INTRODUCTION
Chronic insomnia disorder is a common chronic sleep disorder affecting approximately 10%–15% of adults worldwide and is defined by persistent difficulty initiating or maintaining sleep for at least three months with associated daytime impairment, according to the International Classification of Sleep Disorders, Third Edition (ICSD-3) [1–3]. The ICSD-3 also classifies insomnia disorder as an independent diagnostic entity not better explained by other medical, psychiatric, or sleep disorders, underscoring its neurophysiological underpinnings rather than a purely subjective complaint [1,4].
The physiologic hyperarousal model is widely accepted as a core pathophysiological framework for chronic insomnia disorder. Individuals with insomnia demonstrate increased sympathetic activation, reduced heart rate variability, elevated nocturnal cortisol secretion, and heightened Hypothalamic-Pituitary-Adrenal axis activity, indicating sustained systemic arousal [5–7]. This systemic hyperarousal extends to the central nervous system, where neuroimaging studies show increased basal metabolic activity in regions such as the prefrontal cortex and the default mode network, suggesting a mismatch between sleep-promoting neural networks and persistent wake-like metabolic activation [8].
Although this model predicts elevated basal metabolic rate (BMR) in insomnia, empirical findings have been inconsistent. Early indirect calorimetry studies reported a 7%–10% increase in BMR among insomnia patients [9,10], but subsequent studies have yielded conflicting results—likely due to small sample sizes, methodological variability, heterogeneous diagnostic criteria, inadequate control of comorbid sleep disorders, and variability in measurement timing [11,12]. Differences in measurement type (indirect calorimetry vs. predictive formulas), insomnia phenotypes (sleep-onset vs. sleep-maintenance insomnia), and the lack of integration with objective polysomnography (PSG)-derived sleep measures further contribute to this inconsistency.
Predicted BMR, estimated using demographic and anthropometric variables, provides a scalable and clinically accessible alternative to direct calorimetry and has demonstrated validity across diverse populations [13–15]. However, no study has systematically examined whether predicted BMR is associated with PSG-derived measures of sleep continuity or fragmentation in individuals with chronic insomnia disorder—representing a significant gap in our understanding of insomnia’s metabolic physiology.
Therefore, in this age- and sex-matched cohort study, we aimed to determine whether predicted BMR differs between individuals with chronic insomnia disorder and matched controls, and to evaluate how predicted BMR relates to PSG-derived measures of sleep continuity and fragmentation. Through this approach, we sought to clarify the metabolic characteristics of chronic insomnia disorder and evaluate the potential utility of predicted BMR as a physiological correlate of objective sleep architecture.
METHODS
Study Design and Participants
This study was a single-center, retrospective cohort analysis conducted at the Sleep Center of Samsung Medical Center. A total of 7,338 adults (≥18 years) who underwent both overnight PSG and bioelectrical impedance analysis between 2017 and 2024 were considered as the initial study population. Among these, 389 participants diagnosed with chronic insomnia disorder based on the ICSD-3 criteria were identified as the insomnia candidate group [1]. Individuals who had undergone PSG during the same period but reported no insomnia-related symptoms and did not meet ICSD-3 criteria (n=249) served as the pool for potential controls.
Inclusion criteria were: 1) age ≥18 years, 2) availability of a complete overnight PSG recording, and 3) complete anthropometric and questionnaire data, including Pittsburgh Sleep Quality Index (PSQI), Insomnia Severity Index (ISI), Epworth Sleepiness Scale (ESS), and Beck Depression Inventory (BDI). Participants in the insomnia group were required to meet full ICSD-3 diagnostic criteria, and only those whose symptoms could not be explained by other sleep disorders—such as obstructive sleep apnea (OSA), periodic limb movement disorder (PLMD), or restless legs syndrome—were included.
Exclusion criteria were as follows: 1) OSA, defined as apnea–hypopnea index (AHI) ≥5; 2) Periodic Limb Movements during Sleep (PLMS) index ≥15/hr or evidence of PLMD sufficient to explain sleep fragmentation; 3) psychiatric disorders known to affect sleep architecture (e.g., major depressive disorder, generalized anxiety disorder); 4) neurological disorders (e.g., Parkinson’s disease, epilepsy); and 5) endocrine or metabolic disorders known to influence BMR (e.g., thyroid dysfunction).
Because age and sex are major determinants of predicted BMR, insomnia and control participants were matched 1:1 based on sex and age (±1 year). After matching and applying inclusion/exclusion criteria, a final sample of 225 insomnia–control pairs (n=450) with complete clinical, PSG, and questionnaire data was included for analysis (Fig. 1).
This study was approved by the Institutional Review Board of Samsung Medical Center (IRB No. 2023-05-02) and was conducted in accordance with the Declaration of Helsinki. Given the retrospective nature of the study and use of de-identified data, the requirement for informed consent was waived.
PSG
Overnight PSG was performed following standardized procedures outlined in the American Academy of Sleep Medicine (AASM) Manual for the Scoring of Sleep and Associated Events [16]. Recordings were obtained in dedicated PSG suites at Samsung Medical Center.
