The art and science of paediatric ‘asthma’

Possibly the most important change in the 2003 SIGN/British Thoracic Society guidelines [1] was the recognition that diagnosis of asthma in childhood is essentially a clinical diagnosis based on the presence of typical symptoms and signs together with a clear response to a trial of medication. This may be a significant (>15%) and repeatable improvement in forced expiratory volume in 1 s or, more commonly, a dramatic and unequivocal improvement in symptoms during the first 2–6 weeks after the introduction of inhaled corticosteroids (ICS). The guidelines also emphasise the need to document the basis on which a diagnosis is made and the need for repeated re-assessment of the child, questioning the diagnosis if management is ineffective. The implications of this most rigorous of ‘evidence-based’ guidelines are that there is no test for asthma other than responding to a trial of ‘asthma’ medication and that other inflammatory conditions can produce very similar symptoms. This may result in a misdiagnosis of asthma, which may only be identified by ‘repeated reassessment of the child, questioning the diagnosis if management is ineffective’. Unfortunately, the fact that the diagnosis remains dependent upon clinical acumen results in on-going over-and under-diagnosis of the condition. As there is no reliable index of disease control, on-going management of the condition is also predominantly based on reported levels of morbidity and is confounded by patient factors such as regimen and device compliance [2].

In developed countries asthma remains the most common chronic disease of childhood. Inevitably it has been subject to intense research over the past half century, with groups attempting to understand the origin, nature and natural history of the disease in order to improve diagnosis, treatment and monitoring of this condition as well as to provide children and parents with prognostic information and provide a guide to possibly prevent the condition. The condition is full of paradoxes. We now know that inflammation is a key component but we do not understand which aspects of inflammation are central to perpetuating the clinical symptoms; we know that inhibiting inflammation reduces morbidity but we continue to rely on the 50-year-old inhaler technology to deliver corticosteroids to the lungs [3], hoping to obtain sufficient dose to the lungs without causing too many side-effects – this ‘blunderbuss’ anti-inflammatory approach has been the cornerstone of therapy for more than 35 years [4]; we know that ‘asthma’ in childhood is labile with many children who develop asthma ceasing to have symptoms entirely or intermittently during childhood, but we cannot predict at an individual level as to which children are destined to have persistent or recurrent asthma [5, 6]; despite the inflammatory nature of the condition, diagnosis is still in essence dependent upon the observation that asthma is a disease that gets better with asthma medication, as reflected in current guidelines – measurements of bronchial hyperresponsiveness (BHR), airways eosinophilia or exhaled nitric oxide may be supportive but are not diagnostic; atopy is an important risk factor in many children with asthma and specific allergens may contribute to the levels of chronic morbidity but we continue to use the same therapeutic strategies in those without identifiable allergies or atopic predisposition; therapies that would appear to be ideally suited to those in whom allergic inflammation might dominate such leukotriene receptor antagonists appear no more or less effective in atopic and non-atopic populations.

Against this background of uncertainty, many have attempted to device strategies that will improve the accuracy of diagnosis and permit clinicians to simply and easily monitor disease activity to guide changes – that is to develop key tests such as the diabetologists have to diagnose and monitor type 1 diabetes using blood glucose and HbA1c. It is probably worth remembering that measuring blood glucose in the untreated patient reflects the effect of lack of insulin while HbA1c reflects the impact of therapy over a period of time. The impact of therapy of course depends on a number of inter-related factors such as prescribed dose, dosage regime, regimen compliance (is the drug taken as recommended), technique of administration, dietary fluctuations and intercurrent illness – not too dissimilar from the complexity of therapy in asthma.

The lack of a simple-to-apply diagnostic test together with the lungs' limited repertoire of responses to inflammation has had profound effects on the management of individual patients and on our interpretation of epidemiological and other follow-up studies. Inflammation within the airways will lead to coughing, may stimulate mucus secretions and may cause noisy breathing. As a result, conditions such as recurrent viral infections, persistent bacterial bronchitis [7, 8] and other less-common conditions such as aspiration can all resemble asthma and indeed asthma can co-exist with other pathologies. As a result, misdiagnosis – both under and over-diagnosis – is common. An example of the gross errors that result from mis-interpretation of observations that can occur due to the lack of a simple objective diagnostic test was the dogma that became widely held in the 1980s that ‘wheezy bronchitis’ (wheezing with a viral bronchitis) was simply the early manifestation of asthma and should be treated as asthma. This was based in part on observation that in the mid-1980s asthma was being under-diagnosed in school-aged children, many of whose symptoms were being inappropriately treated with antibiotics [9]. The conclusion that under-treatment of asthma was common was, without good evidence, extrapolated down into the pre-school age group and combined with retrospective studies [10]. This appeared to show that wheezing in early childhood was an early manifestation of asthma to generate a dogmatic position that contradicted the results of studies in both primary and secondary care studies from the early 1960s, which showed that the majority of young children with wheezing precipitated by viruses did not go on to have asthma. Years of research were required to re-invent or, rather, re-establish the concept of ‘wheezy bronchitis’ or non-asthmatic viral-induced wheezing, most commonly observed in pre-schoolchildren.

