Abstract
Introduction There are few data on the usefulness of different tests to diagnose asthma in children.
Aim We assessed the contribution of a detailed history and a variety of diagnostic tests for diagnosing asthma in children.
Methods We studied children aged 6–16 years referred consecutively for evaluation of suspected asthma to two pulmonary outpatient clinics. Symptoms were assessed by parental questionnaire. The clinical evaluation included skin-prick tests, measurement of exhaled nitric oxide fraction (FeNO), spirometry, bronchodilator reversibility and bronchial provocation tests (BPT) by exercise, methacholine and mannitol. Asthma was diagnosed by the physicians at the end of the visit. We assessed diagnostic accuracy of symptoms and tests by calculating sensitivity, specificity, positive and negative predictive values and area under the curve (AUC).
Results Of the 111 participants, 80 (72%) were diagnosed with asthma. The combined sensitivity and specificity was highest for reported frequent wheeze (more than three attacks per year) (sensitivity 0.44, specificity 0.90), awakening due to wheeze (0.41, 0.90) and wheeze triggered by pollen (0.46, 0.83) or by pets (0.29, 0.99). Of the diagnostic tests, the AUC was highest for FeNO measurement (0.80) and BPT by methacholine (0.81) or exercise (0.74), and lowest for forced expiratory volume in 1 s (FEV1) (0.62) and FEV1/forced vital capacity ratio (0.66), assessed by spirometry.
Conclusion This study suggests that specific questions about triggers and severity of wheeze, measurement of FeNO and BPT by methacholine or exercise contribute more to the diagnosis of asthma in school-aged children than spirometry, bronchodilator reversibility and skin-prick tests.
Abstract
Diagnosing asthma in children is most accurately done by using information on triggers and severity of wheeze and by FeNO measurement, methacholine and exercise challenge tests. http://bit.ly/2kDWaRr
Introduction
Diagnosing asthma in children is not straightforward, because we lack a stand-alone diagnostic test. Symptoms (cough, wheeze, breathlessness) are not specific for asthma and interpretation of commonly used diagnostic tests is complicated by the temporal variability and phenotypic heterogeneity of asthma. Thus, diagnostic guidelines suggest diagnosing asthma based on a characteristic pattern of respiratory symptoms, clinical examination, demonstration of reversible airway obstruction assessed by spirometry and airway inflammation measured by exhaled nitric oxide fraction (FeNO) [1–4]. Allergy tests and measurement of bronchial hyperresponsiveness by direct and indirect challenge tests are used as further aids for diagnosis.
However, the diagnostic algorithm proposed by recent guidelines has been questioned for children and there are surprisingly few data available to assess the usefulness of the different tests in the diagnosis of asthma in school-aged children [5]. Systematic literature reviews done for recent guidelines and for the ongoing taskforce of the European Respiratory Society identified only a handful of publications assessing the accuracy of the different tests for children with suspected asthma [2, 3]. Most publications identified by the searches had a case–control design, comparing children with asthma to healthy controls instead of consecutive referrals of children suspected of having asthma. Available studies had included only few diagnostic tests and no detailed history, and asthma diagnosis used as reference standard was often poorly defined or too narrow, for instance including only allergic asthma. Additionally, papers used different cut-offs for positive tests (e.g. for FeNO or forced expiratory volume in 1 s (FEV1)), and it remains unclear which cut-offs are best for children [1–4]. In this study, we assessed the diagnostic accuracy of reported respiratory symptoms and different objective tests to diagnose asthma in consecutive referrals of school-aged children presenting symptoms suggestive of asthma.
Methods
Study population and study design
For this study, we re-analysed data from a clinical study performed in 2007–2008 in Switzerland. It included consecutive first-time referrals to the respiratory outpatient clinics of two paediatric hospitals (St Gallen and Basel) of 6–16-year-old children for evaluation of a possible asthma diagnosis with a history of wheezing, dyspnoea or cough. Children were excluded from the study if they had a known chronic respiratory disease such as cystic fibrosis, or a respiratory tract infection during the 4 weeks prior to the visit. The aim of the initial study had been to compare the results of mannitol challenge tests to exercise challenge tests [6].
Study procedures
All children referred for the first time by general practitioners or primary care paediatricians for evaluation of possible asthma were invited to participate in the study, which included two visits to the hospital within a week (figure 1). At the first visit, all children underwent clinical evaluation, skin-prick testing (unless printed results of a skin-prick test done during the past 2 years were available), measurement of FeNO, spirometry, exercise bronchial provocation tests (BPT), methacholine BPT and bronchodilator reversibility test, in that order. Children who reacted to the exercise challenge and received salbutamol returned for an extra visit within the following few days to perform the methacholine challenge test. Within a week all children repeated the FeNO measurement and performed a mannitol BPT. Between visits, the family completed a questionnaire. Ethical approval was obtained from the local ethics committee and all parents gave informed consent at baseline (EKSG 07/001).
