Urinary Eosinophil Protein X in Childhood Asthma: Relation with Changes in Disease Control and Eosinophilic Airway Inflammation

The aim of this study was to assess cross-sectional and longitudinal correlations between uEPX and other markers of asthma control and eosinophilic airway inflammation. Methods. We measured uEPX at baseline, after 1 year and after 2 years in 205 atopic asthmatic children using inhaled fluticasone. At the same time points, we assessed symptom scores (2 weeks diary card), lung function (forced expiratory volume in one second (FEV1)), airway hyperresponsiveness (AHR), and percentage eosinophils in induced sputum (% eos). Results. We found negative correlations between uEPX and FEV1 at baseline (r = −0.18, P = 0.01), after 1 year (r = −0.25, P < 0.01) and after 2 years (r = −0.21, P = 0.02). Within-patient changes of uEPX showed a negative association with FEV1 changes (at 1 year: r = −0.24, P = 0.01; at 2 years: r = −0.21, P = 0.03). Within-patient changes from baseline of uEPX correlated with changes in % eos. No relations were found between uEPX and symptoms. Conclusion. In this population of children with atopic asthma, uEPX correlated with FEV1 and % eos, and within-subjects changes in uEPX correlated with changes in FEV1 and % eos. As the associations were weak and the scatter of uEPX wide, it seems unlikely that uEPX will be useful as a biomarker for monitoring asthma control in the individual child.


Introduction
Eosinophilic protein X is produced by activated eosinophils and can be measured in blood, stools, and urine (uEPX) [1,2]. Elevated uEPX has been reported in asthmatics with or without allergy [3,4], with great overlap between healthy controls and asthmatics [5,6]. A relation with asthma symptoms and lung function has been shown, and uEPX increases signi�cantly during asthma exacerbations [7][8][9]. A reduction in uEPX has been reported aer starting inhaled corticosteroids [10]. erefore, it has been suggested that measuring urinary EPX may be useful for monitoring the effect of ICS in asthma [3,10,11].
Since measuring uEPX is completely noninvasive and independent of patient's cooperation, it has the potential to be a feasible tool in asthma management in young children. Several studies have demonstrated cross-sectional correlations between uEPX and conventional clinical markers of asthma control and airway in�ammation [3,4,12]. Previously, we reported these cross-sectional data measured 2 Mediators of In�ammation at baseline in part of the present study population [13]. Data on longitudinal changes in uEPX within subjects compared to changes in markers of asthma control or airway in�ammation were examined in only one study in 14 children [3].
We analysed cross-sectional and longitudinal uEPX data from a long-term follow-up study in children with atopic asthma [14]. We hypothesised that changes in uEPX would be related to changes eosinophilic airway in�ammation and could provide information that is additional to symptoms and lung function.

Subjects.
Asthmatic children took part in a randomised controlled multicentre trial (CATO: Children Asthma erapy Optimal) reported in detail previously [14]. Brie�y, 288 children with a documented clinical history of moderately severe asthma were recruited from 15 paediatric clinics including 7 university hospitals in the Netherlands. Al patients had a positive, class 1, or more radioallergosorbent test result for one or more airborne allergens.
During a run-in phase, all patients were treated with 100 or 250 g �uticasone bid depending on their equivalent treatment before the study. Children were selected on the basis of symptoms and/or AHR, and 210 children were randomised to a reference strategy (adjustment of treatment on symptom score) or an AHR strategy (treatment adjusted on the basis of AHR and symptom score). During the 2-year followup study, 3 monthly visits were performed. At each visit symptoms scores (diary card over the last 2 weeks), forced expiratory volume in one second (FEV 1 ), and methacholine challenge results were obtained, and medication (�ve levels: �uticasone 100 or 200 g per day or �uticasone/salmeterol 200/100, 500/100, or 1000/100 g per day (Flixotide Diskus or Seretide Diskus, GlaxoSmithKline, UK) adjusted according to the algorithms of the reference-and AHR strategy.
For the present study, �baseline� is de�ned as the moment of randomisation for the CATO study.

