Science of the Total Environment 520 (2015) 160–167

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

An approach to using heart rate monitoring to estimate the ventilation and load of air pollution exposure☆ Izabela Campos Cozza a,⁎, Dirce Maria Trevisan Zanetta b, Frederico Leon Arrabal Fernandes a, Francisco Marcelo Monteiro da Rocha c, Paulo Afonso de Andre d, Maria Lúcia Bueno Garcia e, Renato Batista Paceli a, Gustavo Faibischew Prado a, Mario Terra-Filho a, Paulo Hilário do Nascimento Saldiva f, Ubiratan de Paula Santos a a

Pulmonary Division, Heart Institute (Incor), Hospital das Clínicas, Medicine School, University of São Paulo, Brazil Epidemiology Department, Public Health School, University of São Paulo, Brazil c Science and Technology Institute, UNIFESP, São José dos Campos, SP, Brazil d Experimental Laboratory of Atmospheric Pollution, Pathology Department, Medicine School, University of São Paulo, Brazil e Clinical Medicine Department, Medicine School, University of São Paulo, Brazil f Pathology Department Professor, Medicine School, University of São Paulo, Brazil b

H I G H L I G H T S • • • •

Ventilation estimate has a good correlation with data obtained in the laboratory. Statistical models were tested to estimate ventilation based on heart rate. Wide variations were found in the regression equations between individuals. A statistical model is suitable for situations of heart rate, under 100 beats.

a r t i c l e

i n f o

Article history: Received 28 September 2014 Received in revised form 30 January 2015 Accepted 13 March 2015 Available online xxxx Editor: Lidia Morawska Keywords: Air pollution Ventilation Heart rate Inhaled load estimate Personal exposure

a b s t r a c t Background: The effects of air pollution on health are associated with the amount of pollutants inhaled which depends on the environmental concentration and the inhaled air volume. It has not been clear whether statistical models of the relationship between heart rate and ventilation obtained using laboratory cardiopulmonary exercise test (CPET) can be applied to an external group to estimate ventilation. Objectives: To develop and evaluate a model to estimate respiratory ventilation based on heart rate for inhaled load of pollutant assessment in field studies. Methods: Sixty non-smoking men; 43 public street workers (public street group) and 17 employees of the Forest Institute (park group) performed a maximum cardiopulmonary exercise test (CPET). Regression equation models were constructed with the heart rate and natural logarithmic of minute ventilation data obtained on CPET. Ten individuals were chosen randomly (public street group) and were used for external validation of the models (test group). All subjects also underwent heart rate register, and particulate matter (PM2.5) monitoring for a 24-hour period. Results: For the public street group, the median difference between estimated and observed data was 0.5 (CI 95% −0.2 to 1.4) l/min and for the park group was 0.2 (CI 95% −0.2 to 1.2) l/min. In the test group, estimated values were smaller than the ones observed in the CPET, with a median difference of −2.4 (CI 95% −4.2 to −1.8) l/min. The mixed model estimated values suggest that this model is suitable for situations in which heart rate is around 120–140 bpm. Conclusion: The mixed effect model is suitable for ventilation estimate, with good accuracy when applied to homogeneous groups, suggesting that, in this case, the model could be used in field studies to estimate ventilation. A small but significant difference in the median of external validation estimates was observed, suggesting that the applicability of the model to external groups needs further evaluation. © 2015 Elsevier B.V. All rights reserved.

Abbreviations: PM2.5, particulate matter less than 2.5 μm; logVE, logarithm of ventilation; CI, confidence interval. ☆ Funded by the Technology Science Ministry–National Counsel of Technological and Scientific Development (Ministério da Ciência, Tecnologia e Inovação–Conselho Nacional de Desenvolvimento Científico e Tecnológico, MCTI–CNPq), Edital 18. Processo CNPq 555223/2006-0. ⁎ Corresponding author at: Pulmonology Division, Heart Institute Hospital das Clínicas, Medicine School, University of São Paulo, Brazil. E-mail address: [email protected] (I.C. Cozza).

http://dx.doi.org/10.1016/j.scitotenv.2015.03.049 0048-9697/© 2015 Elsevier B.V. All rights reserved.

