Association of bedroom particulate matter, sleep quality and next-day physical performance

Abstract

Few studies have investigated the association of exposure to PM_2.5 at the individual level on sleep quality and next-day physical performance, which are both important for human health. To fill the gap, this field study was conducted among 183 young adults who were required to participate in standardized physical fitness test which objectively evaluated their physical performance, with their bedroom environment and sleep quality of the night prior to the fitness assessment were continuously monitored. Multiple linear regression was used to analyse the association between the bedroom environment, sleep quality and next-day physical performance, and to examine the interaction effects of environmental factors. The results show that PM_2.5 was significantly associated with a reduction in both the proportion of deep sleep and the next-day performance of a long-distance running test. Its negative association with long-distance running performance was exacerbated by a high CO₂ level (3,961 ppm) during sleep. These findings suggest the importance of good bedroom air quality (low levels of PM_2.5 and CO₂) for maintaining occupants’ sleep quality and health. For young people, maintaining a clean and well-ventilated sleeping environment that minimizes indoor particulate matter exposure is essential for ensuring good sleep quality and safeguarding physical health.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-37949-2.

Introduction

Air pollution is a major global health risk¹, and fine particulate matter (PM_2.5) is of particular concern because of its ability to penetrate deep into the respiratory tract and affect multiple physiological systems^2,3. Rapid urbanization, industrial activities, and climate-related events have increased particulate pollution worldwide, and outdoor particles readily infiltrate indoor spaces through ventilation and natural air exchange^4–6. In developing countries, indoor PM_2.5 concentrations often approach those outdoors, and indoor–outdoor PM_2.5 infiltration rates can even exceed 1 during certain seasons⁷. The effects may be particularly pronounced in bedrooms⁸, where individuals spend nearly one-third of their time at low metabolic rates⁹, potentially constituting a vulnerable exposure window. Despite this, studies specifically examining particulate matter in bedroom environments remain scarce.

Maintenance of good sleep quality and physical performance are important to long-term physical and mental health¹⁰. Emerging evidence has shown that PM_2.5 exposure is associated with deteriorations in sleep health^11,12. Kim et al.¹³ discovered that exposure to specific heavy metal PM (lead, manganese, cadmium, and aluminum) in the air was associated with reduced subjective sleep quality in college students, as determined by a subjective assessment using the Pittsburgh Sleep Quality Index (PSQI). Using objective sleep monitoring, Thanh Tung et al.¹⁴ demonstrated that exposure to PM_2.5 significantly reduced the proportion of deep sleep, suggesting PM_2.5 may affect the core recovery function of sleep. Mechanistically, PM_2.5 may provoke inflammatory and oxidative responses and alter autonomic balance (reduced heart-rate variability, sympathetic predominance), which can disrupt slow-wave sleep and impair restorative processes^15–20. However, these studies and most of the other investigations that have examined the impacts of PM_2.5 all have the same limitation: they depend on PM_2.5 exposure estimates derived predominantly from outdoor environmental monitoring stations. Consequently, such assessments cannot claim to have correctly assessed an individual’s actual inhalation concentration within a particular environment (e.g., a confined dormitory), diminishing the reliability of causal inference.

In parallel, a growing body of research has investigated the influence of PM_2.5 exposure on physical performance, consistently showing that high PM_2.5 concentrations during exercise impair cardiorespiratory fitness, endurance, and performance^21,22. From the perspective of physiology, previous studies have shown that PM_2.5 exposure can reduce maximal oxygen uptake (VO₂max)²³, impair pulmonary function^24,25, and damage skeletal muscle mitochondria^26,27. These studies have shown that these health effects may persist for several hours after the exposure^28,29, but a fundamental question has been overlooked: does particulate exposure during nighttime sleep influence next-day physical performance, either by decreasing sleep quality or more directly? The available literature on this subject is limited. Of particular note is the reliance on macroscopic exposure data in existing studies of the PM-physical performance association, as well as the lack of any individualized and fine-grained exposure assessment. This has resulted in difficulties in accurately reflecting the relationship between the actual exposure levels of individuals and their physical performance responses.

To overcome these limitations, the present study combines individual-level PM exposure monitoring in bedroom environments with detailed sleep assessment, followed by standardized physical performance testing the next morning. This design enables the examination of the short-term effects of nocturnal particulate exposure on sleep quality and subsequent exercise performance under realistic living conditions. The findings aim to provide scientific evidence for optimizing bedroom air quality management and developing practical interventions to protect both sleep and physical performance from the adverse impacts of particulate pollution.

Method

Approach

A field study was conducted on how the bedroom environment affects sleep quality and physical performance during the transition season in Shanghai, China. The study was conducted during the official school physical fitness tests, specifically on the weekends from 30th March to 21st April and 2nd November to 30th November, 2024. Weekend scheduling ensured alignment with the fixed testing timetable while also reducing variability in students’ daytime academic workloads and extracurricular activities, which tend to fluctuate more on weekdays. To further control for external activity-related variation, all participants were instructed to maintain their regular daily routines, avoid vigorous exercise, and refrain from consuming caffeine-containing beverages for 48 h prior to the monitored night. They were also required to return to the dormitory by their usual evening time and to refrain from late-night social activities, ensuring greater consistency in pre-sleep behaviors. During each study night, subjective and objective assessments of bedroom environment parameters were conducted in each dormitory room, and sleep quality was monitored using a wearable sleep tracker. On the following day, each participant completed the official outdoor physical fitness test, which served as the standardized measure of physical performance. Ambient meteorological variables (including outdoor temperature, relative humidity, wind speed, and global solar radiation) and particulate matter concentrations were monitored to record the outdoor environmental conditions during the testing periods to document the outdoor environment.

