Biomarker-based Asthma Phenotypes of Corticosteroid Response

Abstract

Background

Asthma is a heterogeneous disease with different phenotypes. Inhaled corticosteroid (ICS) therapy is a mainstay of treatment for asthma but the clinical response to ICS is variable.

Objective

We hypothesized that a panel of inflammatory biomarkers i.e. F_ENO, sputum eosinophils and urinary BromoTyrosine (BrTyr) might predict steroid responsiveness.

Methods

The original study, from which this analysis originates, comprised 2 phases: a steroid naïve phase 1 and a 28-day trial of ICS (phase 2) during which times, F_ENO, sputum eosinophils, and urinary BrTyr were measured. Response to ICS was based on clinical improvements including: ≥12% increase in FEV₁; ≥0.5 point decrease in Asthma Control Questionnaire; and ≥2 doubling dose increase in provocation concentration of adenosine 5′-monophosphate causing a 20% fall in FEV₁ (PC₂₀ AMP). Healthy controls were also evaluated in this study for comparison of biomarkers to asthmatics.

Results

Asthmatics had higher than normal F_ENO, sputum eosinophils and urinary BrTyr at steroid naïve phase and after ICS. After 28-day trial of ICS, F_ENO decreased in 82% of asthmatics, sputum eosinophils decreased in 60% and urinary BrTyr decreased in 58%. Each of the biomarkers at steroid naïve phase had utility for predicting steroid- responsiveness, but the combination of high F_ENO and high urinary BrTyr had the best power (13.3 fold; p<0.01) to predict a favorable response to ICS. However, the magnitude of decrease of biomarkers was unrelated to the magnitude of clinical response to ICS.

Conclusion

A noninvasive panel of biomarkers in steroid naïve asthmatics predicts clinical responsiveness to ICS.

INTRODUCTION

Inhaled corticosteroids (ICS) are the mainstay of treatment for asthma. However, a considerable proportion of asthmatics do not respond to ICS based on lung function(1) and/or other clinical outcomes. The variability in response is attributed to different mechanisms underlying the airway inflammation (2–4). Biomarkers relevant to the underlying pathophysiological process, and/or the response to treatment, would be useful in personalizing care of the asthmatic patient (5).

Over the last decade, fractional exhaled nitric oxide (F_ENO) and sputum eosinophils have been used as biomarkers of airway inflammation and predictors of steroid-responsiveness. F_ENO is correlated with airway eosinophilia(6) and associated with airway hyper-responsiveness(7). F_ENO in healthy and asthmatic populations overlap, but F_ENO is higher in asthmatics as compared to healthy individuals(8, 9). Further, studies indicate that high F_ENO in asthmatics indicate an at-risk phenotype for exacerbation and predict clinical response to inhaled or oral corticosteroids(9).

Eosinophils are important effector cells in asthma, and elevated numbers in the sputum and peripheral blood are well-recognized as biomarkers of active atopic inflammation(10). Levels of eosinophilia identify clinical asthma phenotypes, e.g. Eosinophilic (EA) and Non-Eosinophilic Asthma (NEA)(10). There is a relationship between sputum eosinophils and exacerbation on withdrawal of steroids(2, 11). Thus, measurement of sputum eosinophilia represents a possible tool for adjusting asthma therapy to reduce exacerbations, and is related to measures of airflow obstruction and bronchial hyper-responsiveness(12). Upon activation, eosinophils undergo respiratory burst, generating high levels of reactive oxygen species (13), eicosanoids, platelet-activating factor and cytokines. Eosinophil peroxidase (EPO) is unique in its ability to convert respiratory burst generated hydrogen peroxide into hypobromous acid, a reactive brominating oxidant that modifies protein tyrosine residues forming 3-bromotyrosine (BrTyr)(13, 14). Thus BrTyr is a biochemical fingerprint of eosinophil activation, and this highly stable product can be detected in blood and urine.