I.C. Cozza et al. / Science of the Total Environment 520 (2015) 160–167

1. Introduction

2. Methods

Short- and long-term exposures to vehicular pollution are associated with increased morbidity and mortality, primarily due to cardiorespiratory diseases (Brunekreef et al., 2009; HEI, 2010; Krewski et al., 2009). A comparative risk assessment of 21 regions of the world indicated that from 1990–2010, approximately 3.2 million lives (6.1% of all fatalities) were lost and 76 million (3.1%) disability adjusted life years (DALYs) could be attributed to particulate matter (PM 2.5 ) pollution (Lim et al., 2012). The main source of pollution in urban centers is vehicular. To estimate the effects of pollution on health, most studies, including both cohort (Pope et al., 2002) and time series (Santos et al., 2008), have used fixed samplers. Almost 20 years ago, a study conducted in England (Watt et al., 1995) compared the use of personal air sampling versus fixed samplers and showed higher values in the individual samplers. This study was followed by others which compared the use of individual samplers (Brunekreef et al., 2009; Jansen et al., 2005; de Hartog et al., 2010). Since then, several studies have been published to obtain a better individual pollutant exposure estimate; these studies have been used, both for general population (Dons et al., 2012) and for specific groups, such as policemen, taxi drivers, couriers, and traffic agents (Lioy et al., 2011; Tomei et al., 2001; van Roosbroeck et al., 2006). Despite the fact that fixed samplers are still widely used because of their lower costs, under certain conditions their estimates do not appear to differ significantly from personal samplers in individuals who were not exposed to environmental tobacco smoke (Brunekreef et al., 2005). However, most studies do not take into account the performance of different physical activities during the day, which determines the ventilation variation and the inhaled air volume. It is difficult to measure ventilation in field studies. To circumvent this limitation, some studies (Mermier et al., 1993; Samet et al., 1993) have attempted to estimate ventilation using heart rate, which is easily measured by portable heart rate monitors. Although heart rate is influenced by factors such as temperature, time, and stress, heart rate displays a good correlation with oxygen consumption and ventilation (Samet et al., 1993; Zuurbier et al., 2009). This ultimately allows for the estimation of the exposure load per individual and provides an association with biological indicators (blood, lung function, heart rate, etc.). Studies have found (Mermier et al., 1993; Zuurbier et al., 2009) that the equation obtained using the ventilation and heart rate data of each individual provides a more accurate estimate of individual ventilation, but equations developed based on a group of individuals can also be used to estimate ventilation in population groups. However, these studies (Mermier et al., 1993; Zuurbier et al., 2009) did not test for the model's external validity and some doubts still remain about whether the equations developed from cardiopulmonary exercise test (CPET) in the laboratory can be used to estimate ventilation in other individuals with similar activities using 24-hour heart rate monitoring. The results found in the study aforementioned (Zuurbier et al., 2009) could not be compared with the results found in Belgium (Int Panis et al., 2011). Direct ventilation measurements were performed in the field and the study showed that ventilation in cyclists was 4.3 times higher than in car passengers (Int Panis et al., 2010), which was two times higher than ventilation measurements found in The Netherlands (Zuurbier et al., 2009). This could be partly explained by the difference in cycling speed, 12 km/h (Zuurbier et al., 2009) and 20 km/h (Int Panis et al., 2011), but also, the method of direct ventilation measurement may have had high influence. The aim of this study was to develop and evaluate models to estimate ventilation, based on heart rate, which can be used to estimate the inhaled load of pollutants, in field studies.

2.1. Assessed individuals

161

The participants in this study took part in a larger study, titled “Evaluation of pollution toxicity by particulate matter generated by different emission sources: clinical and experimental studies proposition”, with partially published results (Torricelli et al., 2013). This study included male nonsmokers or those who had quit smoking for at least one year and aged between 18 and 65 years. The non-smoking status was confirmed by exhaled CO measures, determined in ppm by a Micro CO Meter, Micro Medical Limited (Rochester, England). Individuals with different exposures to air pollution were selected. Employees of the Forest Institute, which is located in a park 13 km distant from central São Paulo City, Brazil, with low pollution exposure, and traffic controllers and taxi drivers, who worked in the broad central section of São Paulo, with high exposure to vehicular pollution, were recruited during workplace meetings or through newspaper advertisements. The individuals who met the inclusion criteria answered a questionnaire on working conditions, demographic data and comorbidity history, followed by clinical examination. Individuals with medical conditions that would not allow the evaluations to be accomplished were excluded. Within the 65 selected individuals who were evaluated, eight were traffic controllers, 38 were taxi drivers and 19 were park rangers. Five individuals (three taxi drivers and two park rangers) were excluded, because they could not finish CPET: four presented a diastolic blood pressure higher than 120 mm Hg and one subject presented reduction in the systolic blood pressure higher than 20 mm Hg. The studied individuals were divided in 3 groups: the park group, consisted of park rangers; the public street group, consisted of traffic controllers and taxi drivers; and the test group, consisting of 10 individuals randomly selected from public street group to evaluate the consistency of the models evaluated in the study. In this way, 17, 33 and 10 individuals were evaluated in park, public street and test group, respectively. All of the participants agreed to participate in the study and signed an informed consent form. The study was approved by the Ethics Committee for Research Projects Analysis of the Hospital das Clínicas, Medical School, University of São Paulo (CAPPesq 0565/07). 2.2. Study protocol The individuals were selected and evaluated between October 2010 and June 2012. They underwent assessments on two different and consecutive days during the morning. On the first day a technician installed the portable heart rate monitor and the PM2.5 individual sampler for recording in the field over a 24-hour period. On the second day the equipment was removed, they answered a questionnaire concerning the occurrence of any cardiopulmonary complications that could interfere with the tests, underwent a spirometry, and performed the CPET. 2.3. Pulmonary function test Spirometry was performed with a KoKo spirometer (Pulmonary Data Services Instrumentation Inc., Louisville, USA) according to the ATS/ERS recommendations (Miller et al., 2005a, 2005b). We applied the predicted values of normality for a Brazilian population (Pereira et al., 2007) and used the interpretive criteria recommended by the ATS/ERS (Pellegrino et al., 2005). 2.4. Cardiopulmonary exercise test All of the subjects were assessed in the morning. After a five-minute rest, each volunteer was positioned on an electromagnetic brake cycle ergometer (Encore Vmax 29S model, VIASYS, USA). The test began

