Exhaled nitric oxide predicts control in patients with difficult-to-treat asthma

We aimed to evaluate the accuracy of baseline exhaled nitric oxide fraction (FeNO) to recognise individuals with difficult-to-treat asthma who have the potential to achieve control with a guideline-based stepwise strategy. 102 consecutive patients with suboptimal asthma control underwent stepwise increase in the treatment with maximal fluticasone/salmeterol combination dose for 1 month. Then, those who remained uncontrolled received oral corticosteroids for an additional month. With this approach, 53 patients (52%) gained control. Those who achieved control were more likely to have positive skin results (60.4% versus 34%; p = 0.01), positive bronchodilator test (57.1% versus 35.8%; p = 0.02) and peak expiratory flow variability ≥20% (71.1% versus 49.1%; p = 0.04). Conversely, depression was more frequent in those who remained uncontrolled (18.4 % versus 43.4 %; p = 0.01). An FeNO value ≥30 ppb demonstrated a sensitivity of 87.5% (95% CI 73.9–94.5%) and a specificity of 90.6% (95% CI 79.7–95.9%) for the identification of responsive asthmatics. The current results suggest that FeNO can identify patients with difficult-to-treat asthma and the potential to respond to high doses of inhaled corticosteroids or systemic steroids.

A sthma treatment ideally achieves a steady state of no symptoms and no exacerbations with minimum medication. In practice, many patients with severe disease will not be optimally controlled, and asthma surveys indicate that a high proportion of patients remain uncontrolled even while receiving adequate therapy [1,2]. Despite guideline-based asthma management goals, the Asthma Insights and Reality in Europe (AIRE) study of 2,803 European patients showed that 46% of them reported daytime symptoms, 30% had asthmarelated sleep disturbances at least once a week, 25% had unscheduled urgent care visit in the past year, 10% had an emergency room visit and 7% had an overnight hospitalisation [3].
There is no universally accepted definition of difficult-to-treat asthma. However, it is reasonable to consider it present when people have persistent symptoms and frequent exacerbations, despite being treated according to the Global Initiative for Asthma steps 4 or 5. Patients with difficult-to-treat asthma are a cause of concern because of impairment of quality of life, continued decreased lung function and adverse effects of high-dose corticosteroids. Escalation of therapy to maximum doses of inhaled corticosteroids and oral corticosteroids is widely accepted in poorly controlled chronic asthmatics [4]. However, there is significant heterogeneity in the response to steroids and patients may be committed to an ineffective and potentially harmful therapy [5].
In the past years, asthma specialists have been interested in looking for an easy-to-measure and reliable biomarker that could facilitate the assessment of the disease and could avoid the risk of too little or too much treatment. Exhaled nitric oxide fraction (FeNO), an indirect marker of eosinophilic airway inflammation that is easy to perform and reproducible, has been successfully employed for the guidance of asthma management [6][7][8].
This study was designed to focus on the practical problem of the patient with diagnosed asthma whose symptoms are not controlled despite high doses of inhaled corticosteroids and other regular therapy. In this prospective, observational study, we adjusted medication according to a stepwise approach, corresponding to the international guidelines, based on asthma control. The aim was to evaluate the accuracy of baseline FeNO levels to recognise individuals with difficult-to-treat asthma who have the potential to achieve control with the stepwise approach. Control was defined as a score of o20 in the Asthma Control Test (ACT). A secondary objective was to identify determinants of asthma control among variables related to clinical and functional patient characteristics.

Study design
This was an observational and prospective study carried out in the pneumology and allergology units of the Hospital Xeral-Calde, Lugo, Spain. All patients were seen and assessed by the same physician at visit 1 and every subsequent clinical visit. The protocol was run on an outpatient basis.

Patients
Consecutive patients were recruited for protocol evaluation between April 2008 and July 2009 if they had a difficult-to-treat asthma defined as not controlled asthma (ACT score ,20) despite minimal maintenance therapy of long-acting b 2 -agonist and high-dose inhaled corticosteroids (o800 mg beclometasone equivalent) for .3 months. All subjects had been free of exacerbations for at least this period. All of them were regularly followed at an outpatient asthma clinic and they were educated on the nature of the illness and the correct use of inhalers. The patients were advised to bring their maintenance mediation to the hospital and all of them denied nonadherence at the time of the first clinical assessment. Patients treated with oral corticosteroids or omalizumab were excluded. We also excluded those with a smoking history in excess of 10 pack-yrs. Inclusion criteria were relatively broad because the authors wanted to reflect usual clinical practice.