Despite many studies evaluating biomarkers in asthma, none have specifically set out to compare the variance of biomarkers in response to treatment or evaluated the utility of biomarkers in combination to define treatment response phenotypes. Several studies describe correlations between F_ENO and sputum eosinophils(6, 7, 15). However, others show that the sensitivity and specificity of F_ENO as a predictor of sputum eosinophilia are modest, and, indeed, the relationship between F_ENO and eosinophilia appears to be independent of asthma control(16). Furthermore, anti-IL5 therapy decreases sputum eosinophils, but does not affect F_ENO, suggesting that biomarkers can provide unique information about clinical responsiveness and mechanisms of inflammation (5, 17). In this context, the combination of high urinary BrTyr and F_ENO were found to be associated with greater odds of asthma, but only urinary BrTyr was associated with prediction of future asthma exacerbations in both pediatric and adult cohorts(18, 19). Altogether, current data indicate that the information provided by the biomarkers F_ENO, sputum eosinophils and urinary BrTyr, is not necessarily duplicative. They provide distinct insights into the underlying pathophysiological mechanisms of disease and effects of treatment, but perhaps more importantly may provide for biomarker-based phenotyping of clinical responders to treatments.

The purpose of this study was to determine if a panel of inflammatory biomarkers (F_ENO, sputum eosinophils and/or urinary BrTyr) might accurately predict clinical responsiveness to ICS. Biomarkers were measured in steroid-naïve asthmatics in comparison to healthy controls, and again after ICS. Cut-points for each biomarker and combinations of biomarkers that predict steroid responsiveness were determined. Finally, the change in each biomarker in response to ICS was evaluated in order to investigate the correlation of steroid effect on biomarker and clinical response.

METHODS

Study population

Data and samples originated from 46 stable persistent asthmatics between 18 and 75 years old who were enrolled in a previously described study(20). Forty healthy control subjects were enrolled for comparison of baseline values of inflammatory biomarkers. Exclusion criteria for enrollment included respiratory tract infection in the preceding four weeks; more than ten pack year smoking history or smoking in the previous three months; use of oral prednisone in the previous three months; history of life threatening asthma; forced expiratory volume in one second (FEV₁) less than 50% predicted; other pulmonary disease; significant co-morbidity likely to influence the conduct of the study; pregnancy and breast feeding.

Study design

The original study(20), from which this secondary analysis originates, comprised two phases: Steroid Naïve (Phase 1) and Inhaled Steroid Treatment, (Phase 2), an open label trial of inhaled fluticasone (500ug twice daily for 28 days).

Phase 1 – Steroid Naïve

To achieve steroid naïve state, asthmatics were withdrawn from their inhaled corticosteroids and long-acting β-agonists for 28 days or until “loss of control” (LOC) occurred. Individualized criteria for LOC was based on modification of criteria developed by Jones et al (15) and included any one of the following criteria: (1) decrease in mean morning peak expiratory flow (PEF) by ≥10%, (2) decrease in two consecutive am or pm PEFs by >20%, (3) increase in average daily bronchodilator requirement by ≥4 puffs, (4) increase of ≥2 nights in nocturnal wakening because of asthma, or (5) experience of asthma symptoms that are distressing/intolerable(14). At LOC or after 28 days, whichever came sooner, all asthmatics underwent evaluation by measurement of lung functions(21), bronchial hyperresponsiveness to adenosine-5′-monophosphate (AMP)(22), Asthma Control Questionnaire(23) and biomarkers [F_ENO(24), sputum eosinophils(25), and urinary BrTyr measurements(18, 19)].

Phase 2 – Steroid Treatment

During the Steroid Phase, asthmatics were given fluticasone (Flixotide, GlaxoSmithKline, Greenford, UK) 500 micrograms, twice daily by inhalation via a spacer for 28+ days during which they completed a daily diary. After steroid treatment, subjects underwent evaluation by measurement of lung functions(21), bronchial hyperresponsiveness to AMP(22), Asthma Control Questionnaire(23) and biomarkers [F_ENO(24), sputum eosinophils(25), and urinary BrTyr measurements(18, 19)

Defining clinical responsiveness to ICS

Steroid clinical responsiveness was defined as one or more of the following: ≥12% increase in FEV₁ (35); ≥0.5 point decrease in Asthma Control Questionnaire (23), ≥2 doubling dose increase in provocative concentration of AMP causing a 20% fall in FEV₁ (PC₂₀AMP)(22).

Study procedures

A shortened 6-item version of the Asthma Control Questionnaire, a validated questionnaire for assessing asthma control, that excluded measurement of FEV1, was used (23, 26). Each item is scored on a 7-point scale (0–6), and a minimal clinically important change of 0.5 in the mean of the 6 items would justify a change in the patient’s treatment (in the absence of undue side effects or excessive costs)(23).

