Aims: (1) To obtain an overall estimate of variability of personal exposure of Paris office workers to fine particles (PM2.5) and nitrogen dioxide (NO2), and to quantify their microenvironmental determinants. (2) To examine the role of potential determinants of indoor concentrations.
Methods: Sixty two office workers in a Paris municipal administration (all non-smokers) were equipped with personal samplers: passive samplers for 48 hours for NO2 (n = 62), and active pumps for 24 hours for PM2.5 (n = 55). Simultaneous measurements were performed in homes and offices; the local air monitoring network provided ambient concentrations. A time activity diary was used to weight measured concentrations by time spent in each microenvironment in order to estimate exposure concentrations.
Results: On average, PM2.5 personal exposure (30.4 μg/m3) was higher than corresponding in-home (24.7 μg/m3) and ambient concentrations (16.7 μg/m3). Personal exposure to NO2 (43.6 μg/m3) was significantly higher than in-home concentrations (35.1 μg/m3) but lower than the background outdoor level (60.1 μg/m3). Personal exposures to PM2.5 and NO2 were not significantly different from in-office concentrations. PM2.5 and NO2 personal exposures were not significantly correlated. In-home, in-office, in-transit, outdoor time weighted concentrations, and time spent in other indoor microenvironments explain respectively 86% and 78% of personal variations in PM2.5 and NO2. In-home PM2.5 concentration was primarily influenced by exposure to environmental tobacco smoke, and secondly by the ambient level (R2 = 0.20). NO2 in-home concentration was affected mostly by the ambient level and gas cooking time (R2 = 0.14).
Conclusion: While results show the major contribution of in-home and in-office concentrations to both NO2 and PM2.5 personal exposures, the identification of indoor level determinants was not very conclusive.
- personal exposure
- nitrogen dioxide
- indoor concentration
- ETS, environmental tobacco smoke
- PM, particulate matter
- TWC, time weighted concentration
Statistics from Altmetric.com
In epidemiological studies, accurate estimation of exposure is important for evaluation of health risks. Assessment of exposure to air pollution is often based on fixed site measurements provided by the local air quality monitoring network. Because of economic and practical reasons, personal exposure measurements are rare in epidemiological studies.
In the 1980s personal exposure studies were carried out in the United States.1–3 These studies first focused on gaseous pollutants such as nitrogen dioxide (NO2), using passive samplers. Personal exposure to particulate matter (PM),4–6 which requires a noisier and more bulky active sampler, was determined later. Studies were subsequently conducted worldwide,7–31 but only a small number have been done in France.32–34
Personal exposure studies aim to assess the distribution of individual exposure, and to identify and quantify the contribution of different factors such as environmental tobacco smoke (ETS), cooking activities, outdoor concentrations, and traffic. In this context, outdoor and indoor levels are often determined at the same time as personal measurements. Because of practical difficulties, microenvironmental measurements generally take place at only one site: the home, workplace, or school. Moreover, few personal exposure studies have been carried out on randomised populations, especially for PM, which requires an important contribution from participants.
The objectives of this study were: (1) to obtain an overall estimate of variability of personal exposure to fine particles (PM2.5) and nitrogen dioxide (NO2) in a specific population of office workers living in the Paris metropolitan area; (2) to evaluate the relative contribution of different microenvironments to personal exposure by performing measurements in both the home and workplace; and (3) to examine the role of potential determinants of indoor concentrations.
MATERIALS AND METHODS
Personal and indoor NO2 and PM2.5 measurements were conducted simultaneously with the same devices. Indoor measurements were carried out in the home and workplace, the two microenvironments most frequented by city dwellers. Measurements took place from December 1999 to September 2000 (at the rate of two persons per week), and were supplemented by three questionnaires related to microenvironmental characteristics and time activity patterns.
Each participant was equipped with personal samplers for a period of 48 hours for NO2 and 24 hours for fine particles. Outdoor concentrations during the measurement period were provided by the Paris air quality monitoring network.
