Study subjects were selected among outpatients of the Pneumology and Cardiology Departments of the Catholic University Hospital in Rome (UCSC) included in routine clinical follow-up programs. Eligible for the study were residents of Rome, living in census tracts less than 2 km away from one of the six air monitoring stations considered in this study.
A number of clinical criteria was specified for each nosological category. Eligibility for the COPD panel included age from 50 to 80 years, a ratio of forced expiratory volume in one second to forced vital capacity (FEV1/FVC) less than 60%, partial oxygen pressure (SpO2) at arterial blood oximetry = 60–70 mmHg, normal values of carboxyhemoglobin (COHb), normal acid-base balance, no concomitant IHD, no need of oxygen therapy or breathing apparatus, no pacemaker, no cardiac arrhythmias, diabetes, Parkinson disease or chronic alcohol abuse, and no use of either psycho-chemical drugs or long acting bronchodilators; occasional use of short-acting bronchodilators was allowed.
Admittance to the asthmatic panel was allowed to subjects aged 18 to 64 years, positive at the bronchial reactivity test by hypertonic saline solution, with disease in the mild intermittent stage . Occasional use of β-adrenergic stimulants was allowed, but assumption of steroids or other asthma-preventive drugs (either before, or during the study periods) was not.
Participation in the IHD panel was restricted to subjects aged 40 to 64 years, with stable angina or previous myocardial infarction (at least 1 year prior to recruitment), no concomitant COPD, no use of calcium channel blockers, no pacemaker, no atrial fibrillation (other arrhythmias admitted), diabetes, Parkinson disease or chronic alcohol abuse, and no use of psycho-chemical drugs.
In relation to smoking habits, participation in the asthmatic panel was restricted to never smokers. Never smokers, however, were almost absent from the clinical series of COPD and IHD outpatients; former smokers were then admitted to the COPD and IHD panels if they had given up smoking at least 1 year before enrollment (sustained quitters).
The study protocol was approved by the Ethical Committee of the UCSC. Twenty-nine patients with COPD (7 men and 4 women), asthma (5 men and 6 women) or IHD (6 men and 1 woman) gave their written informed consent to undergo repeated clinical examinations for two one-month periods, in the spring and winter of 1999.
Study time period
The time period of interest consisted of 67 days in total, from 24 May to 24 June and from 18 November to 22 December 1999. These periods were chosen based on historical time series analyses of air pollution levels in Rome, due to their high variability in air pollutant concentrations.
Study subjects were scheduled to be examined three days apart, at home (COPD and IHD panels) or at the Pneumology Clinic of the UCSC (asthmatic panel). Forced vital capacity (FVC) and forced expiratory volume in 1 second (FEV1) were measured by spirometry. Spirometries were supervised, and done with the subject in a sitting position and wearing a nose-clip, following the suggestions of the American Thoracic Society . Spirometries were always done in the afternoon (between 4 pm and 8 pm), at least 6 hours after a possible inhalation of short-acting bronchodilators. A heated Fleish tube n. 3 portable spirometer (Biomedin, Italy) was used in the COPD and IHD panels, and a light bell Stead-Wells spirometer (Biomedin, Italy) in the asthmatic group. Between instruments reproducibility, for both FVC and FEV1, was within 30 ml and the calibration procedures were regularly performed . Pulmonary function indices used in the analyses are expressed as the percentage of the predicted values based on the subject- specific sex, age, height and weight .
Only amongst asthmatics we determined concentrations of nitric oxide (NO) in exhaled breath, an indicator of bronchial inflammation , using the analyzer model 280 (Sievers Instruments, USA). Subjects were breathing NOx free air prior to this test. Asthmatics were also asked to fill in a brief daily questionnaire collecting information about the occurrence of asthma attacks and β-2 agonist inhalations.
Study subjects' characteristics
Table 1 describes the characteristics of subjects at entry and the group-distribution of the outcome variables. As expected on the basis of the eligibility criteria, asthmatics were younger than COPD and IHD patients, and the group average values of FVC and FEV1 were sensibly lower among COPD cases compared to both IHD and asthmatic subjects. Overall, the 29 study subjects underwent a total of 449 spirometries. Due to dropouts, a variable number of observations per case was available. The average number of repeated observations was 15 in the COPD panel (ranging from 1 to 32 per subject), 24 in the IHD panel (from 12 to 32 per case), and 9 among the asthmatics (from 6 to 18 per person). All IHD patients were regularly treated with aspirin, statins and nitrates. No COPD or IHD patient made use of bronchodilators (short- or long-acting) during the survey periods, while 7 out of 11 asthmatics reported β-2 stimulant inhalations on one or more of the clinical monitoring days. As to previous smoking habits, all COPD patients were sustained quitters, all asthmatics were never smokers, while in the IHD panel the five male patients were sustained quitters and the single female participant was a never smoker.
Mean daily temperature (T, Celsius), barometric pressure (BarP, mmHg) and relative humidity (RelHum, %) were available from the Rome weather station (Collegio Romano – Ufficio Centrale di Ecologia Agricola). The Department of Environment of the Latium Region provided us with hourly concentrations of nitrogen dioxide (NO2), carbon monoxide (CO), ozone (O3), and sulphur dioxide (SO2) recorded at the fixed sites for air-quality monitoring in Rome. These sites are equipped with continuous inlet samplers. NO2 is determined by chemiluminescence, CO by IR absorption, O3 by UV absorption, and SO2 by UV fluorescence.