Spirometry was performed using a rolling seal spirometer (Sensor Medics Corporation, Yorba Linda, CA) in accordance with ATS/ERS guidelines(21).

Bronchial hyperresponsiveness to adenosine-5′-monophosphate (AMP) was performed using the standardized protocol of Polosa, R., et al.(22). Briefly, on each challenge day AMP doses (ranging from 0.59mg/ml to 300mg/ml) were freshly prepared. Increasing doubling concentrations of AMP were delivered by a nebuliser connected to a breath-activated dosimeter (Morgan, Kent, UK) at 5 minute intervals and spirometry was performed. The provocative concentration that caused a 20% fall in FEV₁ (PC₂₀AMP) was determined by linear interpolation of the dose-response curve. AMP challenges in which a 20% fall in FEV₁ was not achieved were assigned a PC₂₀AMP of 1200mg/ml.

F_ENO was measured using a chemiluminescence analyzer (NiOX MINO; Aerocrine, Stockholm, Sweden) before any forced expiratory maneuvers according to current guidelines at an exhaled flow rate of 50 ml/ second(24).

After sputum induction(27), sputum eosinophil counts were obtained by using the standardized protocol of Fahy et al. (25). Briefly, total cell differential was obtained by counting 400 non-squamous cells. All cell counts were read and confirmed by two trained observers. A cut-point of ≥2% was used to define eosinophilic asthma, <2% to define Noneosinophilic asthma(3).

Urinary bromotyrosine was assayed as previously reported using stable isotope dilution HPLC with on-line electrospray ionization tandem mass spectrometry(18, 19). Briefly, synthetic [13C6]-BrTyr was used as an internal standard and [13C9,15N1]-tyrosine was included to simultaneously monitor for potential artificial generation of analyte. Amino acids in urine were separated over a Prodigy C18 column (150 × 2.0 mm, 5 mm, Phenomenex, Torrance, CA, USA), at a flow rate of 0.2 ml/min using a discontinuous gradient generated with mobile phases comprised of 0.2% formic acid in water (solvent A) and 0.2% formic acid in acetonitrile (solvent B). The mass spectrometer was operated in positive ionization mode. Multiple-reaction monitoring (MRM) was used to detect unique precursorà product transitions of both 79Br and 81Br isotopologues of each bromotyrosine (natural abundance and its isotopologues) using unique mass-charge ratios (m/z) for the molecular cation [MH]⁺ precursor and product ions as follows: [79Br]Bromotyrosine m/z 260à214; [81Br]Bromotyrosine m/z 262à216; [79Br,13C6]Bromotyrosine m/z 266à220; [81Br,13C6]Bromotyrosine m/z 268à222; [79Br,13C9,15N1]Bromotyrosine m/z 270à223; [81Br,13C9,15N1]Bromotyrosine m/z 272à225. Differences in urinary dilution were adjusted for by spot urinary creatinine (Cr) concentration with urinary BrTyr reported in BrTyr ng/mgCr. Under the conditions employed for assay, no artificial bromination was detected, average spike and recovery was 101% and range from 98%–105%; assay precision of <7% was noted across all concentrations ranges examined(18, 19).

Atopy was determent by skin prick testing and defined by at least one positive reaction (a weal of >2mm) to the following allergens: cat pelt, grass mix, and house dust mite (Hollister-Stier Laboratories, LLC Spokane, WA99207, USA)].

Whole blood IgE was measured by Fluorescent Enzyme Immunoassay ImmunoCAP Total IgE (Thermo scientific, www.phadia.com) by Canterbury healthy Laboratories, Christchurch, New Zealand.