The personal exposure study was conducted on Paris office workers recruited in a Paris municipality service in charge of social action, childhood, and health (DASES). A total of 200 subjects were selected by randomisation from the 2100 DASES office workers living and working in Paris or in one of the three peripheral departments covered by the regional air quality monitoring network. After excluding 60 smokers (smoking status interfering with personal measurements), 140 workers were invited to participate. The order of participants was also determined by randomisation.
A sample size of at least 50 persons was previously determined, in order to be able to show a correlation between personal and outdoor measurements of more than 0.40, with α = 0.05 and a power of 0.80. Because of the expected response rate (40%) and the necessity to exclude smokers whose proportion was estimated at one third, 200 subjects were sampled.
Sampling methods and analytical procedures
NO2 measurements were taken from Monday to Wednesday, or from Wednesday to Friday, for 2884 ± 62 minutes (minimum 2595, maximum 3045). During personal measurements, the NO2 badge (Ogawa & Co, Pompano Beach, Florida, USA) was attached to clothing near the breathing zone, while at night it was put on the bedside table. The passive sampler for residential measurement was placed 1.5 metres high in the home living room, and in the office. NO2 was captured on a coated filter and measured by spectrophotometry (λ = 545 nm).
Personal PM2.5 measurements were taken on the first day of the NO2 measurements periods, for 1425 ± 80 minutes (minimum 1110, maximum 1680). For in-office PM2.5 measurements, a timer was used, whereas in-home measurements were only taken when the subject was present at home; he had to turn the pump on when he arrived home and off when he left. The sampling times were respectively 617 ± 41 minutes (600 and 720 in a few cases) in the office, and 862 ± 195 minutes (minimum 510, maximum 924) in the home. PM2.5 measurements were performed using a pump sampling air at a flow rate of 4 l/min (Gillian Instruments, model Gil-Air 5, Sensidyne, Clearwater, Florida, USA). Particles were selected by a GK2.05 cyclone KTL (BGI incorporated, Waltham, Massachusetts, USA) and collected on a filter (EMFAB TX40HI20-WW, Pallflex Putnam, Connecticut, USA). Flow rates were calibrated at the beginning and measured at the end of each personal and indoor measurement. The level of pump noise was reduced by placing the pump in a shell equipped with cork and rubber. During the day the shell was carried in a rucksack, facilitating movement during transport. During the night the personal sampler was located in the living room. Indoor PM2.5 samplers were placed on a table or a desk in the home and in the office.
Particles were analysed by gravimetric method. A microbalance M5P (Sartorius Laboratoire, Palaiseau, France) with 1 μg reading precision, was used in a room where temperature (t) and relative humidity (RH) were controlled (t = 20±1°C, RH = 50±5%). Before and after sampling, all filters were weighed twice; when the difference between two consecutive weighings exceeded 4 μg, a third weighing was performed.
Twenty four hour outdoor PM2.5 concentration was obtained from the only fixed urban background station equipped with a continuous PM2.5 analyser (TEOM, R&P, New York, USA). Forty eight hour NO2 concentration was the arithmetic average provided by the two (of 15) urban background stations respectively closest to the home and workplace and equipped with chemiluminescence analyser. Concentrations in different means of transport (car, subway, bus, bike) were estimated using data from a fixed monitoring traffic station located near the Paris ringroad.
Evaluation of exposure conditions
Three questionnaires were used to describe for each subject: (1) general characteristics of the residential and occupational environment (topographic situation, local traffic, heating and cooking system); (2) time activity diary with 15 minutes resolution; and (3) unusual exposure during measurements (exposure to ETS, cooking activities, etc).
Results from PM2.5 and NO2 samplers were compared with respectively a TEOM analyser equipped with PM2.5 cyclone and a chemiluminescence analyser at an outdoor site. The correlations were highly significant (Pearson correlation coefficient being respectively: r = 0.96 (n = 17) for PM2.5, r = 0.97 for NO2 (n = 49)). On average, the two series of measurements related to each pollutant did not differ significantly (13.9 v 12.8 μg/m3 PM2.5, 54.8 v 56.7 μg/m3 NO2).
