This was a prospective, observational, cross-sectional analytical study based on a cohort of patients with clinical suspicion of obstructive sleep apnea (OSA) who underwent Type I polysomnography between August 2023 and March 2025. Data were prospectively collected from two specialized sleep medicine centers in Colombia’s Caribbean region as part of a broader institutional project on diagnostic approaches to sleep-related breathing disorders. This subanalysis focused exclusively on older adults, defined as individuals aged 60 years or older, with the aim of characterizing sleep efficiency, architecture, and fragmentation in this population.

We included patients ≥60 years old with a complete polysomnographic recording of at least 360 minutes of interpretable data and complete clinical, anthropometric, and laboratory information. Exclusion criteria were a prior diagnosis of other sleep disorders (e.g., chronic primary insomnia, narcolepsy, or periodic limb movement disorder), previous treatment with CPAP or mandibular advancement devices, or major neurological conditions that could interfere with sleep architecture, such as advanced dementia or Parkinson’s disease.

Clinical and demographic variables were extracted from electronic medical records and institutional databases. Data collected included age (both continuous and categorical: 60 - 69, 70 - 79, ≥80), sex, body mass index (BMI, kg/m²), marital status, educational level, and socioeconomic stratum. We also documented chronic comorbidities such as hypertension, type 2 diabetes, dyslipidemia, cardiovascular disease, and smoking history (active, former, or never smoker).

The decision to stratify patients into 60 - 69 versus ≥70 years was based on previous geriatric sleep research, where age 70 is frequently used as a threshold for increased frailty and accelerated decline in restorative sleep stages (e.g., N3 and REM). This cut-off has also been applied in large population-based studies such as the HypnoLaus and Sleep Heart Health Study, which reported steeper deterioration in sleep architecture beyond this age. Similarly, the selection of <10% versus ≥10% of N3 sleep as a categorical variable follows prior literature identifying this threshold as clinically relevant in older adults. Proportions of N3 below 10% have been associated with frailty, functional decline, and increased mortality. These thresholds, therefore, provide a clinically meaningful framework to explore vulnerability markers in this population.

All Type I polysomnographic studies were performed by certified technicians and supervised by sleep medicine specialists, following the technical standards of the American Academy of Sleep Medicine (AASM). The recordings included continuous monitoring of EEG (C3-M2 and C4-M1 channels), EOG, submental EMG, thoracoabdominal respiratory effort, nasal and oral airflow (thermistor and nasal pressure transducer), pulse oximetry, snore sensor, single-lead ECG, and body position.

The following variables were extracted from the PSG reports: total sleep time (TST, in minutes), sleep efficiency (ratio of TST to total time in bed), apnea-hypopnea index (AHI, defined as number of respiratory events per hour of sleep), microarousal index (arousals per hour), average heart rate, minimum and mean oxygen saturation, and percentage of time spent in each sleep stage (N1, N2, N3, REM), expressed as a proportion of TST. All variables were validated through double review and manually curated to eliminate duplicate or incomplete records. The final sample of older adults with valid data comprised 129 patients, representing 27.9% of the initial cohort (n = 462). Data were anonymized to ensure confidentiality.

After structuring the final dataset, we performed data cleaning and tabulation. Missing values were assessed and handled using listwise deletion when the missing rate was <5%. For key continuous variables (efficiency, sleep stages, AHI), distribution was evaluated using Shapiro-Wilk, Kolmogorov-Smirnov, and Anderson-Darling tests, complemented by visual inspection of histograms and Q-Q plots. These assessments informed the selection of appropriate parametric or non-parametric statistical tests.

The older adult subsample (n = 129) allowed estimation of means with an absolute precision of ±5 percentage points for continuous variables and detection of intergroup differences of at least 10 points with ≥80% power, assuming 95% confidence and moderate variances.

The statistical analysis was conducted in three phases. First, univariate descriptive statistics were generated. Continuous variables with normal distribution were reported as mean ± standard deviation; those with non-normal distribution were summarized as median and interquartile range. Categorical variables were described using absolute and relative frequencies.

Second, bivariate analyses were conducted to compare groups based on AHI severity categories (mild: 5 - 14.9; moderate: 15 - 29.9; severe: ≥30 events/hour), sex, and presence of comorbidities. For continuous variables, comparisons used Student’s t-test or Mann-Whitney U test as appropriate; for categorical variables, Pearson’s chi-square test or Fisher’s exact test was used when expected frequencies were <5.

In the third phase, multivariate analysis was performed using multiple linear regression models to identify independent predictors of three key outcomes: sleep efficiency, microarousal index, and N3 sleep proportion. Models were adjusted for age, sex, BMI, and presence of hypertension or cardiovascular disease, predefined as potential confounders. Assumptions of linear models were verified, including normality of residuals, homoscedasticity, and absence of multicollinearity (tolerance > 0.7 and VIF < 2). Statistical significance was set at p < 0.05. Analyses were conducted using Jamovi v2.3.18 (based on R), with partial replication in SPSS v26.0.

