Asthma affects up to one fifth of children in the UK.¹ It is primarily a disorder of allergic inflammation characterised by the presence of lymphocytes, eosinophils, and mast cells.² Our ability to manage asthma and predict exacerbations is currently impeded by the lack of non-invasive markers of airway inflammation. At present we rely on symptoms and lung function measurements. These do not directly reflect airway inflammation^3,4 and do not allow us to predict future asthma exacerbations.⁵ Considerable effort has been devoted to finding a useful inflammatory marker for asthma—for example, serum eosinophilic cationic protein has been shown to correlate with airway inflammation in cross sectional studies.^6,7 However, many markers require a serum sample, results are not immediately available, and correlate poorly with disease severity.⁶ It is also possible to determine a child’s percentage sputum eosinophilia.³ However, this is only successful in 50–75% of children, is labour intensive and, again, the results are not immediately available. For all these reasons, other surrogate measures of airway inflammation in asthma have been investigated.

Exhaled nitric oxide (NO) levels have been studied by a number of groups as a possible inflammatory marker of asthma control.^8–10 Subjects with asthma have higher fractional exhaled NO concentrations (Fe_NO) as a group than non-asthmatic controls.¹¹ Fe_NO increases with worsening asthma control and, in general, correlates with markers of inflammation within the lungs.¹² Studies comparing Fe_NO between groups of treated and untreated asthmatics, while revealing clear differences, demonstrate considerable overlap of data points between the two groups.¹³ However, these studies rely on cross sectional measurements at one time point and do not reflect sequential changes in Fe_NO within individuals.

The aim of this study was to determine whether sequential measurements of Fe_NO in individual children with seasonal allergic asthma provide a clinically useful tool to monitor asthma. Children with seasonal allergic asthma develop allergic asthma with environmental exposure to grass pollen and are generally asymptomatic during the rest of the year. This group therefore provides a unique clinical paradigm to study the sequential effects of allergen exposure on allergic asthma.

METHODS

Study participants

Subjects were recruited from the paediatric chest clinic at St Mary’s Hospital. All had mild to moderate summer asthma and rhinoconjunctivitis, normal winter lung function without treatment, and sensitisation to Phleum pratense. The diagnosis of asthma was based on clinical history, examination, and reversibility with inhaled bronchodilators.¹⁴ Skin testing was performed with negative and positive controls, P pratense, three tree pollens, Dermatophagoides pteronyssinus, Dermatophagoides farina, Alternaria, Cladosporium, Aspergillus, cat and dog (positive ⩾3 mm; ALK, Horsholm, Denmark).¹⁵ Asthma symptoms were treated according to the British Thoracic Society guidelines¹⁶ with inhaled steroids being commenced or increased if there were three episodes of symptoms a week.

Study design

All subjects were assessed at least once before the pollen season when they were clinically well. They were then asked to attend for a clinical assessment at regular 4 weekly intervals during the grass pollen season of 2000. Lung function and Fe_NO levels were measured and daily asthma symptoms were recorded on a scale of 0 (none) to 3 (severe). Bi-daily peak expiratory flows were measured using a mini-Wright meter. An exacerbation of asthma was defined as: (1) three episodes of wheeze, chest tightness or shortness of breath per week; (2) two consecutive peak flows of <80% of the patient’s maximum; or (3) an increase in daily inhaled corticosteroids (⩾100 μg) or systemic steroids. Families were asked to contact the clinic if an exacerbation occurred. Exacerbations were managed using the British Thoracic Society guidelines.¹⁶ Asthma control during each week of the study was defined as: excellent (no symptoms); good (symptomatic on 1 day); acceptable (2 days), or poor (at least 3 days).

The study was approved by the research ethics committee at St Mary’s Hospital. Informed consent was obtained from all the participants and their parents.

Spirometry

Pulmonary function (forced expiratory volume in 1 second, FEV₁) was measured with a paediatric Masterscreen (Jaeger, Wuerzburg, Germany). FEV₁ measurements were standardised by expressing them as a percentage of the child’s best value during the preceding winter. The 95% confidence interval (CI) for the difference in percentage best FEV₁ in subjects measured by different observers in our department is ±7.5%.