All participants resided in university-arranged dormitory buildings with standardized layouts, facilities, and management, which helped minimize variability in living environments. These dormitories were not equipped with mechanical ventilation systems, and during the transition season the thermal environment generally remained within a neutral range, with no use of fans or air-conditioning, as confirmed by field observations and measurements. Importantly, the dormitory setting reflects students’ typical daily living conditions, thereby enhancing the ecological validity of the study. All study protocols were approved by the Ethics Committee of Shanghai Jiao Tong University (No. E20240609C) and followed the guidelines in the Declaration of Helsinki. Each participant signed an informed consent form.

Participants

Prior to conducting the study, a sample size calculation was performed using G*Power. For the model with the largest number of predictors (sleep environment to physical performance), an F-test for linear multiple regression (fixed model, R² deviation from zero) was used, with a medium effect size (f² = 0.15), α = 0.05, power (1-β) = 0.8, and a total of 19 predictors (1 main effect and 18 covariates). The calculated required sample size was 153. When including an additional interaction term, the total number of predictors increased to 20, yielding a required sample size of 157. A total of 183 undergraduate students were recruited, with valid data collected from 163 participants. This indicates that the sample size was sufficient to detect effects of at least medium magnitude.

Participants were recruited through an online questionnaire distributed to all undergraduate students via the university’s official online platforms, including class groups and the student affairs system, ensuring broad accessibility. The questionnaire was open to all students participating in the mandatory physical fitness test, rather than being targeted at specific dormitories or subgroups, thereby minimizing selection bias. Given that the physical fitness test is a university-wide requirement and all undergraduates must participate, the participant population was relatively homogeneous in terms of health status and daily habits, further reducing the likelihood of recruitment bias. Following the initial recruitment, applicants were screened based on the information provided. The inclusion criteria were as follows: an absence of smoking habits and alcohol abuse, and no current or recent use of any medication. Participants were aged 18–22 years. All participants met the study eligibility criteria and complied with the standardized pre-sleep behavioral requirements described above. They were required to attend the scheduled fitness test the following day and to continuously wear a sleep-tracking device throughout the monitored sleep period. In addition, female participants were asked to complete the assessments during their non-menstrual period. These screening and compliance requirements aimed to further reduce individual-level confounding related to health and lifestyle factors.

Measurements

Environment parameters

Indoor environmental data were collected using a portable indoor environment detector that simultaneously measured air temperature, relative humidity, CO₂ concentration and PM_2.5 concentration using integrated calibrated sensors. The specific sensor models, their manufacturers, measurement ranges, and accuracies are detailed in Table 1. Measurements were logged at 5-minute intervals, and participants were instructed to place the device at the head of the bed to capture the micro-environment most relevant to sleep. Meteorological parameters, such as outdoor air temperature, relative humidity, wind speed and solar radiation, were obtained from a micro-meteorological station, with all equipment specifications provided in Table 1. Outdoor PM_2.5 concentrations were derived from a nearby meteorological bureau. The accuracy specifications for environment parameters met the requirements for characterizing environmental conditions in field studies and were sufficient for detecting the variations relevant to this research. All instruments were newly purchased and factory-calibrated according to manufacturer specifications prior to deployment.

Table 1.

Specifications of environment measurement instruments and sensors.

Monitoring scenario	Measurement instrument	Parameter	Sensor	Measurement range	Accuracy
Indoor environment	iBEM indoor environment monitor	Air temperature (Ta)	SHT30(Sensirion AG, Switzerland)	-40 ~ 80 °C	± 0.3 °C
		Relative humidity (RH)	SHT30(Sensirion AG, Switzerland)	0–99%	± 5%
		CO2 concentration	S8-0053(SenseAir, Sweden )	400 ~ 5,000 ppm	± 40 ppm
		PM2.5 concentration	PMSA003-A(Plantower China)	0 ~ 1,000 ug/m3	± 10 ug/m3
Outdoor environment (meteorological parameters)	HOBO H21 microclimate stations	Air temperature	S-THB-M002(HOBO, USA)	−40 °C to 75 °C	± 0.21 °C
		Relative humidity	S-THB-M002(HOBO, USA)	0 to 100%	± 2.5%
		Wind speed	S-WSB-M003(HOBO, USA)	0 to 76 m/s	± 1.1 m/s
		Solar radiation	S-LIB-M003(HOBO, USA)	0 to 1280 W/m2	± 10 W/m2

Sleep quality

Objective sleep quality was monitored by a sleep tracker (Fitbit Alta HR) worn on each participant’s non-dominant wrist. Sleep quality parameters were recorded in real time, including total sleep time (TST), sleep efficiency (SE), time awake, and the duration and proportion of light, deep, and rapid eye movement (REM) sleep. Like other consumer-grade actigraphy devices, the Fitbit Alta HR has certain limitations. Specifically, its sleep–wake detection relies on accelerometry and heart-rate variability rather than EEG, which may reduce accuracy in distinguishing detailed sleep stages compared with clinical polysomnography. Despite these limitations, Fitbit devices have been widely adopted in field-based sleep research^30–32, and were in good agreement with the polysomnogram³³. Given the logistical constraints of large-scale, in-situ monitoring in real-world residential environments, the Fitbit Alta HR provides a practical, non-invasive, and reliable tool for capturing variations in sleep patterns across participants.