Study Measurements and Statistical Analyses

The characteristics of the study population are described using mean and Standard Error. To evaluate significant changes in urinary BrTyr, sputum eosinophils, or F_ENO in response to ICS treatment, paired T-test was utilized to compare values obtained during Steroid Naïve and Steroid Phases. To evaluate urinary BrTyr, sputum eosinophils, and F_ENO as biomarkers for inhaled corticosteroid clinical responsiveness, cut-points were used. At Steroid Naïve Phase, cut-points of sputum eosinophils ≥ 3%(28) and FENO ≥ 35 ppb(29) were based on published data. A cut-point for urinary BrTyr of 0.45 ng/mg Cr was selected as that which maximized the significance of the two-group Wilcoxon rank sum test comparison (based on the values at Steroid Naïve and Steroid Phases). Clinical responsiveness was defined as significant improvements in ACQ, FEV₁ and PC₂₀AMP based on cut points obtained from international guidelines. A comparison of clinical responsiveness between subjects defined by the biomarker cut-points was made by using the Wilcoxon rank sum test. To further describe the relationships observed in the two-group comparisons (two groups: lower or higher than the biomarker’s cut-point), each clinical responsiveness outcome was assessed with respect to a threshold value, yielding a dichotomous form of the outcome. Logistic regression analyses were then used for each biomarker to estimate the odds ratio (OR) for the association between the biomarker level and the likelihood of achieving a clinical response at or above the threshold. Receiver Operating Characteristic (ROC) curves and their corresponding Areas Under the Curve (AUCs) described visually the relationships between dichotomized forms of responsiveness with respect to thresholds of one, two and/or three clinical outcomes, and the continuous versions of urinary BrTyr, sputum eosinophils and F_ENO.

Ethical Considerations and Patient Safety

Ethical approval was obtained from the Lower South Regional Ethics Committee of New Zealand (LRS/06/11/056, LRS/06/12/059)(20).

RESULTS

Subject Characteristics

This study included a subset of 46 asthmatics from a prior cohort, who had completed the steroid withdrawal and steroid phase (ICS treatment). In addition, healthy controls were studied. Baseline characteristics are shown in Table 1. In this cohort, 80.4% of the asthmatics were atopic asthmatics [Table 1], and 65.2% were classified as eosinophilic asthma (EA) [Table 1, Table E1]. Age, gender, weight and height were not significantly different between controls and asthmatics [Table 1].

Table 1.

Study Population

	Asthmatics (Steroid naïve phase)	Healthy Controls
Gender (F/M)	29/17	21/19
Age (yrs)	39.8 (2.1)	35.4 (1.7)
Height (m)	1.70 (0.01)	1.72 (0.01)
Weight (kg)	78.5 (2.5)	86.7 (4.1)
Duration of asthma (yrs)	22.4 (2.4)	N/A
On ICS at baseline (no/yes)	16/30	N/A
BDP equivalent (μg)	1007.7 (93.5)
Atopy (Y/N)#	37/9	11/29*
Serum IgE levels (kU/mL)	451 (163)	86 (21)*
Eosinophilic asthma (Y/N)	30/16
Lung Functions
FEV1 (l)	2.54 (0.12)	3.38 (0.12)*
FEV1 (%)	76.2 (3.0)	95.0 (2.0)*
FVC (l)	3.97 (0.15)	4.2 (0.2)*
FVC (%)	98.5 (2.1)	97.1 (2.3)
FEV1/FVC	0.66 (0.02)	0.82 (0.01)*
PC20 AMP mg/ml	143.48 (51.8)	N/A
Asthma Questionnaires
ACQ (points)	1.60 (0.15)	N/A

Results are presented as mean (SE).

^#Atopy defined by at least one positive reaction (a weal of >2mm) to the following allergens: cat pelt, grass mix, and house dust mite.

^*P<0.05 between control and asthma.

BDP=beclomethasone dipropionate; FEV₁=Forced expiratory volume in 1s; FVC= Forced vital capacity; ICS=inhaled corticosteroid; PC₂₀ AMP= provocative concentration of adenosine 5′-monophosphate resulting in a 20% reduction in FEV₁; ACQ= asthma control questionnaire; EA=Eosinophilic Asthma, defined as more than 2% sputum eosinophils in steroid naïve phase; NEA= Non Eosinophilic Asthma

The effect of inhaled corticosteroids on inflammatory biomarkers

Urinary BrTyr, F_ENO and sputum eosinophils were measured at the Steroid Naïve and Steroid Phases. The inflammatory biomarkers (urinary BrTyr, and F_ENO) were significantly higher in all steroid naïve asthmatics as compared to control subjects (P<0.05, for all comparisons) [Table 2]. The asthmatics had variable changes in biomarker levels with ICS treatment: 82% of the asthmatics decreased F_ENO, 60% decreased sputum eosinophil levels and 58% decreased urinary BrTyr levels after ICS. The asthmatics who did not decrease in biomarkers, did have baseline F_ENO and urinary BrTyr higher than healthy controls, indicating that the lack of decrease in biomarkers was not due to having ‘normal’ levels.