Study procedures. The report form is a standardised form for physicians to note down the clinical diagnosis. BPT: bronchial provocation test; FeNO: fractional exhaled nitric oxide. #: children who received salbutamol after the exercise BPT conducted the methacholine BPT at an additional visit.
Clinical asthma diagnosis (reference standard)
The study physicians, experienced paediatric pulmonologists, completed a physician's report form that included the clinical diagnosis (definite asthma, probable asthma or other disease), at two time points. At the first visit, physicians considered only medical history, clinical examination, allergy tests, FeNO measurement and spirometry. At the second visit, the same physician reported the clinical diagnosis (as definite asthma, probable asthma or other disease) in the second physicians' report form, taking into account all the information available, i.e. medical history, clinical examination, allergy tests, FeNO measurement, spirometry and results of the BPT and bronchodilator reversibility test. For our main analysis, we defined asthma (reference standard) as an affirmative answer to either definite or probable asthma in the second physician's report form. In a sensitivity analysis, we used the first physicians' report form (based on all the information except the BPTs) to define asthma (reference standard).
Assessment of respiratory symptoms and diagnostic testing
The parental questionnaire included the International Study of Asthma and Allergies in Childhood (ISAAC) key questions for lower respiratory symptoms and more detailed questions on wheeze and cough derived from the questionnaires used in the Leicester respiratory cohort studies (supplementary material) [7, 8]. All diagnostic tests were performed according to published guidelines [9–13]. Short-acting β2-agonists were withheld for 8 h, inhaled corticosteroids, leukotriene antagonists and long-acting β2-agonists for 24 h and antihistamines and sodium cromoglicate for >72 h.
Skin-prick test
We performed skin-prick tests using birch, grass, mugwort, alternaria, cat, house dust mites (Dermatophagoides pteronyssinus), histamine and saline. The skin-prick test was considered positive if the allergen wheal size was ≥3 mm, the positive control (histamine) wheal size was ≥3 mm and the negative control (saline) wheal size was <3 mm. These allergens cover 95% of allergies to inhaled allergens in Switzerland [14].
FeNO
FeNO was measured in doublets before spirometry, using the portable multi-gas analyser (NIOX MINO, Aerocrine, Sollentuna, Sweden), in accordance with published guidelines [10] and previous studies using this device [15, 16]. The portable analyser ensures a constant expiratory flow of 50±5 mL·s−1, has an accuracy of ±10% with a minimum ±5 ppb and the quality was controlled by the lung function technician according to the manufacturer's guidelines.
Spirometry
Spirometry was performed using American Thoracic Society (ATS) criteria for paediatric lung function testing and a Jaeger Masterscope (Erich Jaeger, Würzburg, Germany), using JLAB software (version 4.34). Spirometry was performed in triplicate by experienced lung function technicians, who performed quality control during the measurement and recorded the best measurement. The flow–volume curve was then checked by the responsible paediatric pulmonologist. Results are expressed as proportion (FEV1/forced vital capacity (FVC)) and as z-scores based on Global Lung Initiative 2012 reference standards [17].
Bronchial provocation tests
For all BPTs, baseline FEV1 was measured in triplicate using ATS criteria for paediatric lung function testing [9] and the best measurement was recorded. We reported the results of the exercise BPT as the maximum fall of FEV1 compared to baseline, the methacholine BPT as provocation dose causing a 20% decrease of FEV1 from baseline (PD20) and the mannitol BPT as provocation dose causing a 15% decrease of FEV1 from baseline (PD15). After the methacholine BPT, all children were given four puffs of salbutamol 100 µg to test for bronchodilator reversibility. In addition, children received salbutamol if FEV1 had not returned to within 5% of baseline 15 min after the exercise or mannitol BPT, or in cases of dyspnoea. More details on the BPTs have been published before and can be found in the supplementary material [6].
Statistical analysis
For the reported respiratory symptoms and the different tests, we calculated sensitivity, specificity, positive predictive value and negative predictive value, Youden's index (sensitivity+specificity−1), area under the curve (AUC) and their 95% confidence intervals to diagnose asthma, using the final (post-BPT) physicians' diagnosis as reference standard. We did a sensitivity analysis using the first (pre-BPT) physicians' diagnosis. We displayed the cut-off values with the highest Youden's index in our study and those used in the literature. We used STATA software (version 15; College Station, TX, USA) for statistical analysis.