We computed daily city means (24 h values, from 3 pm to 3 pm of the following day) based on data from varying type and number of sites, depending on the pollutant. For NO2 and CO, we calculated 24 h values from five fixed sites, three of which are located in densely populated areas in the center of Rome (Magna Grecia, Fermi, Libia) and two representing background areas (Preneste and Villa Ada). For O3, concentrations recorded at two background fixed sites (Preneste and Villa Ada) were used. For SO2, we used 24 h concentrations recorded at one urban site (Fermi) and one background site (Villa Ada). The NO2 and SO2 series were complete, while average daily concentrations for one single day were missing for CO and O3.
As to particulate matter, we could not use data from the Rome air-quality monitoring network, because PM2.5 was not routinely measured. Therefore, for the specific purposes of this survey, 24 h concentrations of PM10–2.5 and PM2.5 were measured at two selected fixed monitoring sites: Villa Ada and Istituto Superiore di Sanità (ISS). These sites, located about 3.5 km apart, were chosen because, based on historical PM10 monitoring data, are considered representative of low and high traffic areas in Rome, respectively. Air samples were collected by means of dichotomous samplers (Graseby Andersen, model SA 241) operating at 16.7 L/min, with an omni-directional aerosol inlet. This sampler has been designated as reference for PM10 by US EPA . Sampling was carried out from 3 pm to 3 pm of the following day (in order to match the spirometry time schedule). Sixty-two 24 hour samples were collected, with 5 missing observations at the beginning of the winter survey. The dust on the couple of sequential polytetrafluoroethylene (PTFE) filters (polymethylpentane ringed, 2.0 μm pore size, 37 mm diameter; Gelman, USA) was gravimetrically analyzed to obtain average daily concentrations of PM2.5 and PM10–2.5. PM10 concentrations were calculated by adding the concentrations of the sampled fine and coarse fractions. The averages of PM concentrations measured at the two locations were used in the statistical analyses, as our best estimate of 24 h mean ambient concentrations for the Rome neighborhoods the panel participants lived in.
In the PM10–2.5 and PM2.5 samples, the content of selected metals (cadmium – Cd, chromium – Cr, iron – Fe, nickel – Ni, lead – Pb, platinum – Pt, vanadium-V, and zinc – Zn) was determined by atomic absorption spectrometry (AAS). The concentrations used in the analysis were calculated as the ratio of the metal amount in each PM sample to the air volume collected during the sampling.
As a side validation study, we measured indoor PM2.5 concentrations in a total of five homes of three study subjects per survey. COPD cases were preferentially selected for the side study due to their reduced mobility in comparison with asthmatics and IHD patients. Participation in the side validation study was burdensome to the study subjects: a technician had to come every day to change the sampler filter, and the sampler itself was noisy. Thus, we only succeeded in getting consent to participate from one IHD patient for both surveys, and from four COPD cases, for only one survey each. Participants in the side study were representative of the full study group in terms of housing typology (all lived in apartment buildings), floor [basement, first floor (two homes), second floor, fifth floor], and distance from the ISS or Villa Ada PM2.5 monitoring sites (varying from 0.3 to 5 km). Indoor 24 h air samples were collected on 59 days (from 28 May to 24 June and from 22 November to 22 December 1999) by Micro-Environmental Monitors (SKC, model 400) with a single-stage impactor, operating at a sampling flow rate of 10 L/min and equipped with a PM2.5 sampling inlet and PTFE filters (polymethylpentane ringed, 2.0 μm pore size, 37 mm diameter; Gelman, USA). Indoor PM2.5 mass concentrations were gravimetrically determined. The 24 h average concentrations of PM2.5from three homes per survey were used in the reproducibility analysis.
In a previous inter-method reliability study of PM10 measurements in outdoor and indoor air samples in Rome, based on two series of 12 parallel 24 h samples, a very good correlation between MEM and dichotomous samplers was observed (regression 1: y = 1.192x-3.275 - R2 = 0.9506; regression 2: y = 0.998x-1.332 - R2 = 0.9866) . According to the European Standard EN 12341 criteria , the observed values of the determination coefficients are such that the MEM can be considered equivalent to the dichotomous sampler.
Correlations among outdoor pollutant levels, as well as those between indoor and outdoor PM2.5 concentrations, were evaluated by non-parametric tests (Spearman correlation coefficient) applied to variables in the original scale.
Outcome variables in each panel had observations missing, and there was unequal spacing (the interval between observations was not constant). There were a few missing observations in the exposure variables also (1 missing daily mean for both O3 and CO concentrations, and 5 missing daily means for PM2.5 and PM10–2.5). Missing observations, in the exposure or outcome variables, were not replaced with estimates.
The relationships between respiratory function indices and concentrations of air pollutants were analyzed using generalized estimating equations (GEE) for panel data . An autoregressive correlation matrix of lag 1 was assumed, in order to account for possible correlations between repeated measures on the same subject. The statistical package STATA  was used for the analyses (XTGEE; the option "force" was specified in order to allow for unequally spaced observations). All the linear models included the within-subject between-period effect, using the dichotomous season variable (spring and winter). For the COPD and IHD panels, terms for daily mean temperature (°C), relative humidity (%) and day of the week (weekday/weekend) were included in the regression models. For the asthmatics panel, temperature and humidity terms were included in the regression models along with β-2 agonist use (yes/no), while the dummy variable weekday/weekend was not, because only 8 out of 108 spirometries were done on Saturday and none on Sunday. We considered the possibility of a non-linear effect of temperature by introducing a temperature-squared term in the regression models; however, since no evidence of model improvement was found in any of the panel-specific analyses, only a linear term was left. The daily variability of pulmonary function was examined with respect to the mean pollutant concentrations of the previous 24 hours and to the cumulative exposures over the previous 48 and 72 hours. Results from the analyses of lung function indices are reported as changes in percentage of predicted values per 10 μg/m3 increase in pollutant concentrations (except for CO where the unit increase is 1 mg/m3). In order to assess the relative effects of metals, we report changes in lung function per interquartile range of increasing concentrations.