162

I.C. Cozza et al. / Science of the Total Environment 520 (2015) 160–167

with a rest period of 2 min and 2 min of cycling without load, followed by a ramp protocol with a 20-Watt increment per minute increasing until the tolerance limit. Metabolic, ventilatory, and cardiovascular variables were analyzed breath-by-breath. Blood pressure values and oxygen saturation values were determined every 2 min. For each examination, heart rate and ventilation data were extracted and recorded every 3 s from the initial stage until the recovery phase. The tests lasted for 11–16 min, on average. The tolerance limit was defined by the participant when he determined that he had achieved maximum effort, by the technician when he perceived the patient's inability to sustain the effort, or when the following criteria for test interruption, according to the ATS/ACCP (ATS/ ACCP, 2003), arose: chest pain or electrocardiogram changes suggestive of ischemia, complex ectopy, second or third degree heart block, a drop of 20 mm Hg or more in systolic pressure, hypertension (250 mm Hg systolic; 120 mm Hg diastolic), severe desaturation (SpO2 b 80%), sudden pallor, loss of coordination, mental confusion, dizziness, faintness and/or signs of respiratory failure. Ventilation and heart rate data obtained during the CPET for each individual were used to define the equations for ventilation estimation using the heart rate monitor. 2.5. Heart rate recordings in the field Twenty-four-hour heart rate data were obtained and recorded during the usual daily activities of an ordinary workday of the assessed individuals using a portable heart rate monitor (Polar Electro Oy-RS800CX model, Professionite 5, FI-90440 Kempele, Finland). These records were used to estimate ventilation by using the equations designed to analyze the relationship of heart rate to ventilation, obtained from the CPET. 2.6. PM2.5 concentration estimations with individual samplers The daily average PM2.5 concentration for each individual was obtained from the sampler in the filters. The sampler, designed by the Harvard School of Public Health, operates at a flow rate of 4 lpm through an impaction plate to obtain a cutoff of PM2.5. A silicone catheter connects the air inlet positioned at the volunteer shoulder to the sampler inflow. A SKF vacuum pump, model WR-5000 Aircheck, that incorporates a flow control and a chronometer, to estimate the total volume of sampled air, provides the necessary continuous air flow during the 24hour sampling period. A polycarbonate membrane, a Whatman filter with a diameter of 37 mm and 0.8 pore size (part number 110809), installed inside the sampler after the impaction plate retains the particulate matter of the sampled PM2.5. This membrane was weighed before and after the sampling process in an accurate 1 μg Mettler Toledo scale, model UMX2, following a laboratorial protocol developed to control temperature and humidity, allowing to estimate the total mass of particulate matter collected during the sampling period. The estimated average concentration of PM2.5 was obtained by a gravimetric method (total mass divided by total air volume sampled) as described by de Miranda et al., 2012.

2.8. Statistical analyses The data were analyzed and are presented as mean ± standard deviation, or median and 95% confidence intervals. For non-normally distributed variables, the confidence intervals were estimated by the bootstrap technique with the percentiles 2.5 and 97.5 of 10,000 subsamples obtained by random selection of subjects with replacement. The significance level was considered as 5%. Comparisons between variables were made using t-test, Mann–Whitney U test and Fisher's exact test or by confidence intervals evaluation. The ventilation and heart rate records obtained in the CPET were analyzed using two different models. Following the curvilinear relationship between heart rate and ventilation, regression equations were calculated using the natural log transformation of ventilation. In Model 1 (mean of individual equations), linear regressions of the logarithm of ventilation (logVE) (dependent variable) and heart rate were performed for each individual using the minute-by-minute information obtained during the CPET. The means of the estimated intercepts and slopes for each individual in the public street group and the park group were calculated. In this model, the correlation between the two variables was calculated to evaluate the predictive ability of ventilation estimation using heart rate measurements. In Model 2 (mixed model), using a linear mixed effects model with all individuals, the fixed part evaluated the association of the logVE (dependent variable) and heart rate on a minute-by-minute information basis, controlling for the effects of age, height and body mass index. There was no significant effect of these variables on the relation between logVE and heart rate, and no main effects were observed; therefore, they were not included in the final model. The graph of the profiles of logVE as a function of heart rate suggested one line for each group (public street and park). To model the covariance structure, we analyzed the individual and mean profile graphics of logVE according to time for each group and variance component test (Stram and Lee, 1994). For model selection, the Akaike Information Criterion was used. We built a curve with confidence bands for the two groups. The fixed model effect's significance was evaluated using the Wald test. The heart rates obtained during CPET of individuals from the public street and park group was used in the two models described above and constructed for each group, to estimate their respective ventilation. The heart rates of the 10 individuals in the test group, whose data were not considered while setting up the equations, were applied in the two models constructed with the public street group data, for the ventilation to assess the model's external validity. For each group, the difference between estimated and observed values of ventilation was calculated and the 95% confidence interval of the median of measurements is presented to evaluate the model's adequacy. The analysis was performed using S-Plus software v.8 Statistics (Data Analysis Products Division, MathSoft Inc., Seattle, Washington, USA) and R 2.15.3 by means of the nlme computational library (Linear and Nonlinear Mixed Effects Models. Pinheiro J, Bates D, debroy S, Sarkar D and R Core Team. R package version 3.1-110. 2013).

2.7. Inhaled load estimate For each participant, the PM2.5 inhaled load was estimated in μg using the formula below:   Obtained concentration 3 μg=m  estimated ventilation l= min 1000  60 min  24 hours: The concentration (μg/m3) was obtained using the individual sampler, and ventilation was determined by the CPET and estimated through the regression models.

3. Results 3.1. Study population Table 1 presents the subjects' characteristics, exhaled carbon monoxide values and average PM2.5 concentrations in each group. The public street group reported fewer work years (p = 0.03), greater daily work shifts (p = 0.07) and a higher exposure to PM2.5 concentrations (p b 0.001). The remaining variables did not have significant differences. There were no differences between the public street group and the test group in those variables.