Results
Characteristics of the study population
Of the 124 children invited, 111 (90%) were recruited, 84 from St Gallen and 27 from Basel. The median (range) age was 12 (6–16) years and 62% were male. Most children were referred with wheeze and cough (47%) or wheeze without cough (23%). Inhaled medication had been used by 64% prior to referral, including 19% who had used inhaled corticosteroids (table 1). Of the 111 participants, 80 (72%) were diagnosed with asthma after all BPTs were done compared to 94 (85%) before the BPTs. The remaining children were diagnosed with cough unrelated to asthma (8% before BPTs and 13% after BPTs) and with inducible laryngeal obstruction and dysfunctional breathing (6% before BPTs and 7% after BPTs) (supplementary table S1). None of the children were diagnosed with a severe lung disease such as cystic fibrosis [18].
Characteristics of the study participants
Diagnostic accuracy of respiratory symptoms to diagnose asthma
Reported wheeze in the past 12 months had the highest sensitivity (80%) for physician-diagnosed asthma (table 2). Specificity was highest for frequent wheeze (more than three attacks per year) (90%), awakening due to wheeze (90%) and wheeze triggered by pollen (83%), house dust (93%) or pets (99%). Combined sensitivity and specificity was highest for frequent wheeze in the past 12 months (Youden's index 0.34), awakening due to wheeze (0.31) and wheeze triggered by pollen (0.29) or pets (0.28) (table 2).
Diagnostic accuracy of respiratory symptoms in the past 12 months to diagnose asthma
Diagnostic accuracy of tests to diagnose asthma
All 111 children completed skin-prick testing, FeNO, spirometry and BPT by mannitol. BPT by exercise could not be completed in 12 children because of exhaustion (n=7), inspiratory stridor (induced laryngeal obstruction) (n=2), no cooperation (n=2) or technical difficulties (n=1) [6, 19]. Seven patients could not complete BPT by methacholine due to exhaustion and 36 children performed the test during an extra visit a few days later. In four patients the skin-prick test result was not considered valid because the histamine control was not positive. Test results in patients with and without asthma diagnosis are displayed in supplementary table S2.
The cut-off values with the best diagnostic accuracy were <80% for FEV1/FVC, ≤−0.8 z-score for FEV1, ≥10% increase of FEV1 for bronchodilator reversibility test, ≥8% decrease of FEV1 for BPT by exercise, PD20 <0.7 mg for BPT by methacholine, PD15 <635 mg for BPT by mannitol, ≥2 for the number of positive skin-prick tests, ≥8 mm for the cumulative wheal size of skin-prick tests and ≥21 ppb for FeNO (table 3).
Diagnostic accuracy of clinical tests to diagnose asthma
Accuracy overall was best for FeNO, BPT by methacholine and BPT by exercise (AUC 0.80, 0.81 and 0.74, respectively). Accuracy was lower for BPT by mannitol and skin-prick test (AUC ∼0.70), and lowest for spirometry (AUC 0.62 and 0.66 for FEV1 and FEV1/FVC ratio, respectively) (figure 2).
Receiver operating characteristic (ROC) curve of clinical tests to diagnose asthma. Test (unit): skin-prick test (SPT) number positive (decrease of 1 positive SPT); SPT cumulative wheal size (decrease of 1 mm cumulative wheal size); exhaled nitric oxide fraction (FeNO) (decrease of 1 ppb); forced expiratory volume in 1 s (FEV1) (increase of 0.1 z-score); FEV1/forced vital capacity (FVC) (increase of 1%); bronchodilator reversibility (increase of 1% in FEV1); exercise (decrease of 1% in FEV1); methacholine (increase of 0.1 mg methacholine); mannitol (increase of 5 mg mannitol). #: cut-off with maximum combined sensitivity and specificity.
Sensitivity analysis
In the sensitivity analysis with asthma diagnosis based on the pre-BPT report form, frequent wheeze and wheeze triggered by pollen or by pets in the past 12 months had the highest Youden's index, which was in line with the main analysis. In addition, night cough and hay fever had a high Youden's index for the asthma diagnosis pre-BPT (supplementary table S3), but not for the asthma diagnosis post-BPTs (table 2).
