Personal carbon monoxide exposure levels: contribution of ... - Nature

Journal of Exposure Analysis and Environmental Epidemiology (2004) 14, 312–322 r 2004 Nature Publishing Group All rights reserved 1053-4245/04/$30.00

www.nature.com/jea

Personal carbon monoxide exposure levels: contribution of local sources to exposures and microenvironment concentrations in Milan YURI BRUINEN DE BRUIN,a,c PAOLO CARRER,a MATTI JANTUNEN,b OTTO HA¨NNINEN,b GRETA SCOTTO DI MARCO,c STYLIANOS KEPHALOPOULOS,c DOMENICO CAVALLOa AND MARCO MARONIa a

Department of Occupational Health, University of Milan, Via San Barnaba 8, 20122 Milan, Italy Department of Environmental Health, KTL, National Public Health Institute, PO Box 95 Kuopio, Finland c EC Joint Research Centre, Institute for Health and Consumer Protection, Via E. Fermi, 21020 (VA), Ispra, Italy b

In the framework of the EXPOLIS study in Milan, Italy, 48-h carbon monoxide (CO) exposures of 50 office workers were monitored over a 1-year period. In this work, the exposures were assessed for different averaging times and were compared with simultaneous ambient fixed-site concentrations. The effect of gas cooking and smoking and different methods of commuting on the microenvironment and exposure levels of CO were investigated. During the sampling the subjects completed a time–microenvironment–activity diary differentiating 11 microenvironments and three exposure influencing activities: gas cooking, smoking and commuting. After sampling, all exposure and time allocation data were stored in a relational database that is used in data analyses. Ambient 48-h and maximum 8-h distributions were similar compared to the respective personal exposures. The maximum 1-h personal exposures were much higher than the maximum 8-h exposures. The maximum 1-h exposures were as well higher than the corresponding ambient distribution. These findings indicate that high short-term exposures were not reflected in ambient monitoring data nor by long-term exposures. When gas cooking or smoking was present, the indoor levels at ‘‘home-’’ and in ‘‘other indoor’’ microenvironments were higher than without their presence. Compared with ambient data, the latter source was the most affective to increase the indoor levels. Exposure during commuting was higher than in all other microenvironments; the highest daily exposure contribution was found during ‘‘car/taxi’’ driving. Most of the CO exposure is acquired in indoor microenvironments. For the indoor microenvironments, ambient CO was the weakest predictor for ‘‘home indoor’’ concentrations, where the subjects spent most of their time, and the strongest for ‘‘other indoor’’ concentrations, where the smallest fraction of the time was spent. Of the main indoor sources, gas cooking, on average, significantly raised the indoor exposure concentrations for 45 min and tobacco smoking for 30 min. The highest exposure levels were experienced in street commuting. Personal exposures were well predicted, but 1-h maximum personal exposures were poorly predicted, by respective ambient air quality data. By the use of time–activity diaries, ETS exposure at the workplaces were probably misclassified due to differences in awareness to tobacco smoke between smokers and nonsmokers. Journal of Exposure Analysis and Environmental Epidemiology (2004) 14, 312–322. doi:10.1038/sj.jea.7500327

Keywords: air pollution, carbon monoxide (CO), EXPOLIS, exposure assessment, exposure determinants, time–activity diary.

Introduction Environmental epidemiology in the 1990’s has demonstrated that large public health impacts may be associated with present ambient air pollution levels. Causal explanation of air pollution-related health effects must be based on previous exposure (Jantunen et al., 1998). Therefore, in order to develop health protection policies concerning air pollution, it is necessary to assess the levels and sources of exposures. Traditionally, fixed site monitoring is used to assess urban air pollution. Several indoor air studies have shown, however, that ambient levels do not necessarily represent

1. Address all correspondence to: Dr. Y. Bruinen de Bruin, European Commission-Joint Research Centre, Institute for Health & Consumer Protection, Physical & Chemical Exposure Unit, Sector Indoor Air Quality, Human Exposure Modelling and Urban Noise, Via E. Fermi, 21020 Ispra (VA) Italy. Tel: þ 39-0332-789377. Fax: þ 39-0332-785146. Received 12 December 2003; accepted 28 October 2003

population exposures accurately (Cortese and Spengler, 1976; Akland et al., 1985). The exposure of individuals depends on personal time allotted to daily activities in different microenvironments. A more complete picture of human exposure and factors determining this exposure can be obtained by combining the daily activity and microenvironment monitoring data (Brunekreef et al., 1995). Acute health effects due to exposure to high levels of carbon monoxide (CO) have been well studied (Maroni et al., 1995). The best-known health effects of CO are based on its ability to impair the oxygen binding capacity of blood haemoglobin, reducing the oxygen transfer from lungs to the human body. The Air Pollution and Health; A European Approach (APHEA) study results from Athens, Greece, showed evidence of causal association between daily mortality and ambient CO concentrations (Touloumi et al., 1996). In a Canadian study, CO was found to be the strongest predictor of hospitalisation of the elderly due to congestive heart failure in comparison with SO2, NO2, or O3