I.C. Cozza et al. / Science of the Total Environment 520 (2015) 160–167

3.2. Cardiopulmonary exercise testing The group data comparison concerning the variables obtained at the peak of the CPET is presented in Table 1. The VO2 ml·kg−1·min−1, and VO2 l/min, were higher in the park group; p = 0.002 and p = 0.031 respectively, and the VE/VO2 ratio was higher in the public street group; p = 0.031. There were no significant differences between the public street group and the test group. 3.3. Ventilation and heart rate relationship The linear regression equations between heart rate and logVE were calculated for each individual of the 33 public street group and the 17 park group individuals. The mean determination coefficient between heart rate and ventilation was high in both groups: R2 = 0.94 (±0.07) in the public street group and R2 = 0.96 (±0.04) in the park group (Table 2). To compare, we have also estimated the test group linear regression, which resulted in a mean determination coefficient: R2 = 0.95 (±0.04). For model 2 construction, we have first evaluated the individual profile of ventilation logarithm versus heart rate (Fig. 1, top) which suggested a linear relationship between these two variables. A model with intercept random effect (Model 1) and another one, with intercept and slope random effects (Model 2) were evaluated. The comparison between models showed a better fit for Model 2 (public street group: AIC − 49,6 and −266,3 and park group: AIC − 225,5 and − 253,2, for

163

Model 1 and Model 2 respectively). The variance test component also showed that for both groups we should use Model 2, with intercept and slope random effects (p-value b0.001). Table 2 reports the regression coefficients of both models for the public street group and the park group. There is a minor difference between models considering the intercept values. There are no differences in the slopes between models and between groups. Table 3 shows the estimated ventilation values separately for the public street group, the park group, and the test group, as well as the differences between the estimated values and those observed in the CPET. The median differences obtained using Model 1 were different from zero in the three groups. Using Model 2, the ventilation estimate data of the public street and park group did not differ from the ventilation observed in the CPET, on the other hand, in the test group the median difference of the individual calculations was −2.4 l/min. Fig. 1, bottom, shows the individual linear regression lines of the ventilation measurements by heart rate in the public street group and in the park group. These graphics show the curvilinear relationship between ventilation and heart rate and a great variation between individuals. Such variation was lower in the park group. Five individuals of the public street group and four from the park group had gaps in heart rate and ventilation records in the initial resting phase of the CPET. Due to these gaps, their regression lines started with heart rates above 100 bpm. Fig. 2 shows the linear regression lines estimated with the mixed model for the ventilation estimate using the heart rate measurements

Table 1 Individuals' general characteristics, spirometry values, work characteristics, pollutant measures and cardiopulmonary exercise test variables, allocated into groups. Footnote: COex: carbon monoxide in exhaled air; ppm: parts per million; FVC (%) = percentage of predicted forced vital capacity; FEV1 (%) = percentage of predicted forced expiratory volume in the first second; PM2.5 = particulate matter; VE = minute ventilation; VO2 = oxygen consumed volume; VCO2 = carbon dioxide exhaled volume; VE/VO2 = minute ventilation/oxygen consumption ratio; VE/VCO2 = minute ventilation/carbon dioxide production; heart rate (%) = percentage of predicted heart rate; SpO2 = peripheral oxygen aturation; pO2 = oxygen pulmonary pressure; bpm = beats per minute. Values are expressed as prevalence, mean or *median (95% CI); £Fisher Exact test; #T-test; &Mann–Whitney U test. The ventilation and heart rate records are obtained in the CPET.

Subjects characteristics Age (years) Body mass index (kg/m2)* Systolic blood pressure (mm Hg)* Diastolic blood pressure (mm Hg)* COex (ppm)* Heart rate (bpm) Environmental tobacco smoke At home — n (%) At work — n (%)£

Public street group (n = 33)

Park group (n = 17)

Test group (n = 10)

45.0 (41.7–48.3) 27.9 (26.7–29.7) 130.0 (120.0–130.0) 90.0 (80.0–90.0) 2.0 (1.0–4.0) 68.4 (64.9–71.9)

48.5 (44.1–52.9) 27.0 (23.5–28.9) 125.0 (115.0–130.0) 85.0 (75.0–95.0) 1.0 (0.0–1.0) 68.2 (62.4–74)

49.9 (42.0–57.8) 26.9 (24.3–28.8) 115.0 (103.0–130.0) 80.0 (79.5–90.0) 1.0 (0.5–4.5) 63.5 (57.2–74.4)

2 (6) 18 (54.5)

2 (11.8) 5 (29.4)

2 (20) 6 (60)

Spirometry FVC (%) FEV1 (%)* FEV1/FVC (%)

94.3 (90.6–98.0) 91.4 (86.9–95.2) 79.9 (81.2–78.7)

96.9 (91.0–102.8) 95.5 (91.3–99.9) 79.9 (80.9–79.0)

96.3 (84.8–107.8) 89.8 (74.6–98.5) 75.7 (78.4–73.0)

Work characteristics Total work time (years) Work shift (hours)*

8.1 (6.2–11.0) 14.0 (12.0–14.0)

13.5 (9.6–17.5) 9.0 (9.0–9.0)

15.4 (6.7–24.1) 14.0 (7.0–16.0)

Pollutant measures PM2.5 exposure (μg/m3)*

31.7 (30.5–36.3)

19.7 (17.2–24.6)

83.1 (76.3–89.1) 25.1 (22.8–27.3) 2.0 (1.9–2.2) 2.7 (2.5–2.9) 40.5 (23.1–57.9) 31.0 (29.5–32.5) 164.0 (157.0–169.0) 91.4 (88.0–94.7) 196.3 (190.0–206.6) 90.0 (90.0–100.0) 96.0 (96.0–97.0) 103.3 (101.6–105.0)

80.6 (70.5–90.6) 29.6 (26.0–33.2) 2.2 (2.1–2.4) 2.7 (2.4–3.0) 34.9 (24.4–45.4) 29.5 (27.8–31.3) 162.0 (155.0–166.0) 93.6 (89.6–97.9) 190.3 (177.5–203.1) 100.0 (90.0–120.0) 97.0 (96.0–98.0) 102.4 (100.1–104.8)

Cardiopulmonary exercise test Exercise peak values Ve (l/min) VO2 (ml·kg−1·min−1)# VO2 (l/min)*,& VCO2 (l/min) VE/VO# 2 VE/VCO2 Heart rate (bpm)* Heart rate (%) Systolic blood pressure (mm Hg) Diastolic blood pressure (mm Hg)* SpO2 (%)* pO2 (mm Hg)

(24.1–44.6)

