Sex and age differences in depression and anxiety networks among adolescents with idiopathic scoliosis: A network analysis

Abstract

BACKGROUND

Depression and anxiety are prevalent psychological challenges among patients with adolescent idiopathic scoliosis (AIS), affecting individuals across both sex and age groups.

AIM

To explore the network structure of depression and anxiety symptoms, with a focus on identifying differences at the symptom level between sex and age subgroups.

METHODS

A total of 1955 participants diagnosed with AIS aged 10-18 years were assessed using the Patient Health Questionnaire Depression Scale (PHO-9) and the Generalized Anxiety Disorder Scale (GAD-7), and 765 patients exhibiting PHQ-9 or GAD-7 scores ≥ 5 were enrolled in our study. Network analysis and network comparison tests were utilized to construct and compare the depression-anxiety symptoms networks among sex and age subgroups.

RESULTS

The results revealed GAD3 “Excessive worry” and PHQ2 “Sad mood” were the most significant central symptoms in all subgroups, while “Sad mood” had higher strength than “Excessive worry” in the lower age group. In the network comparisons, the female network exhibited tighter connectivity, especially on GAD6 “Irritability” and GAD2 “Uncontrollable worry”, while only PHQ3 “Sleep” and PHQ9 “Suicidal ideation” had differences at the local level in the lower age group.

CONCLUSION

Several interventions targeting excessive worry and sad mood could reduce the risk of depression and anxiety symptoms in the AIS population. Furthermore, specific anxiety symptoms in females, along with sleep disturbances and suicidal ideation in the lower age group, should be addressed at an early stage to prevent significant disruptions in mental health trajectories.

Key Words: Adolescent idiopathic scoliosis; Network analysis; Depression and anxiety symptoms; Age difference; Sex difference

Core Tip: Among 1955 adolescents with idiopathic scoliosis, 765 participants showed significant depression and anxiety symptoms. Network analysis identified "Excessive worry" and "Sad mood" as central symptoms across all sex and age subgroups, with "Sad mood" more prominent in younger individuals. Females exhibited tighter connectivity, particularly for "Irritability" and "Uncontrollable worry". Notably, differences in "Sleep disturbances" and "Suicidal ideation" were observed in the lower age group. Targeting these central symptoms through early interventions is crucial to mitigate the risk of depression and anxiety in this population.

INTRODUCTION

Adolescent idiopathic scoliosis (AIS) is a common spinal deformity, usually presenting with a radiographic Cobb angle of at least 10° and affecting adolescents aged 10 to 18 years[1]. The global prevalence of AIS ranges from 0.47% to 5.2% and increases with age[2]. AIS exhibits a more aggressive prevalence, more severe spinal curvature, and earlier age of onset in females due to differences in physiological (e.g., genetics, growth spurts, and hormone levels) and environmental factors (various stressors)[3,4]. The spinal curvature rapidly increases at all stages from childhood to adolescence, leading to an irreversible structural spinal deformity, which results in more pronounced physical deformity, pain, pulmonary complications, and psychological disturbances[5]. Therefore, AIS has emerged as a major public health problem that demands attention, in part because of its impact on the physical and mental health of adolescents.

Depression and anxiety are common psychological problems in patients with AIS, with prevalence rates of 39.1% and 24.6%, respectively[6]. Most studies have indicated that the progression of AIS requires patients to confront various stressors such as negative self-image, chronic back pain, and sleep disturbances, heightening their vulnerability to psychological disorders[7,8]. Moreover, numerous treatment options, including braces and surgical interventions, have been shown to induce considerable anxiety in over one-third of the patients, primarily because of concerns about treatment efficacy[9,10]. Notably, the formation and development of depression and anxiety symptoms within this population diverge with age and sex and are influenced by individual physical and psychological characteristics. For instance, females with AIS tend to experience greater emotional disturbances linked to negative body image[11]. Furthermore, the severity of depression and anxiety symptoms can escalate from mild to severe as patients transition from childhood to adolescence[12]. Therefore, a comprehensive understanding of the characteristics of depression and anxiety symptoms across different sex and age groups is essential for mental health interventions in patients with AIS.

