A comparison of respiratory particle emission rates at rest and while speaking or exercising.

Background: The coronavirus disease-19 (COVID-19) pandemic led to the prohibition of group-based exercise and the cancellation of sporting events. Evaluation of respiratory aerosol emissions is necessary to quantify exercise-related transmission risk and inform mitigation strategies.
Methods: Aerosol mass emission rates are calculated from concurrent aerosol and ventilation data, enabling absolute comparison. An aerodynamic particle sizer (0.54-20 μm diameter) samples exhalate from within a cardiopulmonary exercise testing mask, at rest, while speaking and during cycle ergometer-based exercise. Exercise challenge testing is performed to replicate typical gym-based exercise and very vigorous exercise, as determined by a preceding maximally exhaustive exercise test.
Results: We present data from 25 healthy participants (13 males, 12 females; 36.4 years). The size of aerosol particles generated at rest and during exercise is similar (unimodal ~0.57-0.71 µm), whereas vocalization also generated aerosol particles of larger size (i.e. was bimodal ~0.69 and ~1.74 µm). The aerosol mass emission rate during speaking (0.092 ng s-1; minute ventilation (VE) 15.1 L min-1) and vigorous exercise (0.207 ng s-1, p = 0.726; VE 62.6 L min-1) is similar, but lower than during very vigorous exercise (0.682 ng s-1, p < 0.001; VE 113.6 L min-1). Conclusions: Vocalisation drives greater aerosol mass emission rates, compared to breathing at rest. Aerosol mass emission rates in exercise rise with intensity. Aerosol mass emission rates during vigorous exercise are no different from speaking at a conversational level. Mitigation strategies for airborne pathogens for non-exercise-based social interactions incorporating vocalisation, may be suitable for the majority of exercise settings. However, the use of facemasks when exercising may be less effective, given the smaller size of particles produced.

Introduction

The global coronavirus disease-19 (COVID-19) pandemic, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has led to one of the most significant public health emergencies of the last century, placing unprecedented challenges on healthcare systems worldwide[1,2]. Public health measures aimed at reducing rates of infection have focussed on social distancing and the use of face coverings. In addition, the cancellation and prohibition of social and cultural events have resulted in marked societal disruption, impacting a plethora of sporting occasions, including the 2020 Olympic Games[3-8]. Concern regarding SARS-CoV-2 transmission during exercise also resulted in the temporary closure of many indoor exercise facilities and curtailed access to sporting and physical group activities[9-14].

Despite this, it is well recognised that exercise is essential to health and wellbeing. Accordingly, the World Health Organisation recommends >150 min of moderate-intensity or 75 min of vigorous intensity physical activity per week[15]. Regular physical activity is associated with improved quality of life, reduced cardiovascular risk and mental health benefits[16-18]. In addition to individual benefits, the sport and physical activity leisure industry is a substantial employer, contributing an estimated $756 billion to the global economy[19]. It is also recognised that regular exercise and improved physical activity status may mitigate the risk of severe COVID-19[20].

Respiratory disease transmission is governed by host, recipient, pathogen and environmental factors[21,22]. SARS-CoV-2 transmission occurs predominantly through respired particles containing the virus[23,24]. Such particulate matter is expelled during exhalation, with an arbitrary distinction made between aerosols and droplets, as particles smaller and larger than 5 µm diameter, respectively[25,26]. Recently, it has been recognised that a more correct distinction should be made separating particles below and above 100 µm, with all particles smaller than 100 µm inhalable and exhibiting similar aerodynamic behaviour with respect to suspension in air[27]. We previously demonstrated that the number and mass concentrations of aerosols released during breathing and vocalising are directly related to the intensity and type of activity performed and that vocalising produces larger particles than breathing, consistent with studies by other researchers[26,28-30]. The intensity-dependence of aerosol generation is likely relevant during sporting activity, given the hyperpnoea associated with vigorous physical activity[31]. However, despite their importance for transmission modelling and evidence-based mitigation interventions, absolute measurements of respiratory aerosol generation during exercise have only been studied in a limited capacity[32-34]. Performing such measurements in an environment with zero aerosol background concentration levels is essential to ensure that every detected exhaled particle arises from respiratory activity, rather than from inhaled particle-laden air[33].