84.0 (68.3–99.7) 26.1 (22.4–29.8) 2.1 (1.9–2.6) 2.7 (2.3–3.1) 39.5 (23.3–55.7) 30.9 (27.9–33.9) 149.0 (123.0–168.0) 84.2 (73.4–95.0) 188.7 (163.8–213.6) 85.0 (80.0–110.0) 96.5 (96.0–97.0) 103.1 (99.9–106.2)

164

I.C. Cozza et al. / Science of the Total Environment 520 (2015) 160–167

Table 2 Regression model results of minute ventilation on heart rate, obtained during cardiopulmonary exercise test in public street group and park group. Footnote: Model 1 = equation with the average regression parameters of each individual regression; Model 2 = mixed model equation; SE = standard error; R2 = determination coefficient. The regression equations of between the minute ventilation measurements on heart rate were calculated using the natural log transformation of ventilation. R2

Model 1

Intercept

Slope

Mean (SD)

Range

Mean (SD)

Range

Mean (SD)

Range

Public street group (n = 33) Park group (n = 17)

0.54 (0.95) 0.40 (0.52)

(−2.00 to 2.25) (−0.82 to 1.06)

0.025 (0.009) 0.025 (0.004)

(0.014–0.057) (0.019–0.041)

0.94 (0.07) 0.96 (0.04)

(0.62–0.99) (0.85–0.99)

Model 2

Intercept (SE)

Slope (SE)

Public street group (n = 33) Park group (n = 17)

0.58 (0.16) 0.51 (0.10)

0.025 (0.002) 0.025 (0.001)

obtained in the CPET for the public street group and the park group as well as the confidence bands of both curves. Even for higher heart rate values, we can observe narrow confidence bands, thus, the mixed model use could be better for heart rates below 120 bpm and for the park group. Table 4 presents the ventilation estimated values for the public street group, the park group, and the test group for both models using the mean heart rate values obtained during a 24-hour period by a portable frequency monitor during the usual daily activities of the assessed individuals. No significant differences were identified between the groups in the 24-hour heart rate recordings.

3.4. Pollutant load inhaled calculations As an example, the models were applied to obtain the inhaled load estimate, from estimated ventilations. The concentration PM2.5 values are obtained in the individual samplers during the assessed days in the three groups. The concentrations of PM2.5 in the park group were significantly lower than those in the public street group (Table 1). Table 5 shows the PM2.5 concentration estimates using the individual samplers and the inhaled load calculations obtained using the ventilation estimates. When comparing the ratio between the inhaled load in the public street group and the park group

Fig. 1. Graphic of individual profiles of logarithm of ventilation as a function of heart rate with lowess curve superimposed (top) and regression lines (bottom) of individuals of public street and park groups. Footnote: Data obtained from the cardiopulmonary exercise test of the public street group and the park group; the thick lines of the graphics of profile (top) are the lowess curve.

I.C. Cozza et al. / Science of the Total Environment 520 (2015) 160–167

165

Table 3 Ventilation results obtained from the cardiopulmonary exercise test and ventilation estimates through the regression models. Footnote: From the heart rate measurements of each group registered during the cardiopulmonary exercise test, ventilation was estimated using the two regression models. amedian difference of the ventilation estimated and ventilation obtained in the cardiopulmonary exercise test calculated for measurement of the individuals, on a minute-by-minute; Model 1 = equation with the average regression parameters of each individual; Model 2 = mixed model equation; heart rate = average heart rate observed during the cardiopulmonary exercise test; bpm = beats per minute. The values are expressed as the mean ± standard deviation or median (95% confidence interval). Groups Cardiopulmonary exercise test Public street group (n = 33) Park group (n = 17) Test group (n = 10) Model 1 Public street group (n = 33) Park group (n = 17) Test group (n = 10) Model 2 Public street group (n = 33) Park group (n = 17) Test group (n = 10)

Heart RATE (bpm) 116.5 ± 28.6 116.0 ± 25.0 104.1 ± 29.2 116.5 ± 28.6 116.0 ± 25.0 104.1 ± 29.2 116.5 ± 28.6 116.0 ± 25.0 104.1 ± 29.2

(524.3/375.3 mg = 1.4) and the ratio between the environmental concentration data in the same groups (31.7/19.7 mg/m3 = 1.6), there was a variation in these proportions, suggesting that ventilation should be taken into account when calculating inhaled load. 4. Discussion In this study, we evaluated different groups of individuals, and we showed that statistical models constructed from heart rate and ventilation data obtained in the laboratory could estimate ventilation satisfactorily for each assessed group. Therefore, these models can be employed in field studies to estimate ventilation, making it possible to estimate the inhaled pollutant load, which can contribute to improving the assessment of the effect of pollutants on health. On the other hand, the external validation showed significant difference between the observed ventilation measured in laboratory (CPET) and the ventilation estimated. Although the median difference was only of 2.4 l/min, this can limit the model employment on population studies. Other studies (Mermier et al., 1993; Zuurbier et al., 2009) have already addressed the issue, but this is the first published study that tested statistical models with an external group (test group). There are limitations on the extensive use of equations constructed using a specific group of people over another group that is not involved in the analysis. However, when we applied on the 24-hour heart rate

Ventilation (l/min) Observed data 30.1 (26.7–35.3) 30.4 (28.1–33.0) 28.0 (22.6–36.2) Estimated data 26.3 (23.3–29.8) 28.7 (26.2–31.4) 21.1 (16.5–25.2) Estimated data 27.2 (24.8–29.7) 28.6 (25.5–32.4) 21.7 (17.0–25.9)