Network analysis assumes that depression and anxiety are interrelated symptoms that influence the progression of other symptoms or diseases[13,14]. Certain active symptoms, which serve as the core of the disease, can promote and sustain the development of psychological issues through positive feedback mechanisms[15]. Unlike latent variable analysis, network analysis allows for the identification of central symptoms and visualization of specific associations between them rather than relying on total symptom scores that weigh all symptoms equally to explain the severity of depression and anxiety[13,16]. While some studies have investigated the relationship between AIS and psychological problems, they have not provided a comprehensive understanding of depression and anxiety symptoms or clarified how these symptoms vary across different situations and individuals[6,7,17]. To the best of our knowledge, few studies have applied a network analysis perspective to examine depression and anxiety symptoms in patients with AIS across sex and age groups. Therefore, this study aimed to: (1) Examine the network structure and central symptoms of depression and anxiety across sex and age among participants with AIS; and (2) Compare overall and local network differences between subgroups based on sex and age.

MATERIALS AND METHODS

Study participants and data collection

Cross-sectional data from 1955 patients were collected at the Spine Health Centre of Shenzhen Second People’s Hospital between April 2021 and May 2024. The inclusion criteria were as follows: (1) Aged 10 to 18 years; (2) Diagnosed AIS with a Cobb angle ≥ 10° by x-ray; and (3) Exhibiting mild to severe depression or anxiety symptoms based on Patient Health Questionnaire Depression Scale (PHQ-9) or Generalized Anxiety Disorder Scale (GAD-7) score of 5 or higher. The exclusion criteria for the study were: (1) The absence of significant demographic information; (2) Patients diagnosed with congenital scoliosis; and (3) Failure to complete the depression or anxiety scales. Written informed consent was obtained from all 787 participants or their parents before distributing the self-administered questionnaire. After excluding 22 questionnaires, 765 valid questionnaires were included in the final analysis, resulting in a response rate of 97.2%. The participants were divided into two groups based on sex (male and female) and age (≤ 14 years including early adolescence, and > 14 years including mid-to late adolescence). This study adhered to the Declaration of Helsinki and was approved by the Ethics Committee of Shenzhen Second People’s Hospital (No. 2023-315-02PJ).

Measurements

The PHQ-9 and GAD-7 are self-administered scales that were used to assess depression and anxiety symptoms over the past 2 wk[18,19]. All items were measured on a 4-point Likert scale: 0 (not at all), 1 (several days), 2 (more than 7 days), and 3 (almost every day). Thus, the PHQ-9 and GAD-7 scores ranged from 0 to 27 and 0 to 21, respectively, with total scores above 5, 10, and 15 indicating mild, moderate, and severe depression or anxiety symptoms, respectively. Both PHQ-9 and GAD-7 had acceptable internal consistency, with Cronbach’s α coefficients of 0.880 and 0.911 in our sample, respectively. Noteworthy, the reliability of the depression and anxiety scales was similarly confirmed in 10-12 years old adolescents with idiopathic scoliosis using Cronbach's alpha coefficients of 0.867 and 0.870, respectively.

Statistical analysis

R version 4.2.1 was used for the statistical description and network analysis. Before constructing the networks, the mean ± SD was calculated for each symptom score from the PHQ-9 and GAD-7. A t-test was used to compare statistically significant differences in symptom scores across sex and age subgroups.

Network estimation

Given the distribution of PHQ-9 and GAD-7 scores, partial correlation networks were constructed using a Graphical Gaussian Model[20]. A graphical lasso-extended Bayesian method was employed to eliminate weaker edges and enhance the interpretability and stability of the network[21]. The simplified network was visualized using the “qgraph” package, and node distribution was adjusted via the average Layout function within the same package[22].

Network density is an indicator of the overall structure and accounts for the proportion of effective edges to the total number of edges connecting all nodes. Edge weights represent the strength of the correlations between two nodes after controlling for the influence of other nodes[23]. Centrality indices (strength, closeness, and betweenness) were used to evaluate the roles of individual nodes. Strength is defined as the sum of a node’s weighted connections to all other nodes, with higher strength indicating greater influence on the network. Closeness represents the reciprocal of the sum of the shortest distances from one node to the other, demonstrating the speed at which it influences other symptoms. Nodes with higher betweenness have a greater capacity to regulate interactions and are often considered a bridge that influences other symptoms[16,20].