For the diagnostic tests, the Youden's index was highest at the same cut-offs for most tests (supplementary table S4 and supplementary figure S1). Cut-offs were higher for FeNO (25 versus 21) and lower for BPT by exercise (6 versus 8), FEV1 (−0.6 versus −0.8) and bronchodilator reversibility (2 versus 10).
The accuracy was higher pre-BPT than post-BPT for spirometry (AUC 0.71 for FEV1/FVC and 0.65 for FEV1 versus 0.66 and 0.62, respectively) and bronchodilator reversibility (AUC 0.72 versus 0.58) and lower for the BPTs (AUC 0.70 for exercise, 0.68 for methacholine and 0.60 for mannitol versus 0.74, 0.81 and 0.68, respectively). Accuracy was best for FeNO measurement, bronchodilator reversibility, FEV1/FVC ratio and BPT by methacholine and by exercise (AUC 0.78, 0.72, 0.71, 0.70 and 0.70, respectively).
Discussion
This is the first study to systematically assess the diagnostic accuracy of reported symptoms and a range of tests in asthma diagnosis in children compared to a defined reference standard (doctor-diagnosed asthma based on all available measurements and information). The main analysis and sensitivity analysis showed broadly comparable results, suggesting that a history of frequent wheeze, awakening due to wheeze and wheeze triggered by pollen or pets, FeNO measurement, BPT by methacholine and BPT by exercise have the best ability to distinguish asthma from no asthma. FEV1, FEV1/FVC ratio and bronchodilator reversibility had low accuracy.
Only three other studies have assessed the accuracy of symptoms to diagnose asthma in school-aged children consecutively referred to paediatric hospitals [20–22]. They all found that reported wheeze was sensitive (range 0.75–0.86), but not specific (0.64–0.73) and that frequent wheeze and awakening due to dyspnoea were specific (0.84–0.90), but not sensitive (0.33–0.54), which is in line with our findings. Symptom definitions differed between studies, especially those for cough, which results in a wide range of sensitivities and specificities that cannot be compared [20–22]. Five other studies assessed the accuracy of diagnostic tests in school-aged children. Woo et al. [23] found that positive skin-prick tests were sensitive, but not specific (sensitivity and specificity 0.68 and 0.32, respectively) and that FeNO had the best cut-off at 22 ppb (0.57 and 0.87, respectively), which was comparable with our study (21 ppb, 0.59 and 0.87, respectively). Grzelewski et al. [24] found that a FEV1/FVC ratio of <80% was specific (0.91), but not sensitive (0.12) for asthma, which is in line with our findings (<79%; 0.90 and 0.46, respectively). For the bronchodilator reversibility test, Galant et al. [25] and Dundas et al. [26] found a 9% increase in FEV1 to be the best cut-off to diagnose asthma, which is in line with our findings (10%); however, they compared children with asthma to healthy children. For BPT by exercise, Avital et al. [27] found an 8% decrease in FEV1 to be the best cut-off for asthma diagnosis, which is the same as we found. For BPT by methacholine, Zaczeniuk et al. [28] reported a best cut-off of <0.7 mg, which was in line with our study. Anderson et al. [29] found a sensitivity of 0.63 and specificity of 0.81 for the widely used best cut-off of <635 mg for BPT by mannitol, while we found a lower sensitivity and higher specificity (0.43 and 0.93, respectively).
The recent National Institute for Health and Care Excellence (NICE) asthma diagnostic algorithm has been questioned in children. Murray et al. [5] tested the algorithm in the Manchester Asthma and Allergy Study, a population-based cohort of 1184 children aged 13–16 years, of whom 89 were symptomatic, but not regularly inhaling corticosteroids. However, the Manchester study relied on parent-reported data to define asthma (reported wheeze and asthma treatment in the past 12 months plus a doctor diagnosis of asthma ever in life) and compared children with asthma to healthy children, leaving out from the analysis all those with possible asthma. In clinical practice we want to distinguish children with asthma from children with respiratory symptoms due to other causes, not from healthy children. If we had applied the NICE algorithm to our clinical population, only four out of the 111 children would have been diagnosed with asthma at the initial visit (FEV1/FVC ratio <70% and bronchodilator reversibility of ≥12%). 106 would have needed 2 weeks peak expiratory flow monitoring followed by a second visit. In addition, we found that less stringent cut-off values had higher sensitivity and specificity than those recommended by the NICE algorithm (FEV1/FVC ratio <80% versus <70%, bronchodilator reversibility ≥10% versus ≥12% and FeNO ≥26 ppb versus ≥35 ppb, respectively). This highlights the need to base diagnostic algorithms for children on clinical studies done in children, rather than in adults.