Differencea (l/min) – – – 1.7 (0.8 to 2.5) −1.8 (−2.3 to −0.9) −2.9 (−4.7 to −2.2) 0.5 (−0.2 to 1.4) 0.2 (−0.2 to 1.2) −2.4 (−4.2 to −1.8)

monitoring, in real life, the differences between the test group and the public street group were small (Table 4). The use of the models may be appropriate, especially for heart rates lower than 120 bpm, reflecting the majority of an individual's everyday activities (Fig. 2). The differences in the relationship between heart rate and ventilation observed among assessed groups may be also associated with different levels of activity and fitness and the presence of comorbidities observed in those different groups. As can be seen in Fig. 1, park group presented best performing in the exercise test (better VO2max, although without difference in VE/VCO2) and had less variation in the correlations and regression lines between ventilation and heart rate. The high correlation between heart rate and minute ventilation found in the present study was similar to the ones found in previous studies (Zuurbier et al., 2009; Samet et al., 1993). In these studies, the authors suggest that equation models estimated for a sample of a group could be used for estimating the ventilation in the total groups, even though they did not test the equation's validity in an external group as we did in this study. The mixed effect model presented a better performance, when comparing the two tested models. It showed very similar ventilation estimates in both, the public street and park group, compared to the ventilation observed in the CPET, which shows a good internal validation. However, in the other model tested, the ventilation estimate was

Fig. 2. Graphics of the ventilation estimates (with the 95% confidence bands) from the heart rate measurements obtained from the cardiopulmonary exercise test for the public street group and the park group, obtained from the mixed model.

166

I.C. Cozza et al. / Science of the Total Environment 520 (2015) 160–167

Table 4 Ventilation estimated in each group using different equation models, using the 24-hour heart rate register. Footnote: Model 1 = equation with the average regression parameters of each individual regression; Model 2 = mixed model equation; HR = 24 hour HR registered during usual daily activities of the subjects (Polar register); bpm = beats per minute. The values are expressed as the mean or *median (95% confidence interval). Estimated ventilation (l/min) and 24 hour heart rate

Heart rate (bpm) Ventilation — Model 1* Ventilation — Model 2*

Public street group (n = 26)

Park group (n = 17)

Test group (n = 7)

80.0 (77.5–85.0) 12.9 (12.2–14.9) 13.2 (12.4–15.2)

75.0 (70.0–88.0) 10.3 (9.1–13.7) 10.9 (9.6–14.3)

74.0 (66.0–92.0) 11.1 (9.1–17.5) 11.4 (9.3–17.8)

different comparing the two groups and the observed values. In the present study, in the mixed model, the intercept (fixed effect) was smaller and the slope was similar to other study findings (Samet et al., 1993; Zuurbier et al., 2009). The significant difference between observed values in the CPET and the ones estimated by the two models, on the test group, suggests a limitation on the external validity of the models (Table 3). The fact that the bands of the regression curves between the heart rate and ventilation are narrow (Fig. 2) and the high correlation between heart rate and minute ventilation (R2 N 0.9), strengthens the possibility of using the equations built into a group, lab data, being used in groups outside, thus allowing a best estimate ventilation and pollutant load in studies with larger numbers of participants, reinforcing what other authors have suggested (Zuurbier et al., 2009). Despite its importance there are few studies that consider the role of ventilation. A study conducted in Ireland (O'Donoghue et al., 2007) compared the exposure and inhaled hydrocarbon loads in individuals who used busses and bicycles for transportation and evaluated the respiration rate. The study found that although the level of exposure in bus users was high, the inhaled load was even higher in cyclists because of the effect of physical exercise on ventilation. The importance of ventilation was also suggested in a study conducted in Augsburg, Germany, that evaluated the association between traffic exposure and symptoms of heart attack, with a greater risk observed in cyclists compared to the users of other means of transportation (Peters et al., 2004). Although other factors such as stress, drugs and weather can influence the HR, it is mainly influenced by the consumption of oxygen, which has a high correlation with ventilation. Because it is easily measured in the field and have low cost, the HR monitoring constitutes a good tool for population studies, especially to evaluate different groups of individuals, with different physical activities and commutes in cities. The inhaled load estimate was also calculated in a larger study called SHAPES project (Systematic Analyses of Health Risks Associated with Physical Activity and Police Cycling) (Int Panis et al., 2011), involving cyclists and individuals who used cars for transportation. Direct ventilation measurements were made using personal portable CPET equipment. However, it is difficult to measure ventilation in field studies, and it is still necessary to assess whether ventilation and load exposure estimates, using equations developed for a particular group, may be superior for assessing biological marker effects than the simple use of PM2.5 concentrations obtained with individual samplers (Zuurbier et al., 2011a, 2011b).