First visit
Patients were classified according to their clinical control (assessed by the ACT). In all cases, medical history, spirometry, bronchodilator test, methacholine test (if not contraindicated), FeNO, and ambulatory peak expiratory flow (PEF) measurement were recorded. A detailed structured clinical history was obtained for all subjects in order to rule out comorbid conditions (obesity, smoking, anxiety, depression, medications, oesophageal reflux, upper airways disease, vocal cord dysfunction, bronchiectasis, chronic obstructive pulmonary disease (COPD) and allergic bronchopulmonary aspergillosis) that could influence asthma control. Vocal chord dysfunction was diagnosed with endoscopy when clinically suspected. Oesophageal reflux was diagnosed by means of medical history (pH metering was not systematically performed). Patients with suspicion of bronchiectasis underwent high-resolution computed tomography. Anxiety and depression were assessed by use of Goldberg's short questionnaire. Allergic bronchopulmonary aspergillosis was investigated in patients with high immunoglobulin E levels.
Patients with suboptimal asthma control (ACT score ,20) with correct inhalation technique received increased therapy with maximal fluticasone/salmeterol combination dose (unless they were treated with such medication) and they were placed to return for a second visit 1 month later.

Second visit
Patients with previous suboptimal asthma control who received increased therapy were assessed for clinical control (via ACT). If they achieved asthma control, new spirometry and FeNO measurements were performed. If they did not achieve control, they underwent another increase of asthma therapy with 30 mg of deflazacort added to their inhaled medication and were placed to return for a third visit 1 month later.

Third visit
Patients were assessed for asthma control. New spirometry and FeNO measurements were recorded.
The time course of the protocol is shown in figure 1.

Measurements
Exhaled nitric oxide A single measurement was undertaken using a portable device (NIOX MINO; Aerocrine AB, Solna, Sweden) at a mouth flow rate of 50 mL?s -1 for 10 s. The sensor on the device was changed periodically, in line with the manufacturer's guidance.

Pulmonary function test
Spirometry was performed using a Datospir 120 spirometer (Sibelmed, Barcelona, Spain) in accordance with American Thoracic Society (ATS)/European Respiratory Society (ERS) guidelines [9]  ASTHMA AND ALLERGY the initial value. Bronchodilator reversibility was calculated as the percentage of change in FEV1 from baseline, 15 min after inhaling 400 mg of albuterol.
Airway hyperresponsiveness to methacholine (expressed as the provocative concentration causing a 20% decrease in FEV1 (PC20) was measured with a 2-min tidal breathing method adapted from COCKCROFT et al. [10]. After an initial nebulised saline challenge, subjects inhaled doubling concentrations, ranging from 0.038 to 19.2 mg?mL -1 of methacholine bromide, at 5-min intervals. Airway hyperresponsiveness to methacholine was defined as a PC20 f8 mg?mL -1 .
Diurnal peak expiratory flow rate (PEFR) was recorded in a daily record card.

Symptom assessment
Asthma control was assessed using the validated Spanish version of the ACT [11]. Patients subjectively evaluated the degree of impairment caused by their asthma during the preceding 4 weeks by responding to five questions using a five-point scale. The ACT is reliable, valid and responsive to changes in asthma control over time. A cut-off score of f19 identifies patients with poorly controlled asthma [12].

Goldberg's anxiety and depression test
This test was designed to be used by non-psychiatrists in clinical investigations. The score is based on responses of ''yes'' or ''no'' to nine depression and nine anxiety items (the full set of nine questions need to be administered only if there are positive answers to the first four), asking how respondents have been feeling in the past month. GOLDBERG et al. [13] considered patients with anxiety scores of o5 or with depression scores of o2 as having a 50% chance of a clinically important disturbance.