Symptom Scores.
Patients �lled in a diary card daily during 2 weeks before each 3-month visit. Cough, shortness of breath, and wheezing during night and day were recorded on a 4 point scale (0 = no symptoms to 3 = severe symptoms interfering with activity or sleep) [15]. e percentage of symptom-free days de�ned as score 0 for cough, wheeze, and shortness of breath was calculated.

AHR and Spirometry.
were measured during each clinic visit [16]. Study medication was stopped 36 hours before lung function and AHR testing. AHR was tested by methacholine challenge using a dosimeter method [17]. e provocation dose causing a 20% fall in FEV 1 from baseline (PD 20 ) was calculated from a log dose-response plot by linear interpolation.

Sputum Induction and
Processing. Sputum induction was attempted in 8 of the 15 participating centres. Sputum was induced according to a standardised method by inhaling an aerosol of hypertonic sodium chloride 4.5% w/v [18,19].
Differential cell counts of cytospins were performed and samples containing more than 80% squamous cells were excluded from the analysis [18].

Urinary Eosinophil Protein X.
Spot samples of urine were collected at randomisation (= baseline), aer a treatment period of 1 year, and aer 2 years (the end of the study), and immediately stored at −20 ∘ C. EPX was determined using a commercial enzyme-linked immunosorbent assay (ELISA) for human EPX in 50-fold diluted samples (Medical and Biological Laboratories, Nakaku Nagoya, Japan). e essay's sensitivity was 0.62 ng/mL. Urinary creatinine levels were measured by the alkaline picrate method (Roche, Mannheim, Germany). Urinary EPX concentrations were expressed as g per mmol creatinine.
2.6. Statistical Analysis. Repeated measurements Anova was used to evaluate associations between the various markers and uEPX levels. e same method was used to assess changes from baseline. With this method (SAS PROC MIXED) the information about the associations at the different study time points can be combined and the method allows for uEPX not being available at all three time points for each patient. In these analyses, uEPX values, % eosinophils in sputum, and �uticasone doses were transformed logarithmically in order to get approximate normal distributions and to reduce the effect of outlying observations, and consequently results for these parameters are expressed as geometric means. Percentage eosinophil values equal to 0 were replaced by 0.1% in order to allow the logarithmic transformation. In a substantial part of the study population, especially at time points T1 and T2, PD 20 was higher than the highest dose methacholine used (1570 g). For that reason PD 20 data were arbitrarily dichotomized in non-AHR (PD 20 > 300 g) and AHR (PD 20 < 300 g). Correlation coefficients calculated at the different time points are Spearman's rank correlations.

Results
Urinary EPX was measured at least once in 205 children. Patient characteristics at baseline are given in Table 1. In total 461, uEPX samples were analysed during the study. During the 2-year treatment period, the geometric mean uEPX signi�cantly decreased from 159 g/mmol to 104 g/mmol ( < 0 001) (Figure 1). Anova showed a signi�cant treatment effect aer 1 year: in the AHR strategy group, uEPX had decreased signi�cantly more than in the reference strategy group, with a 29% lower geometric mean EPX level at that time point ( = 0 03). No treatment effect was found at year 2 ( = 0 2 ). Further analysis showed that at 1 year, there was a signi�cant �uticasone dose effect, and this effect was not found at baseline or at year 2. At 1 year a twofold higher dose was associated with a 25% lower uEPX level (95% CI: 15%-34%; < 0 001 higher dose using Anova, no signi�cant treatment effect remained. (Tables  2 and 3). No correlations were found between uEPX and the percentage symptom-free days or changes in uEPX and changes in percentage symptom-free days. Anova also did not show signi�cant associations. Cross-sectional analysis showed weak negative correlations of the levels of uEPX and FEV (baseline: , ; at 1 year: ; at 2 years: , resp.). Anova showed that the signi�cant associations did not differ between the three time points and adjusting for ICS dose did not change this. Also the changes from baseline of uEPX and FEV were signi�cantly related with each other (at 1 year , and at 2 years; ) ( Figure  2). As compared to baseline, a tenfold increase in uEPX was associated with a decrease of FEV (% pred) of 2.8 percentage points (95% CI : 0.3 to 5.3;

