Conflicts of interest The authors have stated explicitly that there are no conflicts of interest in connection with this article.
Peter Gibson, MBBS, Respiratory Medicine, John Hunter Hospital, Lookout Rd, New Lambton, Newcastle, NSW 2305, Australia. Tel: +61 249213470 Fax: +61 249855850 email: email@example.com
Background: The pattern of granulocyte infiltration can be used to identify different inflammatory phenotypes in asthma. Recognized granulocyte phenotypes using induced sputum are eosinophilic (EA), neutrophilic, mixed granulocytic and paucigranulocytic asthma.
Methods: The recognition and importance of inflammatory phenotype analysis using induced sputum in adult asthma are reviewed using published literature.
Results: Knowledge of inflammatory phenotype is useful because it relates to treatment response, mechanistic pathways involved in disease pathogenesis and future disease risk. The population attributable risk of asthma because of eosinophilic inflammation is about 50%, and conversely, this means that up to 50% of asthma cannot be attributed to eosinophilic inflammation, and represents asthma associated with non-eosinophilic processes. In these patients, bronchial biopsy shows significantly fewer eosinophils in the bronchial mucosa than subjects with EA. This confirms that non-eosinophilic asthma is a consistent pattern/phenotype in the airway lumen and the airway mucosa. A key aspect of asthma inflammatory phenotype analysis is that it can be applied to individual patients.
The underlying principle relates to the association between a clinical response to corticosteroids and the presence of a selective sputum eosinophilia.
Conclusions: Clinically useful applications of induced sputum analysis are the detection of non-adherence to corticosteroid therapy, assessment of adequacy of inhaled corticosteroid therapy, long-term therapy management in asthma, oral corticosteroid dose adjustment in refractory asthma and assessment of occupational asthma.
Please cite this paper as: Gibson PG. Inflammatory phenotypes in adult asthma: clinical applications. The Clinical Respiratory Journal 2009; 3: 198–206.
Asthma is a heterogeneous disease, where, as described in the GINA asthma management guidelines (1), ‘many cells and cellular elements’ are recognized to play a role in disease pathogenesis. Granulocytes such as eosinophils and neutrophils form part of the inflammatory process in asthma, and the pattern of granulocyte infiltration can be used to identify different inflammatory phenotypes in asthma. This information is useful because it relates to treatment response, mechanistic pathways involved in disease pathogenesis and future disease risk (2). Interestingly, inflammatory phenotype does not appear to relate closely to current symptom control.
Eosinophilic asthma (EA), or not, and the triumph of clinical medicine
At the end of the 19th century in Europe, soon after the discovery and use of aldehyde dyes for the recognition of cellular differentiation by Paul Ehrlich, sputum eosinophilia was found to be associated with asthma. This well-described association was considered to be almost pathognomonic of asthma for much of the latter part of the 20th century. Studies using allergen challenge defined the occurrence of delayed physiological changes in the airway, including prolonged increases in airway responsiveness that developed after a single allergen exposure. The asthmatic changes that occurred several hours after cessation of an allergen exposure were later found to be associated with eosinophilic inflammation, and the hypothesis that eosinophilic inflammation was the cause of the airway hyperresponsiveness that typified asthma was born. This hypothesis has served us well, explaining clinical responses to sensitizing agents, driving drug development and culminating in the widespread use of inhaled corticosteroids in combination with long-acting beta2 agonists. These agents target both the eosinophilic inflammation (with corticosteroids) and airway hyperresponsiveness (with long-acting beta-agonists) that typify asthma and have revolutionized the clinical care of asthma (1).