The montage included standard electroencephalography (C3–A2, C4–A1, O1–A2), electro-oculography, submental electromyography, electrocardiogram, respiratory effort belts, nasal pressure transducer, oral–nasal thermistor, pulse oximetry, body position sensor, and bilateral tibialis EMG.
Sleep staging and event scoring were performed manually by certified technologists in accordance with AASM criteria [16]. Extracted PSG parameters included total sleep time (TST), sleep latency (SL), sleep efficiency (SE), wake after sleep onset (WASO), total arousal index, and non–rapid eye movement (NREM)/rapid eye movement (REM) arousal indices.
OSA (AHI ≥5) and clinically significant PLMD (PLMS index ≥15/hr) were excluded to minimize confounding by sleep-disordered breathing or movement-related sleep fragmentation.
Predicted BMR
Predicted BMR was calculated using the Harris–Benedict equations [17], based on age, sex, height, and weight. These equations are widely validated and used in clinical and epidemiologic research for estimating basal metabolic requirements [13–15].
The equations applied were:
Predicted BMR was used instead of indirect calorimetry because it is feasible for large clinical cohorts and reflects a stable metabolic baseline relevant to physiologic hyperarousal mechanisms in insomnia. Lifestyle or medical factors potentially influencing basal metabolism were controlled for through exclusion criteria.
Statistical Analysis
Continuous variables were compared using independent t-tests or Mann–Whitney U tests depending on distributional characteristics, and categorical variables were compared using χ2 or Fisher’s exact tests.
Associations between predicted BMR and insomnia status were evaluated using linear regression. The unadjusted model assessed raw group differences, while the adjusted model included age, sex, and body mass index (BMI) as covariates. Multicollinearity was evaluated using variance inflation factors.
Associations between predicted BMR and PSG parameters were assessed separately in the insomnia and control groups using Pearson correlation coefficients. Partial correlations controlling for age and BMI were also performed. PSG variables included TST, SE, SL, WASO, total arousal index, and NREM/REM arousal indices.
All analyses were two-tailed with statistical significance set at p<0.05. Analyses were performed using R software (version 4.3.3, R Foundation for Statistical Computing).
RESULTS
Baseline Characteristics
A total of 450 participants (225 with chronic insomnia disorder and 225 matched controls) were included after age- and sex-matching. The two groups were comparable in age (36.5±13.1 vs. 36.3±13.1 years, p=0.90) and sex distribution (female 57% in both, p>0.90). Anthropometric measures including height, weight, and BMI also did not differ significantly between groups (all p>0.50).
The predicted BMR was similar between the insomnia and control groups (1,409.5±257.3 vs. 1,426.0± 272.0 kcal/day; p= 0.60), indicating that matched baseline characteristics attenuated metabolic group differences.
Regarding lifestyle factors, caffeine use was lower in the insomnia group (61% vs. 70%, p=0.046), while hypnotic use was markedly higher (32% vs. 7.5%, p<0.001). Smoking, alcohol use, and regular exercise did not differ significantly.
Subjective sleep questionnaires showed substantially higher PSQI, ISI, and BDI scores in the insomnia group (all p<0.001). Objective PSG measures also revealed poorer sleep continuity in the insomnia group, with shorter TST (367.7±66.6 vs. 391.3± 56.5 minutes, p<0.001), lower SE (81.2%±12.9% vs. 87.2%± 10.1%, p<0.001), and longer SL and WASO. The arousal index was also higher in insomnia participants, indicating more fragmented sleep (Table 1).
Association between Insomnia Status and Predicted BMR
In the unadjusted linear regression model, the difference in predicted BMR between insomnia and control groups was not statistically significant (−16 kcal/day; 95% confidence interval [CI], −66 to 33; p=0.50). After adjusting for age, sex, and BMI, predicted BMR remained non-significantly associated with insomnia status (β=−17 kcal/day; 95% CI, −35 to 0.77; p=0.061).
Correlations between Predicted BMR and PSG Parameters
Correlation analyses revealed distinct patterns between the insomnia and control groups (Table 3). In the control group, predicted BMR was not significantly associated with any PSG-derived indices of sleep continuity or fragmentation, including TST, SE, SL, and WASO (all p>0.10). In contrast, individuals with chronic insomnia disorder showed a consistent and significant coupling between metabolic status and sleep continuity. Higher predicted BMR was associated with longer TST (r=0.203, p=0.002) and higher SE (r=0.248, p<0.001), as well as shorter SL (r=−0.196, p=0.003) and reduced WASO (r=−0.195, p=0.003) (Fig. 3). No meaningful associations were observed between predicted BMR and either total arousal index or NREM/REM arousal indices in either group. When adjusting for age and BMI, the correlation between predicted BMR and SE in the insomnia group remained statistically significant (r=0.146, p= 0.030), whereas no PSG variable demonstrated a significant adjusted correlation with predicted BMR in controls (Table 4 and Fig. 4). These findings indicate that metabolic status, as estimated by predicted BMR, is selectively linked to objective sleep continuity only in individuals with chronic insomnia disorder.