Statistical analyses
The baseline characteristics were compared between those who responded to increase of treatment with those who did not achieve control. Comparison of continuous data was made by t-test or Wilcoxon signed rank test, as appropriate. Unpaired t-test was used to compare means between groups and the paired test was employed when repeated measurements were made. Intergroup differences were evaluated by Chi-squared analysis or Fisher's exact test for categorical data.
In order to determine the concentration of FeNO capable of predicting asthma control after stepwise increase of therapy, we calculated sensitivity, specificity, positive predictive power and negative predictive power, with their 95% confidence intervals, for each FeNO measurement. With the resulting data, we generated a receiver operating characteristic (ROC) curve to find the best FeNO cut-off (value with the greatest sensitivity and specificity) and to calculate the area under the curve (AUC), with its 95% confidence interval, to estimate overall diagnostic accuracy. AUCs with a value close to 1 indicated excellent ability to discriminate. A positive likelihood ratio was calculated as sensitivity/(1-specificity) or true-positive rate/ false-positive rate. A positive likelihood ratio reflects increased odds of achieving control after a positive FeNO result. A negative likelihood ratio is (1-sensitivity)/specificity or falsenegative rate/true-negative rate and reflects reduced odds of achieving asthma control after a negative FeNO result. The ROC was fitted using the maximum likelihood fit of a binormal model by employing the web-based calculator JROCFIT [14].
Logistic regression models were used to assess the relationship between predictor variables (duration of the disease, positive reactions to skin-prick tests for common aeroallergens, clinical manifestations of atopic disease, presence of comorbidity, depression, body mass index (BMI), FeNO, FEV1 and daily PEF variability of o20%) and the event of interest (response to therapy).
A value of p,0.05 was considered significant.
Statistical analyses were made using the R statistical package (R Development Core Team).  1).

Patient demographics
Only one patient suffered an exacerbation that required oral steroids, but not hospitalisation, during the study. Another one was diagnosed with pneumonia (and admitted to hospital for this reason) while taking oral steroids between the second and third visit.
Baseline characteristics for the two groups (patients who achieved control and those who did not) are listed in

