Differential Anxiety–Depression–CRP Network Structures Across Insomnia Severity Levels: Evidence From UK Biobank

Anxiety and depressive symptoms are highly prevalent in the general population [1–4]. The World Health Organization (WHO) estimated that approximately 280 million people worldwide suffered from depression in 2019, affecting 5% of adults globally [5], while lifetime anxiety disorder prevalence reaches approximately 34% in the United States [6]. This high prevalence imposes substantial burdens on both mental and physical health. Furthermore, anxiety and depressive symptoms frequently exhibit bidirectional comorbidity with insomnia [7]; [5]. Individuals with insomnia report more severe anxiety and depressive symptoms than those without [8, 9], pointing to potential common underlying mechanisms.

Recent research implicates inflammatory markers, particularly C-reactive protein (CRP), in insomnia, anxiety, and depression. CRP, an acute-phase protein synthesized by the liver in response to inflammation, has emerged as a key biomarker for psychiatric disorders. Elevated CRP levels are consistently observed in patients with anxiety and depression [10–14]. Similarly, individuals with insomnia exhibit elevated levels of CRP, along with other pro-inflammatory cytokines (e.g., IL-6 and TNF-α) [15–17].

Importantly, insomnia severity demonstrates dose–response relationships with both psychiatric outcomes and inflammation. For instance, mild, moderate, and severe insomnia increase the risk of psychiatric diagnosis by 1.7, 2.63, and 5.04 times, respectively [18]. Kelly et al. [19] further demonstrated that insomnia severity predicts depression/anxiety severity independent of covariates (e.g., injury severity in spinal cord injury veterans). Insomnia severity also correlates positively with log-transformed CRP (log-CRP) levels [16]. While bidirectional associations between insomnia and these outcomes exist, the patterns of anxiety–depression–CRP associations within different insomnia severity groups remain unclear [19–21]. Understanding these relationship patterns is crucial for clinical practice and public health, as it could facilitate early identification of high-risk patients and enable more targeted interventions. The interconnected nature of these symptoms suggests that addressing one domain (e.g., insomnia) may yield cascading benefits across others (e.g., mood and inflammation), potentially improving both mental and physical health outcomes simultaneously.

Network analysis offers a powerful framework for this investigation. Unlike conventional approaches relying on latent variables, this method directly models relationships among observed variables, identifying core symptoms that serve as treatment targets and bridge symptoms that connect different symptom clusters [22–26]. From a mental health perspective, identifying these central drivers within symptom clusters informs precision medicine by allowing clinicians to prioritize interventions based on individual network profiles. From a physical health standpoint, mapping the anxiety–depression-inflammatory marker connections across insomnia severity levels provides vital insights into psychophysiological interplay. Critically, network analysis enables statistical comparison of structural differences, including connectivity patterns, among subgroups [27], allowing direct examination of how anxiety–depression–CRP relationships vary across insomnia severity levels [28].

The present study aims to identify core and bridge symptoms within the anxiety–depression–CRP network and to examine the structural differences in this network across groups with different insomnia severity levels in a large cohort from the UK Biobank. The findings may inform more precise identification of therapeutic targets for individuals with different insomnia severity profiles, with important implications for preventing the progression of sleep disturbances into more complex mental and physical health conditions. Understanding these network structures could guide the development of integrated treatment approaches that simultaneously address psychological symptoms and inflammatory processes, potentially improving both mental health outcomes and reducing physical health risks associated with chronic inflammation.

Table 1 presents the demographic characteristics for the general population and three groups categorized by insomnia frequency (never/rarely, sometimes, and usually experiencing insomnia). The sample included variables of sex, age, depression symptoms (PHQ-9), anxiety symptoms (GAD-7), and inflammation marker (CRP, both original and log-transformed values) across different insomnia severity groups.

This study represents the first network analysis examining structural differences in anxiety–depression–CRP networks among groups with varying degrees of insomnia severity. Key findings revealed: (1) significantly different network structures between never/rarely versus usually experiencing insomnia groups, particularly involving energy/appetite-CRP connections; (2) PHQ9−2 (“Depressed mood”) as the most central symptom across all networks; and 3) GAD7−6 (“Irritability”) as a key bridge symptom. These findings reveal that network structures of anxiety, depressive symptoms, and CRP levels differ based on insomnia severity.

Distinct differences in symptom network connections emerged between never/rarely and usually experiencing insomnia groups, particularly regarding energy levels and inflammatory markers. In the usually insomnia group, the edge weights between “PHQ9–4-GAD7–2 (Energy–Uncontrollable worry)”, “PHQ9–4-GAD7–4 (Energy–Trouble relaxing)”, “PHQ9–4-CRP (Energy–CRP)”, and “PHQ9–5-CRP (Appetite–CRP)” were significantly higher than in the group that never/rarely experienced insomnia. This strong association between “feeling tired” and “not being able to stop or control worrying” may be attributed to chronic sleep deprivation, which triggers fatigue while simultaneously compromising stress management capacity through disrupted prefrontal cortex function and heightened amygdala reactivity. This creates a vicious cycle, where persistent anxiety leads to sympathetic nervous system hyperarousal, further disrupting sleep and exacerbating fatigue [52–54].

