Objectives Time series studies have shown adverse effects of outdoor air pollution on mortality and hospital admissions in patients with chronic obstructive pulmonary disease (COPD) but panel studies have been inconsistent. This study investigates short-term effects of outdoor nitrogen dioxide, ozone, sulfur dioxide, particulate matter (PM10) and black smoke on exacerbations, respiratory symptoms and lung function in 94 patients with COPD in east London.
Methods Patients were recruited from an outpatient clinic and were asked to complete daily diary cards (median follow-up 518 days) recording exacerbations, symptoms and lung function, and the amount of time spent outdoors. Outdoor air pollution exposure (lag 1 day) was obtained from local background monitoring stations.
Results Symptoms but not lung function showed associations with raised pollution levels. Dyspnoea was significantly associated with PM10 (increase in odds for an IQR change in pollutant: 13% (95% CI 4% to 23%)) and this association remained after adjustment for other the pollutants measured. An IQR increase in nitrogen dioxide was associated with a 6% (0–13%) increase in the odds of a symptomatic fall in peak flow rate. The corresponding effect sizes for PM10 and black smoke were 12% (2–25%) and 7% (1–13%), respectively.
Conclusion It is concluded that outdoor air pollution is associated with important adverse effects on symptoms in patients with COPD living in London.
- Air pollution
- chronic obstructive pulmonary disease
- COPD exacerbations
Statistics from Altmetric.com
What is the key question?
Does outdoor air pollution affect exacerbations, respiratory symptoms and lung function in patients with COPD?
What is the bottom line?
In patients with COPD living in London, there is evidence for adverse effects of outdoor pollution on symptoms and exacerbations, particularly for PM10, black smoke and NO2.
Why read on?
Ecological studies have shown links between outdoor air pollution and increased COPD mortality/morbidity and this study shows similar effects in individual patients, supporting the case for causality.
Chronic obstructive pulmonary disease (COPD) is increasing in incidence worldwide and is currently the sixth leading cause of death.1 Patients with COPD are prone to acute deterioration in their chronic symptoms. These exacerbations of COPD were first shown to be associated with air pollutants during the London smog episode of 1952.2 There is now clear evidence that air pollution at current levels in London affects mortality3 4 and that air pollution levels in Europe lead to hospital admissions for COPD.5
Panel studies have shown adverse effects at relatively low levels of pollution. One of the earliest panel studies in patients with COPD showed that changes in pollution level were associated with exacerbations of chronic bronchitis.6 A study in Merseyside, UK involving 75 patients with COPD reported effects of ozone (O3) and sulfur dioxide (SO2) on peak expiratory flow (PEF) and respiratory symptoms recorded daily using diary cards.7 A Dutch panel study similarly reported adverse effects of PM10 (particulate matter), black smoke, sulfate and SO2 on PEF, and effects of black smoke on upper respiratory symptoms in 326 symptomatic adults.8 A panel study in Italy reported adverse effects of particles and nitrogen dioxide (NO2) on forced vital capacity (FVC) and forced expiroatory volume in 1 s (FEV1) in 29 subjects with COPD,9 and similar associations between FEV1 and particles were reported in a US COPD panel with 24 subjects.10 A New Zealand COPD panel study in 48 patients found no effect of pollution levels on PEF but observed effects of particles on night-time symptoms and nebuliser use.11 A French study conducted at a similar time to the present study analysed effects of air pollution on COPD exacerbations and reported adverse effects of O3.12
In this paper we examine the effects of a range of air pollutants on COPD exacerbations, respiratory symptoms and respiratory function, including large decrements in PEF, in a panel of 94 patients with moderate to severe COPD selected from a COPD clinic in London and followed over a 2-year period, October 1995–October 1997.
The East London COPD study was established in 1995 as a prospective study of the role of viral infections and environmental factors in COPD exacerbations.13–16 Subjects were patients with moderate to severe COPD attending an outpatient clinic at the London Chest Hospital, London. Recruitment was from October 1995 to October 1997. Inclusion criteria were: FEV1 <70% predicted for age and height, β2-agonist reversibility <15% or 200 ml17 18 and no exacerbations in the previous 4 weeks. Exclusion criteria were asthma, bronchiectasis, carcinoma of the bronchus or inability to complete diary cards. Patients were seen at the clinic monthly during the colder months and 3-monthly during the rest of the year. Patients were also seen acutely at exacerbation and then at a convalescent visit 4–6 weeks postexacerbation.