CO exposure in Milan

(Burnett et al., 1997). Acute accidental and suicidal fatal CO exposures are still common and kill a large number of people annually in Europe (WHO-a, 2000). CO is a product of incomplete combustion, and therefore the highest and most lethal CO exposures are usually caused by unvented (faulty) indoor combustion devices. In population exposures studies, daily CO exposures have also been associated with the total time spent on commuting because most ambient CO emissions are produced by vehicle traffic (Cortese and Spengler, 1976). More recent studies, however, have identified other high concentration microenvironments, such as garages, residences, offices, bars and restaurants, due to their vicinity to traffic sources and/or the presence of indoor sources, such as gas appliances or tobacco smoke (Akland et al., 1985; Nagda and Koontz, 1985; Alm et al., 1994, 2000). The majority of the studies have focused on selected population samples: Cortese and Spengler (1976) on only nonsmoking volunteers; Ziskind et al. (1982) on nine participants in a pilot study; Nagda and Koontz (1985) on office workers, construction workers, and housewives; Lee et al. (1992) on students; Maroni et al. (1996) on office workers; Vellopoulou and Ashmore (1998) on nonsmoking commuters and Alm et al. (1994, 1999, 2000) on preschool children. The U.S. EPA Washington–Denver study was the only large-scale study that used stratified probability samples of the target population (Akland et al., 1985; Jungers et al., 1985). In our study, personal exposure measurements were performed in 1997–1998 as part of the population-based air pollution exposure study (EXPOLIS) Air Pollution Exposure Distributions of Adult Urban Populations in Europe. EXPOLIS studied the relationships between indoor and outdoor urban air pollution and time–microenvironmental–activity data in seven major European cities; Athens (Greece), Basel (Switzerland), Grenoble (France), Helsinki (Finland), Milan (Italy), Oxford (United Kingdom), and Prague (Czech Republic). Similar field study protocols were used in all cities (Jantunen et al., 1998, 1999). Among the six EXPOLIS cities, the 4-million inhabitant metropolitan area of Milan is known as a city with dense traffic having frequently to deal with severe air pollution episodes. The most recent episodes in Milan and the whole North-western Po-river valley of Italy, occurred during the winter season 2001–2002 when the western air circulation was blocked for several weeks due to persistent inversion during anticyclone conditions. In the four winter months, PM10 concentrations exceeded 75 mg/m3 (the Italian level of alarm) during 44 days and twice for over 7 days in sequence. The longest episode lasted 3 weeks with average daily PM10 concentrations reaching up to 217 mg/m3 (Regione Lombardia/Settore Qualita` dell’Ambiente). These concentrations led to traffic restrictions in many Northern Italian cities. Within EXPOLIS, carbon monoxide (CO) was selected to represent exposures to traffic exhaust and indoor combustion sources, such as smoking and use of gas appliances. Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4)

Bruinen de Bruin et al.

Georgoulis et al. (2002) found out that the personal 48-h CO exposure levels as well as the exposure/ambient ratios differ between five EXPOLIS cities. In this paper, we analyse the EXPOLIS-Milan data to (1) assess the exposures to CO of the Milan urban adult office working population, and (2) to evaluate the microenvironmental CO levels and sources affecting these levels. Gas cooking and environmental tobacco smoke (ETS) were the indoor sources considered. The effects of the different means of transportation on the exposure levels were also assessed.

Methods In Milan, since over 75% of the working population operates in offices or similar microenvironments, it was decided to evaluate the exposure for only this category of workers (Jantunen et al., 1999). The office workers in Milan were selected in order to represent the office working population of Milan in terms of sex, smoking habits, and locations of homes and offices, respectively. The exposure monitoring sample consisted of 50 office workers, aged 25–55 years (27 females and 23 males, mean 7 standard deviation age 3978.8), living in 50 homes, and working in 50 office buildings. The population characteristics are described in more detail by Jantunen et al. (1999) and Rotko et al. (2000). Selection of homes and office buildings, representative for the various building tipologies present in the metropolitan area of Milan, was based on a previous study performed by the Institute of Occupational Health of the University of Milan (Maroni et al., 1996). Personal exposures were monitored for 48-h and logged every minute using an electrochemical high-resolution Langan Model T15 DataBear personal monitor (Langan, Inc., San Fransisco, USA) according to KTL Standard Operating Procedures (1997a, b). Each person was sampled once. After sampling, the data were downloaded into a computer using Langan.exe software (Ha¨nninen, 1996). The measurement and data processing methodologies are described in more detail by Ha¨nninen (1996), Ha¨nninen et al. (2002), and Georgoulis et al. (2002). The exposure data were stored in the EXPOLIS database (Ha¨nninen et al., 2002). Specifically for the Milan data, an intermediate database was developed to facilitate modifications for the temperature, zero, span, and calibration. The EXPOLIS-Milan database also contains hourly ambient CO concentrations obtained from the fixed-site city monitoring network (six stations operated by Provincia di Milano/Settore Ecologia). During sampling, the participants completed a 15-min resolution time–activity diary differentiating 11 microenvironments and three activities. The microenvironments were ‘‘home indoor’’, ‘‘work indoor’’, ‘‘other indoor’’, ‘‘home outdoor’’, ‘‘work outdoor’’, ‘‘other outdoor’’, ‘‘walking’’, ‘‘train/metro’’, ‘‘bus/tram’’, ‘‘motorbike’’, and ‘‘car/taxi’’. In 313



Milan, walking mostly consisted of walking from, for example, the subway-, train-, or tram station (partly indoors and outdoors), along streets with heavy traffic (outdoors), to the office (indoors), or vice versa. Since walking involves potential elevated exposures to traffic emissions, this microenvironment best fits into the group that represent different methods of commuting, consisting of the five microenvironments: ‘‘walking’’, ‘‘train/metro’’, ‘‘bus/tram’’, ‘‘motorbike’’, and ‘‘car/taxi’’. The activities recorded were gas cooking in the same room, passive and active ( ¼ self) smoking. Active and passive smoking were combined into ‘‘ETS exposure’’, because the dominant exposure component from direct smoke inhalation of active smoking was not measured, that is, only passive smoking exposure was measured. Simultaneous presence of gas cooking and smoking was processed as a separate activity. Four activity classes were used in the analyses: 1. absence of smoking and gas cooking, 2. presence of gas cooking, 3. presence of smoking, and 4. simultaneous presence of both gas cooking and smoking. The status ‘‘doer’’ was given to each participant that stayed 15-min or longer in a certain microenvironment while exposed to or performed a certain activity. A total of 46 individuals provided data for the subsequent analysis. Data of four participants were lost or could not be used; three participants did not complete their diaries and in one case the monitor failed. The original 1-min exposure data were used to calculate 15-min averages corresponding to the timeactivity diaries. From these, the distributions of the personal 48h averages and running 1- and 8-h maxima were calculated. The population time spent in the three indoor microenvironments was classified into 15-min periods by the four defined activities. The time fraction spent by the participants in each microenvironment performing any activity was combined with the respective CO levels to obtain a timeweighted average (TWA) microenvironment concentration (further referred to as microenvironment/activity). The microenvironment/activity concentrations were compared to corresponding mean ambient CO concentrations based on the six fixed-site monitoring stations. An individual’s exposure depends on the fraction of time spent and concentration in each microenvironment/activity. The contribution of each microenvironment/activity to the average daily population exposure was estimated by multiplying the fraction of population time with the corresponding average microenvironment/activity concentrations. The contribution percentage is calculated by dividing the microenvironment/activity contribution by the sum of all average microenvironment/activity contributions. This procedure was repeated including only the individuals that actually visited a particular microenvironment/activity: the ‘‘doers’’. For example, if a person spent 13 h indoor at home with an exposure of 1.5 ppm, and 9 h indoor at work with an exposure of 2.0 ppm, and 2 h commuting with an exposure of 314