Measurement of exhaled nitric oxide (Fe_NO) levels

Fe_NO levels were measured using the single expiratory breath method with a chemiluminescence analyser (Logan Research LR2500, Rochester, UK).¹⁷ A biofeedback mechanism was used to maintain the expiratory flow rate at 250 ml/s and subjects exhaled against a resistance to prevent upper airway contamination.¹⁸ Measurements were made in a standardised manner with the subject standing without a nose clip. Fe_NO measurements were repeated until two consecutive results within 10% were obtained; this generally required 2–4 attempts. All Fe_NO measurements were undertaken before spirometric testing. The Fe_NO measurement was read from the plateau phase. Exhaled NO values were only discarded if the ambient level was above 100 ppb as ambient NO has been shown not to affect readings when the single breath technique is used.¹⁹ Subjects and medical staff involved in their care were blinded to Fe_NO results. When Fe_NO is measured twice by different observers in our department, the interquartile range for the relative difference is −3.3% to 12.9%.

Pollen counts

Daily atmospheric pollen counts for the North West London area were measured during the pollen season with a Burkard 7 day volumetric spore trap (National Pollen Centre, University of Worcester, UK).

Statistical analysis

Analyses were undertaken using absolute and standardised Fe_NO values. Standardised Fe_NO values were derived by dividing each Fe_NO value by the subject’s mean result from the Fe_NO result from all their asymptomatic time points outside the pollen season. The relationship between Fe_NO and the pollen count was examined using a time series regression model.²⁰ This statistical model calculates the degree of association between Fe_NO and the pollen count at the level of the subject. It allows for multiple sequential observations from individual subjects and enables a subject’s Fe_NO measurements to be compared with their previous measurements. The relationship between Fe_NO and pollen count on the assessment day was determined. As exposure to even small amounts of allergen over a number of days can impair lung function,²¹ we looked to see whether Fe_NO was related to the mean pollen count during different periods leading up to and including the assessment day. In addition, as allergens give rise to a delayed asthmatic reaction,²² the possibility of a lagged relationship between the two variables was investigated by looking at different time periods leading up to but not including one or more days immediately before the assessment. For this analysis the p values were adjusted using a Bonferroni transformation to take into account the multiple comparisons examined. The relationship between Fe_NO and asthma control was assessed using a summary measure for each subject.²³ The slope of the line of best fit for the relationship between Fe_NO and asthma control was calculated for each subject having recoded excellent to poor asthma control as 1–4. We then tested whether the mean slope was significantly different from zero, indicating that Fe_NO was related to asthma control. Again this methodology enables individual subject’s Fe_NO measurements to be compared with their previous ones. Similar comparisons were also undertaken for the percentage best FEV₁.

RESULTS

Forty four subjects (33 male) aged 6–16 years with seasonal allergic asthma were studied sequentially at regular 4 weekly intervals over the grass pollen season (table 1); 38 (86%) had required treatment with inhaled corticosteroids during the previous summer. Pollen counts were initially low (<20) in May, then high (median 58.5, range 3–178) in June and moderate (median 19, range 1–125) in July. A total of 289 weeks of symptom diary scores and Fe_NO measurements were obtained from these 44 subjects during the pollen season. Fe_NO was measured every 4 weeks, giving a median of two measurements (range 1–3) per subject during the pollen season. Twelve subjects (27%) required treatment with inhaled corticosteroids during the pollen season.

Correlation of standardised Fe_NO but not spirometry with pollen count in preceding week

The median Fe_NO level before the pollen season was 6.2 ppb (interquartile range (IQR) 4.7–10.4; standardised Fe_NO 1.0, IQR 0.8–1.1). This is similar to our normal range for healthy atopic children (median 6.2 ppb, IQR 4.3–10.3, 30 subjects aged 7–16 years with allergic rhinitis, insect allergy or food allergy). Fe_NO levels increased significantly during the grass pollen season to 9.2 ppb (IQR 7.6–13.6; standardised Fe_NO 1.3, IQR 1.1–1.9; p<0.002, Wilcoxon sum rank test). The median change for each subject from pre-season to the pollen season was 2.9 ppb (95% CI 1.5 to 5.4). This change was seen despite the fact that 12 subjects received inhaled corticosteroids during the pollen season. Whereas Fe_NO rose during the pollen season, there was no observed decrease in percentage best FEV₁.

The relationship between Fe_NO levels and the pollen count during the 2 weeks before the measurement was investigated for all subjects and a significant relationship was seen for both absolute and standardised Fe_NO levels (table 2). A significant relationship was found between standardised Fe_NO values and the pollen count on the day of measurement (regression coefficient 0.0089, n = 122, p<0.01). The strength of this association improved if, instead of using the pollen count on the day of measurement, data from counts during the preceding weeks were used. The maximum association was observed using the mean pollen count in the 7 days preceding the standardised Fe_NO measurement (regression coefficient 0.0146, n = 122, p<0.001). Similar results were obtained when the analysis was repeated after removing the 18 measurements taken while the subjects were receiving treatment with inhaled corticosteroids (data not shown).