The never/rarely experiencing insomnia group exhibited significantly higher edge weights than the usually experiencing insomnia group for the following connections: “PHQ9–3-GAD7–1” (Sleep–Nervousness), “PHQ9–4-GAD7–1” (Energy–Nervousness), “PHQ9–9-GAD7–5”(Suicidal ideation–Restlessness), and “PHQ9–4-GAD7–6” (Energy–Irritability). A plausible explanation for this result is that individuals who rarely experience insomnia tend to have stronger concerns and negative expectations about sleeplessness, which may intensify their anxiety response when encountering sometimes insomnia, thereby creating a vicious cycle of sleep disturbance.

These specific connectivity patterns reflect the neurobiological consequences of chronic sleep disruption. Persistent insomnia activates the hypothalamic–pituitary–adrenal (HPA) axis, leading to chronic cortisol elevation that simultaneously dysregulates inflammatory resolution and metabolic processes. This explains why CRP selectively connects with energy and appetite symptoms in frequent insomnia: both inflammatory signaling and metabolic regulation converge on shared brain regions, including the hypothalamus and anterior cingulate cortex, which are essential for energy homeostasis and fatigue perception [55]. The stronger energy–anxiety connections likely reflect compromised prefrontal cognitive control under chronic sleep deprivation, creating a cycle where fatigue impairs worry regulation, while persistent anxiety further disrupts sleep [56]. Supporting these findings, previous research analyzing UK Biobank and Netherlands Study of Depression and Anxiety (NESDA) data found that CRP was significantly associated with altered appetite, sleep problems, fatigue, irritability, and worry control, with increased CRP levels specifically correlated with prolonged sleep duration and increased appetite in the NESDA cohort [57].

These network connectivity differences between insomnia groups stem from chronic sleep-related dysregulation of the HPA axis and associated neuroimmune pathways [58, 59]. Beyond cortisol-mediated effects, chronic insomnia creates cascading dysfunction through disrupted neurotransmitter function, particularly serotonin, GABA, and dopamine systems, which further compromises cognitive–emotional regulation. Additionally, chronic insomnia may exacerbate emotional vulnerability, intensifying the interplay between anxiety and depression symptoms and the inflammatory marker CRP [60].

From a clinical perspective, these findings suggest that CRP may serve as a valuable biomarker for identifying individuals whose mood and physical symptoms are interconnected through inflammatory processes, particularly in the context of chronic insomnia. The sparse but selective connectivity of CRP within our networks indicates that while inflammatory markers may not be directly driven by psychological symptoms, their strategic connections with specific somatic symptoms (energy, appetite) in vulnerable populations point to potential targets for integrated interventions addressing both psychological and inflammatory components of distress. This mechanistic understanding highlights the importance of considering sleep-inflammation-mood interactions in clinical assessment and treatment planning.

Despite significant differences in edge weights across insomnia severity levels, all groups demonstrated consistent patterns in the strongest symptom connections. Key consistent connections included GAD7−2 (“Uncontrollable worry”) and GAD7-3 (“Excessive worry”), PHQ9-1 (“Anhedonia”) and PHQ9-2 (“Depressed mood”), and GAD7-1 (“Nervousness”) and GAD7-2 (“Uncontrollable worry”), revealing fundamental and stable connectivity patterns that potentially represent core pathological pathways. While insomnia severity modulates the manifestation intensity and interaction dynamics of specific symptoms (reflected in edge weight variations), it does not alter the essential symptom interconnections.

PHQ9−2 (“Depressed mood”) emerged as the symptom with the highest strength centrality across all networks, with GAD7−2 (“Uncontrollable worry”) and PHQ9−2 (“Depressed mood”) also exhibited high expected influence centrality. These findings align with prior systematic reviews identifying “Sad mood,” “Uncontrollable worry,” and “Worrying too much” as the most central symptoms, and “Sad mood,” “Restlessness,” and “Motor disturbance” as the most frequent bridge symptoms [61]. Similarly, Mullarkey et al. [48] found depressed mood highly central in adolescent depression networks, suggesting this pattern may be consistent across age groups.

The identification of these central symptoms has important therapeutic implications, as both depressed mood and uncontrollable worry are commonly targeted in cognitive behavioral therapy (CBT) approaches [62–65], indicating that existing evidence-based treatments may effectively address the most influential network nodes. These symptoms also correspond with those addressed by mindfulness-based interventions [66, 67], suggesting multiple therapeutic pathways for targeting network-central symptoms. This convergence across studies indicates that both central and bridge symptoms may represent stable therapeutic targets.