NO2 Ogawa samplers enabled all the measurements to be duplicated by placing one filter at each tip of the sampler. A good relation was observed between NO2 duplicated measurements (n = 185; 41.2 v 41.3 μg/m3, p > 0.05; r = 0.92), and the arithmetic mean of the duplicates was considered for each measurement. With regard to PM2.5, it was not possible to ask participants to carry two pumps simultaneously. Thus only some indoor PM2.5 measurements could be duplicated, showing a good repeatability, the deviations between duplications being less than 10% (n = 10; 24.7 v 23.8 μg/m3, p > 0.05; r = 0.87). Moreover, in the Expolis study the authors duplicated several personal 48 hour PM2.5 measurements using the same devices and obtained results of the same order: absolute average differences between duplicate results was 2.1 (SD 2.0) μg/m3.35
Statistical analysis was performed with BMDP software (University of California). Results are expressed in terms of concentration (C) and time weighted concentration (TWC). For each microenvironment, TWC is the product of the concentration, measured or estimated, by time fraction spent in each microenvironment. The distributions of concentration, TWC, and their log transformed values, were tested for normality (Shapiro–Wilk test). Results are expressed as mean (SD), median, minimum, and maximum. Correlations between variables were estimated by Pearson or Spearman coefficients. Linear multiple regression was used to identify and quantify determinants of personal exposure and indoor concentrations. For each pollutant, two models of personal exposure were tested, the independent variables being concentration in the first, TWC in the second.
Population, living environment, and time activity patterns
The study sample consisted of 62 subjects for NO2, and 55 subjects for fine particles, seven women having refused to carry the particle pump because of its noise and the weight of the rucksack.
All volunteers were non-smokers; there were 53 women, nine men. Mean age was 45.2 (10.0) years (minimum 23, maximum 61).
A small majority of subjects lived in Paris (56.4%), with 95.2% of subjects dwelling in an apartment with a mean surface area of 67.6 (28.1) square metres, the remainder living in a house. Exposure to tobacco smoke was not very frequent (11.3%) in the home. Around 60% of volunteers had a room directly exposed to traffic, moderate to heavy in 23 cases. A total of 22.6% of participants owned an individual gas heating system and 56.5% used a gas or mixed cooker.
All workplaces were located in Paris. Participants worked in an office with a surface area of 22.1 (10.0) square metres on average; 62.9% worked in an office that was either not or little exposed to traffic.
During the 24 hour PM2.5 personal measurements, volunteers spent on average 21.7 hours indoors (home: 13.4 hours; office: 6.7 hours; other: 1.6 hours), 1.0 hour outdoors, and 1.3 hours in transport. The most frequently used means of transport were subway (39%) and walking (35%), followed by car (12%), bus (11%), and motorbike or bike (3%). One subject out of two was exposed to tobacco smoke (more than one cigarette a day). When subjects were equipped with the rucksack they spent less time in transport (−0.5 hour) than when they just wore the NO2 passive sampler on the second day, but the difference was not significant (p > 0.05).
Distribution of personal exposure and microenvironmental concentrations
Most of the personal exposure, indoor and outdoor concentrations, and their TWC were log normally distributed. For both pollutants, as no difference in personal exposure was observed between people living in Paris and subjects living in suburbs, or according to season, all subjects were analysed together for the entire measurement period (December 1999 to September 2000).
Table 1 presents a summary of the results. Personal exposure to fine particles was significantly higher (p < 0.0001) than the in-home concentration, which was itself higher (p < 0.0001) than the ambient level. In offices, large variations in concentration were observed, with two very high measurements (265.1 and 162.7 μg/m3 PM2.5), as a result of intensive smoking during sampling. On average, personal exposure to NO2 was not significantly different from occupational concentration, but was significantly higher than in-home concentration (p < 0.001) and lower than background outdoor concentration (p < 0.0001).