Towards the end of the 20th century, however, some inconsistencies started appearing in the dominant EA hypothesis. Coincidentally, these originated from Canada, in fact from the same laboratory that 20 years earlier had provided the revolutionary observation that allergen inhalation could induce increases in airway responsiveness among patients with asthma. These paradigm-shifting observations arose from the careful observation of patients with airway disease, linked to the use of objective measures, in this case the measurement of airway inflammation by sputum cellular analysis, and reflected upon by brave clinical scientists! Dolovich and Hargreave observed patients in their clinic with chronic cough who responded to anti-inflammatory treatment, but did not have airway hyperresponsiveness. They did not have asthma, yet they were successfully treated by asthma therapy. By applying induced sputum analysis, these patients were found to have marked increases in sputum eosinophils; they had ‘asthmatic’ inflammation, but no asthma (3, 4). This broke the then very tight nexus between eosinophilic inflammation and airway hyperresponsiveness. The second discrepant observation was reported in 1995. Turner et al. reported subjects with asthma who had symptomatic disease, frequent use of beta2-agonist, airway hyperrresponsiveness, but had normal levels of sputum eosinophils (5) (Fig. 1). These observations directly challenged the eosinophilic nature of asthma, and now many cases of asthma without sputum eosinophilia are recognized, a condition termed non-eosinophilic asthma (NEA). The remainder of this paper will examine NEA, but perhaps the key observation has already been made, namely that clinical scientists can use induced sputum analysis to investigate patients under their care, and this will provide them with a new and useful dimension to understanding their patients' problems, at times with paradigm-shifting results.
Sorting out the basics
Before describing the details of NEA, it is important to deal with a few basic issues. How much asthma is really eosinophilic? Is NEA just a phenomenon of sputum, or does the airway mucosa behave similarly? Is NEA just a steroid-treated disease? And importantly, is the non-eosinophilic phenotype reproducible?
The best way to address the first question (How much asthma is really eosinophilic?) is to measure the population attributable risk, i.e. the proportion of asthma in the population that can be attributed to eosinophilic inflammation. To do this requires an epidemiological study where both asthma and sputum eosinophilia are measured. These are challenging studies, and few have been reported. Pearce et al. used this approach to re-analyse population studies of asthma and atopy (6). They then applied a similar methodology to three community studies of asthma and airway eosinophilia. The results were both consistent and important (7). The population attributable risk of asthma because of eosinophilic inflammation was about 50%. This meant that 50% of asthma could be attributed to eosinophilia. Perhaps the converse was more important, namely that up to 50% of asthma could not be attributed to eosinophilia. This figure was also similar to the prevalence of NEA that was being seen in clinics.
Is NEA just a phenomenon of sputum, which already had a reputation of being difficult to deal with? To address this question required a careful analysis of airway tissue. This can be achieved by bronchial biopsy, or by analysis of airway tissue obtained at autopsy. A detailed analysis of autopsy material in terms of inflammatory phenotype has not been reported. It is needed to provide further validation of the NEA concept, but will be difficult to do because of the scarcity of cases. A probable confounder that will need to be addressed in such a study is the fact that more severe forms of asthma tend to exhibit a complicated inflammatory phenotype with increases in both eosinophils and neutrophils, termed mixed granulocytic asthma (MGA) (see below). This means that the best study will require large subject numbers, with some ability to characterize severity before death, and preferably cases where asthma was not the cause of death. This leaves bronchial mucosal biopsy as the available means to assess whether NEA occurs in the airway wall, as well as induced sputum. Bronchial biopsy studies have been conducted, and these confirm that the non-eosinophilic phenotype is present in the airway mucosa, as well as the airway lumen. Berry et al. performed bronchoscopy with lavage and biopsy on well-characterized adult asthmatics (8). The diagnosis of NEA was confirmed using stringent criteria, namely sputum eosinophils <3% on two separate occasions, and a non-response to corticosteroid in a placebo-controlled trial. The results showed that subjects with NEA had significantly less eosinophils in their bronchial mucosa than subjects with EA, and that they were not different to healthy controls. This confirms that NEA is a consistent pattern/phenotype in the airway lumen and the airway mucosa.