4.0 ppm, then the contribution of commuting to the average daily exposure of this person is calculated as follows: 2h 4:0ppm 100% ð13h 1:5ppmÞ þ ð9h 2:0ppmÞ þ ð2h 4:0ppmÞ ¼ 17:6% The effects of gas cooking and ETS on the indoor levels were studied. Depending on the exposure status, a code was given for each 15-min period that a person indicated being in an indoor microenvironment. Different codes were given for (1) absence of gas cooking and smoking, (2) gas cooking event, (3) consecutive periods after each gas cooking event, (4) present smoking, (5) consecutive periods after smoking, (6) occurrence of both gas cooking and smoking at the same time, and (7) the corresponding ambient concentration. The consecutive periods after gas cooking or ETS exposure were coded to study how long the levels are affected by either cooking with gas or smoking. The difference of the means were analysed against the mean when indoor sources were absent by ANOVA one-way analyses of variance. Multiple regression analyses were used to test the effect of the activity gas cooking and smoking and their respective tails (the consequtive periods after each activity) and ambient concentrations on the measured indoor CO concentrations. Each code indicating a specific activity class was transformed into classification variables 0 and 1 corresponding to the activity’’s absence or presence. The dependent variable used was the 15-min average indoor concentration. The independent variables were the simultaneous ambient concentration and classification variables.

Results The percentages of population time spent in the various microenvironments are presented in Figure 1. The time spent indoors (‘‘home-’’ þ ‘‘work-’’ þ ‘‘other indoor’’) dominate in comparison to the outdoor (‘‘home-’’ þ ‘‘work-’’ þ ‘‘other outdoor’’) and transportation (‘‘car/taxi’’ þ ‘‘bus/ tram’’ þ ‘‘motorbike’’ þ ‘‘walking’’ þ ‘‘train/metro’’) microenvironments; the study population spent 55.7% of the time at ‘‘home’’, 30% at ‘‘work’’ and 5.1% in ‘‘other Indoor 90.8%

work 0.6% 0.6% work 30.0%

Outdoor 1.7%

home 0.2%

other 5.1%

other 0.9%

train/metro 0.7%

home 55.7%

Transportation 7.5%

bus/tram 1.8%

walk 2.4%

motorbike car/taxi 0.2% 2.4%

Figure 1. Percentages of population time spent in different microenvironments. Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4)


Population time fractions and concentration summary statistics in different microenvironments presented as mean and SD are listed in Table 1. Microenvironments are 20

1-h

Exposure/Concentration [ppm]

indoor’’ microenvironments, summing up to 90.8% of the population time. In all, 1.7% of the time was spent outdoors. The Milan participants spent on average about 7.5% of their time in transportation, corresponding to 1 h 48 min daily. Most of the transportation time is spent ‘‘walking’’ and using ‘‘car/taxi’’ or ‘‘bus/tram’’, less by ‘‘train/metro’’ and ‘‘motorcycle’’. The distributions of the average 48-h exposures, the running 8- and 1-h maximum exposures, and the corresponding ambient levels are shown in Figure 2. The ambient 48-h and maximum 8-h distributions are quite similar compared to the exposures. The 8-h maxima remained below the 8-h air quality standards of U.S. EPA, 1995 (9 ppm) (EPA, 1995), and the WHO-a, 2000, guideline value (10 ppm) (WHO-b, 2000). The maximum 1-h ambient distribution is higher than the maximum 8-h ambient distribution. Similarly, the maximum 1-h personal exposures are much higher than the maximum 8-h exposures. The maximum 1-h exposures are as well higher than the corresponding ambient distribution. These findings indicate that high short-term exposures are not reflected in ambient monitoring data nor by long-term exposures.


10 9 8 7 6 5

8-h

4

48-h

3 2

1 0.9 0.8 0.7 0.6 0.5

1-h 8-h 48-h 2

5

10

20

30

40

50

60

Personal mean±sd ppm 7.3±3.2 3.3±1.5 2.1±0.9 70

80

Ambient mean±sd ppm 5.0±2.7 3.3±1.7 2.1±0.9 90

95

98

Percentile [%]

Figure 2. Distributions of the average 48-h exposures the running 8- and 1-h maximum exposures, and the corresponding ambient levels based on the exposure time corresponding concentration mean of six fixed-site monitors.

Table 1. Population (n ¼ 46 subjects) time fraction spent (mean% and STD%), microenvironment (mE)/activity and corresponding ambient concentrations (ppm).

Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4)

315


subdivided according to the presence of local sources. Within all the indoor microenvironments, most of the time was spent in the absence of indoor sources. During only gas cooking or only with ETS exposure, the ‘‘home-’’ and ‘‘other indoor’’ concentrations were higher compared to the absence of these sources. In comparison with the ambient levels shown in the same table, it can be seen that during presence of ETS at ‘‘home-’’ and in ‘‘other indoor’’ the indoor levels are clearly higher than the ambient levels. A similar increase was not observed at ‘‘work’’ indoors. All traffic exposures were elevated relative to ambient concentrations. These levels were more than double during ‘‘car/taxi’’ driving. Table 2 shows the contribution of microenvironments and activities to the total population’s and doers’ exposure. All indoor microenvironments with all activities summed up contribute 81.8% to the daily population exposure. In all, 9% was contributed during the presence of indoor sources. The contribution to the exposures of the doers is higher than the contribution to the total population, as can be expected. For the ‘‘home-’’ and ‘‘work indoor’’ microenvironments with no sources present, the total population’s contribution exposure percentage is lower compared with the percentage


of the time fraction. This result indicates that the microenvironment/activity concentrations are lower than the average exposure concentrations. This means that the more time spent would be at ‘‘home’’ or at ‘‘work’’ in the absence of indoor sources, the lower the exposure levels would be. In contrast, the ‘‘other indoor’’ exposure contribution with and without indoor sources is higher indicating that the concentrations present in the ‘‘other indoor’’ microenvironments are higher than the average exposure concentrations. The more time would be spend in ‘‘other indoor’’, the higher the exposure levels would be. The total population’s exposure contribution percentages while in transit are high compared with percentage fractions of times spent. This higher concentration is caused by higher exposures using any method of transportation. In the ‘‘other indoor’’ microenvironments, ETS contributed up to almost 24% to the population exposure in this microenvironment. When the ‘‘other indoor’’ was checked in the diaries, the specifications mostly indicated visits to smoky bars or restaurants. This percentage probably reflects these visits. When gas cooking was present the contribution to the doers is 25.5 times higher compared to the total population in

Table 2. Daily microenvironment (mE) and activity contributions to the total population (n ¼ 46 subjects) and doers’ exposure.

316




the ‘‘other indoor’’ microenvironments. This result, however, might not be representative due to the low number of doers (two subjects or 4% of the total sample population) and conclusions must be drawn with care. On average, 16% of the population exposures occur in traffic. Travelling by ‘‘train/metro’’ and ‘‘motorbike’’ caused a rather similar contribution to the doers. (2.8% vs. 3.3%, respectively), but as shown in Table 1, exposure levels during commuting are about 40% higher while ‘‘motorcycling’’ than while travelling by ‘‘train/metro’’. ‘‘Walking’’ and transportation by ‘‘bus/tram’’ and ‘‘car/taxi’’ caused the highest traffic exposure contributions to the total population with ‘‘car/taxi’’ being highest (3.4%, 3.4%, and 8.0% respectively). Table 3 shows statistics of the indoor concentrations during and after gas cooking and/or smoking events. In the ‘‘home indoor’’ microenvironment, levels during gas cooking were significantly higher (Po0.05) than without the presence of indoor sources until 1-h after the event. Looking back to Table 1, however, we saw that compared to simultaneous ambient levels, the gas cooking event levels were not higher. By studying the time–activity data, the cooking times were found to be overlapping with the times of the morning and evening rush hours. The ambient concentrations, measured by fixed-site monitoring stations that were situated nearby streets with heavy traffic, are therefore expected to be positively influenced by the intense rush-hour traffic during cooking times. When smoking was present, the indoor CO concentrations were also significantly higher (Po0.05) than without the

presence of indoor sources until 30 min after the event. The concentration elevation was significantly highest (Po0.05) during simultaneous presence of gas cooking and smoking. In the absence of indoor sources the indoor concentration was significantly lower (Po0.05) than the respective ambient concentration. Applying the Kruskal–Wallis analyses to test several independent samples for (in)dependence, tested at the P ¼ 0.01 level, we found that the indoor levels at ‘‘home’’ during gas cooking until 60 min after each event can be grouped together. According to these analyses, consequently, gas cooking can be considered as an event that affects the indoor concentrations for approximately 60 min. A similar result was found at the ‘‘home indoor’’ microenvironment with ETS exposure lasting until 30 min after the smoking event. In the ‘‘other indoor’’ microenvironment, significantly higher CO concentrations (Po0.05) were found during either gas cooking or smoking (Po0.05). Highest levels were present during gas cooking, but these are based on only six available 15-min concentration data points. The ‘‘other indoor’’ concentrations in the absence of indoor sources were higher than the ambient concentrations. Table 4 shows the results of multiple regression analyses of the log transformed 15-min concentrations in the indoor microenvironments. Only 1/4 of the variation in the ‘‘home indoor’’ concentrations was explained by the chosen independent variables (r2 ¼ 0.24). The independent variables that showed the strongest and statistically significant associations (Po0.01) were ambient concentration, presence

Table 3. Distribution of the average 15 min concentrations during gas cooking and/or presence of ETS in the indoor microenvironments and 15-min periods after the respective activities. Statistical significance, calculated using analyses of variance; Bonferroni and Tukey, are presented. Exposure status (a) Home indoor Absence of indoor sources Gas cooking 1–15 min after 16–30 min after 31–45 min after 46–60 min after ETS 1–15 min after 16–30 min after Gas cooking and ETS Ambient (b) ‘‘Other indoor’’ Absence of indoor sources Gas cooking Presence of ETS Ambient

n. samples

AM

STD

GM

GSD

4112 174 68 62 53 47 155 75 60 16 4649

1.74 2.37a 2.36a 2.51a 2.68a 2.48a 2.86a 2.36a 2.27a 3.36a 2.11a

1.26 1.52 1.38 1.45 1.81 1.38 1.61 1.24 1.33 2.20 1.66

1.34 1.95a 1.94a 1.71a 2.13a 2.04a 2.43a 2.04a 1.92a 2.79a 1.59a

2.20 1.90 1.99 2.04 2.12 2.09 1.80 1.76 1.81 1.88 2.17

321 6 91 275

2.49 6.53b 3.47b 1.68c

2.17 2.48 2.93 2.40

1.79 6.21b 2.74b 1.68c

2.38 1.39 1.95 2.40

a

Difference between levels while a source or sources is/are present and the nonsource concentrations is significant at the 0.05 level. Difference between levels while a source is present and the nonsource concentrations is significant at the 0.05 level. Difference between ambient and the nonsource concentrations is not significant at the 0.05 level (P ¼ 0.76).