Urinary EPX versus Parameters of Asthma Control
) and this decrease did not signi�cantly differ between the two time points ( ). A similar result was found when changes from the previous assessment were evaluated.
At all visits, uEPX levels were higher in patients with AHR (PD methacholine 300 g methacholine) than in patients without AHR. At 1 year this difference was statistically signi�cant, with 13% higher uEPX in patients with AHR ( ) ( , 1 , resp., at baseline and aer 2 years treatment). is effect remained aer adjustment for dose. With regard to percentage eosinophils in sputum a signi�cant correlation with uEPX was found at 2 years ( 1, 1, ). Anova showed a signi�cant overall association between uEPX and percentage sputum eosinophils. A twofold higher value of sputum eosinophils was associated with a 14% increased value of uEPX (95% CI: 6%-22%; 1), and this relation did not signi�cantly differ between the 3 time points ( 11), nor was the relation affected by ICS dose. Although changes from baseline of uEPX and percentage sputum eosinophils did not signi�cantly correlate with each other at the separate time points, a signi�cant relation was found combining both time points using Anova. A within-patient doubling of sputum eosinophils was associated with a mean increase of uEPX of 16% (95% CI: 0.1%-36%; ), and the ICS dose did not affect this relation. e individual variations however are large (Figure 3).

Discussion
We prospectively assessed changes in uEPX in relation to changes in asthma control within subjects over an observation period of 2 years. We found in atopic asthmatic children that changes in uEPX measured aer 1 and 2 years related signi�cantly with changes in FEV 1 and % sputum eosinophils, but not with changes in symptoms. Changes in uEPX correlated with changes in ICS dose.
Con�icting data have been reported about cross-sectional relationships between uEPX and FEV 1 . Previously we described the results of cross-sectional analysis at baseline in the present study population [13]. We found a weak inverse correlation between uEPX and FEV 1 . With the present study, we con�rmed the relation between uEPX and FEV 1 at different time points. Moreover, we showed that changes in uEPX and changes FEV 1 were related. is is in line with the signi�cant correlation between these parameters found by Lugosi et al. in 14 asthmatic children [3]. However the interval of their measurements was much shorter (1-2 months). It has been shown that correlations between eosinophilic airway in�ammation and lung function are weak and may only reach signi�cance in large populations [4,6,20,21]. Clearly FEV 1 is dependent on many more factors than eosinophilic in�ammation alone.
e correlation between uEPX and sputum eosinophils was weak, and this is in line with the relations between sputum ECP, sputum eosinophils, and uEPX described by Mattes et al. [12]. Weak correlations seem to be the rule when different markers of eosinophilic in�ammation are compared [22]. Moreover, uEPX is secreted by activated eosinophils, and it is therefore possible that uEPX provides information about eosinophil activation rather than numbers.
We found no relation between uEPX and symptoms. A signi�cant relation between uEPX and symptoms has been described during acute asthma exacerbations, spontaneously or provoked by ICS withdrawal [23]. Also, signi�cantly higher uEPX levels were found in symptomatic asthmatic children than in asymptomatic children [1,3,4]. However, a 6-month follow-up study in 14 children with mild asthma found, no signi�cant association was found [9]. So the severity of symptoms should be taken into account when correlating uEPX and symptoms. In our study population, the Mediators of In�ammation 5 symptom scores were low and this may explain the apparent discrepancy with these previous studies. It could well be that relations between uEPX and parameters of asthma control could have been in�uenced by other factors as well, including seasonal variations, severity of allergy, and asthma and atopic dermatitis [24,25]. e use of varying doses of ICS is another possibility, for which we have corrected.
What are the clinical implications of our �ndings ? As uEPX does not correlate with symptoms, but does correlate with sputum eosinophils, we could argue that uEPX might provide additional information on activity of eosinophilic airway in�ammation. However, the correlations were weak and the scatter of individual uEPX values wide. erefore, it seems unlikely that uEPX will be useful as a biomarker for monitoring asthma in the individual child.