DISCUSSION
Our results demonstrate, for the first time, that FeNO levels might be predictive of response to a stepwise approach in patients with difficult-to-treat asthma. This study adds to previous research showing a clinical utility of FeNO measurements in asthmatic patients [6][7][8]. It has been demonstrated that FeNO correlates with eosinophilic inflammation measured using bronchial biopsies and induced sputum [15,16]. In addition, previous studies have shown that high numbers of sputum eosinophils were predictive of steroid response [17,18]. This underlines that steroid response is related to particular characteristics of airway inflammation. Conversely, FeNO is reduced by treatment with inhaled corticosteroids [19], but elevated levels of this biomarker were previously observed in patients with severe asthma despite corticosteroid treatment [20]. This might imply either steroid resistant inflammatory processes in the airway, or insufficient doses of anti-inflammatory medication. Theoretically, FeNO measurements might help us to identify individuals with persistent eosinophilic inflammation in which a steroid response is more likely. This hypothesis is supported by our results in a difficult asthma population, indirectly by those of SMITH et al. [8], who found that FeNO measurements provided a means of predicting steroid response in patients with undiagnosed respiratory symptoms, and also by the findings of LITTLE et al. [21], who have shown that response to oral steroids in asthma patients can be predicted in most cases by analysing this biomarker. Even with various expert-derived guidelines that provide asthma treatment strategies, many patients remain suboptimally controlled. In our series, 48% of the patients did not achieve control, as assessed by the ACT questionnaire, despite receiving the best available treatment and optimal management efforts. This figure is in accordance with the Gaining Optimal Asthma Control (GOAL) study, which showed that symptoms were uncontrolled in as many as 38% of patients with moderate-tosevere asthma, despite high doses of salmeterol/fluticasone, good adherence (virtually 100%) and tightly monitored inhalation techniques [22]. The addition of oral prednisolone (0.5 mg?kg -1 ) led to a modest 7% increase in the percentage of well-controlled patients [22]. The most widely accepted explanation for these unsatisfactory findings is the view that the term ''difficult-to-treat asthma'' might include a broad spectrum of inflammatory patterns, not always as responsive to steroids as an eosinophil-associated process could be. In fact, several phenotypes of refractory asthma have been proposed, including those subjects who have persistent eosinophilic inflammation despite steroid treatment, but also those with predominant neutrophilic airway inflammation and those in whom virtually no inflammation is present on bronchial biopsy [23].
We have found a higher proportion of positive skin test results in those patients who achieved control than in those who remained uncontrolled. Although the classification between atopic and nonatopic disease has recently come under scrutiny, the ENFUMOSA (European Network for Understanding Mechanisms of Severe Asthma) study found fewer positive skin-prick tests in severe asthmatics compared with controlled patients, suggesting an association between atopy and the potential for poor/good asthma control with steroid/b-agonist therapy [24]. Positive bronchodilator test and PEF variability .20% were also significantly more common in asthmatics who gained control, possibly reflecting a more reversible clinical situation. Conversely, depression was more frequent in patients who did not achieve control.  It must be noted that we assessed asthma control by administering the ACT. The recently published ATS/ERS consensus about standardisation of outcomes relating to asthma control recommend this kind of composite measure designed to provide numerical comparisons of treatment effect [25]. This brief five-item questionnaire measures several different areas of asthma control, including symptoms, rescue inhaler usage and the impact of asthma in everyday functioning, but, even using this tool, accurate assessment can be difficult and comorbidities might alter the scoring [26]. In two case series, coexisting disorders with asthma-like symptoms were found in 19% and 34% of patients with difficult asthma [27,28]. In such individuals, a variety of comorbid diseases, such as gastro-oesophageal reflux disease, obesity, vocal cord dysfunction and upper airway disease (e.g. seasonal allergies), may overlap with symptoms of asthma, making it difficult to assess control. Particularly, it has been reported that depressive and anxiety disorders were associated with a decreased level of asthma control, including more visits to the doctor or emergency room, inability to do usual activities, and more days of symptoms compared to those without depression or anxiety [29]. We have not found differences in the rate of other comorbid conditions between patients who reached control and those who remained uncontrolled.
Some limitations of this study must be addressed. 1) The sample size was small. 2) It is possible that the treatment periods (1 month) were too short to reach the maximum effect.
In fact, one study demonstrated that asthmatic patients with stable dosing tend to improve further, confirming the benefit of sustained treatment in subjects who have difficulty in achieving control [22]. 3) Airway hyperresponsiveness, a factor that could potentially predict therapeutic response, was not assessed in all of the patients. 4) A selection bias is possible because the study design excluded patients who were taking oral steroids. Thus, the sample might not accurately represent the whole population of difficult-to-treat asthmatics. Conversely, it must be taken into account that the diagnosis of severe asthma still represents a challenge for physicians, and many patients with other entities like COPD could be categorised as ''difficult-to-treat asthma''. However, in our sample there were 71% females, more than 90% were never smokers, hyperresponsiveness was present in almost all of them and mean FeNO value was 43 ppb. All of these facts, taken together, make us feel confident that our patients were truly asthmatics. 5) The investigators were not blinded to the FeNO results. This fact could be a possible source of bias, although we believe that the influence in our results is not relevant because therapeutic decisions were not based on FeNO values, but rather were derived from the ACT score, which is self completed by the patients. 6) Although all of the patients had been regularly followed at an outpatient asthma clinic, they had previously been educated on the correct use of inhalers, they were advised to bring their maintenance mediation to the hospital and all of them denied nonadherence at the time of the first clinical assessment we did not measure adherence objectively. GAMBLE et al. [30] have recently demonstrated that a significant proportion of patients with difficult-to-control asthma remained nonadherent to inhaled or oral corticosteroids. However, patients were unaware that they were being observed and it is well known that patients who agree to participate in research are more likely than nonparticipants to be adherent with their regimen. Anyway, although noncompliance could underestimate the population response to the stepwise approach, it is unlikely to affect the predictive accuracy of FeNO.  Finally, it must be highlighted that no single outcome measure can adequately assess asthma control. The clinical value of composite scores like ACT is limited by the lack of validation in a wider range of settings, particularly in patients with different asthma phenotypes.
The present study may have implications for clinical practice and future research. Such information could be beneficial when advising patients what to expect when deciding to escalate their medication and to employ potentially harmful drugs. Conversely, it is of critical importance to identify patients who are less responsive to steroid treatment and are at risk of developing persistent airway obstruction. These patients should be closely monitored and considered for novel anti-asthma drugs in order to prevent progression of the disease. In addition, attempts at treating by phenotype will aid in the development of a more rational approach to the evaluation of interventions like therapy with omalizumab, mepolizumab, imatimib or anti-tumor necrosis factor-a agents.
In conclusion, the current results suggest that FeNO can identify patients with difficult-to-treat asthma and the potential to respond to high doses of inhaled corticosteroids or systemic steroids.

STATEMENT OF INTEREST
Statements of interest for L.A. Pérez-de-Llano and for the study itself can be found at www.erj.ersjournals.com/misc/statements.dtl