INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: asthma; exacerbation; nitric oxide; eosinophils; bronchial provocation tests

Nitric oxide is a key messenger for cell to cell signaling, and has an important role in the biochemistry of inflammation (1, 2). Exhaled nitric oxide (eNO) has been confirmed as a marker of airway inflammation and is present in higher concentrations in steroid-naive asthma compared with normal control subjects (3). Higher levels of eNO are seen during asthma exacerbations (4), and decreases occur following treatment with both inhaled (5) and systemic corticosteroids (6). Furthermore eNO appears to be sensitive to changes in antiinflammatory treatment, even in the absence of changes in lung function (7). These findings suggest that eNO may be a useful indicator in the longitudinal assessment of asthma control.

Both induced sputum cell counts (8) and responsiveness to hypertonic saline challenge (HSC) (9) have also been investigated as markers of airway inflammation in asthma. Sputum eosinophil numbers increase during asthma exacerbations (10). Conversely, a decrease in the percentage of sputum eosinophils occurs following prednisone treatment (11), and after commencing inhaled corticosteroid (ICS) (12). Du Toit and coworkers have demonstrated a progressive reduction in responsiveness to HSC following initiation of ICS treatment (13). However, the scope for using these techniques to monitor asthma control in clinical practice is limited by the resources required for repeated measurements. In contrast the measurement of eNO is quick and easy to perform, thus lending itself to repeated measurements over time. Although single measurements of eNO have been used to assess airway inflammation in asthma (14), the usefulness of eNO in the longitudinal assessment of asthma control has not been extensively investigated. To be clinically useful eNO would need to correlate with known markers of asthma control as well as airway inflammation, and be responsive to changes in these parameters over time. If this were the case, then eNO measurements could be used to confirm poorly controlled asthma and to predict imminent deterioration. There might also be a role for eNO in optimizing antiinflammatory therapy such as has been done using measurements of airway hyperresponsiveness (AHR) (18).

The aim of our study was to evaluate the predictive and diagnostic value of eNO in unstable asthma and to correlate this with sputum eosinophils and AHR to hypertonic saline. Using a model of steroid withdrawal described by Gibson and coworkers (19), we aimed to induce a deterioration in asthma control in the majority of patients thus enabling us to assess these parameters in the context of increasing degrees of airway inflammation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Patients with mild to moderate asthma, confirmed at our research screening clinic using ATS criteria (20), and who had been taking ICS therapy for at least 6 mo were recruited. The dose of ICS was unchanged for at least 6 wk. Patients were excluded (because the study involved withdrawal of ICS treatment) if they had a history of acute asthma requiring hospital admission, asthma characterized by sudden attacks, or used oral prednisone during the previous 3 mo.

Study Design

ICS treatment was stopped following a 2- to 4-wk run-in during which the maintenance dose remained unchanged. Patients were then reviewed weekly until loss of control (LOC) developed or for a maximum of 6 wk. When LOC occurred, patients were seen within 24 h. The visit at which ICS therapy was stopped was designated visit 1, the final visit of the study was designated visit F, and the visit immediately prior to the final visit was designated visit P (penultimate).

Criteria for LOC were as follows:

1. A fall in the mean (over last 7 d) morning peak expiratory flow rate (PEFR) of greater than 10% from baseline, or a fall in either morning or evening PEFR on two consecutive days to 80% of baseline or less, or

2. Mean daily bronchodilator use of greater than three puffs more than during run-in, or

3. Nocturnal wakening with asthma symptoms on three nights or more per week greater than during the run-in, or

4. Asthma symptoms that were disagreeable or distressing.

Study Procedures

Diurnal PEFR, bronchodilator use, and symptom scores were recorded in a daily record card. Measurements obtained at each study visit are shown in Table 1. eNO was measured prior to all other study procedures using a calibrated chemiluminescence analyzer with on-line measurement of single exhalations according to a standard protocol (21, 22), with the exception of flow rate (250 ml/s). eNO levels were read at the plateau corresponding to 70-80% of the CO₂ curve. Spirometry was measured using a rolling seal spirometer.

AHR to hypertonic saline (4.5%) was measured using a modified standardized protocol (23, 24). Spirometry was performed 1 min after saline nebulization, and patients were encouraged to produce sputum between inhalations. The challenge was discontinued when a 20% fall in FEV₁ occurred or a cumulative inhalation time of 20 min was reached. The PD₁₅ was calculated as the cumulative dose of saline causing a 15% fall in FEV₁.