At recruitment, baseline measurements were made of height, weight, FEV1, FVC and PEF by rolling seal spirometer (Sensor Medic, Yorba Linda, California, USA), reversibility to 400 μg of inhaled salbutamol, and arterialised earlobe blood gases.19 Patients recorded symptoms and lung function daily on diary cards. Patients measured PEF indoors after morning medication using a Mini-Wright peak flow meter (Clement Clarke International, Harlow, UK). They recorded any increase in chronic symptoms during the previous 24 h. Symptoms were categorised as major (dyspnoea, sputum purulence, sputum amount) or minor (nasal discharge/congestion, wheeze, sore throat, cough). Patients recorded bedroom temperature on waking using a 1°C Thermax temperature strip (Thermographic Measurements, Burton, UK). From March 1996 onwards, patients recorded the number of hours spent outdoors. FEV1 and FVC were also measured in a subsample of 28 patients using a hand-held spirometer (Micro Medical, Rochester, UK). At the outset of the study, patients were taught how to measure and record lung function and symptoms, and were reminded and/or re-educated when they visited the clinic.
Patients were asked to attend clinic if their symptoms worsened. Exacerbations were identified by symptoms recorded on the diary cards or from the history when patients presented to the physician (TAS), according to the criteria modified from Anthonisen et al20 of any two major symptoms or one major and two minor symptoms on two consecutive days.13 The first of the 2 days was taken as the onset of the exacerbation. Symptoms present continuously for >5 days prior to the possible onset of an exacerbation were discounted. Patients experiencing an exacerbation were given appropriate medication.
A less severe but more common form of exacerbation was also defined a priori as a fall in PEF of 10 l/min for ≥2 days plus a reported increase in dyspnoea. This was termed a ‘symptomatic fall in PEF’. The rationale for using this is that patients with COPD, particularly frequent exacerbators, have high psychosocial and depression scores13; however, the presence of symptoms together with a change in lung function is more suggestive of an airway effect than psychological effects.
Outdoor air pollution data
Hourly measurements of NO2, O3, SO2 and PM10 were obtained from the national air quality monitoring network station at Bloomsbury Square, central London via the UK National Air Quality Information Archive (http://www.airquality.co.uk/). At the time of the study this was the only monitoring station in London for these pollutants. The following pollutant measures were derived: maximum hourly NO2, maximum 8 h moving average O3, 24 h mean SO2 and PM10. The completeness criterion was the availability of at least 75% of the data used to calculate the summary measure for each day. The PM10 data from this monitor are very closely correlated with the North Kensington site data for the same period (r=0.95) and so it was reasonable for the Bloomsbury data to represent background levels. For black smoke there was a network of monitors at that time (unlike for other pollutants) and so we used 24 h average black smoke data from the monitor nearest each patient's home. Missing pollution data were imputed using a standard method.21 We chose a priori to analyse 1 day lags (previous day) for all pollutants based on the results of time series studies available at the time.
Brief details of statistical methods and rationale are given here, but fuller details are given in the online supplement. Symptoms and exacerbations were analysed as binary incidents—that is, the first day of an episode was the ‘event’. To distinguish between new episodes of symptom worsening and the continuation of a current episode, a new episode was recorded when that symptom was not recorded in the previous 7 days.
Generalised estimating equations (GEEs) were used for analysis, taking account of variation within and between individuals. Effects of each symptom or exacerbation were modelled separately, with each pollutant level in turn as the main explanatory variable. Control was made for daily temperature (average of minimum and maximum) and season (analysed in four groups—spring, summer, autumn and winter), with estimates of variance robust against misspecification of correlation structure, which was assumed to be independent. Results were obtained as ORs and 95% CIs.
All lung function data (PEF, FEV1 and FVC) were analysed as deviations from the individual mean to allow for variation between individuals. Lung function data were analysed in three ways. First they were analysed by season and year (4 seasons×2 years) to investigate seasonal effects, adjusting for indoor temperature and time spent outdoors using methodology described in previous reports.22 23 Secondly, they were analysed using GEEs with Normal errors to summarise effects across the entire time period. Analyses used a first-order autoregressive correlation structure and controlled for temperature and season as described above for symptoms. In addition, a linear time term was included to allow for decline in lung function. Thirdly, lung function was analysed as a binary variable following the method of Hoek.24 For each patient, an adverse event, ‘large peak flow decrement’, was defined if lung function was >20% below that individual's median value. These binary events were also modelled using GEEs as described above for the symptom data.