Network stability and accuracy

The “bootnet” package was employed to evaluate the accuracy and stability of the network[24]. First, a nonparametric bootstrap approach (2000 bootstrap samples) was used to assess the accuracy of the edge weights by calculating 95% confidence intervals (CIs). Narrower bootstrap CIs suggest more reliable and precise edges. Second, the case-dropping bootstrap method (2000 bootstrap samples, α = 0.05) was employed to evaluate the centrality stability, specifically strength. A correlation stability (CS)-coefficient above 0.25 was considered an acceptable stability in most small samples (n < 500), while a value exceeding 0.5 represented higher stability, indicating that the network could preserve the original structure even after discarding 50% of the subjects.

Network comparison

The “Network Comparison Test” (NCT) package was used to perform network difference based on sex (male vs female) and age groups (lower age vs higher age), including: (1) Network structure; (2) Global strength; and (3) The strength of edges and nodes[25]. The network structure test assumes that there is a significant difference in at least one set of edge strengths between the two networks. Global strength measured the overall network connectivity, with larger values denoting tighter connections among all nodes. In addition, edge and node strengths were used to reflect local features, especially after detecting differences in network structure. The Bonferroni-Holm method was applied to control multiple testing problems at the local level.

RESULTS

Sample characteristics

All demographic and diagnosis information among total and subgroups based on sex and age are shown in Supplementary Table 1. The total scores for depression and anxiety in the entire population were 7.91 ± 4.60 and 7.18 ± 4.27, respectively. We did not identify any statistically significant differences in total depression and anxiety scores between groups based on gender or age (P < 0.05).

The means and standard deviations of all symptoms are reported in Table 1. In the sex-grouped sample, 175 males (22.9%) and 590 females (77.1%) met the criteria for mild depressive or anxiety symptoms. Statistically significant differences were only observed in the means of the PHQ2 and GAD5 (P < 0.05). Within aged-subgroups, 476 participants in the lower age group (≤ 14 years, 62.2%) and 289 participants in the higher age group (> 14 years, 37.8%) reached the threshold for depression or anxiety symptoms. Only the mean differences in the PHQ7, PHQ9, and GAD7 were statistically significant (P < 0.05).

Table 1 Characteristics of depression and anxiety symptoms of the participants.

Items	Symptoms	Male (n = 175)		Female (n = 590)		P	Lower age (n = 476)		Higher age (n = 289)		P value
		M	SD	M	SD		M	SD	M	SD
PHQ1	Anhedonia	1.02	0.73	1.12	0.77	0.126	1.12	0.81	1.08	0.68	0.490
PHQ2	Sad Mood	0.94	0.70	1.08	0.77	0.029	1.07	0.79	1.02	0.69	0.423
PHQ3	Sleep	1.09	0.97	1.13	1.00	0.648	1.13	1.03	1.11	0.93	0.757
PHQ4	Fatigue	1.20	0.81	1.18	0.85	0.833	1.19	0.90	1.19	0.74	0.957
PHQ5	Appetite	0.82	0.92	0.80	0.87	0.857	0.84	0.92	0.75	0.81	0.203
PHQ6	Guilty	1.00	0.92	1.08	0.91	0.321	1.10	0.95	0.99	0.85	0.096
PHQ7	Concentration	0.66	0.83	0.65	0.81	0.864	0.59	0.82	0.76	0.79	0.003
PHQ8	Motor	0.53	0.80	0.49	0.77	0.594	0.53	0.83	0.44	0.68	0.106
PHQ9	Suicidal ideation	0.43	0.77	0.43	0.72	0.954	0.48	0.78	0.35	0.64	0.012
GAD1	Nervousness	1.13	0.68	1.23	0.78	0.120	1.18	0.71	1.24	0.72	0.372
GAD2	Uncontrollable worry	0.91	0.76	1.05	0.86	0.058	0.99	0.82	1.06	0.79	0.272
GAD3	Excessive worry	1.10	0.81	1.16	0.86	0.438	1.11	0.87	1.21	0.80	0.082
GAD4	Trouble relaxing	1.10	0.87	1.00	0.85	0.205	1.04	0.89	1.00	0.79	0.592
GAD5	Restlessness	0.81	0.85	0.62	0.79	0.008	0.66	0.84	0.66	0.76	0.961
GAD6	Irritability	1.15	0.86	1.22	0.88	0.393	1.25	0.94	1.13	0.81	0.074
GAD7	Feeling afraid	0.95	0.97	0.91	0.90	0.643	1.00	0.97	0.8	0.81	0.003

Lower age: ≤ 14 years old; Higher age: > 14 years old. PHQ: Patient Health Questionnaire Depression; GAD: Generalized anxiety disorder; M: Mean.