b c


317



Table 4. Results from multiple linear regression of log transformed 15 min indoor microenvironment CO concentrations (dependent variable) using the coded gas cooking, presence of ETS, consecutive 15-min periods (tails), and corresponding ambient concentration as independent classification variables. Dependent variables Home indoor R ¼ 0.49, Adj. R2 ¼ 0.24, Po0.01 Independent variables Intercept (ppm) CO ambient concentrations (ppm) Presence of gas cooking (yes ¼ 1, no ¼ 0) 1–15 min after gas cooking event (yes ¼ 1, no ¼ 0) 16–30 min after gas cooking event (yes ¼ 1, no ¼ 0) 31–45 min after gas cooking event (yes ¼ 1, no ¼ 0) 46–60 min after gas cooking event (yes ¼ 1, no ¼ 0) Presence of ETS (yes ¼ 1, no ¼ 0) 1–15 min after ETS event (yes ¼ 1, no ¼ 0) 16–30 min after ETS event (yes ¼ 1, no ¼ 0) 31–45 min after ETS event (yes ¼ 1, no ¼ 0) Presence of both gas cooking and ETS (yes ¼ 1, no ¼ 0)

b (SE(b))

Work indoor R ¼ 0.36, Adj. R2 ¼ 0.13, Po0.01

P

b (SE(b))

1.100 (1.120) 0.453 (0.013)

o0.01 o0.01

0.084 (0.025)

o0.01

0.074 (0.037)

o0.05

0.093 (0.038)

o0.05

0.099 (0.042)

o0.05

0.069 (0.045)

P o0.01 o0.01

0.244 (0.025)

o0.01

0.025 (0.022)

0.249

0.162 (0.036)

o0.01

0.016 (0.025)

0.524

0.127 (0.039)

o0.01

0.004 (0.029)

0.892

0.010 (0.035)

0.788

0.229 (0.080)

b (SE(b)) 1.180 (1.03) 0.874 (0.029)

P o0.01 o0.01

0.111 (0.073)

0.132

0.030 (0.028)

0.280

0.121

o0.01

of gas cooking until 45 min after cooking (Po0.05), presence of ETS until 30 min after the smoking, and simultaneous presence of gas cooking and ETS. For the ‘‘work indoor’’ microenvironment, only 13% of the variation in the concentration can be explained by the regression model. Ambient concentration was the only variable with statistically significant association (Po0.01) with the ‘‘work indoor’’ microenvironment concentration. For the ‘‘other indoor’’ microenvironment almost 80% of the concentration variation was explained by the regression model (r2 ¼ 0.79), although the only statistically significant parameter (Po0.01) was the ambient concentration.

Discussion Comparisons of the EXPOLIS Milan CO data with the other EXPOLIS cities, are described in more detail by Georgoulis et al. (2002).

Representativeness of Total Working Population The limited number of subjects and selection of participants should be kept in mind when drawing conclusions that concern the entire target population. The majority of the 318

1.070 (1.02) 0.555 (0.030)

‘‘Other indoor’’ R ¼ 0.89, Adj. R2 ¼ 0.79, Po0.01

Milan working population, approximately 75%, exist of office workers (Jantunen et al., 1999). The results of our office workers presented in this paper are not expected to reflect the remaining nonoffice working population; certain individuals have jobs related to elevated CO exposures, like street workers, policemen, taxi drivers (exposed to traffic emissions), bar, and restaurant personnel (ETS exposure). Also, some activities were rather uncommon in certain microenvironments, such as gas cooking in the ‘‘other indoor’’ microenvironment, causing conclusions to be drawn with caution. Detailed characteristics of the population sample have been described by Rotko et al. (2000).

Environmental Tobacco Smoke (ETS) In our study, elevated indoor levels were found in the presence of ETS in both the ‘‘home-’’ and ‘‘other indoor’’ microenvironments. This is consistent with other studies that found increasing personal exposures during ETS exposure or when smoking was present in the investigated microenvironments (Ziskind et al., 1982; Akland et al., 1985; Ott et al., 1988; Siegel, 1993; Vellopoulou and Ashmore, 1998; Klepeis, 1999; Alm et al., 2000). Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4)


At the ‘‘work indoor’’ microenvironment, we were not able to detect an elevation of the CO concentrations when smoking was present; they actually appeared to decrease when the subjects reported exposure to ETS. This finding contradicts the known CO emissions from smoking and earlier findings, including the EXPOLIS findings in indoor microenvironments (Georgoulis et al., 2002). Therefore, the workday average CO exposure of self-reported nonsmokers vs. smokers (who would presumably be more exposed to ETS than nonsmokers) was checked. As expected, the average exposure of smokers was indeed higher, by 60% (data not shown). Additional analyses of the time–activity data showed that during the working hours smokers reported less time with ETS exposure than nonsmokers. One could speculate that while some nonsmokers are quite sensitive to ETS and report even marginal exposures, smokers, on the other hand, are often only weakly aware of ETS. Especially 30 min following an active smoking event, when the smoker is exposed to a significantly elevated ETS level, but his or her awareness of ETS is lowest, might be misclassified as nonETS. Analysing the smokers’ and nonsmokers’ self-reported ETS workplace exposure data (data not shown), we found that the smokers’ ETS exposure is 29% lower than that of nonsmokers. In contrast, in the ‘‘home indoor’’ and in the ‘‘other indoor’’ microenvironments, the smokers’ exposures to ETS were 20% and 41% higher, respectively, than those of nonsmokers. In the homes, the situation is obviously more clear cut; the non-ETS exposed individuals mostly do not smoke, do not live with a smoking partner or request the smoking partner/visitor to smoke outside. People are more selective to enter certain ‘‘other indoor’’ microenvironments, like smoky bars or restaurants, involved with elevated exposures. Assuming that this explanation is correct, it should also be kept in mind when analysing other ETSrelated exposures, like PM2.5, benzene, and NO2.