If a 20% fall in FEV₁ occurred salbutamol was administered and sputum induction continued until an adequate sputum sample was obtained. The whole specimen (sputum plus saliva) was analyzed using a standardised method (25). Cell viability was assessed by the trypan blue exclusion test, and a cell count was performed by hemocytometer. Cytospin slides were stained with May-Grunwald-Giemsa stain and a total of 400 nonsquamous cells were counted on two occasions. Where the difference between the two counts was greater than 10% for any cell type then the count was repeated twice more and the mean for all four was recorded.

Ethical Considerations and Safety

Each patient's asthma control was monitored closely throughout the study. All patients were provided with an individualized self-management plan, an emergency card, and a supply of prednisone tablets. Patients had 24-h access to one of the study investigators via the hospital paging system. For ethical reasons, LOC criteria included symptoms that were disagreeable or distressing irrespective of PEFR changes. Ethical approval was obtained from the Otago Ethics Committee and informed consent was obtained from all study participants.

Statistical Analysis

Sources of variability in eNO measurements during run-in were estimated using variance components methods (26). Associations between eNO and other measurements were analyzed using the rank correlation coefficient because it does not depend on the measurement scale. Regression methods were used to compare airway inflammation parameters in patients who lost control with those who did not, and to adjust the comparisons for baseline differences. The prognostic and diagnostic utility of eNO was evaluated and compared with other measures using methods for constructing receiver-operator characteristic (ROC) curves (27). Exhaled nitric oxide and PD₁₅ measurements were analyzed using logarithmic transformations to remove the skew in these data; other measurements were analyzed without transformation. The results of these analyses did not change following the transformations.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Run-in Data

Seventy-eight patients entered the study. Demographic data are given in Table 2. The mean eNO during run-in was 9.38 ppb (95% reference range 2.72-32.35). The coefficients of variation for eNO were measured over four visits during the run-in phase of the study. The within-patient within-sitting coefficient of variation was 4.1% and the within-patient between-sitting variation was 10.5%. eNO was not related to ambient NO (range 0-234 ppb, p = 0.23) and so no corrections were made for ambient NO measurements.

Loss of Control

Sixty patients (77.9%) developed loss of control according to predetermined study criteria. The median time to LOC was 17 d (95% CI: 14, 28). Twenty-two patients developed LOC within 1 wk of corticosteroid withdrawal. The frequencies with which LOC criteria were met were fall in PEFR, 40; increased bronchodilator use, 15; increased nocturnal waking, 10; distressing symptoms, 39. Nine patients had LOC on the basis of distressing symptoms alone, two on the basis of increased reliever use alone. Thirty-two patients fulfilled two or more criteria at the time of LOC. Sputum induction was performed in all patients at visits 1 and F. Adequate sputum samples were obtained in 71 of 77 patients at visit 1 (92%), and in 54 (90%) and 15 (88%) of the LOC and no LOC groups at visit F, respectively.

The LOC group experienced a 2.16-fold increase in eNO between visits 1 and F, which was significantly greater than the 1.44-fold increase for the no LOC group (p = 0.004) (Table 3). There were also significant differences between LOC and no LOC groups for the fall in mean morning PEFR (13% versus 1%, p < 0.0001) the decrease in FEV₁ (mean fall of 11.9% predicted compared to 2.6% predicted, p = 0.001), the increase in sputum eosinophils (4.73-fold increase compared with 2.05-fold, p = 0.044), and the decrease in saline PD₁₅ (0.8 doubling doses compared with 0.03 doubling doses, p = 0.001).

Correlations between Study End Points

There were highly significant correlations between the changes in eNO that occurred between visits 1 and F and the changes in symptoms, lung function, sputum eosinophils, and AHR to HSC that occurred over the same period (Table 4). In general, the correlations between single measurements of eNO and these same parameters were lower than those seen with changes over time and were not consistently significant (Table 4).