Physical performance

The university mandated that all undergraduate students must undergo a fitness test at least once per academic year. This test comprised a 1000-metre run (for male students) or 800-metre run (for female students), a 50-metre short-distance run (sprint), one-minute of pull-ups (for male students) or sit-ups (for female students), and a standing long jump. The test is designed to assess various physical abilities, including aerobic endurance, speed, core muscle stability, and body coordination and balance.

The fitness tests are administered following national and university protocols, with standardized procedures, trained examiners, and calibrated equipment. These tests are part of the official academic evaluation, ensuring consistent administration and participant engagement. Individuals whose physical condition was temporarily compromised (e.g., due to illness or injury) were permitted to request postponement, further ensuring the validity of the performance measurements. Raw performance metrics (time, counts, distance) were converted to standardized scores using the national student physical fitness scoring rubric (0–100 scale) prescribed by the Ministry of Education. These officially recorded standardized scores were used as the indicators of next-day physical performance in all analyses.

As a field-based assessment, the fitness test provides objective measures of functional performance under real-world conditions, although it does not capture detailed physiological variables such as VO₂max or muscle oxygenation. Its strength lies in the standardized administration and broad national implementation, which enhances reproducibility and comparability across participants and studies.

Subjective questionnaire

To characterize habitual sleep patterns over the past month, the Pittsburgh Sleep Quality Index (PSQI)—a widely validated and reliable instrument for assessing sleep quality in young adults—was administered at recruitment³⁴. Information on participants’ regular sleep routines was also collected. These baseline sleep measures were subsequently included as covariates to account for inter-individual differences that might influence associations among the sleep environment, and sleep quality. During the study period, participants completed brief online questionnaires each evening and morning. To ensure temporal consistency, the evening questionnaire was required to be completed within 1 h before bedtime, and the morning questionnaire within 1 h of awakening; time stamps were used to verify adherence. In the evening questionnaire, participants reported their diet, exercise, and daily activities, including the Physical Activity Rank Scale-3 (PARS-3), a validated tool for quantifying habitual activity levels in Chinese college populations³⁵. Participants also provided self-assessments of physical performance and fitness³⁶. These indicators were used as covariates to adjust for baseline fitness differences that could confound the relationship between sleep environment and physical performance. The morning questionnaire captured participants’ estimated sleep duration, perceived bedroom environmental conditions during sleep, and sleep quality, assessed using the Groningen Sleep Quality Scale (GSQS), an established and validated measure in observational sleep research³⁷. As illustrated in Fig. 1, subjective evaluations of the bedroom environment were rated on an 11-point scale³⁸. To enhance data quality, all questionnaires were pilot-tested prior to deployment to ensure clarity and feasibility. Responses submitted outside the designated time windows or missing key items were identified and excluded from night-level analyses. Although such criteria were predefined, no data were excluded for these reasons in the present study.

Fig. 1.

The scales used to collect ratings of the bedroom environment.

Data analysis

Multiple regression

The present section details the sequence of steps undertaken in the data analysis. Firstly, for this pre-processing of the data, the median method was employed to estimate missing values, while the outliers were substituted based on the interquartile range (IQR) method to enhance model robustness. In analysing the association between bedroom environment and physical performance, outdoor meteorological parameters were incorporated into the model as control variables to exclude the interference of meteorological factors in the sports environment on physical performance. Multiple regression analyses were conducted to investigate the associations between bedroom environmental factors, sleep quality, and next-day physical performance. Linear models were first fitted for PM_2.5 and CO₂ to assess their direct relationships. Based on prior evidence suggesting a non-linear association, a quadratic term was included for relative humidity (RH). For temperature (Ta), which ranged from 16 to 26 °C during the transitional season, the functional form was uncertain; therefore, the inclusion of a quadratic term was determined using the Akaike Information Criterion (AIC) to select the best-fitting model. In all models, covariates were selected a priori based on theoretical relevance and prior evidence regarding determinants of sleep and physical performance. Individual characteristics³⁹ (e.g., gender, body mass index) and lifestyle habits^40,41 (including napping behaviour, all-nighters, insomnia symptoms, tea/coffee or alcohol consumption, and habitual exercise frequency) were adjusted to account for baseline behavioural and physiological differences across participants. For models predicting physical performance, additional covariates were included to capture baseline fitness status (e.g., PARS-3, self-assessed physical performance and fitness levels). Because all physical performance tests were conducted outdoors, meteorological variables were incorporated to control for environmental conditions that may influence outcomes. Multicollinearity was inspected using variance inflation factors (VIFs) among different predictors. As polynomial terms (e.g., X and X²) are mathematically dependent and therefore exhibit structural—not problematic—collinearity, VIF diagnostics were applied only to distinct explanatory variables rather than to polynomial expansions of the same variable.

In order to investigate the hypothesis that other environmental factors may moderate the causative pathways of PM_2.5 effects on sleep quality and physical performance, this study further introduced interaction terms to construct multiple linear regression models. PM_2.5 was used as the primary independent variable, with air temperature (Ta), relative humidity (RH), and CO₂ being regarded as the moderating variables. This approach resulted in the construction of three models, each containing interaction terms. The following acronyms were employed: PM_2.5 × Ta, PM_2.5 × RH, and PM_2.5 × CO₂, respectively. In order to enhance the comparability of the results, all continuous variables were centred prior to analysis. The interaction effects were then tested for significance (p < 0.05) in order to determine the presence of moderating effects. The minimum, median, and maximum values were used to represent the low, medium, and high levels of the moderator variables, in order to examine the interaction trends.