Table 2.

Measurements of F_ENO, sputum eosinophils and urine BrTyr in steroid naïve and steroid treated Asthmatics and Controls

	Controls	All Asthma			Asthmatic with decrease in Biomarkers				Asthmatic with no decrease in Biomarkers
		Steroid Naïve	Steroid treatment	Paired T-test	Steroid Naïve	Steroid treatment	Paired T-test	Responder (%)	Steroid Naïve	Steroid treatment	Paired T-test
FENO (ppb)	15.5 (1.1)	56.5 (5.6) *	25.8 (2.3)	<0.0001	61.9 (6.5) *	22.7 (1.7)	<0.0001	82	29.9 (7.2) *	39.8 (9.5)	0.03
BrTyr (ng/mg Cr)	0.28 (0.04)	0.53 (0.1) *	0.49 (0.06)	0.43	0.63 (0.2) *	0.45 (0.1)	0.006	58	0.44 (0.1) *	0.56 (0.07)	0.006
Eosinophils (%)	N/A	18.9 (3.3)	9.9 (2.6)	0.012	26.5 (4.6)	9.0 (2.7)	<0.0001	60	4.8 (2.2)	12.8 (4.9)	0.08

Results are presented as mean (SE).

^*p<0.05 for all comparison to healthy controls including steroid naïve, steroid treated, with or without decrease in biomarkers

The changes in biomarkers in asthmatics treated with ICS was not concordant (Figure 1). More than 50% of asthmatics had both decrease in F_ENO and sputum eosinophils, but 30% of asthmatics had significant drop in F_ENO but no change in sputum eosinophils [Figure 1a]. There was less concordance between F_ENO and BrTyr, but most asthmatics dropped one of the two of these biomarkers with ICS, i.e. only 3.8% do not decrease either F_ENO or BrTyr [Figure 1b]. Urinary BrTyr and sputum eosinophils had the least concordant change [Figure 1c].

Figure 1. Changes in biomarkers in asthmatics treated with ICS.

The percent of asthmatics that decrease biomarkers with ICS treatment for 28 days proportionately shown by size of colored circles [blue, F_ENO; green, sputum eosinophils; red, urinary BrTyr], and percent of asthmatics that do not have decrease in either of the two biomarkers in each panel is shown by the black circles.

Biomarkers as predictors of Clinical Response to ICS

Receiver operating characteristic curve analyses were used to assess the utility of biomarker measurements during the Steroid Naïve Phase for the prediction of clinical responsiveness to steroid. Clinical responsiveness was defined by improvements in clinical outcomes (≥12% increase in FEV₁ and/or ≥0.5 point decrease in ACQ and/or ≥2 doubling dose increase in PC₂₀AMP). Figure 2 shows the ROC curves for F_ENO, sputum eosinophils and urinary BrTyr biomarkers as predictors of steroid response based on improvement in at least two of the three clinical outcomes. Furthermore, likelihood ratio tests confirmed that clinical responsiveness to steroid was associated with F_ENO (Wilcoxon P=0.004), sputum eosinophils (Wilcoxon P=0.001) and urinary BrTyr (Wilcoxon P=0.03). The odds ratio for dichotomized steroid responsiveness with improvement in 2 or 3 clinical outcomes with respect to baseline sputum eosinophils ≥3% was 9.20 (improvement in two clinical outcomes: 95% CI 2.31 – 43.09; p=0.001) and 10.50 (improvement in three clinical outcomes: 95% CI 1.73 – 203.7; p=0.002) respectively. Asthmatic individuals with baseline F_ENO values ≥35 ppm had 10.5 fold greater likelihood of response to inhaled steroids as measured by improvement in at least three clinical outcomes [OR: two clinical improvements: 3.43 (95% CI 0.93 – 413.54, p=0.004); three clinical outcomes: 10.50 (95% CI 1.73 – 203.7, p=0.014)]. The odds ratio with respect to baseline urinary BrTyr ≥0.45 ng/mg Cr was 6.22 (improvement in two clinical outcomes: 95% CI 1.22 – 47.94, p=0.031) and 1.5 (improvement in 3 clinical outcomes: 95% CI 0.3–7.3, p=0.619). Asthmatics with high F_ENO (≥35ppm) and high urinary BrTyr levels (≥0.45 ng/mg Cr) were 13 times as likely to have a positive response to steroid treatment as measured by two clinical outcomes [OR: two clinical outcomes, 13.3 (95% CI 2.1 – 130.6, p=0.007); three clinical outcomes: 7.2 (95% CI 0.9 – 153.5, p=0.076); Figure 3].