Predictive and Diagnostic Values

The ability of eNO measurements to predict upcoming LOC was assessed in three ways: first, using the baseline eNO measurement (visit 1), second, using the measurement of eNO at the visit immediately prior to LOC (the penultimate visit, visit P), and third, using the change in eNO that occurred between visit 1 and visit P. Analysis of receiver-operator curves (ROCs) demonstrated parity between these different approaches. The curves were similar whether absolute or proportional changes in eNO were used. The sensitivities, specificities, and positive and negative predictive values at relevant cut points also showed that no one prognostic indicator was clearly superior (Table 5). Specific eNO cut points evaluated included 10 ppb, 15 ppb, and a 60% increase over the baseline mean (the upper limit of the 95% reference range for weekly variation in eNO over the run-in period).

The prognostic utility of eNO was also compared with other indices of airway inflammation, specifically single measurements of responsiveness to hypertonic saline (PD₁₅ less than 12 ml) and sputum eosinophils (greater than 4%) obtained at visit 1 (Figure 1, top panel, and Table 5). Compared with eNO, no measurement was clearly superior. Similarly, the prognostic value of changes in FEV₁% predicted, daily PEFR variation, symptom scores, and bronchodilator use between visit 1 and visit P was evaluated. None of these clinical parameters was found to be superior to eNO in predicting LOC (Figure 1, bottom panel).

View larger version (15K): [in this window] [in a new window]   Figure 1.   (A) Receiver-operator curves for eNO (proportional change between visit 1 and visit P), percentage eosinophils at visit 1, and saline PD15 at visit 1 for predicting upcoming loss of control. (B) Receiver-operator curves for the changes in eNO, FEV1, peak expiratory flow rate (PEFR), bronchodilator use, and symptom score between visit 1 and visit P for predicting upcoming loss of control.

The ability of eNO to diagnose LOC was also assessed using eNO measurements at visit F (Table 6). Both the single measurements of eNO at visit F and the change between visit 1 and visit F were evaluated. As for the assessment of prognostic utility, there was no clearly superior eNO measurement, and the performance of eNO was comparable to that based on sputum eosinophil counts and saline PD₁₅ measurements.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This is the largest longitudinal study to date in which the utility of eNO measurement in asthma management has been assessed. Our results document the usefulness of eNO for predicting and diagnosing poorly controlled asthma compared with other currently used markers of airway inflammation and clinical parameters. Regardless of the way in which eNO measurements were analyzed (absolute values, absolute changes, or proportional changes from baseline) the results were similar. eNO was associated with a positive predictive value (PPV) of between 80 and 90% for predicting and diagnosing LOC. On the whole, changes in eNO over time had higher PPV, sensitivities, and specificities both for predicting and diagnosing LOC than did single measurements. For example, an increase in eNO between visit 1 and visit P of more than 60% over baseline had a PPV for predicting LOC of 83% (sensitivity 50%, specificity 65%). A similar increase between visit 1 and visit F had a PPV for diagnosing LOC of 87% (sensitivity 68%, specificity 65%). In comparison, for a single measurement of greater than 15 ppb obtained at visit 1, when patients were still taking inhaled steroids, the PPV was 88% (sensitivity 25%, specificity 88%) for predicting LOC within 1 wk.

Overall, using the cut points selected, these outcomes reflect poor sensitivity but good specificity for eNO measurements. Nevertheless, they compare favourably with the usefulness of the other, more elaborate techniques of sputum induction and hypertonic saline challenge used in this study to assess deteriorating asthma. For example, we found that a doubling dose increase in saline PD₁₅ was marginally better than the other measurements for diagnosing LOC, with a PPV of 95% (sensitivity 41%, specificity 94%). This compared with a PPV of 84% for a 4% change in eosinophils (sensitivity 51%, specificity 64%) and with a PPV of 91% for an increase in eNO from baseline to LOC of 10 ppb (sensitivity 48%, specificity 82%). However, as a procedure, measuring eNO has the advantage of being quick and easy to perform, making it a suitable test for use in the clinical rather than the research setting. This becomes all the more important given our finding that changes in eNO have higher sensitivity and specificity for changes in clinical status than do single measurements, implying the need for repeated tests. Interestingly, eNO proved to be comparable to other more conventional measurements such as FEV₁, peak flows, symptom score, or daily reliever use in predicting upcoming LOC (see Figure 1) despite the fact that some of these other parameters were used in the definition of LOC.