Data observed in other studies (O'Donoghue et al., 2007; Nwokoro et al., 2012) have shown how relevant the influence of different daily activities is on ventilation that influences the inhaled pollutant load. These observations give plausibility to the assumption that individual exposure and not just environmental concentration estimates may create more accurate studies on the effects of pollution in health. The use of individual samplers, as used in our study, may better represent the exposure of a particular individual and thus could better estimate health effects. This better approach has been suggested by several studies (Brook et al., 2011; Jacobs et al., 2010) that have compared the records of individual and fixed samplers and identified differences between the sampled values (21.9 ± 24.8 μg/m3 and 15.4 ± 7.5 μg/m3, respectively) with measurements in the same region. The study (Brook et al., 2011) also showed that personal monitoring was able to identify the PM2.5 effects on blood pressure, which was not identified when using the fixed environmental samplers on the same period. With the use of individual samplers, this study also observed differences in PM2.5 exposure among the groups, as well as in the estimated inhaled loads by using a heart rate monitoring in the field; this study highlighted the important role of activity on the ventilation and consequently on the inhaled load of the pollutants. The individual PM2.5 concentrations registered in the public street group ranged from 8.5 μg/m3 to 84.0 μg/m3, with a median of 31.7 μg/m3; this result was higher than the amounts observed in the park group that ranged from 9.8 μg/m 3 to 38.7 μg/m 3 , with a median of 19.7 μg/m 3 . Although our records were conducted for 24 h with equipment that did not allow for the exposure concentrations to be apportioned to shorter time windows (e.g., sleeping time, commuting time, working time, resting time, and exposure time to environmental tobacco smoke in residences), the average values were still noteworthy. The significant differences observed between groups in some variables on CPET is modest and may be associated with the fact that the park group individuals were more active, with regular daily walking in their work environment, between 2 and 4 km. Thus, the park group VO2 ml·kg 1·min− 1 amounts were higher, and as has been well documented in the literature, the practice of regular moderate exercise promotes improved aerobic performance (Marshall et al., 2001). The difference in the relationship between the concentrations obtained (31.6/19.7 μg/m3 = 1.60) and the ratio of the inhaled loads (524.3/375.3 μg/m3 = 1.40) between the public street group and the park group suggests the importance of considering ventilation to better estimate the ratio of the real exposure and inhealed load and its possible effects. As mentioned before, another limitation of this study was the sampler model used for PM2.5 concentration measurements, because it was not possible to separate registers in different periods, such as work or rest, daytime or nighttime. The only possibility was to register the whole recording period, or 24 h. Thus, the estimates made using the 24-hour average concentration, are inaccurate for calculating the PM2.5 load inhaled described in the article to illustrate the potential use of the model. To solve that, in further studies, another PM monitor model that fractionates records by time could be used, enabling more appropriate estimates of inhaled load.

Table 5 Environmental PM2.5 concentration and estimated inhaled load for 24 h using the estimated ventilation. Footnote: Data aobtained through individual samplers; busing ventilation estimates by the equation with the average regression parameters of each individual regression; cusing ventilation estimates by the mixed model equation. The values are expressed as the median (95% confidence interval). PM2.5

Public street group (n = 33)

Park group (n = 17)

Test group (n = 10)

Concentration (μg/m3)a Inhaled load (μg)b Inhaled load (μg)c

31.7 (30.5–36.3) 514.8 (398.8–650.4) 524.3 (424.1–662.2)

19.7 (17.2–24.6) 356.8 (219.0–440.8) 375.3 (231.5–458.8)

29.8 (24.1–44.6) 490.2 (323.3–587.4) 497.7 (330.8–631.9)

I.C. Cozza et al. / Science of the Total Environment 520 (2015) 160–167

5. Conclusion The mixed effect model is suitable for ventilation estimate, with good applicability in the same group of individuals who were used to model construction. In this case, the model could be used in field studies to estimate ventilation in usual daily activities. A small median difference between observed and measured ventilation in external validation group was observed, suggesting that the applicability of the model to external groups needs further evaluation. Disclosure statement The authors declare that there are no conflicts of interest to disclose. References ATS/ACCP Statement on cardiopulmonary exercise testing, 2003. American Thoracic Society; American College of Chest Physicians. Am. J. Respir. Crit. Care Med. 167, 211–277. Brook, R.D., Bard, R.L., Burnett, R.T., Shin, H.H., Vette, A., Croghan, C., et al., 2011. Differences in blood pressure and vascular responses associated with ambient fine particulate matter exposures measured at the personal versus community level. Occup. Environ. Med. 68, 224–230. Brunekreef, B., Janssen, N.A., de Hartog, J.J., Oldenwening, M., Meliefste, K., Hoek, G., et al., 2005. Personal, indoor, and outdoor exposures to PM2.5 and its components for groups of cardiovascular patients in Amsterdam and Helsinki. Res. Rep. Health Eff. Inst. 127, 1–70 (discussion 71–9). Brunekreef, B., Stewart, A.W., Anderson, H.R., Lai, C.K., Strachan, D.P., Pearce, N., Group, I.P.S., 2009. Self-reported truck traffic on the street of residence and symptoms of asthma and allergic disease: a global relationship in ISAAC phase 3. Environ. Health Perspect. 117, 1791–1798. de Hartog, J.J., Boogaard, H., Nijland, H., Hoek, G., 2010. Do the health benefits of cycling outweigh the risks? Environ. Health Perspect. 118 (8), 1109–1116. de Miranda, R.M., de Fatima Andrade, M., Fornaro, A., Astolfo, R., de Andre, P.A., Saldiva, P., 2012. Urban air pollution: a representative survey of PM(2.5) mass concentrations in six Brazilian cities. Air Qual. Atmos. Health 5, 63–77. Dons, E., Int Panis, L., Van Poppel, M., Theunis, J., Wets, G., 2012. Personal exposure to black carbon in transport microenvironments. Atmos. Environ. 55, 392–398. HEI Panel on the Health Effects of Traffic-Related Air Pollution, 2010. Traffic-related air pollution: a critical review of the literature on emissions, exposure, and health effects. HEI Special Report 17. Health Effects Institute, Boston, MA. Int Panis, L., Geus, D.B., Vandenbulcke, G., Willems, H., Degraeuwe, B., Bleux, N., et al., 2010. Exposure to particulate matter in traffic: a comparison of cyclists and car passengers. Atmos. Environ. 44, 2263–2270. Int Panis, L., Meeusen, R., Thomas, I., Geus, D.B., Vandenbulcke-Passchaert, G., Degraeuwe, B., et al., 2011. Systematic Analysis of Health Risks and Physical Activity Associated with Cycling Polices Belgian (SHAPES). p. 117 (Brussels, Belgium). Jacobs, L., Nawrot, T.S., de Geus, B., Meeusen, R., Degraeuwe, B., Bernard, A., et al., 2010. Subclinical responses in healthy cyclists briefly exposed to traffic-related air pollution: an intervention study. Environ. Heal. 9, 64. Jansen, K.L., Larson, T.V., Koenig, J.Q., Mar, T.F., Fields, C., Stewart, J., Lippmann, M., 2005. Associations between health effects and particulate matter and black carbon in subjects with respiratory disease. Environ. Health Perspect. 113, 1741–1746. Krewski, D., Jerrett, M., Burnett, R.T., Ma, R., Hughes, E., Shi, Y., et al., 2009. Extended followup and spatial analysis of the American Cancer Society study linking particulate air pollution and mortality. Res. Rep. Health Eff. Inst. 140, 5–114 (discussion: 115–36). Lim, S.S., Vos, T., Flaxman, A.D., Danaei, G., Shibuya, K., Adair-Rohani, H., et al., 2012. A comparative risk assessment of burden of disease and injury attributable to 67 risk

