a positive association between premenarchal status and overcorrection nighttime bracing failure. According to univariate linear regression analyses, the failure of the Providence brace was significantly correlated with low flexibility and premenarchal status. R² was 0.23 for flexibility and reached 0.39 by adding menarchal status. Only one low-quality study evaluated brace compliance as a predictive factor for treatment outcome [37]. The evidence suggesting that poor brace compliance is associated with treatment failure was inconclusive. The evidence of treatment failure indicated for initial curve magnitude, curve type, IBC, age, Risser sign, curve apex, and sex was conflicting (Table 4).

In a high-quality study on 102 adolescents with AIS with curves between 20° and 42°, D’Amato et al. [8] found that in cases with a curve of >35°, 77% of patients treated with a Providence brace experienced curve progression of >5°. In cases with a curve of <35°, 34% of patients experienced curve progression of >5°. Davis et al. [33], in their high-quality study, found that the initial curve magnitude was inversely related to Providence bracing success when the curve was between 25° and 40°. They divided patients into two groups based on curves of <35° and >35°. Patients with a curve of <35° experienced a 70% success rate, while those with a curve of >35° had a 41% success rate. Ohrt-Nissen et al. [35] and Lee et al. [38] found no relationship between the initial curve magnitude and overcorrection nighttime bracing success.

Curve flexibility evaluation provides guidance related to therapeutic plans, so it is essential in AIS treatment [42–44]. Measuring curve flexibility is also important for estimating the bracing outcome in AIS. Because patients with greater flexible curves experience a higher IBC, curve flexibility may affect the final outcome of bracing [35,45]. The results of this systematic review suggest that there is strong evidence of a positive association between curve flexibility and overcorrection nighttime bracing success [8,35]. However, limited evidence indicates no association between curve flexibility and overcorrection nighttime bracing failure. Failure of overcorrection nighttime bracing may increase if the patient has low flexibility and premenarchal status. Various techniques of evaluating curve flexibility have been proposed, such as supine lateral bending radiographs (SLBRs), supine, prone, sitting with lateral bending, prone with lateral bending, and fulcrum-bending radiographs [36,42,46]. However, there is a lack of consensus among researchers about which method of assessing flexibility is the best. He and Wong [46] measured spinal flexibility in the following positions: prone, supine, sitting with lateral bending, and prone with lateral bending. The found that the prone position is the best for measuring the flexibility of the spine [46]. Ohrt-Nissen et al. [35] assessed curve flexibility in a Providence brace using the SLBR method and defined it as [(Cobb angle in standing position–Cobb angle in supine lateral bending position)/Cobb angle in standing position]×100. The authors found that curve flexibility can be an important predictive factor for the outcome of using a Providence brace in AIS. In contrast, some researchers have found that the SLBR method depends on the patient’s effort and the radiologist’s expertise, which limits the method’s reliability and constancy [42,47,48]. Moreover, if a patient has low back pain, the accuracy of the test might decrease because of failure to use high forces [47]. Future studies on this topic are therefore recommended.

We found strong evidence of an association between a high Risser sign and overcorrection nighttime bracing success. The association between this factor and overcorrection nighttime bracing failure was also strong [33,34]. The Risser sign is generally applied as a maturity index in AIS with reference to iliac apophysis ossification. However, due to the radiographic parallax, it is less apparent on postero-anterior radiographs. Moreover, the Risser sign can be distinguished after the peak height velocity, it is not a good index for evaluating the amount of growth remaining, and it correlates differently in boys than in girls [3,49]. To determine accurate maturity during adolescence, Sanders et al. [50] developed a reliable skeletal maturity scoring system based on patterns of the digital skeletal age of the hand. These patterns are highly correlated with the peak height velocity. None of the studies included in this systematic review used the Sanders system. Further studies that take this variable into account to determine the success or failure of overcorrection nighttime bracing in AIS need to be undertaken.

According to a high-quality study by Ohrt-Nissen et al. [35], moderate evidence indicates that premenarchal status is associated with Providence brace failure. The definition of evidence used in our review is similar to that used by van den Bogaart et al. [23]. However, based on their results, inconclusive evidence suggests that menarchal status is not associated with full-time bracing failure [23]. Further studies that measure premenarchal status should be undertaken to develop a full understanding of the impact of this variable on the outcome of overcorrection nighttime bracing in AIS.

IBC, which describes the difference between the in-brace Cobb angle and the initial out-of-brace Cobb angle, is one of the main factors for the long-term efficacy of full-time bracing [23,24]. Studies have reported a strong relationship between a lack of IBC and full-time bracing failure [23,24]. However, in our review, the evidence of the success or failure of overcorrection nighttime bracing being related to IBC was inconclusive and conflicting. A possible explanation could be differences in IBC measurement methodologies. Ohrt-Nissen et al. [35] measured IBC as [(Standing out-of-brace Cobb angle–standing in-brace Cobb angle)/standing out-of-brace Cobb angle]×100. D’Amato et al. [8] and Simony et al. [39] measured IBC as [(standing out-of-brace Cobb angle–supine in-brace Cobb angle)/standing out-of-brace Cobb angle]×100. In contrast, Trivedi and Thomson [41] measure IBC as [(standing out-of-brace Cobb angle–prone in-brace Cobb angle)/standing out-of-brace Cobb angle]×100. These differences in methodology make it difficult to determine whether a higher IBC is due to more flexible and therefore correctible curves or better fitting of the brace.