We also examined the relationship between the grass pollen count and each subject’s percentage best FEV₁. No correlation between lung function and pollen count was apparent either on the day of measurement or for the mean count in the preceding 7 days (data not shown). Exclusion from the analysis of data from subjects on inhaled steroids did not reveal any relationship between pollen count and spirometric tests that could have been masked by treatment.

Correlation of standardised Fe_NO but not spirometry with asthma control

Analysis of asthma control data revealed 208 excellent patient weeks, 41 good weeks, 25 acceptable weeks, and 15 poor weeks. To examine the relationship between Fe_NO and asthma control, the slope of the relationship between them was calculated for each subject. This methodology takes into account the multiple measures from each subject and enables each subject’s Fe_NO measurements to be assessed in terms of their other measurements. For Fe_NO (logarithmically transformed) and asthma control the mean slope was 0.058 (95% CI 0.017 to 0.099), which was very similar to the standardised Fe_NO (logarithmically transformed) value (0.058 (95% CI 0.016 to 0.101)). The slope for both absolute and standardised Fe_NO is significantly greater than zero, which suggests that Fe_NO increases with poorer asthma control. The slope for the relationship between percentage predicted FEV₁ and asthma control was 2.70 (95% CI −0.18 to 5.57), suggesting that there is no relationship between FEV₁ and asthma control in these subjects. When measurements taken while subjects were on inhaled steroids were removed, there was no significant relationship between asthma control and absolute or standardised Fe_NO or FEV₁.

A post hoc analysis of the data suggested that Fe_NO may rise before an exacerbation of asthma. Twenty three subjects had a total of 32 asthma exacerbations during the pollen season within 4 weeks after an Fe_NO reading, eight of which were observed in the first 5 days after the Fe_NO reading. Four subjects required treatment with oral prednisolone. The median standardised Fe_NO level during the 5 day period before an exacerbation was 1.93 (IQR 1.6–2.4), higher than the median standardised Fe_NO level for all subjects during the pollen season (see above). Unlike the standardised Fe_NO values, neither the absolute Fe_NO nor the percentage best FEV₁ changed in the week before an exacerbation.

DISCUSSION

The clinical usefulness of Fe_NO as a surrogate marker of airway inflammation in asthma is limited by a paucity of longitudinal data in individual patients. In this study we have further evaluated the clinical role of this marker by sequentially following a group of children with seasonal allergic asthma through the grass pollen season. Like other investigators,¹⁰ we found that Fe_NO increases with environmental allergen exposure. We have extended these observations to show that Fe_NO also correlates with preceding allergen load and that the timing of exposure is important. In addition, we have shown that Fe_NO levels increase with worsening asthma control. Our results show that spirometry does not significantly correlate with allergen load or with asthma control, even when it is standardised for each individual.

Fe_NO values are usually presented in absolute terms (parts per billion). When presented in this way, the range of normal values is wide.^10,11,24 Furthermore, there are suggestions that levels vary with many factors including age, sex, body size, and atopic status.^12,24,25 In this respect, Fe_NO behaves similarly to other lung function measures such as FEV₁; this may explain the considerable overlap seen between different groups of patients in cross sectional studies.¹³ To overcome this we normalised values to each subject’s mean Fe_NO level when they were asymptomatic before the pollen season. Data from our clinic suggest that Fe_NO values have minimal variability in these circumstances. In a total of 148 measurements in 16 subjects the median difference from each subject’s mean Fe_NO as a proportion of the mean Fe_NO was −0.037 (IQR −0.488 to 0.587).

The value of using standardised Fe_NO data is illustrated in a number of ways. Firstly, the mean Fe_NO during the grass pollen season was within our normal range. However, when intersubject variability is removed by standardising seasonal values to each subject’s pre-seasonal level, exhaled NO is seen to be significantly increased during the pollen season. Secondly, using statistical techniques which, like standardised Fe_NO, enable an individual’s observations to be compared with their previous measurements, we found relationships between Fe_NO and both the pollen count and asthma control. It therefore seems that levels of exhaled NO provide better clinical information when compared with a child’s previous exhaled NO level than when compared with a population based normal range. This approach may not be useful for children with symptoms that persist throughout the year as it may not be possible to measure a baseline Fe_NO when they are symptom free. Further studies are also needed to test the general usefulness of this approach in other asthmatic populations.