Vaccination, social distancing and other factors have reduced SARS-CoV-2 transmission in certain contexts. However, an improved understanding of airborne viral transmission and mitigation strategies remains relevant to all respiratory pathogens, including SARS-CoV-2. This study investigates the generation of respiratory aerosol during exercise, comparing these emissions to those produced by breathing at rest and by speaking at a conversational level, both activities upon which current transmission mitigation guidance are based. We hypothesised that exercise would generate more aerosol than at rest or during the conversational speech, due to the increased ventilation rates and effort associated with exercise.

We demonstrate that the size of aerosol particles generated at rest and during exercise is similar, whereas vocalisation also generates aerosol particles of a larger size. The aerosol mass emission rate during speaking and vigorous exercise is similar, but lower than values measured during very vigorous exercise. Mitigation strategies for airborne pathogens, deemed appropriate for non-exercise-based social interactions incorporating vocalisation, may be suitable for the majority of exercise settings.

Methods

Study design and participants

“The Investigation of ParticulatE Respiratory Matter Release During perFormance and Exercise to infOrm Guidance in the SARS-CoV-2 PandeMic” (PERFORM-2) is an observational study investigating respiratory aerosol generation during a range of activities impacted by the COVID-19 pandemic, including singing and playing musical instruments[28,29]. In this study, 25 healthy individuals, who were free from significant cardiovascular or respiratory illness, were recruited to include a range of athletic abilities across both sexes.

Participants attended on a single occasion for testing. The participants exhibited no COVID-19 symptoms, were lateral flow negative, and refrained from vigorous exercise, smoking, consuming alcohol, or eating heavily for 4 h prior to testing[35]. Written informed consent was provided by all participants and study approval was granted by the Public Health England Research Ethics and Governance of Public Health Practice Group (PHE REGG, NR0221). Written informed consent for publication of the images depicting the experimental setup was obtained. The sample size of 25 participants was chosen to allow detection of differences in the magnitude of 2.5 with a power of 0.9, using sampling variances from a previous study (PERFORM-1) in calculations and allowing for a potential dropout rate of 10%[28].

Study protocol

The study protocol is illustrated conceptually in Fig. 1b. Following familiarisation with the experimental setup, participants underwent a maximal cardiopulmonary exercise testing (CPET) on a cycle-ergometer [CORTEX MetaLyzer 3B-R3 + Wattbike Atom (Next Generation) cycle-ergometer or VyaireTM Medical Vyntus CPX + VIAsprint 200 P W/BP Serial Ergometer system] to voluntary exhaustion, per CPET international guidelines, to characterise exercise capacity and ventilatory response[35].

Fig. 1

Experimental setup and study protocol.

a Experimental setup: cycle-ergometer-based exercise testing with concurrent ventilatory and aerosol measurements via APS-mask, from sampling line inserted into sampling hole at the tip of the nose. Informed consent for publication of the image was obtained. b Experimental design: maximal CPET followed by a two-stepped exercise test, post 1-h interval. c Typical time series of aerosol number concentration data sampled periodically via APS-mask during the two-stepped exercise test. Horizontal scale adjusted to align with corresponding activities in b. APS aerodynamic particle sizer, CPET Cardiopulmonary exercise testing.

Following at least 1 h of rest, participants then completed a second exercise test where both ventilatory and aerosol measurements (see detail below) were concurrently evaluated using an adapted facemask (see Fig. 1a). First, ventilation and aerosol generation at rest was measured over 1 min. Then, ventilatory and aerosol measurements were made whilst participants vocalised a set text at a constant pace of at least 70 dBA (A-weighted decibels). The sound level was measured by a sound level metre placed 30 cm from the mouth and 70 dBA was selected as a target for participants, to enable comparison with previous measurements of conversational speaking in the 70–80 dBA range, while also accounting for attenuation attributable to the CPET mask (see Supplementary Figures 5 and 6)[28,36,37]. Next, participants were instructed to begin exercising, with two fixed periods of constant work, prescribed from the CPET and selected to replicate work intensities of vigorous intensity (80% of the anaerobic workload, for ~6 mins) and very vigorous (anaerobic workload +30% (maximal workload minus anaerobic workload), for ~4 mins) gym-based exercise (see Fig. 1b). Aerosol measurements were taken following 2 min of vigorous exercise and following 30 sec of very vigorous exercise[38]. Perceived exertion was assessed at rest, during vigorous exercise and during very vigorous exercise using the BORG CR-10 Scale[39]. CPET data were analysed using Cortex MetaSoft® Studio Version 5.12.0 (Cortex system) and SentrySuite® software V. 320 (VyaireTM Medical system).