DISCUSSION
This age- and sex-matched cohort study investigated whether predicted BMR, derived from anthropometric measures, differs between individuals with chronic insomnia disorder and matched controls, and whether predicted BMR is associated with objective sleep architecture. The main findings were: 1) predicted BMR did not differ between insomnia and control groups after rigorous matching, and 2) significant correlations between predicted BMR and PSG-derived sleep continuity indices—including TST, SE, SL, and WASO—were observed only in the insomnia group. These results suggest that while predicted BMR is not a discriminative marker of insomnia per se, it may reflect metabolic coupling with sleep stability within the insomnia phenotype.
The absence of group-level differences in predicted BMR is consistent with the growing understanding that physiologic hyperarousal in insomnia does not uniformly manifest as a global increase in resting metabolism. Although early indirect calorimetry studies reported 7%–10% higher metabolic rates in insomnia populations [9,10], subsequent research has been inconsistent due to small sample sizes, methodological variability, heterogeneous diagnostic criteria, and insufficient control of comorbid sleep disorders [11,12]. Furthermore, hyperarousal encompasses autonomic, endocrine, cortical, and cognitive domains [4–7], and recent conceptual frameworks emphasize that these components do not necessarily converge on a simple hypermetabolic phenotype.
Our finding that higher predicted BMR corresponds to better sleep continuity only in individuals with insomnia aligns with recent models emphasizing metabolic heterogeneity rather than uniform hypermetabolism. Wassing et al. [18] demonstrated that impaired slow-wave generation and reduced energy restoration—rather than sympathetic arousal alone—may drive sleep fragmentation in a subset of insomnia patients. Their model suggests that individuals with a low metabolic tone may have insufficient energetic support for stable sleep-wake transitions, leading to increased arousability and vulnerability to fragmentation. In this context, higher predicted BMR in our insomnia sample may reflect a more resilient metabolic state that partially compensates for instability within sleep–wake regulatory networks.
Similarly, Richardson [19] proposed physiologically distinct high- and low-metabolic models of insomnia, highlighting heterogeneity in arousal, metabolic activity, and sleep-maintenance characteristics among patients. This aligns closely with our results showing that lower predicted BMR is linked to poorer SE, longer SL, and greater WASO—patterns consistent with a fragile sleep-wake homeostatic system. Thus, predicted BMR may index an individual’s capacity for maintaining nocturnal metabolic homeostasis, which becomes clinically relevant only within the dysregulated physiological environment of insomnia.
Autonomic–metabolic coupling may further explain the group-specific associations observed. Insomnia is characterized by sympathetic predominance and reduced heart rate variability [5], which interact with metabolic tone to influence sleep stability. Grimaldi et al. [5] showed that autonomic dysregulation is closely linked to impaired sleep homeostasis in insomnia, suggesting that altered autonomic control may contribute to sleep fragmentation and non-restorative sleep. This provides a mechanistic basis for why predicted BMR relates strongly to sleep continuity within the insomnia group—where autonomic dysregulation is prevalent—but not in healthy controls.
Neuroimaging studies provide additional support. Insomnia is associated with altered prefrontal and thalamocortical metabolic activity [8], and these networks appear sensitive to baseline energetic support. In metabolically vulnerable individuals, even subtle reductions in basal metabolic drive may destabilize sleep-promoting circuits. Thus, the selective presence of predicted BMR–sleep continuity coupling in insomnia suggests a network-level vulnerability heightened by metabolic insufficiency, rather than a universal metabolic elevation predicted by classical hyperarousal theory.
This study has several strengths. First, the use of strict ageand sex-matching controlled for the two strongest determinants of BMR [13–15]. Second, PSG was used to quantify objective sleep structure, avoiding the limitations of self-reported sleep measures. Third, the exclusion of OSA, PLMD, major psychiatric illness, and neurologic disorders minimized confounding effects. Finally, this is one of the first studies to evaluate metabolic–sleep coupling in a large insomnia cohort using predictive BMR, offering novel insight into the metabolic physiology of insomnia.
Several limitations should be noted. Predicted BMR is an estimate and may diverge from directly measured metabolic rate. Although validated population-level tools [13–15], predictive formulas do not capture short-term metabolic fluctuations or physiologic hyperarousal biomarkers such as heart rate variability, cortisol rhythms, or body temperature [4–7]. Additionally, the retrospective single-center design may introduce selection bias, and residual confounding from lifestyle or stress factors cannot be fully excluded. Finally, mechanistic pathways linking metabolic tone to sleep continuity were inferred from existing literature rather than directly measured.
Despite these limitations, our findings suggest that metabolic vulnerability may meaningfully influence sleep stability in individuals with insomnia. This supports emerging models that emphasize physiologic heterogeneity and neuro–metabolic fragility rather than a uniform hypermetabolic state. Future studies integrating direct calorimetry, autonomic indices, endocrine markers, and multimodal neuroimaging are needed to validate metabolic subtypes and elucidate mechanisms linking metabolic tone with sleep continuity. Predicted BMR, although indirect, may serve as a practical and non-invasive indicator of metabolic vulnerability contributing to sleep fragmentation within insomnia phenotypes.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation 
Print 