Gas Cooking Indoor air levels were elevated when gas appliances were used. Similar findings have been reported by several other studies (Brunekreef et al., 1982; Akland et al., 1985; Nagda and Koontz, 1985; Ott et al., 1988; Alm et al., 1994; Moriske et al., 1996; Klepeis, 1999; Naeher et al., 2000; Wallace, 2000). Higher personal CO exposures have been observed especially when gas appliances were malfunctioning (Alm, 1999). In the cases of gas cooking in the ‘‘other indoor’’ microenvironment and simultaneous gas cooking and smoking in the ‘‘home’’ microenvironment, too few 15-min measurements are available to draw specific conclusions. Transportation We observed elevated exposure levels during all means of transportation; the exposures were lowest in ‘‘trains/metros’’ and highest in ‘‘cars/taxi’s’’. ‘‘Car/taxi’’ caused the highest Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4)


traffic exposure contribution probably because of the proximity of car air intakes to the exhaust pipes of other cars in dense traffic. Other studies have indicated similar increases in exposures due to traffic (Cortese and Spengler, 1976; Zikind et al. (1982); Akland et al., 1985; Nagda and Koonz, 1985; Ott et al., 1988; Maroni et al., 1996; Vellopoulou and Ashmore, 1998).

Microenvironment and Ambient Concentrations The fixed-site stations are located near high traffic streets. Therefore, the levels exceed the average indoor ‘‘home’’ and ‘‘work’’ concentrations in the absence of gas cooking and smoking. The contribution of the ambient concentration due to penetration into the indoor microenvironments, however, was not directly assessed in this work. Neither microenvironment concentrations were adjusted according to the ambient concentrations. The geographic distribution of the limited number of locations of indoor microenvironments visited by our few subjects around the nearest monitoring stations is not expected to result in useful evaluation. To assess these penetration factors, simultaneous and local outdoor data would be needed to compare with each indoor microenvironmental measurement. Such CO data, however, was not measured within the framework of EXPOLIS. Owing to the locations of the fixed-site stations in Milan, the ambient data were not considered for such calculations. Penetration of outdoor CO into indoor microenvironments should be taken into account in assessing the exact impact of gas cooking or presence of ETS on the CO levels in indoor microenvironments. In the ‘‘other indoor’’, all outdoor (‘‘home-’’ þ ‘‘work-’’ þ ‘‘other outdoor’’), and transportation (‘‘car/ taxi’’ þ ‘‘motorbike’’ þ ‘‘bus/tram’’ þ ‘‘train/metro’’ þ ‘‘walking’’) microenvironments, the average CO concentrations were higher than the corresponding ambient levels, indicating the presence of nearby sources. Levels in the ‘‘other indoor’’ microenvironments are probably elevated because of smoking. Especially when the ‘‘other indoor’’ microenvironments existed of bars and restaurants, also their respective locations, frequently close to high traffic streets, become important. This is expected to cause the levels to be higher than the average ambient concentrations, also when indoor sources were absent. The study subjects spent just a small fraction of their time in the outdoor microenvironments. Therefore, we cannot draw conclusions relative to the ambient concentrations. 24-h Average Exposure Exposure to tobacco smoke is rather common in Milan. In total, 42% of the population was either actively smoking or exposed to ETS in the ‘‘home’’ and in the ‘‘other indoor’’ microenvironments. In all, 58% of the total population reported ETS exposure at the workplace. In a study of Maroni et al. (1996), the reported daily CO exposure contribution of Milan office workers were 50.1%, 319


32.9%, and 17.0%, respectively, in ‘‘home-’’, ‘‘work indoor’’, and ‘‘transportation’’ microenvironments. In total, five microenvironments were defined differentiating ‘‘home’’ and ‘‘work indoor’’, ‘‘bus’’, ‘‘tram’’, and ‘‘metro’’. Our work found comparable contribution percentages of these microenvironments: 48.4, 27.0, and 16.1%, respectively. In EXPOLIS, however, 11 microenvironments were defined, causing the exposures to be distributed among a higher number of microenvironments. When the analyses were confined to the doers’ exposures, the microenvironment/activity contributions are, as one might expect, higher than those of the total population. An extreme example is that during gas cooking in the ‘‘other indoor’’ microenvironment, the exposure of the ‘‘doers’’ was about 26-fold higher. Since gas cooking exposure in ‘‘other indoor’’ is based on rather few 15 min measurements from only 2 participants, this increase is not considered to be representative. To assess the exposure significance of a particular activity, the doer’s values, however, are the ones that should be used. As an example, people do not spend 0.2% of their time or acquire 0.3% of the CO exposure during ‘‘motorbike’’ driving. Instead, the doers spend in average 2% and the nondoers 0% of their time driving a ‘‘motorbike’’. As well, the doers acquire 3.3% and non-doers 0% of their exposures while driving a ‘‘motorbike’’.

Exposure Determinants The contribution percentages depend on both the time spent and concentrations experienced in the various microenvironments. The exposure concentrations in traffic are among the highest of all microenvironments. Therefore, the time spent using any methods of commuting is more significant for exposure than the time spent in the indoor and outdoor microenvironments. Still, the total daily CO exposure is mostly determined at home under generally relatively low CO concentrations, because in this microenvironment such a dominating fraction of time, in average 56%, is spent. Prediction of Population Exposures using Fixed Site Monitoring Data For decades urban air pollution has been assessed using ambient fixed site monitoring. The Total Exposure Assessment Methodology Study (CO-TEAM Akland et al., 1985; VOC-TEAM Wallace et al., 1988; PTEAM Thomas et al., 1993; Clayton et al., 1993) assessed, among other issues, the use of ambient air monitoring to predict population exposures. It was found that the correlation between personal exposures and ambient air concentrations was low (r2 ¼ 0.2) (Wallace, 1992). Also in a study in Athens, in which the CO exposures of commuters were assessed, the concentrations at the nearest fixed-site city station proved to be poor estimators of the commuters’ personal CO exposures (Vellopoulou and Ashmore (1998)). The authors explained the low correlation 320