Mediation effect

To examine whether sleep quality mediates the relationship between the bedroom environment and exercise performance, a three-step regression approach was used to test for mediation effects. The analytical framework followed the causal steps outlined by Baron and Kenny, with the indirect effect quantified by the product of coefficients method⁴². This approach was chosen because the study involved observational, continuous variables and aimed to quantify indirect pathways rather than estimate causal effects under experimental conditions. The classical Baron and Kenny procedure provides a clear structural decomposition of the total, direct, and indirect effects in such settings. The three regression equations are specified as follows:

1. Total effect of the independent variable on the dependent variable:

   1

   where Y denotes next-day physical performance, X represents environmental exposure (e.g., PM_2.5, CO₂), and “c” represents the total effect of X on Y.

2. Effect of the independent variable on the mediator:

   2

   where M represents the sleep quality indicator (e.g., TST, SE), and “a” represents the effect of X on the mediator.

3. Effect of the mediator on the dependent variable, controlling for the independent variable:

   3

   where “b” estimates the effect of the mediator M on the outcome Y, and “c’” represents the direct effect of X on Y after accounting for the mediation pathway.

   The indirect (mediated) effect of X on Y through M is quantified as the product of coefficients:

   4

   Bootstrapping with 5,000 iterations was used to estimate confidence intervals for the indirect effect, and mediation was considered significant if the 95% confidence interval did not include zero.

Results

Basic information

Characteristics of the subjects

Figure 2 describes the demographics and living habits of the 163 participants included in the analysis. The average age was 18.9 years, the mean BMI was 21.3, and the average PSQI score was 5.5. Gender was balanced (male-to-female ratio ≈ 1:1). Most participants led healthy lifestyles. Alcohol and tea/coffee consumption were prevalent, alongside individual differences in sleep-wake routines and physical exercise. To reduce inter-subject variability, consumption of these beverages was prohibited during the study. Alcohol consumption and the consumption of tea and coffee were prevalent, while there were certain individual differences in sleep routines and levels of physical exercise. In order to reduce the inter-subject variance, although some subjects reported habitual consumption of alcohol or tea/coffee, they were all prohibited from consuming these beverages during the experimental period.

Fig. 2.

Information about the study population and their lifestyle habits (N = 163).

Environmental parameters, sleep quality and physical performance

Figure 3 summarizes the objective and subjective environment measures. The mean bedroom Ta, RH, CO₂ and PM_2.5 concentrations were 21.3 °C, 62%, 1,437 ppm, 16 µg/m³, respectively. The average overall comfort score was 7.3, indicating a generally comfortable environment. Perceived temperature (mean 5.1) and humidity (mean 4.9) were near neutral, while air freshness, air velocity, lighting, and noise were rated at 6.5, 1.7, 2.6, and 2.9 respectively, suggesting acceptable subjective environment. Meteorological parameters included outdoor air temperature (20.3 ± 4.0 °C), RH (63 ± 17%), wind speed (0.2 ± 0.2 m/s), solar radiation (250 ± 111 W/m²) and PM_2.5 (23 ± 13 µg/m³). These were used as control variables in subsequent correlation analyses, to adjust for outdoor environmental influences on physical performance.

Fig. 3.

Objective measurements and subjective evaluation of environment parameters.

Table 2 presents both objective and subjective sleep quality scores. Objective measures were based on National Sleep Foundation (NSF) criteria⁴³, which primarily distinguish the general population from sleep disorder patients. Except for a slightly inadequate TST (20.9% rated as “inappropriate”), most objective sleep indicators fell within acceptable ranges, and subjective sleep quality was generally rated as satisfactory.

Table 2.

Objective measurements and subjective evaluation of sleep quality.

Sleep quality indexes		All data Mean ± SD, (Q1, Q3).	Sleep quality recommendations
			Appropriate, n (%)	Uncertain, n (%)	Inappropriate, n (%)
Objective parameters	Total sleep time (min)	408.7 ± 57.3 (367.0, 444.5)	63 (38.7%)	66 (40.5%)	34 (20.9%)
	Sleep efficiency (%)	87.1 ± 3.0 (85.0, 89.3)	122 (74.8%)	41 (25.2%)	0 (0.0%)
	Time awake (min)	60.8 ± 17.0 (48.0, 70.0)	/	/	/
Duration of sleep stages	REM sleep (min)	92.5 ± 28.3 (74.0, 109.0)	/	/	/
	Light sleep (min)	235.2 ± 49.9 (201.0, 260.0)	/	/	/
	Deep sleep (min)	80.9 ± 20.9 (69.0, 95.0)	/	/	/
Proportion of sleep stages	REM sleep proportion (%)	22.6 ± 5.9 (18.8, 26.7)	/	163 (100%)	0 (0%)
	Light sleep proportion (%)	57.4 ± 8.0 (52.0, 62.3)	/	/	/
	Deep sleep proportion (%)	20.0 ± 5.1 (17.3, 23.5)	/	161 (98.8%)	2 (1.2%)
Subjective sleep indicators	GSQS	3.7 ± 1.2 (3.0, 4.0)	/	/	/

* REM: Rapid Eye Movement; GSQS: Groningen Sleep Quality Scale; Sleep quality recommendations for young adults (18–25 years): Total sleep time: 7–9 h is considered appropriate; 6 h or 10–11 h is classified as uncertain; <6 h or > 11 h is considered inappropriate. Sleep efficiency: ≥85% is appropriate; 65–84% is uncertain; ≤64% is inappropriate. REM sleep proportion: No consensus recommendation is available. For descriptive classification, > 40% is considered inappropriate, ≤ 40% is uncertain. Deep sleep proportion: No consensus recommendation is available. For descriptive classification, > 5% is considered inappropriate, ≤ 5% is uncertain. The symbol “/” indicates that the NSF does not provide a recommended value for this parameter.