Some associations were observed for effects of single pollutants on certain symptoms and so we decided post hoc to fit selected multipollutant models to try to disentangle these effects and aid interpretation.
All effect estimates are presented in two ways: (1) representing a unit change in pollutant level (ppb or μg/m3 as appropriate) and (2) scaled to an IQR change in that pollutant level to aid interpretation. All analyses used Stata v11.
Summary statistics for subjects and exposure data
A total of 125 patients were recruited, of whom 31 were excluded for the following reasons: not continuously resident in London (13), in the study for <3 weeks (10), patient errors in data recording (1), misplaced diary sheets (1), dementia (1), asthma (1), precancerous illness (1), psychiatric disorder (1) and died (2). Median duration of follow-up was 518 days (range 21–709). Shortened length of follow-up was due to either late recruitment, moving away or death. Fifteen patients had their series of measurements truncated as it was suspected that use of the measuring instrument was unreliable during the first few weeks. Further patient characteristics are reported in table 1 and show that the series predominantly comprised older males with obstructive lung disease. Mean numbers of symptoms over the period varied considerably among individuals but were on average quite low, suggesting that the condition of some patients was reasonably stable. Mean values for pollutants were: NO2 1 h max, 51.4 ppb; O3 8 h average, 15.5 ppb; SO2 24 h average, 7.5 ppb; PM10 24 h average, 37.7 μg/m3; black smoke 24 h average, 10.1 μg/m3 (table 2). Data from eight different black smoke monitors were used although the majority of patients (63%) lived nearest to one monitor (Stepney), which was close to the hospital.
Effects on lung function
When lung function was analysed by season, very few statistically significant effects were observed. There was a significant adverse effect of O3 on PEF in summer 1996 but not in summer 1997. For FEV1, there was one significant association for SO2 in spring 1996 but this was not replicated the following year. For FVC, significant negative associations were observed for NO2, PM10 and black smoke in summer 1996 and SO2 in spring 1996, but none of these seasonal effects was replicated in 1997 (tables E1–E3 in the online data supplement).
These non-significant findings were confirmed when lung function was analysed over the 2 years combined (table 3). Most regression coefficients were positive not negative, and one (PEF and NO2) was statistically significant. There was no consistency in findings for PEF, FEV1 and FVC, again providing no support for any adverse effect of pollution on mean respiratory function (table 3). There was no evidence for effects of any pollutant on large PEF decrements (table 4), with all ORs close to 1.0 and non-significant.
Effects on exacerbations and symptoms
All ORs for effects of pollutants on exacerbations were >1.0, suggesting adverse effects, but all were non-significant (table 4). Effect sizes were bigger for symptomatic fall in PEF where PM10 and black smoke showed significant adverse effects. The remaining ORs for NO2, O3 and SO2, while not significant, were nonetheless consistently >1.0. ORs for a symptomatic fall in PEF for a change in pollutant level equivalent to the IQR were 1.12 (95% CI 1.02 to 1.25, PM10) and 1.07 (1.01 to 1.13, black smoke). Dyspnoea was significantly associated with PM10 (OR 1.13; 95% CI 1.04 to 1.23) for a rise in PM10 across its IQR. No other symptom–pollutant combinations were significant, except O3 which showed a protective association with nasal discharge/congestion (table 5).
In multipollutant models, the effect size for PM10 on symptomatic fall in PEF remained similar and borderline significant after adjustment for other pollutants (table 6). In contrast, the effects of NO2 and black smoke were weaker after controlling for PM10. For dyspnoea, the effect of PM10 was slightly stronger and remained significant after adjustment for either NO2, black smoke or both (table 6). There was no evidence of any adverse effect of NO2 or black smoke after allowing for PM10.