However, our identification of “Uncontrollable worry” (GAD7−2) as having high expected influence centrality differs from Contreras et al. [23], who found worry symptoms showed moderate rather than high centrality in anxiety disorder networks. This may reflect differences in sample characteristics (clinical vs. population-based) or measurement instruments. The prominence of worry symptoms in our general population sample may reflect its fundamental role in cognitive control processes, as theoretical models suggest that uncontrollable worry reflects dysfunction in prefrontal-limbic circuits essential for cognitive–emotional regulation. When these regulatory mechanisms are impaired, the inability to control worry could have cascading effects on multiple anxiety and depression symptoms.

Our analysis identified GAD7−6 (“Irritability”) as a bridge symptom connecting the anxiety–depressive–CRP network. These findings regarding bridge symptoms are consistent with some previous studies [68, 69]. However, our results are inconsistent with yet other studies [70–73]. This inconsistency may be attributed to multiple factors. Differences in sample populations could play a role, such as studies involving disabled elderly individuals [73], adults with disabilities [72], or youth survivors of childhood sexual abuse [70]. Additionally, the use of different scales to measure the same symptoms across studies may also contribute to these discrepancies [71].

The neurobiological basis for irritability's bridge function may explain its therapeutic potential. Irritability likely stems from its dependence on emotion regulation circuits, particularly orbitofrontal and ventromedial prefrontal cortex regions critical for behavioral inhibition. This symptom involves altered dopaminergic signaling and increased noradrenergic activity, creating a neurobiological profile that overlaps with both anxiety (noradrenergic hyperactivation) and depression (dopaminergic hypofunction) pathways [74–76]. This dual neurochemical involvement positions irritability as an ideal therapeutic target. Consistent with this mechanistic understanding, existing literature supports cognitive–behavioral techniques such as anger management strategies, mindfulness-based stress reduction, and emotion regulation skills training [77–79]. Furthermore, given the observed differences across insomnia severity levels, cognitive–behavioral therapy for insomnia (CBT-I) may be an important area for future investigation ([7, 20, 21]). Future research should investigate whether interventions targeting these specific symptoms can effectively reduce overall symptom severity and prevent comorbid conditions.

CRP's low predictability (1.7%) within the anxiety–depression symptom network is both theoretically expected and clinically meaningful. As a systemic inflammatory marker, CRP reflects complex physiological processes influenced by numerous factors beyond psychological symptoms, including physical health conditions, medications, genetic factors, and lifestyle variables [16, 17]. Rather than indicating a limitation, this low within-network predictability suggests that CRP functions as a bridge between psychological symptomatology and broader physiological systems, consistent with psychoneuroimmunology frameworks [57].

A notable strength of this study lies in its large sample size and examination of how the network structure of anxiety–depression symptoms and CRP levels differs across different insomnia severity levels, offering valuable insights into the underlying psychosomatic mechanisms. Nevertheless, the present investigation also has several limitations. First, although the UK Biobank represents a longitudinal cohort, the variables examined in this analysis were assessed at a single time point, precluding longitudinal examination and causal inference. Future experimental studies are warranted to establish causal relationships and evaluate targeted interventions. Second, reliance on self-reported instruments may introduce response bias, despite their established psychometric properties and clinical utility. Third, we did not exclude participants with preexisting psychiatric conditions, which may confound the interpretation of network structural differences across insomnia severity groups. Future investigations should incorporate comprehensive psychiatric histories to elucidate the independent contribution of insomnia severity to network architecture.

Our findings underscore how insomnia severity is associated with distinct patterns of connectivity among mental health symptoms and inflammatory processes, offering a novel framework for understanding the complex interrelationships among these factors. The identification of central and bridge symptoms provides valuable targets for intervention. These findings have significant implications for developing precision medicine approaches, where treatment strategies can be tailored based on individual insomnia severity profiles and central symptom networks. From a public health perspective, early insomnia detection and intervention may serve as an effective strategy for preventing the escalation of anxiety-depression-inflammation network connections.

We would like to thank all participants in this study.

This study was funded by the National Key Research and Development Program of China (Grant 2021YFC2501500), the National Natural Science Foundation of China (Grant 82271525 82071488), the Natural Science Foundation of Guangdong Province (Grant 2023A1515010456), and the Nanfang Hospital Clinical Research Project of Southern Medical University (Grant 2021CR009).

Data and materials are available from the corresponding author upon reasonable request.

The authors declare no conflicts of interest.

Xue Luo: Writing – review and editing, writing – original draft, validation, supervision, methodology, formal analysis, conceptualization. Shuqiong Zheng: writing – review and editing, methodology. Yihong Cheng: writing – review and editing, validation, methodology, conceptualization. Shuai Liu: validation, methodology, conceptualization. Shufei Zeng: resources, methodology. Leqin Fang: resources, methodology. Shixu Du: writing – review and editing, methodology. Weimin Li: formal analysis, methodology. Hangyi Yang: data curation, methodology. Zhiting Huang: data curation, methodology. Bin Zhang: writing – review and editing, writing – original draft, validation, supervision, project administration, methodology, conceptualization.

Additional supporting information can be found online in the Supporting Information section.