As illustrated in Table 3, the subjects’ physical performance in five fitness tests was evaluated: standing long jump, long-distance running, sprint, sit-ups and pull-ups. For the standing long jump, the mean score was 79.5 ± 18.7, with a pass rate of 92.9% and an excellent-good rate of 61.2%. The long-distance running test had an average score of 72.4 ± 14.0, a pass rate of 80.6%, and an excellent-good rate of 30.0%, suggesting that while the majority met the standard, those demonstrating high-level performance were relatively few. Sprint performance was most consistent (79.1 ± 10.7) with a 98.8% pass rate, and 38.9% excellent-good rate. Sit-ups averaged 77.0 ± 13.2 (96.0% pass, 44.0% excellent-good). Pull-ups showed the weakest performance, with the lowest mean score (27.1 ± 34.1), pass rate (26.5%), and the excellent-good rate (12.0%). The subjects demonstrated proficiency in the sprint, standing long jump, and sit-ups events, exhibiting high pass rates and excellent-good rates. The long-distance running event exhibited moderate performance, and the pull-ups exhibited considerable room for improvement.

Table 3.

Objective measurements of physical performance.

Items		Score	Sample Size	Excellent ≥ 90	Good [80, 90)	Pass [60, 80)	Fail < 60
Standing long jump	All	79.5 ± 18.7 (72.0, 90.0)	157	54 (34.4%)	42 (26.8%)	50 (31.8%)	11 (7.0%)
	Female	83.9 ± 19.0 (78.0, 95.0)	74	37 (50.0%)	17 (23.0%)	16 (21.6%)	4 (5.4%)
	Male	75.5 ± 17.5 (68.0, 85.0)	83	17 (20.5%)	25 (30.1%)	34 (41.0%)	7 (8.4%)
Long-distance race	All	72.4 ± 14.0 (66.0, 80.0)	160	15 (9.4%)	33 (20.6%)	96 (60.0%)	16 (10.0%)
	Female (800 m)	73.8 ± 16.6 (72.0, 80.0)	74	11 (14.9%)	17 (23.0%)	38 (51.4%)	8 (10.8%)
	Male (1000 m)	71.3 ± 11.2 (66.0, 78.0)	86	4 (4.7%)	16 (18.6%)	58 (67.4%)	8 (9.3%)
Sprint	All	79.1 ± 10.7 (72.0, 85.0)	157	29 (18.5%)	32 (20.4%)	95 (60.5%)	1 (0.6%)
	Female	76.1 ± 10.1 (70.0, 80.0)	74	8 (10.8%)	16 (21.6%)	49 (66.2%)	1 (1.4%)
	Male	81.8 ± 10.6 (74.0, 87.5)	83	21 (25.3%)	16 (19.3%)	46 (55.4%)	0 (0.0%)
Sit-ups	Only for female	77.0 ± 13.2 (68.0, 85.0)	75	15 (20.0%)	18 (24.0%)	39 (52.0%)	3 (4.0%)
Pull-ups	Only for male	27.1 ± 34.1 (0.0, 60.0)	83	8 (9.6%)	2 (2.4%)	12 (14.5%)	61 (73.5%)

Association between bedroom environment and sleep quality

To ensure clarity and avoid overextending the main text, only statistically significant associations between environmental factors and sleep quality are reported in this section; full model results, including non-significant findings, are provided in the Appendix 1. The maximum VIF among all non-polynomial predictors was 1.37, indicating no problematic multicollinearity. As shown in Fig. 4, the association between the bedroom environment and sleep quality was analysed using regression analysis. Only the results that were found to be significant are presented. The findings indicate a negative correlation between PM_2.5 and the proportion of deep sleep (DS) , with a significance level of p = 0.03. There was a positive correlation between CO₂ and light sleep (LS) proportion (p = 0.04). Ta was significantly and positively correlated (p = 0.03) with DS proportion. In summary, exposure to PM_2.5 was associated with a reduction in the proportion of DS, while elevated CO₂ levels were associated with an increase in the proportion of LS. Within the observed temperature range of approximately 16–26 °C, a rise in Ta was associated with an increase in the proportion of DS. No other significant correlations between bedroom environment and sleep quality were observed.

Fig. 4.

Regression analysis of bedroom environmental factors on sleep quality. The regression formula in the figure only shows the coefficients of the predictor X (PM_2.5, CO₂, Ta), while the effects of other covariates are held at their sample means and incorporated into the intercept b. In other words, the equation represents the estimated effect of X on Y while controlling for the average values of all other covariates. In addition, when analysing the effect of a specific environmental variable, the other environmental variables were included in the model as covariates. Other covariates included gender, BMI, PSQI score, presence of chronic diseases, and lifestyle habits (all-nighters, napping, insomnia, and tea/coffee or alcohol consumption).