In a group of 21 children with seasonal allergic asthma Baraldi et al¹⁰ showed that Fe_NO increases during the grass pollen season. We have extended these observations to look at the temporal relationship between Fe_NO and allergen exposure and found that Fe_NO correlates best with the pollen count in the previous 7 days (table 2). This may be explained by the dual response following allergen exposure consisting of a transient early phase and a more prolonged late phase inflammatory response 4–8 hours later. Fe_NO has been found to be increased only during the late phase asthmatic reaction.²⁶ With Fe_NO increasing during the late phase response, it is not surprising that it correlates better with the mean pollen count in the days before its measurement than the count on the day of measurement.

Within this study we have restricted our experimental model to focus on the relationship between Fe_NO, pollen exposure, and asthma control. In doing this we have used pollen counts from only one site in North West London although the residences of the subjects were spread across this area of London. While the trends of exposure for all subjects will be similar within this small geographical area, these values may not accurately reflect personal pollen exposure. Furthermore, we have not taken into account the many other factors that influence the allergenic potential of grass pollen including the use of inhaled corticosteroids and meteorological parameters such as rainfall, humidity, and thunderstorms. These meteorological factors seem to exert their influence by inducing the release of submicron allergenic particles from the large 15–40 μm pollen grains, thereby allowing them to penetrate further into the lower airways.²⁷ Despite these factors, we have still been able to demonstrate relationships between exhaled NO and both pollen exposure and asthma control indicating that, within a longitudinal model, these relationships are very robust.

The effects of inhaled or oral corticosteroids on Fe_NO measurements are unclear. Many paediatric and adult studies have clearly shown high Fe_NO levels in untreated asthma that decline with steroid therapy.^8,28 Conversely, other studies have found increased levels in association with asthma symptoms despite corticosteroid therapy.²⁹ The data presented here suggest that Fe_NO levels are related to allergen exposure, whether or not subjects are taking inhaled corticosteroids. It should be remembered, however, that only a quarter of the subjects required treatment with inhaled corticosteroids during the study and then only for a relatively short period of time.

This study has a number of limitations. Firstly, the method for measuring Fe_NO differs from the current American Thoracic Society recommendations.¹² An exhalation flow of 250 ml/s was used as originally recommended by the only guidelines available when the study commenced.¹⁷ Using a lower flow rate is unlikely to have altered the pattern of the results, but it would have increased the measured Fe_NO values increasing the differences seen with changing allergen exposure or asthma control.³⁰ Secondly, we have chosen children with grass pollen induced asthma as a clinical model. While there is no reason to believe that relationships seen in this group will not be replicated in children with perennial allergic asthma, this work needs to be validated in this important group of patients, particularly where viral infections are the key trigger for exacerbations. It is possible that, in this group, baseline Fe_NO values would need to be measured when they were asymptomatic after treatment but this awaits further study. In addition, our study group was relatively small and contained a predominance of boys. Although there is no reason to believe that there are sex differences in Fe_NO levels, it is important that this study is replicated in a larger group with more girls. Lastly, the potential relationship that we highlight between an exacerbation and raised Fe_NO measurement in the preceding week was part of an exploratory analysis of the data. Given the small number of standardised Fe_NO values within 5 days of an asthma exacerbation and the fact that it was a post hoc analysis of the data, we have not presented a formal statistical analysis. The presence of such a relationship is, however, supported by the results of Jones et al³¹ who found that Fe_NO measurements predicted exacerbations of asthma in adults from whom inhaled corticosteroids had been withdrawn. The value of a raised standardised Fe_NO as a predictor of future exacerbations should be tested in further studies.

We have presented data which show that Fe_NO levels are related to prior environmental allergen exposure and asthma symptoms and potentially predict future exacerbations. In contrast, lung function in this study does not reflect either allergen load or asthma severity. Sequential Fe_NO measurements may therefore be a more useful clinical tool than spirometry to monitor asthma control. This study provides preliminary evidence that sequential standardised Fe_NO measurements may be useful for monitoring children with asthma in a clinical setting. Access to a sequential, non-invasive, rapidly available measure of airway inflammation may allow us to reduce treatment to the minimum required to suppress inflammation and, furthermore, to respond to pre-symptomatic increases in inflammation. The advantage of using a measure of airway inflammation in directing asthma management has recently been demonstrated in a randomised controlled adult study using sputum eosinophilia in which the intervention group had better asthma control without increasing inhaled corticosteroid usage.³² More frequent sequential assessment of Fe_NO in children will allow us to determine whether such measurements will translate into improved patient outcome in the clinical setting and whether a rise in standardised Fe_NO can successfully predict an asthma exacerbation.