Aerosol measurements (0.54–20 μm diameter)

Aerosol measurements were accomplished using an aerodynamic particle sizer (APS; TSI Inc. model 3321; 1 L min−1 sample flow rate, 4 L min−1 sheath flow rate, 1-second sampling interval), which measures the number concentration and size distribution of aerosol particles in the 0.54–20 µm diameter size range. The APS size range overlaps with that associated with the vast majority of respiratory aerosol by number, allowing definitive characterisation of the size distributions typically ascribed as the bronchiolar and laryngeal modes[26]. This study was conducted in a laminar flow operating theatre with sufficient air changes per hour within the ultraclean ventilation canopy to ensure the background aerosol number concentration within the APS size range was 0 particles cm−3. Consequently, aerosol detected by the APS can be confidently attributed to the participant, with background aerosol concentrations returning to 0 cm−3 during sampling pauses. Room temperature was controlled at 18 °C, with relative humidity (RH) 40%.

Aerosol measurements were taken in two sampling configurations. In the main configuration (see Fig. 1a), aerosol was sampled from a modified CPET facemask (referred to as APS-mask). The silicone mask (Hans Rudolph 7450 Series V2) was adapted with a 6 mm sampling hole cut at the tip of the nose to facilitate the passage of the aerosol sampling line (while sampling aerosol in this configuration) or with a tight-fitting silicone bung (while performing ventilatory measurements alone). The sampling port site was selected to avoid the collection of water droplets pooling in the facemask. In the second sampling configuration, aerosol was sampled without a CPET mask present using a 3D-printed funnel positioned 15–20 cm from the participant’s mouth (referred to as APS-cone; the sampling funnel is visible in Supplementary Figure 1). As demonstrated by the example time series in Fig. 1c, all APS-mask and APS-cone measurements were made during two, 30-second periods unless the participant struggled to maintain the activity, in which case the measurement period was shorter. The data presented are based on the mean values for each 30-second sampling period. Results reported in the main body of this manuscript were all taken in the APS-mask configuration, as this configuration was the more robust aerosol sampling approach. Supplementary Methods 1 provides detailed validation of this aerosol sampling methodology.

Aerosol number concentrations and size distributions were extracted directly from the time-averaged APS data. APS data were analysed using the Aerosol Instrument Manager v10.3 (TSI inc) software. Aerosol mass concentrations were calculated based on the mean diameter of each size bin. A particle density of 1 g cm−3 is assumed, as aerosol is generated in the respiratory tract at very high RH (>99%)[40,41]. Additionally, our recent work based on analysis of the sampling of aerosols through the collection funnel and into the APS shows that the full-size distributions (0.54–20 µm) reported here can fully equilibrate in size to the sampling RH, with sufficient time from exhalation to size measurement by the APS instrument[42]. Although our previous study suggests the RH for the measured size distributions remains high, we cannot unambiguously state the RH at which our size distributions are measured and this will be the subject of a future study. A comparison in terms of aerosol mass concentration assumes the potential dose transmitted by an infected individual scales with particle volume.

A key aspect of this study is the concurrent measurement of aerosol concentration and ventilation, with the APS sampling directly from the CPET facemask. The aerosol measurements allow quantification of the aerosol concentration in the expiratory jet. The ventilation measurements permit quantification of the total flow rate of the expiratory jet (see Table 1). When combined, the separate aerosol and ventilation measurements enable estimation of the absolute number of particles and amount of aerosol emitted during each activity, per unit time, which are the absolute aerosol number and mass emission rates, respectively. Estimates of emission rates are important to achieve as they enable absolute comparisons across activities with differing rates of ventilation of the number and mass of respiratory particles emitted by a participant per second.

Table 1

Demographic and exercise physiology data.

