Diagnosis of asthma in symptomatic children based on measures of lung function: an analysis of data from a population-based birth cohort study

Summary Background Concerns have been expressed about asthma overdiagnosis. The UK National Institute of Health and Care Excellence (NICE) proposed a new diagnostic algorithm applying four lung function measures sequentially (ratio of forced expiratory volume in 1 s [FEV1] to forced vital capacity [FVC] <70%, bronchodilator reversibility ≥12%, fractional exhaled nitric oxide [FeNO] ≥35 parts per billion, and peak expiratory flow variability >20%). We aimed to assess the diagnostic value of three of the tests individually, and then test the proposed algorithm in symptomatic children. Methods We used follow-up data at age 13–16 years from the Manchester Asthma and Allergy Study, a prospective, population-based, birth cohort study. We initially present results for the whole population, then by subgroup of disease. To simulate the situation in primary care, we included participants reporting symptoms of wheeze, cough, or breathlessness in the previous 12 months and who were not on regular inhaled corticosteroids. We used an epidemiological definition of current asthma, defined as all three of physician-diagnosed asthma, current wheeze, and current use of asthma treatment, reported by parents in a validated questionnaire. We assigned children with negative answers to all three questions as non-asthmatic controls. We also measured spirometry, bronchodilator reversibility, and FeNO at follow-up; data for peak expiratory flow variability were not available. We calculated the proportion of participants with a current positive lung function test at each step of the algorithm, and recorded the number of participants that met our definition of asthma. Findings Of 1184 children born into the cohort, 772 attended follow-up at age 13–16 years between July 22, 2011, and Nov 11, 2014. Among 630 children who completed spirometry, FEV1:FVC was less than 70% in ten (2%) children, of whom only two (20%) had current asthma. Bronchodilator reversibility was positive in 54 (9%) of 624 children, of whom only 12 (22%) had current asthma. FeNO was 35 or more parts per billion in 115 (24%) of 485 children, of whom 29 (25%) had current asthma. Only four of 56 children with current asthma had positive results for all three tests (spirometry, bronchodilator reversibility, and FeNO). Conversely, 24 (43%) of the 56 children with current asthma were negative on all three tests. FEV1:fvc (p=0·0075) and FeNO (p<0·0001), but not bronchodilator reversibility (p=0·97), were independently associated with asthma in multivariable logistic regression models. Among children who reported recent symptoms, the diagnostic accuracy of the algorithm was poor. Interpretation Our findings challenge the proposed cutoff values for spirometry, the order in which the lung function tests are done, and the position of bronchodilator reversibility within the algorithm sequence. Until better evidence is available, the proposed NICE algorithm on asthma diagnosis should not be implemented in children. Funding UK Medical Research Council.


Selection of children with recent symptoms.
Children were defined as having recent symptoms if they answered yes to any of In the past 12 months has your child had wheezing or whistling in the chest?
Does your child usually have a cough during the day apart from with colds?
Does your child usually have a cough at night apart from with colds?
Does your child cough after exertion apart from with colds?
Does your child cough when he / she is excited apart from with colds?
Does your child cough on exposure to cold air apart from with colds?
Of the 481 children with three measures of lung function, 189 had symptoms of either cough or wheeze in the previous 12 months. The distribution of symptoms is shown in Table 1  Of the 26 on regular ICS who were excluded from some analysis, 18 reported both cough and wheeze and 8 reported wheeze in the last 12 months.

Sensitisation
We ascertained sensitisation to allergens by skin prick testing (house dust mite [Dermatophagoides pteronyssinus], cat, dog, grasses, moulds, milk, peanut and egg [Bayer, Elkahrt, IN, USA]). Sensitization was defined as a mean wheal diameter of 3 mm or greater than that elicited by the negative control

Statistics
Descriptive statistics were calculated for the main diagnostic tests in the group of patients with spirometry available (N=630). Frequencies and percentages were used to summarise binary or categorical variables and comparisons were made using chi-squared tests. Means, standard deviations and 95% confidence intervals (CIs) were calculated for normally distributed, continuous variables and comparisons were made using independent samples t-tests. FeNO was log normally distributed; results are presented as geometric mean (GM) and 95% CI. Exploratory analyses investigated whether rhinitis affected FeNO (one-way ANOVA, appendix).
The number of children with values above the cut-offs for each of the diagnostic tests considered in the algorithm were presented along with whether they met our definition of asthma. Due to the expected differences between boys and girls, results are presented separately where appropriate. Independent samples t-tests and chisquared tests were used to compare variables of interest between the genders. Relationships between the variables considered in the algorithm were assessed using correlations.
In the group of children that met our definition of asthma or were non-asthmatic (excluding those with possible asthma), the variables in the algorithm were assessed using sensitivities, specificities, positive predictive values (PPVs), negative predictive values (NPVs), and areas under receiver operating characteristic curves (AUROCs).
A multivariable logistic regression model, assuming a linear functional form for the predictors, was used to investigate the importance of the considered variables. This analysis was repeated in the subgroup of symptomatic children not on regular ICS with all required lung function tests available. The diagnostic algorithm was tested in this subgroup of children to approximate its ability when being used in the clinical setting.
All analyses were performed using SPSS 22 and a 5% significance level was used throughout the paper.

BDR
BDR did not differ significantly between boys and girls (p=0.21, Table 2). An increase in FEV 1 of ≥12% from baseline was observed in 54 children (8.7%), all of whom showed an improvement of 200mls or more. Of the 54 children with BDR≥12%, 12 had asthma (8 girls), 16 had possible asthma (5 girls), and 26 were non asthmatic (12 girls).  Table 3). FeNO was higher in children with rhinitis, but subgroup analysis revealed that this was only significant in those who did not have asthma (Tables 4 and 5

Diagnosing asthma in children with recent symptoms
Of the 481 children with spirometry, BDR and FeNO measures available, 189 reported symptoms of cough or wheeze in the last 12 months

Text relating to Figure 2b: FEV 1 /FVC<LLN in main manuscript:
Eight of the 89 children had obstructive spirometry using LLN criteria, five of whom had BDR≥12%, meeting the criteria for asthma. Three of these children met our definition of asthma. All three remaining children with BDR<12% had FeNO<35ppb, which would trigger a referral for specialist assessment. Among 81 children with FEV 1 /FVC≥LLN, 22 had FeNO≥35ppb and would be diagnosed with asthma, or suspected asthma depending on the results of PEFR monitoring; 13 of these children met our criteria for asthma. FeNO was <35ppb in 59 children (17 of whom met our criteria for asthma), and if a PEF diary had shown 20% reversibility would fall in to the suspect asthma part of the algorithm, in which tests should be repeated at 6 weeks.