This study was approved by the Research Ethics Committee of Universidad del Sinú—Cartagena Campus (Approval No. 05-2023) as part of a registered protocol. All patients in the original cohort provided written informed consent, and this subanalysis was conducted using anonymized data, in compliance with the Declaration of Helsinki and Colombian ethical regulations (Resolution 8430 of 1993, Ministry of Health).

Table 1 summarizes clinical and polysomnographic characteristics by sex. While age was similar between men and women, meaningful differences in sleep structure were observed. Women showed a higher proportion of N1 sleep, indicating a greater tendency toward light sleep. In contrast, men had slightly higher percentages of N3 sleep, which is considered physiologically restorative. Minimum oxygen saturation was lower in men, consistent with their greater burden of severe OSA. Figure 1 presents these sex-based differences via boxplots, highlighting greater dispersion of N3 sleep percentage in women, which may reflect more heterogeneous physiological responses to upper airway collapse.

Data are presented as mean, standard deviation (SD), median, minimum, and maximum values. Variables include demographic data (age, BMI, neck circumference), sleep parameters (AHI, oxygen saturation, microarousals, total sleep time [TST]), and distribution of sleep stages (N1, N2, N3, REM).

Table 2 shows a clear pattern of sleep disruption with increasing AHI severity. Patients with severe OSA had lower proportions of N3 and REM sleep, and higher microarousal indices. These alterations reflect a vicious physiopathological cycle in which respiratory events disrupt sleep continuity, prevent entry into deep stages, and trigger frequent autonomic activations. The median microarousal index increased from 18.6 events/hour in the mild group to 30.4 in the severe group, while minimum oxygen saturation dropped from 84.1% to 78.5%. Figure 2 illustrates the linear relationship between age and sleep fragmentation (Spearman r = 0.34), suggesting a synergistic effect of aging and OSA severity on sleep quality.

Table 3 compares polysomnographic variables between age groups (60 - 69 vs. ≥70 years), showing progressive deterioration of sleep architecture in the older group. The mean N3 percentage was significantly lower in those ≥70 years (10.1%) compared to the younger group (13.7%), with higher N1 sleep and more microarousals. These findings are consistent with known cerebral aging processes—such as ventrolateral preoptic nucleus atrophy, circadian desynchronization, and decreased slow-wave activity.

Values represent mean, standard deviation (SD), median, minimum, and maximum for each sleep stage (N1, N3, REM), microarousal index, and minimum oxygen saturation. Progressive disruption in deep and REM sleep, along with increased microarousals and lower oxygen saturation, was observed with increasing AHI severity.

Data include sleep stage percentages (N1, N3, REM), microarousal index, and minimum oxygen saturation. Older age was associated with increased sleep fragmentation and reduced deep (N3) and REM sleep, suggesting a deterioration of sleep architecture consistent with physiological aging.

Figure 3 , a Spearman correlation heatmap, revealed inverse associations between AHI and deep sleep stages (REM: r = −0.31; N3: r = −0.27), and positive correlations between age and microarousals. This suggests that both aging and OSA severity contribute to increased fragmentation and loss of restorative sleep, although multivariate models attenuated this effect.

To explore independent predictors of N3 sleep, we conducted a multiple linear regression analysis, as shown in Table 4 . Although none of the covariates reached statistical significance, there was a negative trend between age and N3 sleep (β = −0.17; 95% CI: −0.36 to 0.02; p = 0.077). This association, although consistent with previous evidence, was weak and explained little variance (adjusted R² = 0.11), which highlights the limited explanatory power of the selected predictors. These findings underscore that the decline in deep sleep among older adults is likely influenced by additional, unmeasured variables. In particular, neuroinflammation and physical activity deserve attention: neuroinflammation has been linked to aging-related cognitive impairment and may directly alter slow-wave generation, while regular physical activity is a well-established modulator of sleep continuity and depth. Both factors are especially relevant in older adults, where systemic inflammation and reduced mobility are frequent. Incorporating these variables in future studies could help clarify biological mechanisms and identify modifiable targets for intervention. The regression coefficients and confidence intervals for this model are displayed in Figure 4 (forest plot).

Lower N3 sleep was associated with older age, higher AHI, increased microarousals, and lower minimum oxygen saturation—suggesting this group represents a more functionally and physiologically vulnerable phenotype.

To evaluate the potential role of N3 sleep as a physiological marker, we constructed Table 4 by comparing patients with reduced deep sleep (<10%) to those with preserved levels (≥10%). The group with low N3 sleep was significantly older, had higher AHI, greater sleep fragmentation (higher microarousal index), and lower minimum oxygen saturation. These findings support the hypothesis that reduced deep sleep is not an isolated feature but a reflection of overall disease burden and impaired sleep continuity. Within geriatric sleep medicine, the concept of a functional marker refers to a parameter that transcends descriptive measurement, reflecting physiological reserve and vulnerability to decline. In this context, the proportion of N3 sleep may serve as such a marker, identifying older adults with diminished resilience and higher risk of adverse outcomes.