		Variables	Mean ± SD
Demographics		Sex	13 males 12 females
		Age/years	36.4 ± 14.9
		BMI	23.8 ± 4.1
Maximal CPET		Maximal work rate/W	304 ± 78
		HR %max/%	96% ± 0.04
		VE/L min−1	120.13 ± 45.83
		BF/bpm	48 ± 12
		VT/L	2.48 ± 0.68
		VO2 max% predicted/%	130 ± 26
		VO2/kg/ml kg−1 min−1	42.4 ± 11.01
		RER	1.24 ± 0.1
Two-stepped exercise test	Rest	HR %max	43 ± 0.05
		VE/L min−1	11.44 ± 3.93
		BF/bpm	15 ± 4
		VT/L	0.85 ± 0.4
		VO2 max% predicted/%	16.3 ± 5.1
		RER	0.84 ± 0.07
		BORG CR-10 scale	“Nothing at all” 0.06 ± 0.22
	Speaking	HR %max	45 ± 0.06
		VE/L min−1	15.1 ± 5.58
		BF/bpm	23 ± 5
		VT/L	0.93 ± 0.32
	Vigorous exercise	Work rate/W	152 ± 52
		HR %max/%	79 ± 0.06
		VE/L min−1	62.62 ± 17.94
		BF/bpm	29 ± 5
		VT/L	2.22 ± 0.64
		VO2 max% predicted/%	97.9 ± 22.8
		RER	0.94 ± 0.05
		BORG CR-10 Scale	“Somewhat strong” 4.16 ± 1.37
	Very vigorous exercise	Work rate/W	226 ± 66
		HR %max	92 ± 0.05
		VE/L min−1	113.61 ± 38.73
		BF/bpm	47 ± 12
		VT/L	2.46 ± 0.62
		VO2 max% predicted/%	126.9 ± 24.1
		RER	1.06 ± 0.04
		BORG CR-10 scale	“Very strong” 8.46 ± 1.78

n = 25 participants. Maximal CPET values are those achieved maximally. Body mass index (BMI), oxygen uptake (VO2 kg−1), heart rate (HR), minute ventilation (VE), breathing frequency (BF), tidal volume (VT), respiratory exchange ratio (RER). VO2 max% predicted/% calculated for rest, vigorous, and very vigorous exercise relative to maximal CPET[39].

Aerosol emission rates were estimated based on the synchronous measurements of both aerosol concentrations and VE for the four activities investigated. Due to the concurrent nature of these measurements, these values were determined independently for both repeats of each of the four activities, prior to any statistical analysis.

The aerosol number emission rates are calculated by Eq. (1):

The mass emission rates are calculated by Eq. (2):

Statistical analysis

The aerosol data are clustered by activity: for each participant there are eight sets of measurements representing the four different activities (rest, speaking at 70–80 dBA, vigorous exercise and very vigorous exercise) and two replications of each per participant. For each response variable considered (number concentration, mass concentration, number emission rate and mass emission rate) a positively skewed histogram was observed, with log transformation producing an approximately symmetric distribution that can be represented by a normal distribution. Note that aerosol generation and aerosol size distributions are lognormally distributed across participants for all activities, consistent with the previous studies[25,28,29]. Given the clustering, a two-level random effects model was used to fit each (logged) response variable (with measures nested within participants) using MLwiN v3.05[43]; fixed effects were included for activity type, to be able to compare activity types while adjusting for participant differences. Initially, activity type was assessed using a likelihood ratio test and then each pair of activities was compared using (two-sided) Wald tests to identify which activities significantly differ. Missing data were assumed to have occurred at random.

Table 2

Summary of aerosol concentration, emission rate and size distribution data.

Activity	Aerosol number concentration/cm−3 (IQR)	Aerosol Mass Concentration/µg m−3 (IQR)	Aerosol number emission rate/s−1 (IQR)	Aerosol mass emission rate/ng s−1 (IQR)	Aerosol modal diameter/Dp/µm (SE)
Rest	0.03 (0.01−0.07)	0.02 (0.01−0.07)	6.5 (2−14)	0.003 (0.001−0.01)	0.57 (±0.04)
Speaking	0.26 (0.21−0.31)	0.40 (0.25−1.22)	58.5 (43−98)	0.092 (0.06−0.23)	0.69 (±0.01) and 1.74 (±0.10)
Vigorous exercise	0.12 (0.06−0.25)	0.17 (0.06−0.25)	145 (46−285)	0.207 (0.05−0.36)	0.59 (±0.06)
Very vigorous exercise	0.24 (0.16−0.50)	0.42 (0.24−0.66)	625 (230−1003)	0.682 (0.31−1.28)	0.71 (±0.02)

Median aerosol number and mass concentrations and emission rates and size distribution modes were obtained from breathing at rest, speaking, vigorous exercise and very vigorous exercise across n = 25 participants.