A main strength of our study is that it represents a real-life situation in everyday paediatric practice. With the clinical design, it reflects the typical mix of patients in a paediatric outpatient clinic. All children were first-time referrals for evaluation of possible asthma, which is the patient group the diagnostic tests are intended for. Therefore, the study population is representative of daily clinical practice, in contrast to many published studies that selectively include well-defined moderate-to-severe asthmatics comparing them to healthy controls and excluding patients with unclear degrees of airway reactivity. In addition, our patients had an extensive array of examinations for lung function, BPT and allergy, which allowed us to assess the accuracy of different symptoms and diagnostic tests simultaneously.
An important limitation of this study was that the reference standard for asthma diagnosis (the final diagnosis by the physician) took into account the results of the patient history and diagnostic tests for which the accuracy was assessed. However, as there is no single objective test to diagnose asthma and be used as a comparator, the clinician's judgement, taking into account the full history, examination and test results, is the best we can do. The sensitivity analysis using the physicians' diagnosis before BPTs were performed, showed comparable results. However, the small differences highlight the dependence of the physician's diagnosis on the array of tests performed. The reference diagnosis of asthma was made by experienced paediatric pulmonologists (three in Basel and two in St Gallen), trained in Switzerland, who met several times prior to and during the study to standardise their procedures and minimise centre-specific effects. In this study we restricted analysis to basic clinical tests. The advantage of this approach is that most of these tests are available in clinical routine. However, future studies should also evaluate the diagnostic accuracy of newer techniques such as component-resolved IgE diagnostic, multiple-breath or single-breath washout techniques.
Our findings, which need to be replicated in other populations of patients, will help to propose a more evidence-based paediatric diagnostic algorithm, which incorporates both information on symptoms and objective measures. This might be helpful in reducing the need for trials of asthma treatment, which can be costly, time consuming and can lead to misdiagnosis and overtreatment. Our study is therefore an important contribution to the small body of evidence about the value of different tests for the diagnosis of paediatric asthma on which guidelines should be based. Mild paediatric asthma is a disease with highly variable activity and paroxysmal clinical manifestation. It seems unlikely that any test performed at a specific time point will be accurate enough to either prove or exclude reactive airway disease. Future studies should ideally be larger, to allow analysing the value of combination of several tests, and focus on children newly referred for evaluation of possible asthma, and be referenced to a clearly defined and robust reference diagnosis.
Our results suggest that, until more evidence is available, diagnosis of asthma in school-aged children should rely primarily on reported triggers and severity of wheeze and results of FeNO, and, if available, methacholine and exercise challenge testing which were most accurate in our study.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary figure and tables ERJ-01326-2019.Supplement_1
Supplementary methods ERJ-01326-2019.Supplement_2
Supplementary questionnaire ERJ-01326-2019.Questionnaire
Shareable PDF
Supplementary Material
This one-page PDF can be shared freely online.
Shareable PDF ERJ-01326-2019.Shareable
Acknowledgements
We thank all participants and lab technicians of the pulmonology department in the children's hospitals in Basel and St Gallen for their assistance in our study, Marie-Pierre Strippoli (Institute of Social and Preventive Medicine (ISPM), Bern, Switzerland) for her work on the study at baseline. We thank Christopher Ritter (ISPM) for his editorial assistance.
Footnotes
This article has supplementary material available from erj.ersjournals.com.
Author contributions: C.E. Kuehni and J. Barben conceptualised and designed the study. D. Trachsel and J. Barben supervised data collection. C.C.M. de Jong analysed the data and drafted the manuscript. E.S.L. Pedersen and M. Goutaki supported the statistical analysis and gave input for interpretation of the data. All authors critically revised the manuscript and approved the final manuscript as submitted.
Conflict of interest: C.C.M. de Jong has nothing to disclose.
Conflict of interest: E.S.L. Pedersen has nothing to disclose.
Conflict of interest: R. Mozun has nothing to disclose.
Conflict of interest: M. Goutaki has nothing to disclose.
Conflict of interest: D. Trachsel has nothing to disclose.
Conflict of interest: J. Barben has nothing to disclose.
Conflict of interest: C.E. Kuehni has nothing to disclose.
Support statement: This work was supported by the Swiss National Science Foundation (grant 32003B_162820), AstraZeneca (Switzerland), the Lung League St Gallen and the Schmidheiny Foundation (Heerbrugg, St Gallen). Funding information for this article has been deposited with the Crossref Funder Registry.
- Received July 5, 2019.
- Accepted September 3, 2019.
- Copyright ©ERS 2019