This study examined the short-term effects of ambient air pollutants on the risk of TIA and minor stroke in a cohort of patients attending TIA services in five UK centres. To date, this is the first study to have specifically investigated the effect of ambient air pollutants on the occurrence of TIA and minor stroke. Its main finding is that NO may be associated with the onset of TIA and minor stroke. A statistically significant association between the occurrence of TIA and minor stroke and exposure to NO with a 3-day lag was observed for the Manchester dataset. In Liverpool, there was no significant positive association between TIA and minor stroke and any of the pollutants, but negative association were observed with PM10 and NO at lag 0 and 3 respectively. The study also found effect modification by age, gender and season. Measured pollutants have had more pronounced effects in those above 65 years old, in male and in cold months than in participants below 65 years, female and warmer months respectively.
Reported regression results of the study are based on statistically significant associations, but a statistically non-significant association does not necessarily mean that a relationship does not exist. This study has found several marginally non-significant associations between various pollutants and TIA in both cities, where the lack of significance might be attributed to the small study size in the two centres.
The observed association between NO and TIA and minor stroke could be attributed to chance, due to multiple testing undertaken in the study. We have performed a total of 48 logistic regression analysis and with a conventional 5% significance level, one would expect some of these estimates would achieve this level of ‘significance’ occurred by chance. Thus, the interpretation of the results needs caution and the results need to be corroborated by other studies.
The findings of earlier studies that investigated associations between different subgroups of stroke mortality and morbidity are inconsistent, with some studies reporting increased risk of stroke mortality [11, 28–30], and morbidity [2, 4–6, 31–33], while others found no association [16, 18]. These variations may be attributed to the concentration level (and exposure gradient), composition and distribution of the exposure parameter, the degree of exposure misclassification, type of studied pollutant, underlying health status of study population and the method of ascertainment of study outcome.
Although several studies have investigated associations between ambient air pollutants and stroke, fewer have studied these pollutants on TIA [4–6]. A recent US study found an association between same-day and previous day exposure to fine particles and ozone and ischaemic stroke, as well as TIA . A Canadian study found a positive association between same-day exposures to SO2 and increased emergency department visits for TIA in the warm season . Similarly, a French study reported a significant association between ozone exposure and risk of TIA at lag of one day . While our study has found a positive association of NO with TIA, none of these three studies have investigated the effect of NO. Furthermore, the effects of NO in our study was observed at lag of 3 days, while the other studies found association on the same day or at lag of one day.
Some studies have found associations between ambient air pollutants and certain stroke subtypes in towns where the concentrations of air pollutants were relatively high. Tsai and others found associations of particles and gases with ischaemic and haemorrhagic stroke when the ambient temperature was above 20°C in Kaohsiung, Taiwan . While most studies have found stronger evidence for the association of ambient air pollutants with ischaemic stroke, a few have reported positive associations of haemorrhagic stroke with ambient PM and gases  or total suspended particles .
Nitrogen oxide emissions occur largely from motor vehicles in the form of NO, which is a major primary pollutant and precursor of NO2. NO typically comprises about 95% of oxides of nitrogen (NOx) from a combustion source . Because of the high correlations among measured pollutants in the current study, it is difficult to separate their individual effects; however, NO may be associated with TIA and minor stroke more than the other pollutants as its effect persisted in a two pollutant model. However, we are unable to explain the contradictory negative effect of NO in Liverpool. Perhaps differences in the overall pollution profile and composition might be responsible for the observed effect. The biological mechanism by which NO might be associated with an increased risk of stroke or TIA has not been elucidated so far and there is little information from published literature. Various studies have linked NO2 to increased morbidity [18, 35], yet related substances like NO are uncommonly investigated in most epidemiological studies. Those few studies that have included NO as an exposure metric have highlighted its positive associations with the respective health outcomes [31, 36] as well as a higher correlation with UFP  which suggests that NO is an important pollutant to be considered in epidemiological studies.
NO, through a previously unrecognized mechanism, may be responsible for the observed association. NO is endogenously generated and involved in various regulatory and inflammatory functions of the vascular endothelium. It is possible that ambient NO influences bioavailability of endogenous NO. Inhibition of NO and the concomitant changes are known to contribute to essential hypertension, myocardial ischaemia, atherogenesis, thrombosis, insulin resistance and heart failure . A few experimental studies have shown that exposure to nanoparticles and diesel exhaust, which contains significant amounts of NO and UFP, is associated with impairment of vascular function and vasoconstriction, possibly through inhibition of endogenous NO synthesis [38–40]. A recent experimental study which has compared the vasculotoxic responses to various combustion source pollutants has found a significant role of monoxide gases of NO and CO in vascular toxicity .
It is possible that NO is a surrogate for other unmeasured pollutants such as ultrafine particles (UFP) . A study in three UK conurbations, including Manchester, has shown a strong correlation between UFP and NOx (r = 0.85) and CO (r = 0.78) . A Finnish study reported a significant association between UFP and reduced peak expiratory flow rate (PEFR), a measure of lung function . In that study, UFP was highly correlated with NO (r = 0.77) and the effect of UFP on PEFR diminished in a two-pollutant model when paired with traffic-related gaseous pollutants CO, NO and NO2.
One strength of our study is that the diagnosis of TIA or minor stroke was confirmed by a consultant stroke physician or neurologist independent of the study. Most previous studies have used hospital admission or discharge diagnostic code data to establish diagnosis which carries a significant risk of misclassification [44, 45]. The study has a number of limitations. Firstly, participants were recruited at first presentation to secondary care services which in some cases was as late as 6 weeks and this might lead to recall bias. This was standard clinical care in the UK at the time (2003-7) but may also have introduced selection bias as higher-risk patients may have had a subsequent stroke prior to presenting to TIA services. Participants included in this study may therefore likely to represent lower- risk cases of TIA and minor stroke. There could be a differential effect of pollution on low and high risk cases which would affect the generalizability of the results. Secondly, the exposure data were obtained from fixed central site monitors which do not necessarily reflect personal exposure, especially for traffic related gaseous pollutants that are not spatially homogenous like NO and NO2 and this may have led to exposure misclassification. Additionally, as most individuals spend most of their time indoors, use of exposure data from outdoor central monitors will introduce exposure measurement error. However, such error is likely to be non-differential and would be expected to underestimate the observed effects . Study numbers are also relatively low, particularly as for geographic reasons they had to be divided into two populations.