167

factors and risk factor clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study. Lancet 380, 2224–2260. Lioy, P.J., Fan, Z., Zhang, J., Georgopoulos, P., Wang, S.W., Ohman-Strickland, P., et al., 2011. Personal and ambient exposures to air toxics in Camden, New Jersey. Res. Rep. Health Eff. Inst. 160, 3–127 (discussion 129–51). Marshall, P., Al-Timman, J., Riley, R., Wright, J., Williams, S., Hainsworth, R., Tan, L.B., 2001. Randomized controlled trial of home-based exercise training to evaluate cardiac functional gains. Clin. Sci. (Lond.) 101, 477–483. Mermier, C.M., Samet, J.M., Lambert, W.E., Chick, T.W., 1993. Evaluation of the relationship between heart rate and ventilation for epidemiologic studies. Arch. Environ. Health 48, 263–269. Miller, M.R., Crapo, R., Hankinson, J., Brusasco, V., Burgos, F., Casaburi, R., et al., 2005a. General considerations for lung function testing. Eur. Respir. J. 26, 153–161. Miller, M.R., Hankinson, J., Brusasco, V., Burgos, F., Casaburi, R., Coates, A., et al., 2005b. Standardisation of spirometry. Eur. Respir. J. 26, 319–338. Nwokoro, C., Ewin, C., Harrison, C., Ibrahim, M., Dundas, I., Dickson, I., Mushtaq, N., Grigg, J., 2012. Cycling to work in London and inhaled dose of black carbon. Eur. Respir. J. 40, 1091–1097. O'Donoghue, R.T., Gill, L.W., McKevitt, R.J., Broderick, B., 2007. Exposure to hydrocarbon concentrations while commuting or exercising in Dublin. Environ. Int. 33, 1–8. Pellegrino, R., Viegi, G., Brusasco, V., Crapo, R.O., Burgos, F., Casaburi, R., et al., 2005. Interpretative strategies for lung function tests. Eur. Respir. J. 26, 948–968. Pereira, C.A., Sato, T., Rodrigues, S.C., 2007. New reference values for forced spirometry in white adults in Brazil. J. Bras. Pneumol. 33, 397–406. Peters, A., von Klot, S., Heier, M., Trentinaglia, I., Hörmann, A., Wichmann, H.E., Löwel, H., 2004. Exposure to traffic and the onset of myocardial infarction. N. Engl. J. Med. 351, 1721–1730. Pope, C.A., Burnett, R.T., Thun, M.J., Calle, E.E., Krewski, D., Ito, K., Thurston, G.D., 2002. Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. JAMA 287, 1132–1141. Samet, J.M., Lambert, W.E., James, D.S., Mermier, C.M., Chick, T.W., 1993. Assessment of heart rate as a predictor of ventilation. Res. Rep. Health Eff. Inst. 59, 19–55 (discussion: 57–69). Santos, U.P., Terra-Filho, M., Lin, C.A., Pereira, L.A., Vieira, T.C., Saldiva, P.H., Braga, A.L., 2008. Cardiac arrhythmia emergency room visits and environmental air pollution in Sao Paulo, Brazil. J. Epidemiol. Community Health 62, 267–272. Stram, D.O., Lee, J.W., 1994. Variance components testing in the longitudinal mixed effects model. Biometrics 50, 1171–1177. Tomei, F., Ghittori, S., Imbriani, M., Pavanello, S., Carere, A., Marcon, F., et al., 2001. Environmental and biological monitoring of traffic wardens from the city of Rome. Occup. Med. (Lond.) 51, 198–203. Torricelli, A.A., Novaes, P., Matsuda, M., Braga, A., Saldiva, P.H., Alves, M.R., Monteiro, M.L., 2013. Correlation between signs and symptoms of ocular surface dysfunction and tear osmolarity with ambient levels of air pollution in a large metropolitan area. Cornea 32, e11–e15. van Roosbroeck, S., Wichmann, J., Janssen, N.A., Hoek, G., van Wijnen, J.H., Lebret, E., Brunekreef, B., 2006. Long-term personal exposure to traffic-related air pollution among school children, a validation study. Sci. Total Environ. 368, 565–573. Watt, M., Godden, D., Cherrie, J., Seaton, A., 1995. Individual exposure to particulate air pollution and its relevance to thresholds for health effects: a study of traffic wardens. Occup. Environ. Med. 52, 790–792. Zuurbier, M., Hoek, G., van den Hazel, P., Brunekreef, B., 2009. Minute ventilation of cyclists, car and bus passengers: an experimental study. Environ. Heal. 8, 48. Zuurbier, M., Hoek, G., Oldenwening, M., Meliefste, K., Krop, E., van den Hazel, P., Brunekreef, B., 2011a. In-traffic air pollution exposure and CC16, blood coagulation, and inflammation markers in healthy adults. Environ. Health Perspect. 119, 1384–1389. Zuurbier, M., Hoek, G., Oldenwening, M., Meliefste, K., van den Hazel, P., Brunekreef, B., 2011b. Respiratory effects of commuters' exposure to air pollution in traffic. Epidemiology 22, 219–227.