Similarly, it would appear that currently one of the most common causes of over-diagnosis and/or treatment of asthma is failure to diagnose persistent bacterial bronchitis. PBB, which if untreated for years or decades is likely to eventually lead to bronchiectasis. This was extremely well characterized in the 1950s [11] but the widespread use of antibiotics led to a dramatic reduction and it was largely forgotten. The recent, appropriate dramatic reduction in antibiotic prescribings for viral respiratory tract infections has led to a resurgence in cases, as this reduction has not been accompanied by delayed prescribing for coughs that do not resolve. PBB, similar to wheezy bronchitis, superficially resembles asthma, with parents reporting a persistent night-time cough, exercise-induced symptoms with viral exacerbations. Asthmatics with PBB as a co-morbidity pose a particular challenge and this is almost certainly far more common in ‘difficult asthma’ than is generally recognized. Those undertaking bronchoscopies as part of the assessment of ‘difficult asthma’ express surprise at the frequency of positive bacterial cultures [12, 13] but tend not to report acting on this information.

These unfortunate collective exercises in amnesia have been compounded at an individual patient and an epidemiological level by failure to use terms as central and basic as wheeze. Various studies involving children from infancy to adolescence clearly indicate that wheezing is grossly over-reported, with 30–50% of all wheeze reported by parents not being wheeze [14–16]. The SIGN/BTS guideline recognizes the problem by stating that ideally wheeze should be heard by a healthcare professional. However, even this is far from ideal, with trained doctors failing to agree if wheeze is present in up to 40% of cases when listening to a particular child at the same time [17].

At an individual level, this can have significant implications for individual patients and it certainly calls into question some of the conclusions of epidemiological studies. These probably significantly over-estimate the true incidence of wheeze and it may be that true recurrent wheeze in early childhood has greater predictive significance than some of the recent large cohort studies might suggest [18]. The importance of accurate clinical assessment with a clear description of phenotypes is clearly seen in the follow-up of children with ‘acute bronchiolitis’ and/or respiratory syncytial virus (RSV) infection. The term ‘bronchiolitis’ is used to describe quite different phenotypes of illness in different countries with the United States favouring ‘first episode of wheeze with an apparent viral infection’ and the United Kingdom favouring lower airways obstruction due to a virus with widespread crepitations on auscultation (they may occasionally wheeze as well). Both forms of illness are associated with later recurrent respiratory morbidity but the patterns are quite different and do not dependent on the virus initiating the acute illness. A recent study [19] of infants admitted to hospital with RSV lower respiratory tract infection (LRTI) indicated that those in the former group have higher levels of morbidity, an excess of atopy and excess of ICS use at follow-up. In this ‘American bronchiolitis’ group, some patients will fit the ‘wheezy bronchitis’ phenotype and other will be ‘asthmatic’. In contrast those with ‘UK bronchiolitis’, consistent with previous studies among this ‘phenotype’, had an excess in symptoms but far less than those in the ‘American bronchiolitis’ group. These were predominantly with intercurrent infections in the first year or two after the original LRTI. They were no more atopic than controls and did not have an excess of ICS use. This would suggest that it is the host response rather than the virus that is predictive of future ‘asthma’.

Having focused much of our research over the past two decades on the importance of inflammation in asthma, it is not surprising that ‘inflammometry’ [20] – the measurement of a marker of inflammation – has been the focus of attention for the past decade and is advocated by some as the ultimate answer to the diagnostic and therapeutic challenges we face. However, none of the current candidates are ideal and at best are supportive of the clinical process. These enthusiasts often dismiss physiological effects and particularly clinical effects [20] as being too far from the site of action, but of course we still do not understand the processes that initiate and perpetuate ‘asthmatic inflammation’ and, perhaps more importantly, which aspects of inflammation are key to controlling the disease. The two candidates that have received most attention over recent years have been eosinophil numbers in sputum and fractional exhaled nitric oxide (FeNO), but their use in the diagnosis and management of asthma remains unclear. Many feel that airways eosinophilia is a critical aspect of asthma [21] but it remains to be seen whether their presence is indeed central. Studies with anti-IL-5 antibodies virtually eliminated circulating and airways eosinophils but had no significant impact on clinical outcomes or airways hyperresponsiveness [21, 22] while adults who have ‘out-grown’ asthma continue to have raised eosinophils in sputum despite normal lung function and an absence of symptoms [23]. Treatment of these asymptomatic subjects leads to reduced BHR but has no clinical impact or effect on quality of life [24]. Moreover, the presence of eosinophils in the airways of children with ‘difficult asthma’ has no predictive value in terms of response to steroid therapy [25, 26].