Cumulative exposures to PM2.5 and NO2, assessed by personal measurements (respectively 30.4 and 43.6 μg/m3) were greater than the indirect estimates from time activity data and microenvironmental concentrations, the sum of the TWC being respectively 26.3 for PM2.5 and 38.8 μg/m3 for NO2. It appears that home TWC (TWCH; 13.7 μg/m3 for PM2.5 and 19.7 μg/m3 for NO2) and working TWC (TWCW; respectively 10.1 and 12.3 μg/m3) play a large part in cumulative personal exposure, whereas influence of transport (TWCT; 2 and 6.3 μg/m3) and outdoor TWC (TWCO; 0.5 and 2.3 μg/m3 respectively) is minor.
Despite a good correlation between NO2 and PM2.5 urban background levels (n = 55; r = 0.69; p < 0.001), no relation was observed between personal exposure to these two pollutants (n = 53; r = 0.12; p = 0.38). NO2 and PM2.5 concentrations were correlated neither in-home nor in-office (respectively n = 54; r = 0.06; p = 0.69, and n = 55; r = 0.05; p = 0.74).
Determinants of personal exposure
As the residuals of the model using non-transformed data were normally distributed, only non-transformed data were used.
The stepwise regression model showed that in-home, in-office, and outdoor concentrations explained 80% of variations in personal exposure to PM2.5 (table 2). Using TWC, five variables—TWCH, TWCW, TWCT, TWCO, and time spent indoors other than in-home or office—contributed significantly to this personal exposure level (table 2). They explained 86% of personal exposure variations with a major contribution from TWCH and TWCW (69%), followed by time spent in other indoor microenvironments, and finally by in-transit TWCT and outdoor TWCO.
In-home, in-office NO2 concentrations, and time spent in transport were the three variables selected in the stepwise multiple linear regression, and accounted for 75% of NO2 personal exposure variations. Using TWC, the model including TWCH, TWCW, TWCT, TWCO, and time in other indoor microenvironments, did not really increase the coefficient of determination (R2 = 0.78). However, all these variables were significantly associated with NO2 personal exposure (table 3).
Determinants of indoor concentrations
Considering all the participants, including those exposed to ETS, in-home concentrations were influenced primarily by the number of cigarettes smoked in the home, and secondly by ambient concentration (R2 = 0.20). In office buildings, the impact of tobacco smoke was higher (R2 = 0.95) than in the home because of a larger number of sites with smokers (n = 13) and a higher smoking intensity, whereas the contribution from the outdoor level alone was very low (2%). We did not show the impact of any other factors.
In homes without ETS (n = 48), outdoor PM2.5 level, person density, and local traffic were significantly associated with in-home PM2.5 concentrations and explained 39% of their variation (table 4), mainly owing to outdoor concentrations. Cooking time and cleaning activity had no significant contribution. In offices not exposed to tobacco smoke (n = 42), the regression model showed a significant influence of outdoor level and person density. These factors explained 20% of in-office PM2.5 concentration variations.
A small part (R2 = 0.14) of in-home NO2 concentration variations could be explained by NO2 ambient level and the time spent cooking with gas (table 5). The other factors such as ETS, local traffic, and floor height, did not have a significant influence. Twenty four per cent of variations in in-office NO2 concentrations could be explained by two factors: outdoor level and floor height. There was an inverse relation between this last variable and in-office concentration (table 5).
This study was carried out on a randomised sample of Paris municipal administration office workers; 85.3% were women, a percentage which does not differ from the DASES population composed of more than 80% women. The sex ratio in this administration is consistent with French statistics that indicate a high proportion of women in health social services (77.2%). The study sample can thus be regarded as representative of this field of activity.36
Our subjects spent a long time (approximately 90%) in indoor microenvironments. Time spent outdoors was low. A similar finding was observed by other teams,6–8 especially Sexton and colleagues,4 who reported that subjects from Waterbury stay indoors for a slightly higher percentage of the time (more than 94%).
