Exhaled Nitric Oxide and Clinical Phenotypes of Childhood Asthma

Bruno Mahut; Séverine Peyrard; Christophe Delclaux


Respiratory Research. 2011;12(65) 

In This Article


Design of the Study

La Berma Cohort This single centre cohort conducted in a secondary care out-hospital clinic enrolls asthmatic children since 1997. Since 2008, exhaled NO and clinical events were recorded. Levels of asthma control were systematically assessed using only two levels of GINA guidelines during past three months:[11] controlled versus partially/uncontrolled asthma (omitting lung function since PFT were obtained without treatment). Severe exacerbations, according to ATS/ERS definition,[12] and the number of days (1) with symptoms (GINA guidelines)[11] and (2) with systemic steroid were specifically recorded. This cohort has been declared to our regulatory agency for computer data collection (Commission Nationale Informatique et Libertés, n°1408710), and approval from the Ethics Committee of French learned Society of Pulmonology - SPLF was obtained (CEPRO 2009/019). All children and parents were informed of the prospective recording of clinical and physiological data.

Patients and Criteria of Selection From the Cohort We selected a sample of children, meeting the criteria of clinical (episodic symptoms of airflow obstruction with excluded alternative diagnoses) and functional (documented bronchodilator response based on FEV1 or sRaw)[13] diagnosis of asthma and who satisfied a full description of their asthma: these 28 variables (see Table 1) are categorized as (1) anthropometrics, (2) past history, (3) parental smoking (more than 5 cigarettes per day), level of control, treatment, and (3) pulmonary function. All data were those specifically determined at the time of only one visit, corresponding to routine evaluation in France. These variables allowed the assessment of three domains of asthma severity: level of current prescribed treatment, level of current baseline control of asthma and immediate past burden of asthma exacerbations, accordingly to Bush and Saglani.[14] The population included in the current retrospective, post hoc, database design study overlaps to some extent with the populations of children published previously.[2,6,15–17]

Exhaled NO (FENO, 0.05)

Exhaled NO was measured online, using the Nitric Oxide Analyzer (NIOX; Aerocrine AB; Solna, Sweden: measurement at a constant 50 mL/s expiratory flow rate: FENO,0.05). Measurements were performed according to the ERS/ATS guidelines before pulmonary function tests.[18]

Pulmonary Function Tests (PFT)

All PFT were performed without inhaled treatment (bronchodilator or LABA/ICS association) on the day of the measurement, by the same operator (BM). Spirometry and plethysmographic measurement of specific airway resistance and thoracic gas volume were performed according to international guidelines and as previously described.[13,15,19] The bronchodilator response to salbutamol 400 μg: (post minus baseline)/baseline FEV1 was systematically assessed. Reference values were based on equations edited by Zapletal,[20] as commonly done in Europe.[21]

Statistical Analysis

First Approach Potential explanatory variable were: age, gender, height, atopy, tobacco exposure, control, treatment and pulmonary function tests. The association between the different explanatory variables and FENO was examined in a multiple linear regression model using the procedure for general linear models with log-transformed FENO values as the dependent variable. The multivariate analysis was performed with a backward selection method and variables with P values of less than 0.05 were retained in the FENO model.

Second Approach We used the same approach than Haldar and colleagues.[22] Briefly, a cluster analysis methodology was applied to define homogeneous groups of patients. Principal component analysis (PCA) is a mathematical procedure that uses an orthogonal transformation to convert a set of observations of possibly correlated variables into a set of values of uncorrelated variables called principal components. The number of principal components is usually less than the number of original variables (data reduction). To obtain reliable results, the minimal number of subjects providing usable data for the analysis should be five times the number of variables being analyzed (28 × 5 = 140). This transformation is defined in such a way that the first principal component has as high a variance as possible, and each succeeding component in turn has the highest variance possible under the constraint that it be orthogonal to (uncorrelated with) the preceding components. Then we requested a rotation of the resulting factors which follows completion of the analysis of the data.[23] It has been shown that the relaxed solution of k-means clustering, a common method of cluster analysis,[24] specified by the cluster indicators, is given by the PCA principal components, and thus PCA facilitates k-means clustering to find near-optimal solutions.[25] Cluster analysis allows the partitioning of data into meaningful subgroups (phenotypes), when the number of subgroups and other information about their composition may be unknown. We hypothesised that FENO measurement could be associated with one of these subgroups of asthmatic children. First, variables for the cluster analysis were selected using a principal component analysis (PCA). When considering variable selection for the cluster analysis, our aims were (1) to choose variables that were measured in clinical practice and contributed to the clinical evaluation of asthma, and to avoid choosing different variables that were representative of the same aspect of the disease as this would introduce further bias when the cluster analysis was performed. We thus performed PCA of our 28 commonly measured clinical variables. Orthogonal varimax rotation was performed and the results are summarized in Table 2. To avoid weighting the analysis, we selected only one parameter that was representative of each factor. Two additional variables were also included (see Table 2 legend). Then, a uniform cluster analysis methodology was applied accordingly to Haldar and colleagues (dendrogram for estimation of the number of likely clusters that was further prespecified in a k-means cluster analysis).[22] Finally, characteristics of clusters were compared using analysis of variance for continuous variables or Kruskal Wallis Rank test (non normal values) and χ2 test for proportions. Statistical analyses were performed using MedCalc 11.3.8 (Mariakerke, Belgium) and OpenStat (version 5) softwares.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.