partly by the fact that the monitoring station was near the city centre, while the participants moved between different urban and suburban microenvironments with different CO concentrations. In addition, the authors pointed out that the relationships between the personal exposures and ambient concentrations vary among the subjects, due to their different activity patterns and the geographic characteristics of their home locations. Both these reasons may well be appropriate explanations also for our findings. In Milan, all fixed site monitoring stations are located in the vicinity of streets with high traffic. The second explanation might very well be applied for both Milan home and work locations. The 8-h exposure maxima resulted in rather similar distributions as the respective ambient data. Similar results were reported by Cortese and Spengler (1976) and Akland et al. (1985). This, indicates that the averages from the six fixed site monitors in Milan are suitable for estimating the 8-h exposure distributions for, at least, our sample population. This suitability, however, does not provide information about the source apportionment. No linearity-based estimates can be made concerning the effectiveness of any exposure limiting measure. For example, the effect of reducing vehicle emissions by 10% does not intend that the overall exposures will also be reduced by 10%. The contribution due to indoor sources, such as gas cooking and smoking, will remain. The suitability is also remarkable due to the fact that they were quite inaccurate for predicting the exposure of a specific individual at a specific time. The levels of the running 1-h exposure maxima were found to be much higher than the simultaneous ambient maxima concentrations, indicating the existence of elevated shortterm exposures, for example, during smoking, gas cooking or due to time spent in traffic. Consequently, the ambient CO monitoring network data from Milan appear to be appropriate for estimating the health effects of 8-h maximum exposures, but not for the shorter, 1-h, maxima. Only exposure and/or indoor monitoring or modelling will provide such data (WHO-a, 2000).

Regression Analyses using 15-min Indoor Concentrations For a nonreactive pollutant, such as CO, the relationship between the interior and exterior concentrations as a function of time in a no-source microenvironment follows well-established laws that depend on the outdoor concentrations, air exchange rate, and the volume of the microenvironment. In the real-world settings, microenvironmental air exchange rates are finite, and ambient levels are equal to the readings inside the microenvironments only if the averaging time is long enough for the two to reach equilibrium (Ott et al., 1988). The importance of the ambient concentrations was lowest in the ‘‘home indoor’’ microenvironment, followed by the ‘‘work indoor’’ microenvironment, and followed by the ‘‘other indoor’’ Journal of Exposure Analysis and Environmental Epidemiology (2004) 14(4)


microenvironment as most important. These differences of importance were probably caused by increased influence of traffic emissions in the respective microenvironments. The home indoor CO levels were significantly elevated during and after gas cooking (until 45 min after each gas cooking event) and smoking (until 30 min after each smoking event).

Conclusions From this work, the following conclusions were drawn: 1. Most of the CO exposure occurred in indoor microenvironments, because this is where the study population spent over 90% of their time. 2. For the indoor microenvironments, ambient CO was the weakest predictor for ‘‘home indoor’’ concentrations, where the subjects spent most of their time, and the strongest for ‘‘other indoor’’ (than ‘‘home’’ and ‘‘work’’) concentrations, where the smallest fraction of the time was spent. Of the main indoor sources at ‘‘home’’, gas cooking, on average, significantly raised the indoor exposure concentrations for 45 min and tobacco smoking for 30 min. 3. Exposure to tobacco smoke of Milan office workers in indoor environments was rather common (42% are exposed at home and 58% at work). 4. The highest CO exposure levels were experienced in street commuting: travelling by ‘‘car/taxi’’ resulted in a higher contribution to the population exposure than ‘‘walking’’, travelling by ‘‘bus/tram’’ or by ‘‘motorbike’’. 5. The exposure contributions from some activities (like ‘‘motorbike’’ driving) were not important on the population level; however, they were important on an individual level. 6. The 1-h maximum personal exposures were poorly predicted by respective ambient air quality data. 7. By the use of time–activity diaries, ETS exposures at the workplaces were misclassified due to differences in awareness to tobacco smoke between smokers and nonsmokers.

Acknowledgments This work has been supported by EC. Environment and Climate 1994–1998 Programme Contract N ENV4-CT960202 (DG 12-DTEE) and C.E. Ispra n. 18161-2001-07 F1ED ISP IT.

References Akland G.G., Hartwell T.D., Johnson T.R., and Whitmore R.W. Measuring human exposure to carbon monoxide in Washington DC



Denver Colorado during the winter of 1982–83. Environ Sci Technol 1985: 19: 911–918. Alm S., Reponen K., Mukala K., Pasanen P., Tuomisto J., and Jantunen M.J. Personal exposures of pre-school children to carbon monoxide: roles of ambient air quality and gas stoves. Atm Environ 1994: 28: 3577–3580. Alm S. Personal exposures of preschool children to carbon monoxide and nitrogen dioxide in urban air, Academic Dissertation. KTL, Kuopio Finland, 1999. Alm S., Jantunen M.J., and Vartiainen M. Urban commuter exposure to particle matter and carbon monoxide inside an automobile. J Exposure Anal Environ Epidemiol 1999: 9(3): 237–244. Alm S., Mukala K., and Jantunen M.J. Personal carbon monoxide exposures of pre-school children in Helsinki, Finland: levels and determinants. Atm Environ 2000: 34: 277–285. Brunekreef B., Smit H.A., Biersteker K., Boleij J.S.M., and Lebret E. Indoor carbon monoxide pollution in The Netherlands. Environ Int 1982: 8: 193–196. Brunekreef B., Hoek G., and Janssen N. Time–activity patterns in air pollution epidemiology, In:Ackermann-Liebrich U., Viegi G., Nolan C. (Eds.). Time–activity Patterns in Exposure Assessment, Air Pollution Epidemiology Reports Series, Report no. 6, (EUR 15892 EN). European Commission Directorate General XII, Office for Official Publications, Luxembourg, 1995 92pp. Burnett R.T., Dales R.E., Brooks J.R., Raizenne M.E., and Krewski D. Association between ambient carbon monoxide levels and hospitalisations for congestive heart failure in the elderly in 10 Canadian cities. Epidemiology 1997: 8: 162–167. Clayton A., Perrit R.L., Pellizari E.D., Thomas K.W., Whitmore R.W., Wallace L.A., Ozkaynak H., and Spengler J.D. Particle total exposure assessment methodology (PTEAM) 1990 study: distributions of aerosol and elemental concentrations in personal, indoor, and outdoor air samples in a Southern California community. J Exposure Anal Environ Epidemiol 1993: 3: 227–250. Cortese A.D., and Spengler J.D. Ability of fixed monitoring stations to represent personal carbon monoxide exposure. J Air Pollution Control Assoc 1976: 26: 1144–1150. EPA. National Air Quality and Emissions and Trends Report, US Environmental Protection Agency, Research Triangle Park, EPA454-95-005 1995. Georgoulis L.B., Ha¨nninen O., Samoli E., Katsouyanni K., Ku¨nzli N., Bruinen de Bruin Y., Alm S., and Jantunen M.J. Personal CO exposures in five European cities and its determinants. Atm Environ 2002: 36: 963–974. Ha¨nninen O. Langan.EXE F Databear download program Instruction sheet developed for use in the EXPOLIS study. Project software and document, KTL, Kuopio, Finland, 1996. Ha¨nninen O., Alm S., Kaarakainen E., and Jantunen M. The EXPOLIS Databases. Publications of the National Public Health Institute, B13/2002 KTL, Kuopio. ISBN: 951-740-321-6 (printed), ISBN: 951-740-322-4 (pdf-file), ISSN: 0359-3576, 408pp. 2002. Jantunen M.J., Ha¨nninen O., Katsouyanni K., Kno¨ppel H., Ku¨nzli N., Lebret E., Maroni M., Saarela K., Sram R., and Zmirou D. Air pollution exposure in European cities: The ‘‘EXPOLIS’’ study. J Exposure Anal Environ Epidemiol 1998: 8: 495–518. Jantunen M.J., Katsouyanni K., Kno¨ppel H., Ku¨nzli N., Lebret E., Maroni M., Saarela K., Sra´m R., and Zmirou D. Final Report: Air Pollution Exposure in European Cities: the EXPOLIS Study. Publications of KTL B16/1999:127, pp. 4 Annexes, 1999. Jungers R.H., Akland G.G., and Sauls H.B. A human exposure assessment study using personal monitoring of carbon monoxide. Direct and indirect methods. Environ Int 1985: 11: 413–418. Klepeis N.E. An introduction to the indirect exposure assessment approach: modelling human exposure using microenvironmental