Overall this study provided evidence for adverse effects of outdoor pollution on symptoms and exacerbations in patients with COPD living in London, particularly PM10, black smoke and NO2. Most symptoms gave ORs >1 but very few associations were statistically significant. Effect sizes were mostly larger for symptomatic fall in PEF compared with COPD exacerbations and gave significant ORs for NO2, PM10 and black smoke. Symptomatic falls in PEF events were more common than COPD exacerbations and so significance is partly due to increased statistical power. Dyspnoea was associated with higher levels of PM10 but was not significantly associated with any other pollutants. Multiple pollutant models showed that the association between PM10 and dyspnoea was stronger after adjustment for other pollutants, although this analysis was post hoc and conducted to aid interpretation of the findings. In general, symptoms are highly variable in COPD and the appearance of shortness of breath on the diary card may reflect psychological as well as mechanical effects on the airway. The occurrence of shortness of breath with a fall in PEF is suggestive of a mechanical effect on the airway by some stimulus.
When estimated effect sizes were scaled to IQR increases in pollutant level, it was evident that estimated effect sizes were considerable: the odds of a symptomatic fall in PEF increased by 13% when PM10 increased across the IQR, and a similar size effect was observed for dyspnoea. These increases in odds represent a substantial increase in risk, if associations were real and causal, and are stronger than observed effects of raised pollution on COPD hospital admissions.5
Effects on neither mean lung function nor the binary large PEF decrements showed consistent trends, despite evidence from other panels that the binary outcome is more discriminating than mean PEF.8 24 This may reflect the high variability of PEF and/or that lung function was measured after taking medication. Importantly, the findings of adverse effects on symptoms but not mean lung function are consistent with results of ecological studies showing associations with acute events such as death, hospital admission and general practitioner consultations. They lend support to the hypothesis that effects of outdoor air pollution are greater among the very vulnerable.
Findings in this London study were consistent with those of Harré in New Zealand11 who reported associations between particles and symptoms with similar effect sizes to ours; however, they also found no associations with lung function. Trenga10 reported effects of PM2.5 on FEV1 in all adults but no effects on PEF. PM2.5 data were not available when our study was conducted but since PM10 is dominated by small particles, the comparison with our study is reasonable. The adverse effects of O3 on exacerbations demonstrated in the Parisian study12 were stronger than in our study.
The estimated relationships between air pollution and symptoms and lung function in this COPD panel may have been affected by the time patients spent outdoors. Subjects recorded this for part of the study only and so, although we did adjust, full adjustment was not possible.
When this study was designed we chose to limit the use of multiple pollutant metrics to avoid overtesting. In particular we used just one measure for NO2—1 h maximum—partly because of the belief that the peak drives health effects. However, the correlation between 1 h maximum NO2 and average daily NO2 was very high at 0.90, and so the choice would seem unlikely to matter. When our study was conducted, there was a network of black smoke monitors and so we used the monitor nearest to the patient's home to estimate their exposure. The data on the other pollutants all came from a single monitor in central London but the correlation with another monitor for PM10 was high (0.95), suggesting that this was not unreasonable. Even so, the use of multiple monitors may explain the weaker associations observed with black smoke. We modelled exposure to pollution using previous day pollutant level as others have done, and have not looked at long lags25 which would most probably have produced stronger associations.
Daily diary data provide a powerful tool tool to investigate effects of air pollution within individuals but are resource intensive and often panels are only able to include a relatively small sample and/or a short follow-up time. The strength of this study was the relatively large sample, 94, and the lengthy 2-year follow-up period. Since compliance was very good, missing data were minimal.
Recent WHO guidelines for PM10 are 20 and 50 μg/m3 for annual and daily averages, respectively. The levels of exposure for this panel were a little higher. The WHO guidelines were largely based on ecological time series studies and cohort data on mortality. This study therefore lends support to the guideline.
In conclusion, in patients with COPD living in London, there is evidence for adverse effects of outdoor pollution on symptoms and exacerbations, particularly for PM10, black smoke and NO2. The ORs of up to 1.17 for an IQR increase in pollutant level represent substantial effects which would have important public health implications if shown to be real and causal. This deserves further investigation in larger panels.
Funding The East London COPD study was funded by the British Lung Foundation. The statistical analysis for the air pollution analyses was funded by the UK Department of Health.
Competing interests None.
Ethics approval This study was conducted with the approval of the East London and City Health Authority Ethic Committee.
Provenance and peer review Not commissioned; externally peer reviewed.