Network structure

All networks demonstrated relatively strong connections with high densities ranging from 40/120 to 77/120. Specifically, the networks for both female (77/120) and older age groups (77/120) were more strongly connected than the other networks (Supplementary Tables 2-5). Only edges with strengths above 0.08 were displayed in our networks (Figure 1). Some edges exhibited similar and strong weights across all networks (e.g., PHQ1–PHQ4, GAD2–GAD3), whereas others displayed noticeable differences in strength (e.g., PHQ7–PHQ8).

Figure 1 Networks of depression and anxiety symptoms. A: The male subgroup with adolescent idiopathic scoliosis (AIS) (n = 175); B: The female subgroup with AIS (n = 590); C: The lower age group with AIS (n = 476); D: The higher age group with AIS (n = 289). Thicker edges indicated stronger associations. Partial correlation coefficients are labeled between nodes (Strength ≥ 0.08). Positive correlations are shown in green, while negative correlations are depicted in red. PHQ: Patient Health Questionnaire Depression; GAD: Generalized anxiety disorder.

In males, GAD3 “Excessive worry” (Strength = 0.97) was the most significant symptom, with slightly higher strength than PHQ2 “Sad mood” (Strength = 0.80), followed by GAD4 “Trouble relaxing” (Strength = 0.68) and GAD1 “Nervousness” (Strength = 0.64). In the females, GAD3 “Excessive worry” (Strength = 1.08) was an important central symptom, closely followed by PHQ2 “Sad mood” (Strength = 1.08). Additionally, GAD6 “Irritability” (Strength = 0.95) and GAD2 “Uncontrollable worry” (Strength = 0.93) displayed high strength values (Figure 2A).

Figure 2 Standardized estimates of centrality for depression and anxiety symptoms. A: Sex subgroups; B: Age subgroups. PHQ: Patient Health Questionnaire Depression; GAD: Generalized anxiety disorder.

In the lower age group, GAD3 “Excessive worry” (Strength = 1.07) emerged as the second most common symptom after PHQ2 “Sad mood” (Strength = 1.09) according to the centrality assessment, followed by the PHQ4 “Fatigue” (Strength = 0.95) and GAD4 “Trouble relaxing” (Strength = 0.91). In the older age group, GAD3 “Excessive worry” (Strength = 1.18) particularly warranted careful consideration, as it showed a slightly higher strength than PHQ2 “Sad mood” (Strength = 1.16). In addition, some anxiety symptoms deserved adequate attention, especially GAD6 “Irritability” (Strength = 0.94) and GAD1 “Nervousness” (Strength = 0.88) (Figure 2B).

Network stability and accuracy

Bootstrapping analysis revealed that stronger edges exhibited narrower CIs across all networks, indicating greater stability (Supplementary Figures 1 and 2). The CS coefficients of strength were used as criteria to measure network accuracy (Supplementary Figures 3 and 4). In the sex subgroups, males had acceptable CS coefficients (0.297), whereas female exhibited higher strength values (0.673). Within the age subgroups, younger and older patients demonstrated comparable CS coefficients, with values of 0.595 and 0.361, respectively.

Network comparison

NCT revealed significant differences in global strength between males and females (global strength: Male: 3.46 vs female: 6.79; test statistic = 3.32, P = 0.02; network structure: M = 0.27, P= 0.08). Females showed higher global strength, suggesting tighter connections between symptoms, particularly in the GAD2 “Uncontrollable worry” (P < 0.001), GAD6 “Irritability” (P = 0.02) and GAD4 “Trouble relaxing” (P = 0.04) nodes. However, we did not detect differences in global strength (lower age: 6.77 vs higher age: 6.52, test statistic = 0.24, P = 0.33) or network structure (M = 0.124, P = 0.99) across the age subgroups. At the local level, PHQ3 “Sleep” exhibited higher strength (P= 0.02), while PHQ9 “Suicidal ideation” demonstrated the highest closeness in the younger age group compared to the older age group (Supplementary Figures 5 and 6).