Is NEA just a steroid-treated eosinophilic disease? This is a common question which can be addressed from two different perspectives. Firstly, does the absence of eosinophils merely represent that the patients have had corticosteroid treatment, and that they were once eosinophilic? While this is almost certainly true in a proportion of cases, it does not apply to the majority. We know this because studies describing NEA have consistently included a proportion of their subjects who were not taking corticosteroids, and as such, had NEA but were steroid naïve. In fact, the first report of NEA by Turner et al. stratified the NEA subjects by steroid use, and reported normal eosinophils in steroid naïve NEA (5) (Fig. 1). But, the issue won't go away, and perhaps that is because a key question is to know exactly what proportion of NEA patients are steroid-treated EA. To answer this question requires a corticosteroid reduction study, and these studies remain to be reported. My guess is that 30% of NEA represents treated EA.
As mentioned earlier, there is another perspective on this issue. That is, to ask the question in a different way, and in fact in a way that has more clinical applicability. A key aspect of asthma inflammatory phenotype analysis is that it can be applied to individual patients (2). This is a fundamental difference with studies that analyse subject groups, and report average results. For example, while it is true that, on average, people with asthma have increases in airway eosinophils, when individuals from the groups of asthmatics are examined, it is seen that many do not have eosinophil levels increased above the normal range. This important subtlety is hidden within group summary data, but can be revealed by individual patient analysis, as provided by inflammatory phenotype analysis. It can also be usefully applied to the question of whether NEA represents steroid-treated EA. In this situation, when faced with a patient who remains symptomatic despite corticosteroid therapy, the clinical issue becomes what might be causing the symptoms and will increasing corticosteroid help. If a NEA pattern is identified in a symptomatic patient on corticosteroids, then additional corticosteroids are unlikely to be beneficial. Whether or not NEA represents steroid-treated EA is not relevant here, as the question when applied to an individual patient concerns the probability, or not, that more corticosteroid will result in a beneficial outcome, and analysis by individual phenotype provides that information.
Is the non-eosinophilic phenotype reproducible? This is an important question, and unfortunately there are too few data. As the question above, it can also be addressed from two perspectives. Several studies have looked at reproducibility of sputum cell counts. Many studies from many different laboratories have reported that induced sputum cell counts are reproducible (9–11). Additionally, the reproducibility of the particular phenotype classification has been examined, and the agreement between classifications of phenotypes over several weeks has been found to be over 70% (12). The second perspective concerns using the phenotype analysis to address the likely contribution of eosinophilic inflammation to current symptoms and response to corticosteroid treatment. This does not require long-term reproducibility, but merely an association between the presence of a particular inflammatory phenotype and a subsequent treatment response. Several RCTs will be described below that establish that inflammatory phenotype analysis can be used to guide treatment with positive clinical results.
Description of phenotypes
The definition of NEA is symptoms of current asthma, increased variability of airflow, occurring in a patient with normal levels of induced sputum eosinophils. Four distinct inflammatory phenotypes can be identified based on the presence (or absence) of increased levels of eosinophils and neutrophils (Table 1). These phenotypes are termed eosinophilic (increased eosinophils); neutrophilic (increased neutrophils); MGA (increased both eosinophils and neutrophils); and paucigranulocytic (normal levels of eosinophils and neutrophils) (12). The relative frequency of these phenotypes will depend on subject selection methods. In a tertiary referral adult respiratory clinic, where most patients are treated with inhaled corticosteroid, the relative proportions are eosinophilic 41%, paucigranulocytic 32%, neutrophilic 28% (Fig. 2). MGA is typically seen in complicated asthma, or refractory asthma. In other settings, the proportions may be different, e.g. in steroid naïve populations the frequency of EA is higher.
Table 1. Inflammatory phenotypes in asthma
Induced sputum cut-points
The phenotypes were originally described using induced sputum samples processed using the plug selection method with cellular dispersion using dithiothreitol (selection–dispersion method). Several other sputum induction and analysis methods are available (11); however, because there is good agreement between cell counts obtained using the whole-sample method and the selection–dispersion method (13), and between induced and spontaneous sputum samples, these other sputum sampling techniques are also likely to give reliable detection of inflammatory phenotypes.