Association between bedroom environment and next-day physical performance

For consistency with the previous section, only significant associations between environmental factors and physical performance are summarized here, while the complete set of model results is available in the Appendix 2. The maximum VIF among all non-polynomial predictors was 4.81, indicating no problematic multicollinearity. As shown in Fig. 5, the present study investigated the association between the bedroom environment and physical performance. A substantial negative correlation was found between PM_2.5 and long-distance running performance. The relationship between RH and long-distance running performance exhibited a typical nonlinear trend. The fitted quadratic regression model shows that the relationship was statistically significant (p = 0.01). The optimum performance was observed when the RH was approximately 62%, and a low or excessive level of humidity was associated with a decline in performance. It is evident that both the PM_2.5 and RH of the bedroom environment tend to predict the next-day performance of the long-distance running test. The findings of the study show that PM_2.5 had a linear association with a reduction in performance, while RH had an inverted “U”-shaped optimal range association. No other significant correlations between bedroom environment and physical performance quality were observed.

Fig. 5.

Regression analysis of bedroom environmental factors on next day physical performance. The regression formula in the figure only shows the coefficients of the predictor X (PM_2.5, RH), while the effects of other covariates are held at their sample means and incorporated into the intercept b. In other words, the equation represents the estimated effect of X on Y while controlling for the average values of all other covariates. In addition, when analysing the effect of a specific environmental variable, the other environmental variables were included in the model as covariates. Other covariates included gender, BMI, presence of chronic diseases, lifestyle habits (all-nighters, napping, insomnia, and tea/coffee or alcohol consumption), the PARS-3 ranking, self-assessed physical performance and fitness levels, and outdoor meteorological parameters (temperature, humidity, wind speed, and solar radiation).

Interaction of particulate matter with other environmental factors

As shown in Fig. 6, The PM_2.5 × CO₂ interaction was significant for running performance (p = 0.03), showing a stronger negative association at high CO₂ (3961 ppm) than at low CO₂ (445 ppm). Conversely, although a negative trend was observed for PM_2.5 × CO₂ in the DS proportion model (p = 0.20), this did not reach statistical significance. In summary, elevated levels of CO₂ in the bedroom may exacerbate the adverse effects of particulate matter pollution on next-day physical performance. However, the moderating effect on DS proportion remains uncertain, indicating the necessity for further research to elucidate the synergistic effects of composite pollution exposure on physical performance.

Fig. 6.

Moderating effects of bedroom environmental factors on the association between particulate matter and sleep & next day physical performance.

Mediating role of sleep quality between bedroom environment and physical performance

A mediation analysis was performed, including various bedroom environmental factors as independent variables, sleep quality indicators as mediators, and physical performance outcomes as dependent variables. Detailed path coefficients, confidence intervals, and p-values for all tested relationships are provided in Appendix 3. As indicated in the appendix, no significant mediating effects were observed.

Discussion

Effects of the bedroom environment on sleep

In this study, it was found that indoor PM_2.5 was negatively correlated with the proportion of DS, with a standardized effect size of β = − 0.19 (95% CI: -0.36, -0.02; R² =0.10). As the levels of PM_2.5 pollution increased, the proportion of DS tended to decrease. This result was also found in a field study conducted by Kang et al. in Shanghai³⁸. These findings are consistent with previous epidemiological evidence on the negative effects of outdoor PM_2.5 on sleep health: Li et al. collated data on air pollution, demographics, and other relevant factors from the UK Biobank, and found that reducing exposure to PM_2.5 increased sleep duration and reduced the risk of sleep disorders⁴⁴. Furthermore, in a stratified mixed-effects model analysis that included all environmental variables and adjusted for sleep duration as well as for a variety of demographic and behavioural variables, a decrease in sleep efficiency was found with increasing levels of PM_2.5, temperature, CO₂ and noise⁴⁵. Cao et al. claimed that PM_2.5 may induce substandard sleep quality, a conclusion that was reached following a comprehensive review of ten papers from fifteen countries⁴⁶. An experimental study in mice demonstrated that PM_2.5 causes oedema, inflammation and irritation of the upper airways⁴⁷, which may result in narrowing of the airways and an increased risk of obstructive sleep apnoea (OSA), which can in turn precipitate sleep health problems. Furthermore, exposure to PM_2.5 has been demonstrated to alter serotonin levels⁴⁸, which is one of the most significant brain chemicals involved in regulating the sleep-wake cycle⁴⁹.

Consistent with previous research findings, the present study observed a positive correlation between CO₂ concentration and the proportion of LS proportion (standardized β = 0.17, 95% CI: 0.00, 0.34; R² = 0.15). Fan et al. found that an increase in ventilation rate (i.e., a decline in the concentration of CO₂) has been demonstrated to result in a substantial decrease in the proportion of LS and an increase in the duration of DS⁵⁰. Kang et al. conducted an experiment with healthy young adults, in which they observed that exposure to 1000 ppm CO₂, as opposed to 750 ppm, led to a significant reduction in sleep efficiency and an increase in wakefulness duration⁵¹. It has been hypothesised that elevated levels of CO₂ with concentrations greater than 1200 ppm result in a reduction in the respiration rate during sleep, accompanied by an increase in the apnoea hypoventilation index (AHI), which is characterised by an increased frequency of micro-arousals or transient interruptions in breathing^52,53. These respiratory events, while too small to fully awaken the individual, may disrupt deep or REM sleep and transition into shorter, more easily disturbed stages of LS, thereby increasing its percentage. The physiological mechanisms associated with the effects of CO₂ on sleep are not only related to changes in the respiratory system but may also be mediated by inducing changes in the autonomic nervous system^53,54. The percentage of LS was found to be positively correlated with LF/HF, respiratory rate, and AHI, suggesting that when the sympathetic nervous system is active, the maintenance of a LS state is more probable.