DISCUSSION

In the current study, we investigated the network structure of depression and anxiety symptoms among patients with AIS across different sex and age groups. We calculated the centrality indicators and edge strengths, and performed network comparisons at the global and local levels to assess sex- and age-related differences. Our study revealed that GAD3 “Excessive worry” and PHQ2 “Sad mood” were high-strength symptoms provoking depression and anxiety across all subgroups, similar to a previous study of 811 AIS participants[26]. We also identified differences in the centrality of anxiety symptoms between sex subgroups, specifically for GAD2 “Uncontrollable worry” and GAD6 “Irritability”, offering new insights into the prevention and treatment of anxiety within this specific population.

The central nodes for depressive and anxiety symptoms in patients with AIS were similar in terms of sex distribution. Both sexes shared GAD3 “Excessive worry” and PHQ2 “Sad mood” as the highest strength central symptoms, and it occupied the same position in both male and female networks. Meanwhile, PHQ2 “Sad mood” was the symptom with the highest betweenness and closeness centrality, serving as a key connector among various depression and anxiety symptoms[27]. “Excessive worry”, which is recognized as a core diagnostic feature of generalized anxiety disorder in Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition, DSM-5), also appeared to have specificity in patients with AIS[28,29]. One potential explanation is that individuals with AIS are excessively concerned about disease-related risks to their physical health, which may contribute to a sustained negative emotional state[30]. Additionally, “Sad mood” not only demonstrated strong node strength but also exhibited the highest distance weight and strongest moderating ability, playing a significant role in activating and maintaining depression and anxiety networks in both male and female subgroups[31]. As a standard diagnostic indicator of depression, our findings of “Sad mood” align with other clinical and non-clinical studies on depression and anxiety networks[29,32]. Several studies have hypothesized that adolescents with AIS suffer from stress due to changes in body image and somatic pain, often preferring to avoid excessive attention from peers, which can reduce self-esteem and increase their vulnerability to sadness during adolescence[8,33,34].

In sex-distributed subgroups, several anxiety symptoms exhibited significant differences in node strength. Specifically, GAD6 “Irritability” and GAD2 “Uncontrollable worry” displayed higher strengths in the female group, consistent with findings in children with ADHD[35]. Some intersex studies have shown that irritability symptoms are linked to a wide range of mental health challenges in females and demonstrated a moderate positive correlation over time[36,37]. Irritability is influenced by hormone levels and hinders effective emotion regulation, which can lead to increased rumination and depression symptoms in females[38]. In addition to physiological conditioning, females with AIS have heightened concerns about their physical condition being affected by spinal disorders, as evidenced by a decrease in self-esteem and social self-confidence, resulting in a stronger psychological burden[39,40]. Notably, the global strength and the strength of nearly all nodes were lower in males compared than in females. This may be partially attributed to regularization, which removed numerous low-strength edges. However, it could also reflect weaker connections among symptoms in males, suggesting that isolated symptoms are unlikely to develop into serious psychological problems[41]. Therefore, early interventions, such as mindfulness-based interventions and cognitive behavioral therapy (CBT) are recommended for females with AIS who are prone to significant anxiety symptoms in the initial stages[42,43].