The principle of inflammatory phenotype analysis is that it is applied to the analysis of individuals, rather than groups of subjects. Consequently, whereas on average there are increased eosinophils in asthma compared to controls, an analysis of individual subjects identifies elevated eosinophils in up to 50% of individuals with asthma. This means that the recognition of inflammatory phenotypes will be sensitive to the cut-points used for definition of phenotypes. Two approaches have been taken to the definition of cut-points. The first uses values above the normal range, whereas the second uses values associated with a particular outcome, e.g. response to corticosteroid. For example, EA has been described as an eosinophil count of >1% or of >3%. The former cut-point represents the upper limit of normals, whereas the latter represents a cut-point that is both above the normal range and is associated with a favourable response to corticosteroid therapy. This was clearly reported in a study by Meijer et al. who conducted a randomized double dummy, placebo-controlled trial of corticosteroid therapy in 120 adult asthmatic subjects with unstable asthma (14). The response was assessed by change in lung function as FEV1%predicted, and results were stratified by baseline sputum eosinophil count (Fig. 3). As expected, subjects with elevated sputum eosinophils had a significant improvement in lung function with corticosteroid. However, there were two important additional observations made. Firstly, the degree of lung function improvement was greatest in those subjects with the highest sputum eosinophil percentage, and secondly, there was no significant improvement in lung function in those subjects with sputum eosinophils less than 3%. This study clearly showed a relationship between the intensity of sputum eosinophilia and clinical response to corticosteroid in asthma. It is one of several studies that justifies an outcome-based definition for EA/NEA, and increases the clinical utility of the classification.
A relevant issue is whether a differential percentage cell count or an absolute eosinophil count should be used to define EA. It is possible to construct a ‘normality-based’ definition of both eosinophil percentage and absolute cell number by choosing a relevant value for the upper limit of normal. The absolute cell number definition gives similar result to a percentage-based definition when normals and asthmatics are compared. However, problems arise in airway diseases where the total cell count is increased, such as it may be in cases of airway disease complicated by bronchiectasis, some cases of chronic obstructive pulmonary disease (COPD) and bacterial infection. In these cases, there is a significant increase in total cells, which is predominantly neutrophils, but also includes the other leucocytes present in the airway, such as eosinophils and macrophages. In these cases, the absolute number of eosinophils may be as high as it is in asthma, yet the percentage eosinophils remains low, and below 3%. In these cases, a response to corticosteroid is associated with a selective eosinophilia, i.e. a relative increase in eosinophils above 3%, which can be recognized by a percentage-based definition.
A consensus is emerging that the most useful cut-point for EA is >3%, where an outcome-based cut-point is used for the detection of EA. This allows recognition of both EA and NEA (sputum eosinophils <3%), and permits the application of phenotypes to individual patients (EA or NEA), with a strong evidence base supporting likely treatment responses (positive responses to corticosteroid in EA, minimal responses in NEA).
Further analysis of induced sputum cell counts can be used to identify subtypes of NEA, such as neutrophilic asthma or paucigranulocytic asthma. At present, the recognition of neutrophilic asthma is based on a value of sputum neutrophils above the upper limit of normal, at 61% (12). This is largely because there are only limited data relating airway neutrophil levels to outcomes. However, this may require more investigation in the future. Age is a relevant determinant of airway neutrophil levels, and this will need to be considered and may result in a modification of neutrophil cut-points in the future (15). A further issue relates to the use of absolute vs differential cell counts to recognize neutrophilic asthma. Because many of the clinically important neutrophilic asthma cases have an elevated total cell count, incorporation of this feature into the definition by using absolute neutrophil count may emerge as a more robust definition. There is also recent report of a relationship between absolute neutrophil count and the degree of fixed airflow obstruction (16). Thus, it may be possible to develop an outcome definition for NA that relates to a clinical outcome, as has been done with EA. This would provide a robust and clinically meaningful definition for NA.