321


measurements and the recent National Human Activity Pattern Survey. Environ Health Persp 1999: 107(2): 365–374. KTL. CO monitoring. EXPOLIS Standard Operation Procedure. KTL, Kuopio, Finland, 1997a. KTL. CO monitoring and QA. EXPOLIS Standard Operation Procedure. KTL, Kuopio, Finland, 1997b. Lee K., Yanagisawa Y., Spengler J.D., and Billick I.H. Measurement of personal carbon monoxide exposures by mailed passive samplers. J Air Waste Manage Assoc 1992: 42: 1212–1213. Maroni M., Seifert B., and Lindvall T. Indoor Air Quality. A Comprehensive Reference Book. Elsevier Health Sciences, Amsterdam, 1995 ISBN: 0444816429. Maroni M., Carrer P., Cavallo D., Alcini D., Basso A., Preti G., Vercelli F., Visigalli F., and Bersani M. Characterisation of personal exposure to air pollutants of subjects living in Milan. Proceedings of the 7th International Conference on Indoor Air Quality and Climate, Nagoya, Japan, 1996. Moriske H.J., Drews M., Ebert G., Menk G., Scheller C., Scho¨ndube M., and Konieczny L. Indoor air pollution by different heating systems: coal burning, open fireplace and central heating. Toxicol Lett 1996: 88: 349–354. Naeher L.P., Leaderer B.P., and Smith K.R. Particulate matter carbon monoxide in highland Guatemala: indoor and outdoor levels from traditional and improved wood stoves and gas stoves. Indoor Air 2000: 10(3): 200–205. Nagda N.L., and Koontz M.D. Microenvironmental and total exposures to carbon monoxide for three population subgroups. J Air Pollution Control Assoc 1985: 35: 134–137. Ott W.R., Thomas J., Mage D., and Wallace L. Validation of the simulation of human activity and pollutant exposure (SHAPE) model using paired days from the Denver, CO, carbon monoxide field study. Atm Environ 1988: 22(10): 2101–2113.

322


Rotko T., Oglesby L., Ku¨nzli N., and Jantunen M.J. Population sampling in European air pollution exposure study, EXPOLIS: Comparisons between the cities representativity of the samples. J Exposure Anal Environ Epidemiol 2000: 10: 355–364. Siegel M. Involuntary smoking in the restaurant workplace. J Air Waste Manage Assoc 1993: 270(4): 490–494. Thomas K.W., Pellizari E.D., Clayton A., Whitaker D.A., Shores R.C., Spengler J., Ozkaynak H., Froehlich S.E., and Wallace L.A. Particle total exposure assessment methodology (PTEAM) 1990 study: method performance and data quality for personal, indoor and out-door monitoring. J Exposure Anal Environ Epidemiol 1993: 3: 203–226. Touloumi G., Samoli E., and Katsouyanni K. Daily mortality, and ‘‘winter type’’ air pollution in Athens, Greece F a time series analysis within the APHEA project. J Epidemiol Commun Health 1996: 50(1): 47–51. Vellopoulou A.V., and Ashmore M.R. Personal exposures to carbon monoxide in the city of Athens: commuters exposures. Environ Int 1998: 24: 713–720. Wallace L., Thomas J., Mage D., and Ott W. Comparison of breath CO, CO exposure and Coburn model prediction in the US. EPA Washington–Denver (CO) study. Atm Environ 1988: 22: 2183–2193. Wallace L.A. Recent field studies of personal and indoor exposures to environmental pollutants. Ann NY Acad of Sci 1992: 641: 7–16. Wallace L. Real-time monitoring of particles, PAH, and CO in an occupied townhouse. Appl Occup Environ Hygiene 2000: 15(1): 39–47. WHO-a. Guidelines for Air Quality, WHO document. Geneva, WHO/ SDE/OEH/00.02. 2000. WHO-b. Air Quality Guidelines for Europe, 2nd edn. WHO Regional Publications, European Series, No.91. ISBN 92 890 1358 3. Ziskind R.A., Fite K., and Mage D.T. Pilot field study: carbon monoxide exposure monitoring in the general population. Environ Int 1982: 8: 283–293.