FeNO is generally elevated in atopic asthmatics and falls rapidly with the introduction of ICS. However, elevated FeNO is observed in atopic subjects without any evidence of asthma and levels can fall to ‘normal’ levels despite on-going symptoms while dose-response studies have suggested a dissociation between FeNO levels and measures of eosinophilic inflammation and BHR [27]. When compared with conventional management, the use of FeNO to monitor childhood asthma permitted a small reduction in ICS dose but had no impact on clinical parameters [28, 29], and it is possible that the impact on drug dose was more a result of the design of the study than a true difference. These observations make it necessary to question whether FeNO will prove to be the definitive ‘non-invasive’ test that is needed and whether the eosinophil is a key component of the disease or is simply a peripheral marker of something more fundamental. Defenders of eosinophils point out that anti-IL-5 therapy reduces but does not eliminate all eosinophils in the airways wall [30], and they speculate that although free of symptoms, adults who have outgrown their asthma may be experiencing on-going damage that will ultimately cause irreversible harm.

To date, none of the proposed markers of inflammation have been able to reliably and reproducibly diagnose asthmatic patients more effectively than an experienced clinician applying the rule that probably asthma only becomes a firm diagnosis of asthma when ‘asthma’ therapy clearly and unequivocally leads to a dramatic improvement. It is important to remember that FeNO and any other proposed marker for disease activity reflects a complex constellation of factors including disease activity and the effectiveness of therapy [31, 32]. Effectiveness of therapy depends on subjects using inhalers as prescribed (regimen compliance), and when they are used, using them effectively (device compliance). Failure to use them effectively may be due to lack of competence (current devices are not intuitive to use as patients need to be trained regularly) or contrivance – knowing how to use a device but contriving to use it in an ineffective manner [2, 3]. Studies based on ‘inflammatory’ markers tend to see the output from the measurement as a marker of disease severity or activity using the data to alter prescribed doses when the level may simply reflect levels of compliance, be it regimen or device. At best, it is one of the number of tools that an experienced clinician can use to try and optimize treatment; it is not a tool that will replace the expertise of an experienced clinician. In all these studies, ‘compliance’ is the elephant in the room that is ignored or skirted around without addressing. At present, the only intervention shown to have a significant impact on regimen compliance is to use datalogger devices with the facility to provide audible reminders [33] – an approach that is probably cheaper and certainly more effective than education.

The lack of specificity and sensitivity on an individual patient basis of monitoring parameters, such as eosinophil levels in sputum, is highlighted in the study of Lovett et al. [34], published in this issue of the journal. Their study found that children aged 6–17 years with eosinophilic asthma (eosinophils in sputum >2.5%) were more likely to report symptoms and use β₂-agonists when reassessed 5 years later, suggesting that elevated levels of eosinophils predicted more troublesome and persistent disease. However, it is difficult to exclude confounders in that the majority of children were on inhaled steroids when first seen – though apparently not statistically significant, the eosinophilc group was more likely to be on ICS (94% vs. 77%), was on higher doses (median BDP equivalent 750 vs. 400) and had more oral steroid use (15% vs. 5.6%). Was this group generally ‘more severe’ when first seen, or less compliant. Without assessment of compliance, it is impossible to tell if the ‘eosinophilc group’ was more likely to have poor compliance and if so this may well explain the difference in symptoms and β-agonist use 5 years later, which are as likely to reflect compliance as disease severity. Interestingly, at follow-up ICS use was very similar in each group (60% vs. 58%) and the non-eosinophilic group required more oral steroids (21% vs. 36%), and while again this did not quite reach statistical significance, it might suggest that this group is experiencing more exacerbations despite lower levels of interval symptoms. Certainly on an individual basis it would be impossible to use induced sputum results in primary school-age children to provide prognostic information. These tests are labour intensive and as in other similar studies satisfactory samples were only obtained in about three quarters of subjects, suggesting that even in those potentially able to co-operate it is far from an ideal test.

Until measurement of compliance is built into studies such as those above, the results generated will continue to be subject to significant levels of uncertainty. The complex interactions of disease activity, prescribed medication regimen and device compliance and the potential presence of co-morbidities make it very difficult to disentangle key variables. We still desperately need a reliable, simple-to-apply diagnostic test and a simple-to-use monitoring tool that can be combined with objective assessments of compliance. The present candidates such as FeNO, sputum eosinophilia and BHR using challenges such as dry powder mannitol still have much to prove before they can be recommended in evidence-based guidelines – until then the effective diagnosis and management of the vast majority of asthmatic children remains an art rather than a science and unfortunately over-diagnosis, over-treatment as well as under diagnosis and ineffectual treatment will continue.