It should be noted that the quest for phenotype definitions based on clinically relevant outcomes is important, and corresponds to the approach used in other chronic diseases such as hypertension and hypercholesterolemia. It stands in contrast to the normality-based approach recommended for airflow obstruction, such as in COPD. If successful, may be able to avoid some of the difficulties that the definition of COPD is facing regarding the appropriateness (or not) of age adjustment.
Clinical features of inflammatory phenotypes
The clinical features of individuals with different inflammatory phenotypes do not permit precise recognition based on clinical features alone. At a group level, there are some differences in clinical features that are worth commenting on. Patients with neutrophilic asthma tend to be older, with an increased prevalence of fixed airflow obstruction, have a lesser prevalence of atopy and a greater prevalence of past smoking. People with EA tend to be younger and have more severe airway hyperresponsiveness (12, 17).
When assessed using high-resolution computed tomography (HRCT) scanning, there is evidence of bronchial wall thickening in neutrophilic asthma, and earlier studies have reported an association between eosinophilic airway inflammation and bronchial wall thickening in HRCT scanning (17). This suggests that both airway pathologies are capable of inducing a bronchiolar abnormality associated with thickening of the airway wall. The features in neutrophilic asthma differ from COPD, where there is less thickening of bronchial walls and a greater degree of emphysema present.
Association with mechanistic pathways
It is now possible to associate some inflammatory phenotypes with specific mechanistic pathways. This approach permits the recognition of endogeneous asthma phenotypes, or ‘endotypes’(18). The EA phenotype is best recognized at a mechanistic level. In EA, allergens interact with IgE to elicit a T-helper type 2 cytokine response. This is characterized by utilization of signal transduction pathways using Jak and Stat kinases, and results in the release of eosinophil active cytokines such as IL-5. These cytokines promote eosinophil growth, differentiation and activation, and the result is an influx of eosinophils into airway tissue with release of eosinophil mediators. These are capable of modulating airway hyperresponsiveness, and the resulting inflammatory cell activation contributes to increases in airway responsiveness. In EA, there is also prominent release and activation of matrix metalloproteinase-9 (19). This potent enzyme has many functions and is believed to participate in airway wall remodelling. In NA, the released MMP-9 is not active, but is complexed with antiproteases. Parenthetically, effective therapies have been developed for each step of the EA mechanistic pathway, yet asthma persists, further underscoring the existence of other non-eosinophilic pathways leading to the development of clinical asthma.
Whereas EA represents a specific up-regulation of an acquired immune response, neutrophilic asthma exhibits dysregulation of the innate immune pathway (20). Early studies of the inflammatory features of NEA identified increases in airway neutrophils and the neutrophil active chemokine IL-8 (21). There was also evidence of prominent neutrophil activation with increased levels of free and active neutrophil elastase. This pattern of response is typical of activation of the innate immune system. This is an ancient and evolutionary conserved arm of the immune system. The response is activated not by antigen, but by highly conserved molecular patterns, typically found on microorganisms (termed pathogen-associated molecular patterns), or released with damage to tissue (termed damage-associated molecular patterns). These interact with innate immune receptors, e.g. toll-like receptors, usually in association with chaperone molecules, such as collectins like mannose-binding lectin, or CD14, and result in activation of specific innate immune signal transduction pathways. The nuclear transcription factor, NFkB, is then activated, translocates to the nucleus and promotes gene activation of cytokine genes such as IL-8. IL-8 is the most potent neutrophil chemotactic factor for the lung, and leads to neutrophil influx and activation. The products of neutrophil activation, such as LTB4, can then increase airway responsiveness. This pathway is well mapped out in model systems. Based on the similarities between innate responses and the IL-8/neutrophil responses in NA, we hypothesized that NA represented a dysregulation of the innate immune response in the airway and developed assays for sputum to test this. We compared markers of innate immune activation in the different asthma inflammatory phenotypes, as well as healthy controls and bronchiectasis. The bronchiectasis group was included as a positive control group, because they are known to have active neutrophilic inflammation and bacterial colonization, and should have active innate immune responses. The results confirmed evidence in increased gene activation for the innate immune receptors TLR2 and TLR4 in neutrophilic asthma, supporting the presence of innate immune dysregulation in NA (20).