Figure 2.

ROC analysis describes ability of each biomarker i.e F_ENO, sputum eosinophils and urinary BrTyr to classify asthmatics who respond to inhaled corticosteroid as measured by 2 out of 3 clinical outcomes (≥12% increase in FEV₁ and/or ≥0.5 point decrease in ACQ and/or ≥2 doubling dose increase in PC₂₀AMP).

Figure 3.

ORs and 95% Cl for the association between a high level of biomarker (F_ENO ≥ 35ppm; sputum eosinophils ≥ 3%, BrTyr ≥ 0.45 ng/mg Cr) or biomarker combination [F_ENO AND BrTyr] and two positive clinical outcomes in response to ICS. Results shown represent the ORs (filled circles) and 95% Cl (lines) of having a steroid response versus having no steroid response. Asterisk indicates P<0.05 as determined by likelihood-ratio χ² test.

To evaluate if the mechanism of clinical response might be attributed to a biomarker inflammation pathway, we evaluated the numbers of asthmatics within subgroups defined by levels above the cut-points that are clinically relevant to each steroid response and who also experienced a decrease in biomarker. Interestingly 90% of the steroid-naïve asthmatics with F_ENO ≥ 35 ppm decreased F_ENO after ICS; 85% with sputum eosinophils ≥ 3% decreased sputum eosinophils after ICS; and 59% with urinary BrTyr ≥ 0.45 ng/mg Cr decreased urinary BrTyr after ICS. Although the majority of asthmatics did experience a decrease in biomarker, there was no association between the magnitude of decrease in biomarkers and magnitude of clinical response. However, the study may be underpowered to detect this type of association and a larger study may confirm such an association.

DISCUSSION

This is the first study to compare the utility of a panel of biomarkers, which identify the presence of atopic inflammation and oxidative stress, for prediction of clinical response to steroids. The effect of ICS on inflammatory biomarkers i.e. sputum eosinophils, F_ENO and urinary BrTyr was not uniformly concordant, although there were substantial parallel decreases among biomarkers. Each of the biomarkers had utility for predicting steroid-responsiveness; the combination of high F_ENO and high urinary BrTyr had particular power to predict a favorable clinical response to ICS with either improvement in ACQ, FEV₁ or airway reactivity.

This study allows direct comparison among biomarkers in response to ICS treatment. The majority of asthmatics had experienced a drop in F_ENO with ICS and this was by far the most consistent biomarker to decrease as compared to either sputum eosinophils or urinary bromotyrosine. F_ENO is traditionally considered a surrogate marker of eosinophilic asthma(6). Likewise, the end-products of protein bromination, e.g urinary BrTyr, is considered to reflect eosinophil activation(13, 14). The concept that the eosinophil plays a role in asthma is long established, as is the clinical benefit of corticosteroids in eosinophilic asthma in association with suppression of eosinophilic inflammation(10, 30). Treatment strategies based on sputum eosinophil counts lead to decrease in exacerbations without increased treatment, at least in adults(2, 31, 32). Despite the fact that these biomarkers provide information on atopic inflammation, there was only partial concordance in response to ICS of the three inflammatory biomarkers in this study. This may be related to the fact that F_ENO correlates with bronchial mucosal eosinophilia rather than luminal (sputum) eosinophilia (33). Epithelial inducible nitric oxide synthase (iNOS) has been shown to be the main determinant of F_ENO(34) and it is postulated that the decrease in F_ENO seen following steroid treatment may be due to an inhibitory effect of steroid on iNOS induction, perhaps irrespective of airway eosinophilia. Atopy may be an important contributor to F_ENO, independent of eosinophils, so that F_ENO may remain elevated in atopic individuals despite steroid suppression of eosinophilic airway inflammation(35). Finally, F_ENO measurement does not allow an estimate of alveolar and airway contributions(36) such that localization of this inflammatory signal within the lower respiratory tract is not possible. Other studies suggest discordance between biomarkers. Notably, the administration of mepolizumab in refractory eosinophilic asthma led to a decrease in sputum eosinophils, but no change in F_ENO(17). A study in children found marked discordance in the longitudinal relationship between sputum eosinophils and F_ENO(37). Finally, in another study, no correlation was found between urinary BrTyr and F_ENO(19). It is possible, therefore, that sputum eosinophils, F_ENO and urinary BrTyr are not as closely linked as previously thought, and that they may represent biomarkers of different underlying pathophysiological mechanisms.