Initial curve magnitude is calculated from the Cobb angle at the initiation of brace treatment. Overcorrection nighttime bracing is most effective when the initial curve is <35° [8,34]. Five high-quality studies [8,33–35,51] and four low-quality studies [37–40] on adolescents with initial curves of 20° to 45° revealed conflicting evidence associating the initial curve magnitude and overcorrection nighttime bracing outcomes.

There is conflicting evidence regarding the association between the initial curve magnitude and overcorrection nighttime bracing failure. Simony et al. [39] and Davis et al. [33] found that a higher initial curve magnitude is associated with a higher rate of Providence nighttime brace failure, while Bohl et al. [37] and Ohrt-Nissen et al. [35] concluded that there is no association between the initial curve magnitude and Providence nighttime brace failure. According to a systematic review by van den Bogaart et al. [23], moderate evidence indicated that the initial curve magnitude is not correlated with full-time bracing failure.

We found conflicting evidence of a relationship between curve apex and overcorrection nighttime bracing success or failure. A high-quality study suggested that for curves with an apex at T10 or lower, the overcorrection nighttime bracing success rate is significantly higher than for curves with an apex at T6–T9 (69.4% versus 40.6%, respectively) [33]. In another high-quality study of patients treated with a Providence brace [8], the treatment success rate was 79% for those with a curve apex of <T8 and 61% for those with a curve apex >T8. Two low-quality studies by Lee et al. [38] and Bohl et al. [37] found no significant relationship between the curve apex and overcorrection nighttime bracing success or failure.

Curve type is considered a risk factor for bracing failure in AIS [52]. Hawary et al. [24] suggested that the type of curve is a potential risk factor for full-time bracing failure. Preliminary studies on the effectiveness of overcorrection nighttime bracing have hypothesized that these braces are more effective for single lumbar or thoracolumbar curves than for double or thoracic curves [34,40,53]. However, our results found that the evidence of an association between overcorrection nighttime bracing success or failure and curve type is inconclusive and conflicting. In high-quality studies by Ohrt-Nissen et al. [35] and Katz et al. [34], double major curves and single thoracic curve types were identified as a risk factor for overcorrection nighttime bracing failure. Additionally, in high- and low-quality studies [33,37], no relationship was found between curve type and Providence nighttime bracing failure. Furthermore, two published studies have reported that curve type has no significant relationship with Charleston nighttime bracing success [38,41].

The evidence of a relationship of age and sex with overcorrection nighttime bracing success or failure is conflicting. This finding is consistent with that of van den Bogaart et al. [23], who reported conflicting evidence of an association between full-time bracing failure and age and sex.

The role of this parameter in the outcome of overcorrection nighttime bracing has not received much attention and has been evaluated only in one low-quality study [37]. Evidence of an association of poor brace compliance with overcorrection nighttime bracing failure is inconclusive. Compliance with brace wearing may be an important predictive factor for the effectiveness of full-time bracing. Hawary et al. [24] found that a lack of brace compliance is a primary factor for full-time bracing failure. Additionally, moderate evidence indicates that poor brace wearing is associated with full-time bracing failure [23]. Future studies should continue to objectively examine the relationship between compliance and overcorrection nighttime bracing success or failure.

This systematic review had several limitations that need to be considered when interpreting the results. First, strong evidence indicates that increased curve flexibility and a higher Risser sign are associated with overcorrection nighttime bracing success. However, the number of studies evaluating the role of flexibility and skeletal maturity in overcorrection nighttime bracing outcomes in AIS is insufficient. Future studies are therefore recommended. Second, studies evaluating predictive factors for overcorrection nighttime bracing outcomes in AIS are few. In this systematic review, we highlighted the importance of all radiologic and clinical parameters that potentially may play a role in overcorrection nighttime bracing outcomes. However, some questions remain. Therefore, additional longitudinal studies are needed to assess all potential predictive factors for overcorrection nighttime bracing outcomes. Third, the heterogeneity of the measurement methodology of variables, definitions of treatment outcomes, the process of evaluation, and the manner of reporting results, across the included studies, precluded the ability to perform a meta-analysis. Fourth, the results only included articles published in English. Perhaps including studies in other languages would alter the results. Finally, our conclusions are based on statistics of retrospective studies that are prone to numerous prejudices that may indicate differences noticed between predictive factors and overcorrection nighttime bracing outcomes.