Relation to clinical management and treatment responses
Induced sputum cell counts have been used to aid clinical management in a variety of ways. The underlying principle relates to the association between a clinical response to corticosteroids and the presence of a selective sputum eosinophilia. This is established to be the case in EA (14), but is also relevant in NEA induced by cigarette smoking (22) and eosinophilic disorders with asthma-like symptoms occurring in the absence of airway hyperresponsiveness (3, 23).
Detection of non-adherence
In asthma, corticosteroids are effective at suppressing eosinophilic inflammation. This indicates that the persistence of sputum eosinophilia in mild to moderate asthma suggests either non-adherence or inadequate corticosteroid dosing (24) (Table 2).
Table 2. Potential clinical applications of induced sputum analysis in asthma
Detection of non-adherence to corticosteroid therapy
Assessment of adequacy of inhaled corticosteroid therapy
The use of serial induced sputum cell counts to guide asthma therapy has been studied in three randomized controlled trials (25–27), and summarized in a systematic review (28). These studies used a clinical guideline-based algorithm to adjust treatment, and compared this with treatment adjustment based on sputum eosinophil counts. In the sputum algorithm, corticosteroids were administered to suppress sputum eosinophils, and doses were reduced when eosinophils were below 3%. Additional bronchodilator therapy was added when symptoms were present and eosinophils were suppressed. The results were consistent across the studies and showed a significant reduction in asthma exacerbations when asthma was managed using sputum cell counts. As an important proof of the concept, the Jayaram study characterized the inflammatory pattern seen during exacerbations in both treatment arms, and observed almost complete suppression of eosinophilic exacerbations with the sputum strategy during the first year of the study. This is an amazing result which emphasizes the importance of a sputum-guided approach to therapy in asthma (Fig. 4). The absence of effect on non-eosinophilic exacerbations indicates the need for further interventions to target neutrophilic inflammation in asthma.
Refractory asthma is an uncommon problem, but these patients are responsible for a significant disease burden. Many patients with refractory asthma require maintenance oral corticosteroid (OCS) to maintain control or prevent severe exacerbations. OCS dose adjustment is usually achieved using symptoms and lung function measures, but these measurements are known to have a poor relationship to underlying airway inflammation. Pizzichini et al. showed that serial measurement of sputum eosinophils could be a useful measure to monitor inflammation in refractory asthma, and could be an aid to adjustment of OCS dose (29). During OCS dose reduction to establish the minimum maintenance dose, sputum eosinophils increased 6 weeks before clinical deterioration, indicating a role for serial monitoring of sputum eosinophils in refractory steroid-dependant asthma.
Assessment of occupational asthma
Exposure to occupational sensitizers represents an important cause of asthma in adults. Recognition and management can be difficult. The ‘gold-standard’ assessment requires specific inhalation challenge in a controlled situation. The access to such facilities is very limited, and serial induced sputum cell counts have emerged as a clinically useful and cost-effective alternative to aid in the diagnosis of occupational asthma (30, 31).
A key limitation to the application of induced sputum is the perceived difficulty associated with the technique. This belief is widespread, and has prevented the uptake of the technique, despite the very strong evidence base supporting its use. The equipment required is not expensive, and the skills required can be learnt. They require the intersection of two distinct disciplines: respiratory function testing to perform the bronchial provocation and sputum induction, and secondly, laboratory pathology to perform the sputum processing, staining and cell counts. The equipment required is not extensive, and technical staff can perform the cytological techniques required for cell counts. Attempts to automate the process have either not been successful or not been widely adopted.
Induced sputum analysis is a useful tool to investigate and aid in the clinical management of patients with airway diseases. It can provide the clinician with a new window on the patient's problem, suggesting alternative pathologies, or permitting changes to treatment for airway disease. It is also supported by a strong evidence base. The restricted application of the technique remains unexplained, and application of induced sputum in clinical practice does require active acquisition of new skills by practitioners.