Participation rate was consistent with those reported by other authors in such studies. Measurements were performed in satisfactory conditions for NO2, and even for fine particles in spite of constraints related to the sampling device. For this reason measurements of personal exposure to PM2.5 were limited to 24 hours.
A slight behavioural modification, also described by other authors,9 was noted during our PM personal measurements. This modification could distort the personal exposure estimate, but has no effect on the study of exposure determinants. Nevertheless, technological progress to reduce the noise and bulk of the PM sampler could facilitate studies and decrease measurement bias.
Simultaneously with personal measurements, indoor measurements of fine particles were only recorded when subjects stayed in dwellings or premises used for professional purposes. This strategy accurately measured the indoor PM2.5 concentrations to which subjects were exposed.
Conversely, passive samplers for NO2 measurements were used continuously for 48 hours. Thus indoor concentrations reflect pollution throughout the entirety of the measurement period, even when participants were not present in their office or residence. Participants were not asked to block the passive sampler when they left the premises, in order to ensure good analytical sensitivity and to avoid mistakes and omissions.
Ambient levels were estimated from data provided by background stations of the Paris air quality monitoring network. Outdoor NO2 concentration could be estimated accurately for each residence and workplace as a result of the high density of NO2 stations. For PM2.5, only one urban background station was equipped with a PM2.5 analyser. This does not matter, however, as fine particle ambient distribution is spatially homogeneous in urban areas, as shown in Basel37 and American cities.38
Our PM2.5 results are consistent with levels reported in the literature, as shown by personal measurements: 30.4 for Paris office workers versus 24.3 μg/m3 and 28.3 μg/m3 respectively for adults and children in Amsterdam,10,11, 21.6 μg/m3 in Boston,12 24 μg/m3 in Zurich,8 and 21.9 and 36.7 μg/m3 in Grenoble,32 respectively in summer and winter. Indoor levels measured in Paris were 24.7 μg/m3 in homes and 34.5 μg/m3 in offices, whereas other teams have reported levels varying from around 17 μg/m3 in Amsterdam classrooms and Boston homes to 23 μg/m3 in Zurich homes, and 35.1 μg/m3 in Grenoble homes during the winter period.
The level of personal exposure to PM2.5 was greater than the in-home concentration, which in turn was higher than the ambient level during the same period. In previous studies this relation has already been described in healthy subjects,4,5,9,13,14 whatever the considered particles (PM10, PM2.5, etc), even when ambient level is high.14 Rojas-Bracho and colleagues12 found mean personal PM2.5 concentrations of individuals with chronic obstructive pulmonary disease to be lower than corresponding outdoor concentrations; they attributed these lower exposures to the low activity level of these patients. Sarnat and colleagues15 also observed reduced personal exposure and indoor levels in comparison to outdoor concentrations in non-smoking older subjects, possibly because of limited exposure to indoor PM sources. Using TWC, our cumulative directly measured personal exposure is greater than indirect estimation. This difference, called “personal cloud” by Wallace,38 has yet to be explained. Particulate matter resuspending, and exposure in buildings other than the home and workplace, such as canteens, restaurants, and shopping centres (not measured here), could contribute to this personal cloud. It must also be noted that exposure related to transport has not been measured, but only estimated from air quality data provided by a traffic site. Janssen and colleagues9 have found that one hour spent in a vehicle increases personal PM10 concentration by 5.4 μg/m3.
Most NO2 studies16–20,33 show the same relation: personal exposure is situated between indoor levels and outdoor concentrations. This relation is found in our population: on average, personal exposure (43.6 μg/m3) is higher than in-home concentration (35.1 μg/m3) but lower than ambient level during the same period (60.1 μg/m3).