Sputum eosinophil count has utility in the prediction of steroid response(2, 6, 28, 38, 39), in the prediction of loss of control(2, 40), and in titrating steroid dose to minimize exacerbations(32). However, sputum induction is unpleasant for the patient, sputum processing and cell analysis is a time consuming procedure requiring technical expertise, and the test is not widely available. Furthermore, the performance characteristics of sputum eosinophil count for the prediction of steroid responsiveness are modest. Combining sputum eosinophil count and F_ENO may improve this prediction. In this study, the finding of high sputum eosinophils was a good predictor of steroid responsiveness, whether defined by improvement in two (AUC 0.848) or three (AUC 0.749) clinical outcomes. Both sputum eosinophils and F_ENO individually, but not urinary BrTyr, had AUCs >0.7 indicating clinically significant utility in prediction of steroid responsiveness, whether defined by improvement in two or three clinical outcomes. Previously, we have shown that urinary BrTyr, a non-invasive marker of oxidant stress and eosinophil activation, is increased in asthma and predicts exacerbation in both pediatric and adult asthma populations(18, 19). The combination of high F_ENO and urinary BrTyr was associated with the greatest likelihood of clinical response to ICS. This suggests that these two biomarkers used in combination may have superior clinical utility in the prediction of steroid responsiveness, thus avoiding the need for sputum induction, processing and analysis, and the associated discomfort for the patient.

A limitation of this study is that the ICS trial was not placebo controlled. This was for reasons of safety. Seventy per cent of participants demonstrated loss of control within 28 days after steroid withdrawal. It would have been unethical to allow these individuals to proceed for a further 28 days beyond the point of loss of control on a placebo treatment. We do concede that, as a consequence, some of the changes seen in association with steroid treatment may be explained to some degree by regression to the mean. In particular, we cannot exclude the possibility that changes seen in sputum eosinophils did not reflect regression to the mean or variation with time, although the latter in particular seems unlikely given the reproducibility of sputum cell measurements over time(41). In addition, we were unable to assess the predictive power of biomarkers for other important outcomes, such as risk of exacerbations. Previous studies suggest that sputum eosinophils(42), F_ENO(9) and, at least in children, urinary bromotyrosine(19), may have a role in assessing the risk of exacerbation in asthma. Finally, although the majority of asthmatics did experience a decrease in biomarker, there was no association between the magnitude of decrease in biomarkers and magnitude of clinical response. However, the study may be underpowered to detect this type of association and a larger study may confirm such an association. Future studies will focus on evaluation of biomarker panels for assessment of risk of exacerbation and whether magnitude of change in biomarkers might predict the magnitude of clinical benefit with treatments.

Supplementary Material

Table E1: Baseline characteristics at the end of Steroid naïve phase for all subjects who completed Steroid treatment phase based on Eosinophilic Asthma

KEY MESSAGE.

This study shows that

• Asthmatics have variable and nonconcordant decreases in sputum eosinophils, F_ENO and urinary BrTyr in response to ICS.

• Each biomarker at baseline was predictive of clinical steroid-responsiveness; but the combination of high F_ENO and high urinary BrTyr had the greatest power to predict a favorable response to ICS.

ABBREVIATIONS

AMP

Adenosine 5′-Monophosphate

ACQ

Asthma Control Questionnaire

BDP

Beclomethasone DiPropionate

BrTyr

urinary bromotyrosine

EA

Eosinophilic Asthma

F_ENO

fractional exhaled NO

FEV₁

forced expiratory volume in one second

ICS

Inhaled Corticosteroid Therapy

LOC

Loss of Control

EA

Non Eosinophilic Asthma

PC₂₀AMP

Adenosine 5′-MonoPhosphate Provocative Concentration that causes a 20% fall in FEV₁