In both age subgroups, GAD3 “Excessive worry” and PHQ2 “Sad mood” emerged as the most central symptom clusters in the depression and anxiety networks of patients with AIS, which was confirmed by the findings from Tao and Cai on Chinese adolescents at various developmental stages[44,45]. In the lower age group, “Sad mood” was the strongest central symptom and was slightly higher than GAD3 “Excessive worry”. These results are distinct from those in early, middle, and late adolescence, suggesting that “Excessive worry” occupies a primary position[46]. A study by Vogel et al[47] highlighted that while sad mood in early adolescence is an often easily overlooked precursor, it predicted the risk of developing depressive symptoms in mid- to late adolescence. However, sadness can rapidly develop once triggered and may not be observed until friends and social activities almost completely disappear[48]. For patients with an early diagnosis of spinal disorders, various adverse life events such as pain and peer attention, which serve as triggers for sadness, play a significant role in the development and worsening of depression and anxiety[49,50]. Meanwhile, they are more sensitive to teacher-student and peer relationships owing to inadequate social support, which increases the risk and severity of sad mood[51]. With growing age, GAD3 “Excessive worry” became the most prominent central symptom over “Sad mood” and played a critical role in maintaining the stability of the depression and anxiety symptom network, consistent with the results of Tao in Chinese adolescents during late adolescence[44]. In addition to physiological factors, academic stress and family education are non-negligible factors affecting higher adolescents. For example, excessive studying leading up to college entrance exams have been shown to diminish adolescents’ subjective well-being and self-efficacy[52,53]. These findings highlight potential new intervention strategies that focus on the development of spinal disorders and emotional changes in this population during early adolescence to prevent scoliosis-induced psychological disorders. Furthermore, necessary family support should be provided to older adolescents to help alleviate excessive academic anxiety.

Age-related network comparisons revealed local differences between the two nodes. Surprisingly, PHQ3 “Sleep” showed higher strength in the lower age group, indicating that adolescents with scoliosis experience more severe sleep disturbances at an earlier age. This finding is in contrast to a previous study that concluded that older adolescents were overwhelmed by competition for entrance exams, implying permanent fatigue and insomnia[54]. A reasonable hypothesis suggests that spinal disorders trigger insomnia and daytime sleepiness in early adolescence, not only directly affecting adolescents' daily life and academic performance, but further increasing the risk of depression (HR = 7.27) and anxiety (HR = 2.54) by disrupting the normal sleep-wake cycle[55]. PHQ9 “Suicidal ideation” has higher closeness in younger age groups and closer proximity to other nodes of depression and anxiety, which was also identified by Xu et al[56] as the bridge connecting psychological symptoms in adolescents. Previous studies have shown that adolescents’ depression and anxiety networks are highly activated but have low connectivity, while suicidal ideation has lower strength and higher closeness to these symptoms[57]. This suggests that suicidal ideation may serve as an early warning sign of serious psychological issues in individuals who respond positively to this symptom[58]. When strongly influenced by external events, depression and anxiety symptoms in adolescents may rapidly become interconnected, thereby increasing the risk of future suicidal behaviors[59]. Consequently, the early identification and appropriate management of suicidal ideation in lower age groups are crucial for improving overall mental health and well-being.

Our study has some limitations. First, it was based on a cross-sectional design, precluding our ability to capture changes in symptoms over time. Future research could employ longitudinal network analysis to address the potential masking of dynamic features across different sexes and age groups[60]. Second, all self-administered questionnaires were collected at a single hospital throughout the study, limiting our ability to generalize our findings to all patients with AIS. It may also introduce bias related to participant origin. More participants should be recruited from various hospitals to expand the questionnaire's scope as far as capacity allows. Finally, certain unobserved variables, such as additional demographic indicators and lifestyle factors, could potentially influence depressive and anxiety symptoms in patients with AIS. Future studies could include these variables as covariates or integrate them as new nodes into the mental health network[61].

CONCLUSION

Overall, this study is the first to clarify the central symptoms and characteristics of mental health across sex and age groups, which could provide more precise data to support early personalized management and treatment of AIS. Our study indicated that GAD3 “Excessive worry” and PHQ2 “Sad mood” were the most significant central symptoms in all subgroups, while “Sad mood” had higher strength than “Excessive worry” in the lower age group. In the network comparisons, female network exhibited tighter connectivity, especially on GAD6 “Irritability” and GAD2 “Uncontrollable worry”. Only PHQ3 “Sleep” and PHQ9 “Suicidal ideation” had differences at the local level in the lower age group. Several interventions targeting excessive worry and sad mood, such as CBT, emotion regulation training, family support, and psychological education, can reduce the risk of depression and anxiety symptoms in this population.

ACKNOWLEDGEMENTS

The authors extend their sincere appreciation to the research staff for their invaluable contributions to data collection and to all participants for their voluntary participation in this study.