The present study also found a significant positive correlation between increased Ta and the proportion of DS in the 16–26 °C indoor temperature range, during the transition season (standardized β = 0.19, 95% CI: 0.02, 0.36; R² = 0.10). As demonstrated in Appendix 4, the subjective comfort scores of the subjects in this study were predominantly high across all temperature ranges, and the majority of thermal sensation polls fell within the neutral range. This finding suggests that the bedroom environment did not induce a significant thermal load. Meanwhile, studies have indicated that, in thermally cool to neutral environments, moderately elevated temperatures can augment the proportion of DS, reduce sleep latency, and enhance overall sleep efficiency^55–57. This phenomenon may be closely related to the body’s thermal regulation mechanism. A comfortable warm environment has been shown to promote skin vasodilation and accelerate core body temperature decline. This, in turn, has been shown to trigger the initiation of non-rapid eye movement (NREM) sleep stages and enhance DS⁵⁸.

Effect of the bedroom environment on next day physical performance

Although a substantial body of literature has documented associations between PM_2.5 and impaired endurance performance in the context of same-day or during-exercise exposure, these studies do not directly address the effects of nocturnal exposure during sleep. For example, analyses of large-scale marathons have shown that higher PM_2.5 on race day is associated with slower completion times^59,60, and multi-day lag models have reported cumulative associations between PM_2.5 and subsequent athletic performance⁶¹. These studies primarily concern immediate inhalation during physical activity or accumulated exposure across several days, rather than exposure restricted to the sleep period.

In contrast, the present study specifically examined indoor particulate exposure during the night immediately preceding the physical fitness test and its association with next-day performance (standardized β = -0.34, 95% CI: -0.57, -0.12; R² = 0.34). We observed that higher bedroom PM_2.5 concentrations were significantly associated with poorer long-distance running scores. This finding extends prior research by suggesting that nocturnal PM_2.5 exposure—occurring during a period critical for physiological recovery—may also be linked to diminished next-day endurance capacity. Mechanistically, cumulative respiratory burden (e.g., reduced pulmonary function and systemic inflammation) has been proposed to mediate longer-term impacts of PM_2.5 on endurance^22,29,62. Furthermore, the present study also found that under conditions of elevated CO₂ concentrations, the negative association between particulate pollution and aerobic capacity may be further exacerbated. This synergistic impairment may be related to their combined effects on the respiratory system, increasing respiratory load and systemic strain, which could contribute to diminished endurance performance^63–65.

RH in the bedroom environment was also found to have an inverted U-shaped relationship with next-day long-distance running performance: optimal endurance was observed at around 60% humidity, while both lower and higher relative humidity levels were associated with reduced performance. This effect may be explained by the impact of RH on respiratory function and thermoregulation. High RH can impede sweat evaporation and affect mucosal function⁶⁶, whereas low RH may increase trans-epidermal and mucosal water loss, potentially causing mild dehydration^67,68. Future laboratory-based intervention studies are required to validate these findings and explore the underlying physiological pathways in greater depth.

The selective sensitivity of long-distance running to nocturnal PM_2.5 and suboptimal RH is biologically coherent: endurance performance depends critically on sustained pulmonary and cardiovascular function and on efficient thermoregulation^69,70, whereas short-duration anaerobic or strength-based tasks (e.g., sprinting, jumping, sit-ups, pull-ups) rely more on immediate neuromuscular power and phosphagen/glycolytic energy systems that are less directly constrained by modest changes in respiratory gas exchange^71,72. Hence, exposures that principally affect airway mechanics, gas exchange, or fluid balance (e.g., PM_2.5, CO₂, RH) are expected to exert larger effects on aerobic capacity than on brief, high-intensity efforts.

Mediating effects of sleep quality

Previous research has consistently shown that the bedroom environment affects sleep quality, and that sleep quality, in turn, influences next-day physical performance. In line with this, our results indicated that higher bedroom PM_2.5 concentrations were associated with reduced deep sleep proportion and poorer long-distance running performance. Based on this theoretical framework, the present study hypothesized that sleep quality would mediate the relationship between the bedroom environment and exercise performance. However, the mediation analysis did not reveal a statistically significant indirect effect.

Several considerations help contextualize this finding. First, while sleep—particularly deep sleep—is an important component of recovery, other physiological pathways—such as respiratory load, cardiovascular strain, or systemic inflammation—may simultaneously contribute to impaired next-day performance, producing a predominant direct effect of environmental exposure that can overshadow the mediating role of sleep. Second, the magnitude of the sleep-mediated effect may be relatively small in healthy young adults with robust recovery capacity, such that it does not reach statistical significance even when measurable changes in deep sleep are present.