Our results confirm earlier findings that eNO levels are elevated in unstable asthma (4, 6), and point to it as being a useful marker of airway inflammation. Previous studies have yielded inconsistent data regarding the correlation between eNO and sputum eosinophils because of the confounding effect of inhaled steroid use (15, 16, 28). This was not the case in the present study in which data were obtained from patients in whom maintenance inhaled steroid therapy had been temporarily withdrawn. Highly significant correlations between eNO and sputum eosinophils (rank correlation 0.62, p < 0.0001) and saline PD₁₅ (rank correlation 0.41, p < 0.0008) at a single point in time (visit F) were obtained, although the correlations for symptoms and lung function were nonsignificant (see Table 4). More importantly, when measured longitudinally the changes in eNO correlated significantly not only with changes in the other markers of airway inflammation but also with measurements of airway caliber and symptoms (see Table 4). These findings provide additional support for the use of eNO measurements as a tool in the assessment of airway inflammation and long-term asthma control.

Our study was designed to assess the usefulness of eNO in a clinical context. For this reason, loss of control criteria were prospectively based on a combination of peak flow and symptom changes as used in clinical practice. Evidence that these criteria were relevant and appropriate is provided by the significantly greater changes in FEV₁, sputum eosinophils, and saline PD₁₅ seen in the LOC group. Thus the subsequent comparisons between LOC and no LOC groups were valid. During our study eNO was measured using an exhalation flow rate of 250 ml/s, which is significantly more than current guidelines suggest (22). Our study was commenced before these guidelines were published. eNO has been shown to be flow dependent (29), and at lower flow rates the differences between healthy and asthmatic individuals are increased (30). Thus it is possible that at a lower expiratory flow rate (e.g., 50 ml/s) differences between patients who did and did not experience LOC might have been greater. If this were the case, then the sensitivities and specificities for eNO as a diagnostic test may be better than we have reported.

A number of investigators have sought to evaluate the role of eNO in assessing long-term asthma control. In the present study we chose to withdraw ICS in an attempt to induce a deterioration in asthma control and mimic an exacerbation. This implies that our findings may not strictly apply to patients receiving maintenance ICS. However, the usefulness of eNO in the presence of ICS is supported by Stirling and coworkers who found higher levels of eNO in patients with greater asthma severity irrespective of steroid use (17). Further, Kharitonov and coworkers have demonstrated that a simple reduction in the dose of ICS increases eNO (7) even in the absence of significant change in peak flow variability or spirometry. Using a study design similar to our own, but with incomplete withdrawal of ICS, Jatakanon and colleagues (31) reduced the dose of inhaled budesonide to 200 µg/day (less than one-fourth of the usual maintenance dose) in 15 patients with asthma.

Patients were followed for a maximum of 8 wk. Just under half developed an exacerbation. There was a parallel increase in eNO and sputum eosinophils, which correlated over time with changes in FEV₁ and -agonist use. The authors suggested that changes in sputum eosinophil numbers were superior to eNO in predicting loss of asthma control. However, because of the small study numbers a quantitative assessment of the predictive values of the changes in eNO and sputum eosinophils was not possible. In another longitudinal study, Baraldi and coworkers (32) measured eNO in children with asthma before, during, and after the pollen season. A rise in eNO was seen during the pollen season consistent with the presumed increase in airway inflammation that occurred with allergen exposure. This occurred despite the fact that over one-third of the children were taking inhaled steroids and in the absence of changes in FEV₁. These authors also concluded that eNO might be useful in the longitudinal assessment of asthma.

Asthma is a complex disease whose symptoms are dependent on the severity of airway inflammation, airway remodeling, and AHR. There is no firm agreement as to whether therapeutic intervention ought to be aimed at controlling symptoms alone, optimizing lung function, or minimizing airway inflammation, and hyperresponsiveness (33). A recent study has provided evidence that when antiinflammatory therapy is tailored to improve AHR, clinical outcomes are also improved (18). Our results suggest that a similar approach may be valid for eNO. Our data indicate that at an exhalation flow rate of 250 ml/s, an absolute value for eNO of 15 ppb or greater, or an increase of more than 10 ppb or 60% over baseline, is a useful threshold for the detection of ongoing airway inflammation, and that it also positively predicts the advent of breakthrough symptoms. Unfortunately the absence of these changes does not preclude the possibility of deteriorating asthma. Further studies are needed to confirm that the clinical application of eNO measurement is worthwhile in optimizing asthma management.