Prevalence of scoliosis in children and adolescents: a systematic review and meta-analysis
Background: The understanding of the prevalence and early predictive factors of scoliosis in children and adolescents is limited, which poses challenges to developing preventative strategies. This systematic review and meta-analysis aimed to clarify the prevalence and predictors of scoliosis among children and adolescents.
Methods: We conducted a comprehensive search in PubMed, Cochrane, Embase, and Web of Science through October 2023. The quality of included studies was evaluated using the Joanna Briggs Institute scale or the Newcastle-Ottawa Scale. Subgroup analyses were performed to examine different types of scoliosis and specific demographic groups.
Results: From 32 studies encompassing 55,635,351 children and adolescents, we identified 284,114 cases of scoliosis, resulting in a prevalence rate of 3.1% (95% CI: 1.5%–5.2%). This rate varied by gender, degrees of scoliosis severity, and between idiopathic vs. congenital forms. Notable predictors included gender, age, Body Mass Index (BMI), race, environmental factors, and lifestyle choices.
Conclusion: Scoliosis is a significant condition affecting a minority of children and adolescents, particularly adolescent girls and individuals who are overweight. It is recommended that guardians and schools enhance educational efforts towards its prevention.
Systematic Review Registration: https://www.crd.york.ac.uk/, Identifier CRD42023476498.
1 Introduction
Scoliosis is a three-dimensional (3D) spinal deformation characterized by a lateral curvature across one or more segments of the spine, coupled with vertebral rotation, resulting in core deviation and sagittal progression (1). Severe scoliosis can result in significant health complications, such as cardiovascular issues, reduced pulmonary function, chronic pain, and psychological distress (2).
Scoliosis can be categorized etiologically into idiopathic, congenital, and neuromuscular types. Adolescent idiopathic scoliosis (AIS), the most common form, has an unknown cause, which is reflected in the term “idiopathic” (3, 4). Depending on when it manifests, idiopathic scoliosis may be classified by age brackets: Infantile (ages 0–3 years), Juvenile (ages 3–10 years), Adolescent (ages 10–18 years), and Adult (ages above 18 years) (5). AIS is typically detected during childhood and adolescence, especially during rapid growth periods. Treatment options include observation, wearing orthotic braces, and surgery in more severe cases. Congenital scoliosis is caused by abnormal development of the spine during the embryonic stage, which may involve vertebral non-segmentation, abnormal shape, or abnormal quantity. This type of scoliosis is often associated with genetic factors and may coexist with other congenital abnormalities, so it is usually identified at birth or in early infancy (6). Treatment methods primarily involve surgical correction, especially in cases of severe or rapidly progressing curvature.
Currently, there is limited research available regarding the worldwide prevalence of scoliosis in children and adolescents, and a comprehensive investigation of associated risk factors is lacking. Consequently, the development of preventive strategies for children and adolescents faces considerable challenges. Therefore, the objective of this systematic review and meta-analysis was to provide a detailed description of the prevalence and early predictors of scoliosis in adolescents globally, and to offer evidence-based guidance for the detection and prevention of scoliosis.
2 Methods
2.1 Study registration
This meta-analysis was conducted in adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (The PRISMA 2020) guidelines, and our protocol was registered with PROSPERO (ID: CRD42023476498).
2.2 Eligibility criteria
For our systematic review, we strictly included cohort or cross-sectional studies, while expressly excluding the following studies: (1) Studies that were based solely on subject self-reporting of disease diagnosis or assessments derived from specific scales that lacked clinical validation; (2) Case reports, meta-analyses, reviews, and guidelines; (3) Studies that had a sample size of fewer than 20 cases.
2.3 Data sources and search strategy
We systematically searched databases including PubMed, Cochrane Library, Embase, and Web of Science. The search was conducted from the inception of each database up to October 2023, without geographical restrictions. Details of the search are presented in Supplementary Table S1.
2.4 Study selection
All identified literature was imported into EndNote, where duplicate publications were initially filtered out automatically and manually. Studies relevant to our topic were selected by reviewing titles and abstracts, and then the full-texts of potentially relevant articles were downloaded and read. Finally, original research articles that met the inclusion criteria were selected upon full-text review. The literature screening process was independently carried out by two researchers. In the event of any discrepancies, a third researcher was consulted to discuss and make a final decision.
2.5 Data extraction
Before initiating data extraction, we devised a standardized template to systematically collect relevant data. The details gathered included DOI/PMID, first author, year of publication, type of study, author’s nationality, patient demographics, period of sample collection, age group included, scoliosis categories, scoliosis diagnostic criteria, total participant count, number of scoliosis cases, along with age, gender, and independent correlation factors.
Data extraction was independently carried out by two reviewers. In instances of disagreement, a third reviewer was consulted to contribute to the resolution process.
2.6 Risk of bias in studies
The original studies encompassed in this systematic review were either cohort or cross-sectional studies. For the cohort studies, we evaluated their quality using the Newcastle-Ottawa Scale (NOS), a renowned assessment tool designed to judge the quality of non-randomized studies in meta-analyses (7). The NOS examines three domains through eight items, allocating up to one point each for most questions, with the exception of the comparability category, which has a potential for 2 points. Studies achieving a score between 7 and 9 are considered high quality, whereas scores from 4 to 6 signify moderate quality. For cross-sectional studies, the appraisal of quality was conducted using the Joanna Briggs Institute (JBI) scale (8).
The risk of bias in these studies was assessed independently by two researchers. In cases of discrepancy, a third researcher was enlisted to reach a consensus.
2.7 Synthesis methods
Data analysis for this meta-analysis was conducted using R software (version 4.2.2). Prior to performing the meta-analysis of prevalence, data underwent transformation based on predefined criteria: (1) No transformation was necessary if the average prevalence rate across all samples was between 20% and 80%; (2) Logit transformation was applied when the rate was under 20% or exceeds 80%; and (3) The double arcsine transformation method was applied in cases with a significant number of extreme values (0% or/and 100%) (9). The choice of model was guided by the level of heterogeneity, which was determined by the I2 index. A random-effects model was employed when I2 exceeded 50%; otherwise, a fixed-effect model was used. In circumstances of substantial heterogeneity, sensitivity and subgroup analyses were conducted to investigate potential sources of variability. Funnel plots were constructed to visually inspect for publication bias across the studies, complemented by statistical assessment via Egger’s test. In instances where publication bias was detected, the trim-and-fill method was applied to evaluate its influence on the meta-analysis outcomes. A P-value of less than 0.05 was considered indicative of statistical significance.
3 Results
3.1 Study selection
We retrieved a total of 5,531 records. Following the elimination of duplicates, 2,840 records were screened by their titles and abstracts. Of these, 2,766 records were excluded for not meeting our inclusion criteria. A thorough assessment of the full texts for the remaining 74 articles resulted in further exclusion of 42 articles for various reasons including unavailability of full text (n = 8), duplicate or serial publications (n = 7), and incomplete data (n = 27). Consequently, 32 studies were included in our meta-analysis (8–39) (Figure 1).