To further examine whether the null mediation finding could be attributable to sample size, a post hoc power analysis was conducted using Monte Carlo simulation⁷³. Assuming a medium indirect effect size of 0.39⁷⁴, the analysis indicated that the current sample size of 163 provides approximately 0.95 statistical power to detect an effect of that magnitude. These results suggest that if a true mediating effect exists, it is likely smaller than the conventional medium threshold and therefore beyond the detectable range given the current study design and measurement precision.

Limitations and practical implications

The R² values of our regression models ranged from approximately 0.10 to 0.35. Although these values may appear modest, such effect sizes are common in field-based environmental studies, where multiple unmeasured or difficult-to-control individual, behavioral, and environmental factors contribute to outcome variability⁷⁵. Moreover, the standardized regression coefficients reported here (absolute values 0.17–0.34) indicate associations of modest magnitude that are nevertheless meaningful within this context. Nevertheless, these findings should be interpreted cautiously at the level of individual models.

This study has several limitations that should be acknowledged. First, the sample in this study consisted exclusively of undergraduate students, thus caution should be exercised when generalizing the results to other populations, such as children, older adults, or elite athletes. All participants were young adults with relatively homogeneous health status and lifestyle patterns, which reduces population heterogeneity but may also introduce sample bias.

Second, although gender ratios were reported and gender was included as a covariate in all regression models, gender-stratified analyses were not conducted. Stratification would substantially reduce statistical power within each subgroup and could yield unstable or uninterpretable estimates. While gender differences were not the primary focus of the study, their potential relevance is acknowledged, and future studies with larger or more balanced samples should further explore gender-specific associations.

Third, the assessment of physical performance was restricted to five indicators from the school’s standardized fitness test, which did not include more sensitive physiological measures. This study did not collect physiological data such as blood oxygen saturation, blood pressure, or heart rate variability prior to physical performance testing, limiting insights into the physiological mechanisms through which environmental factors may affect physical performance.

Fourth, although several key covariates were included—such as gender, BMI, lifestyle habits, habitual sleep routines, subjective sleep quality (PSQI), baseline fitness (exercise frequency and self-rated fitness), and outdoor meteorological conditions—some potentially relevant factors could not be directly measured. These unmeasured factors include detailed ventilation behaviors (e.g., timing and duration of window opening), short-term indoor activities, additional physiological indicators, psychological stress, screen time, dietary intake, and habitual sleep debt. While objective indoor measurements (e.g., CO₂ and particulate concentrations) captured part of the variation related to ventilation, residual confounding cannot be fully ruled out. Participants were instructed to avoid tea, coffee, and other beverages that could acutely affect sleep on the monitored night and to maintain regular sleep–wake routines. Nevertheless, habitual variations in diet and accumulated sleep debt prior to the study were not fully controlled.

Another limitation is that this study did not investigate the sources or chemical composition of indoor PM_2.5. As a result, we cannot determine the extent to which the observed associations reflect outdoor infiltration, indoor generation, or specific particle constituents that may have differential biological effects.

Finally, the research question focused on how the bedroom environment during single night influences sleep that night and physical performance the following day; therefore, single-night measurements directly addressed this question. However, a one-night assessment may not fully capture the cumulative effects of long-term exposure.

Future research could address these limitations by: (1) considering random sampling or stratified recruitment strategies to further reduce the possibility of bias; (2) incorporating multi-night or long-term monitoring to examine lagged and cumulative effects;3) broadening the scope of physiological indicators (e.g., inflammatory markers, measures of cardiorespiratory fitness) to explore underlying mechanisms in greater depth; and 4) conducting particle chemical speciation/source apportionment to clarify the origins and constituent-specific health effects of indoor PM_2.5.

Conclusion

This study conducted an exploratory, observational field study in university dormitories in which bedroom environmental conditions (PM_2.5, air temperature, relative humidity, and CO₂) were continuously monitored while objective sleep measures and standardized next-day physical fitness tests were obtained. In this sample of healthy undergraduate students, higher nocturnal PM_2.5 concentrations were associated with a lower proportion of deep sleep and with poorer next-day endurance performance, as reflected by long-distance running results; these negative associations were evident even at relatively low exposure levels (Q1 = 6 µg m⁻³, Q2 = 25.5 µg m⁻³). Elevated indoor CO₂ concentrations appeared to amplify the adverse association between PM_2.5 and endurance performance, suggesting a possible synergistic effect of combined exposure to particulate matter and poor ventilation.

These findings highlight that bedroom environment (especially PM_2.5) may play a meaningful role in supporting both sleep quality and next-day endurance performance in young adults. Although causality cannot be established in this observational study, the observed associations suggest that strategies to reduce indoor particulate matter and maintain adequate ventilation—such as air filtration or improved airflow—could potentially benefit sleep and physical performance in dormitory settings. Future research should test these interventions in controlled or larger-scale studies to confirm effectiveness and clarify underlying mechanisms.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

X.L. conceived and designed the study, performed key experiments, analysed the data, and drafted the manuscript. T.J. contributed to study design, coordinated data collection, and assisted with statistical analyses and interpretation. R.G. contributed to data collection, preprocessing, and statistical analyses. C.G. developed and implemented the computational and analytical methods, and assisted with interpretation of the results. P.G. contributed to experimental design, and quality control of the data. L.L. supervised the overall project, conceived and designed the study, provided critical feedback, secured funding, and revised the manuscript for important intellectual content. All authors discussed the results and implications, and reviewed and approved the final manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (No. 52178081) and the National Key R&D Program of China (2022YFC3803202).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on request.

Declarations

Competing interests

The authors declare no competing interests.