One of the most consistent findings in our analysis was the stepwise degradation of sleep architecture with increasing AHI severity. As shown in Table 2 , patients with severe OSA (AHI ≥30) had significantly lower proportions of N3 and REM sleep, alongside a higher number of microarousals and lower nadir oxygen saturation. These results are consistent with prior reports and reflect the recognized pathophysiological cascade in OSA: more frequent respiratory events lead to heightened autonomic activation, more fragmented sleep, and suppression of the deeper sleep stages [20] . The SEPAR guidelines stress that the AHI alone is insufficient to classify clinical severity in OSA and recommend integrating additional parameters such as oxygen desaturation, sleep fragmentation, and architecture metrics. In our study, this progressive disorganization was especially marked in the severe OSA group, in which N3 sleep averaged less than 9% of total sleep time, and REM sleep remained under 14%. These figures are consistent with the findings by Kimoff and colleagues, who described how increased respiratory burden disrupts the slow oscillations that define N3 sleep [21] .

Sleep fragmentation also increased in parallel with OSA severity. Patients with severe OSA had a median of 30.4 microarousals per hour, compared to 18.6 in those with mild OSA. This finding illustrates the direct effect of respiratory events on sleep stability and underscores the importance of considering arousal metrics when making clinical decisions. It has been proposed that the burden of microarousals may account for the variability in subjective sleep quality and daytime symptoms, even among patients with similar AHI scores.

In the age-group comparison ( Table 3 ), patients ≥70 years exhibited a more deteriorated sleep architecture than their counterparts aged 60 - 69 years. This difference was reflected in a lower proportion of N3 sleep (10.1% vs. 13.7%), a higher proportion of N1 sleep, and an increase in microarousals. The loss of deep sleep with aging has been widely documented and is associated with a reduction in the amplitude of cortical slow waves, thalamic dysfunction, and decreased synaptic plasticity [22] . In Figure 4 , this redistribution is clearly visualized through a stacked bar chart: the older group concentrates more time in lighter sleep stages, with lower representation of restorative phases. This trend not only reflects a physiological process but can also be exacerbated by comorbid factors such as frailty, polypharmacy, or cardiovascular disease, all of which are common in this population.

REM sleep, which also decreases with age, showed an average decline of nearly 3 percentage points in the ≥70 group. This stage is involved in emotional consolidation, autonomic regulation, and brain metabolism, and its reduction has been linked to cognitive impairment, depressive symptoms, and increased cardiovascular mortality [23] [24] . Longitudinal studies have demonstrated that the progressive loss of REM sleep in older adults predicts the risk of dementia, independently of AHI, reinforcing the need for a comprehensive evaluation of sleep stages in this population [25] .

A multiple linear regression model ( Table 4 ) was used to identify independent predictors of N3 sleep proportion. Although no variables reached statistical significance, a negative trend between age and N3 sleep was observed (β = −0.17; p = 0.077), which is consistent with physiological expectations. This finding aligns with the work of Redline et al., who documented an approximate 1% annual decline in N3 sleep beginning at age 60, even in individuals without respiratory disorders [26] . The low adjusted R² confirms that unexplored domains—including systemic inflammation, lifestyle, and neurocognitive decline—could play decisive roles. Figure 4 (forest plot) visually represents these findings, showing that neither BMI nor AHI had a significant association with deep sleep in the adjusted model. This is consistent with previous studies suggesting that upper airway collapse is more closely related to sleep fragmentation than to changes in the relative distribution of sleep stages. The determinants of N3 sleep appear to be largely neurological and hormonal [27] .

Stratifying patients by N3 sleep proportion (<10% vs. ≥10%) allowed us to identify a subgroup with a more adverse clinical profile. Those with reduced N3 sleep were older, had higher AHI values, more microarousals, and lower minimum oxygen saturation. These findings suggest that the loss of deep sleep may not simply reflect aging, but also a cumulative physiological burden. Recent studies have proposed that N3 sleep could serve as a functional biomarker in geriatrics, correlating with physical decline, systemic inflammation, and frailty [28] . Ancoli-Israel et al. reported that institutionalized older adults with N3 sleep proportions below 10% had increased risks of falls, greater dependence in basic activities of daily living, and higher 12-month mortality [28] . Our results are in line with this evidence, showing that patients with reduced N3 sleep also had a greater objective burden of respiratory disease and more pronounced sleep disruption. International guidelines now recommend considering sleep efficiency and architecture when making clinical decisions, especially in geriatric populations, and our findings strongly support this approach.