Compared to other personal exposure studies,4,7,8,16,20,33 personal determinants are well identified and their weight is very high: they explain 86% of PM2.5 variations and 78% of those of NO2. For both pollutants, indoor TWCs are the major determinants of personal exposure. TWCH and TWCW explain 69% of PM2.5 personal variations, and 63% of NO2 variations. These proportions are the highest published in the literature, indoor concentrations accounting for 25–30%,4 34–46%,7 or 50%8 of variations in personal particulate exposure levels. TWCT, TWCO and time in other microenvironments increase our coefficient of determination in similar part for both pollutants. These very similar results for both pollutants are observed despite the fact that there is no correlation between NO2 and PM2.5 personal exposure or between NO2 and PM2.5 indoor concentrations. But it must be noted that the measurement times for both pollutants are different. This can also suggest that, as discussed later, indoor concentration determinants are different for fine particles and nitrogen dioxide. Contrary to this, outdoor sources for both pollutants do not really differ in the Paris area, where they are dominated by traffic emissions. Consequently a high correlation was observed between NO2 and fine particles ambient levels. Sarnat and colleagues15 noted an analogous phenomenon in Baltimore.
It should be noted that we were able to obtain good estimations of personal exposure from just the microenvironmental concentrations. Finally, the improvement in the determination coefficient associated with the use of TWC seems to be minor when compared to the difficulty of collecting accurate information about time activity patterns.
Identifying indoor concentration determinants in our study was more difficult, as PM2.5 indoor concentrations are largely influenced by exposure to tobacco smoke. Even in our population, where this exposure was low in frequency and intensity, its influence was very strong, especially in offices. Our results are consistent with those of Phillips et al in several towns21–28 and of Clayton and colleagues,6 who showed that in a home with one or more smokers, PM individual exposure and indoor levels were significantly increased, even during the night. Wallace38 and Janssen and colleagues9 consider that one cigarette is responsible for a 24 hour increase of 2.3 μg/m3 PM10 and 1–2 μg/m3 fine particles respectively. When there is no exposure to ETS, in-home concentration is influenced by ambient level, person density, and proximity traffic (R2 = 39%). In offices not exposed to ETS, two factors appeared to influence PM2.5 concentration: ambient level and person density; these only explain 20% of the concentration. The person density, expressed as the ratio between the number of persons in the room and the surface area, can be regarded as a surrogate of particle resuspending. The negative borderline association between person density and in-office PM2.5 concentration is rather surprising and could be caused by chance. Local traffic was not significant, perhaps because of the fact that the workplaces were not very numerous, with several participants working in the same building. Moreover, traffic was estimated by subjects themselves in four categories (zero, low, medium, heavy), and this appreciation is subjective and inaccurate. Although not significant in our model, cooking time and cleaning activity could also play a role. This contribution, shown by Clayton and colleagues6 and Lioy and colleagues14 for PM10, has not been quantified for PM2.5. However, Te Chang and colleagues29 suggest that the high hourly PM2.5 exposures they observed could be the result of fine particles emitted from grilling and other cooking activities.
Factors influencing NO2 indoor concentrations are different from those observed for fine particles. We confirmed that gas cooking influences NO2 in-home pollution, and using a precise questionnaire we were able to quantify the influence of time using the gas cooking. Usually, authors consider a dichotomous less accurate variable (gas cooking: yes/no).17–20,30,31 In-home concentrations also depend on ambient levels as a result of exchanges between indoor and outdoor air, but no other factor could be identified. ETS did not have a significant contribution to NO2 concentration. There are contradictory results in the literature about this influence.18,30,33,34 An inverse relation was observed between in-office concentration and floor height, and we suppose that this factor could be a surrogate of traffic exposure, which is poorly evaluated by self reported car densities—living on a higher floor implying a lower exposure to traffic.
Building and furniture characteristics are important because they may influence NO2 absorption, but we did not quantify this parameter.
Finally, a similar and primary contribution of indoor microenvironmental concentrations of both pollutants on personal exposure was observed in this study. Transport and outdoor exposure, on the other hand, play a secondary role in personal exposure, although outdoor air quality is indirectly implicated as a result of exchanges between interior and exterior NO2 and fine particles. Identification and quantification of indoor determinants were not very conclusive, and merit further investigation.
The authors thank all the participants in the study.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.