In this community based study of adult asthmatics, in cross-sectional assessment the dominant asthma phenotypes were EA (40%) and PGA (52%). EA was associated with increased airway obstruction, ACQ7 score and BDR. In the nested CIT study, alteration in asthma treatment resulted in more than half of the asthmatics changing phenotype; in particular, sub-optimisation of treatment was associated with a non-significant increased prevalence of EA and NA, whilst optimisation of treatment was associated with an increase in PGA. Treatment optimisation was associated with clinical improvement, reduced eosinophil numbers and increased neutrophil function, statistically significant only in EA.
In our sample of adult asthmatics we found NA in only 8%, consistent with other studies [10, 11]. However, some previous studies have reported that NA is common [5, 13]. We speculate that the higher prevalence of the neutrophilic phenotype found in such studies may be in part due to specific environmental exposures associated with particular occupations or heavy industry [24,25,26]. Furthermore, NA has previously been reported in association with aging [5, 27]. In agreement, we observed NA only in older individuals, with 3/4 asthmatics classified as NA at any assessment being >60 yrs. of age. Also, consistent with others [28, 29] we found significantly lower FEV1, FEV1/FEV, and higher ACQ7 scores and BDR in EA.
PGA (which made up the majority of NEA) was associated with significantly higher eosinophil percentages than non-asthmatics despite levels being below the 2% cut-off used to define EA. Therefore, PGA - as defined in this and other studies - may not necessarily be indicative of different asthma pathology, but may represent a less evident form of EA, overlapping phenotypes [30], or be due to treatment effect (discussed further below).
Whilst some studies have shown that NEA is relatively stable over time [5], 52% of asthmatics in our study altered phenotype following changes in treatment. Similar findings have been previously reported. For example, Hancox et al. found that NEA/EA changes occurred in 50 to 100% of adult asthmatics (n = 54) in response to changes in ICS therapy [11]. Similar findings have also been described in children [12]. ICS use may therefore contribute to misclassification of NEA, as it may reduce sputum eosinophil numbers below the commonly used cut-point of 2%. Our findings, although not reaching statistical significance, thus support previous suggestions that inflammatory phenotype classification based upon a single assessment may not be valid for all asthmatics [31].
Although some previous reports have shown ICS to be less effective in NEA compared with EA [4, 32, 33], we observed improvements in ACQ7 score in the NEA group with optimised treatment. Other studies have also shown that symptoms improve in NEA with ICS treatment, although EA remains associated with a better response [10, 11]. As noted above, it is possible that some NEA subjects have a degree of ICS-responsiveness, but have undetectable (in sputum) eosinophilic inflammation. Alternatively, ICS may act through suppression of non-TH2 mediated pathways in NEA, such as epithelial cytokine production [34].
ICS use has previously been associated with an increase in sputum neutrophils and neutrophil survival [10], possibly via impairment of apoptosis [35]. We saw no significant changes in airway neutrophil percentages with treatment alterations. However, 33% (3/9) of individuals with PGA changed to NA after a reduction in ICS dose. This contrasts with previous findings suggesting that asthma exacerbations after tapered ICS withdrawal are associated with eosinophilic inflammation [36, 37]. However, one previous report suggested that sudden ICS withdrawal, as conducted in our study, may result in neutrophilic exacerbations [38]. It is therefore possible that the nature of airway inflammation post-ICS withdrawal may be dependent upon the kinetics of withdrawal.
We observed that neutrophil function (oxidative burst and phagocytosis) was enhanced with treatment optimisation, with the greatest effect observed in EA. Although earlier asthma studies have investigated sputum phagocyte function [39, 40], we believe that this is the first study assessing neutrophil function in response to changes in asthma treatment. The association between improved neutrophil function and treatment optimisation is intriguing. It is possible that improved neutrophil phagocytosis may be one of the mechanisms by which asthma treatment leads to a reduction in eosinophilic inflammation, possibly through improved efferocytosis [15]. Alternatively, improved neutrophil function may have affected the airway microbiome [41] as suggested by significantly reduced bacterial endotoxin (a strong pro-inflammatory agent) levels (Table 2). It is also possible that impairment of neutrophil apoptosis (described above) may have altered neutrophil maturation status, leading to the altered functional phenotype observed. However, larger longitudinal investigations are required to confirm these findings and more comprehensively determine the effect of medication on neutrophil function; in particular, whether the functional changes observed are directly relevant to the improvement of asthma symptoms, whether treatment leads to alteration of peripheral neutrophil function (blood samples were not available for assessment in the current study) or if treatment is indirectly affecting neutrophil function through modulation of other cell populations, such as macrophages or eosinophils. Also, although not addressed here, there is a possibility that the hypertonic saline challenge or sputum sample processing procedure may directly result in altered or activated neutrophil phenotype. Previous studies have reported increased expression of the activation markers CD11b and CD66 on sputum neutrophils when compared with blood or broncho-alveolar lavage neutrophils [42, 43], but it is not clear if this is due to the airway environment or sampling procedure. However, in the current study, as all samples were processed in the same manner, it is unlikely that the observed differences could be due to either saline challenge or sputum processing.
There were some limitations to this study. Firstly, although asthmatics reported a doctor’s diagnosis of asthma and recent symptoms, as we did not use objective tests (such as bronchodilator reversibility or airway hyperreactivity) to confirm asthma diagnosis, it is possible that some misclassification may have occurred. However, we consider that any bias introduced as result will be minimal given that this approach, which has been used in many previous studies [14, 16,17,18, 26], generally compares well with more clinical definitions of asthma [16, 17]. Also, in some cases it has been shown to be better than more objective measures such as airway hyperreactivity [18]. Indeed, there are a number of issues with objective testing for confirmation of asthma diagnosis in a community based setting, particularly given the inherently variable nature of the condition, and when the majority of asthmatics are not treatment naïve (76% of asthmatics in the current study were using ICS at the time of assessment). This (amongst other reasons) has led to recent recommendations that asthma be considered on the basis of symptoms rather than pathophysiology [44]. Secondly, as the majority of asthmatics were undergoing ICS therapy, and there were safety concerns about total ICS withdrawal in some cases in the CIT study, it is possible that ICS treatment may be “cloaking” physiological responses, as well as eosinophilic inflammation. Thirdly, the relatively large number of cells required to investigate cell function meant that only samples with larger cell yields could be assessed. For this reason, respiratory burst was assessed in relatively few samples (n = 7 for respiratory burst and 16 for phagocytosis assessment). This may have resulted in selection bias. However, our data suggest that there is no significant difference between sputum total cell count between asthmatics and non-asthmatics, and therefore this is unlikely to be a major issue. Fourthly, treatment changes were limited to a 4–6 week period (similar to previous studies [10, 11] and varied somewhat among CIT study participants as asthma management was considered on a case-by-case basis i.e. subjects did not receive a uniform reduction or increase in ICS dose (Additional file 1: Table S1), and in cases in which participants were using combined ICS/LABA inhalers, LABA was also altered. All participants were also using SABA as required. This may have led to some variation in the effect on (particularly eosinophilic) inflammation and/or asthma control, as there are reports that some SABA (in particular terbutaline) may have a permissive effect on airway inflammation, and their use may be associated with an increase in airway eosinophils [45]. However, the evidence for this is mixed, and may not be the case for all SABA [46, 47]. Fifthly, the asthmatics studied were recruited from the general population, and as such had varying levels of asthma control and severity at assessment which makes the interpretation of the results more difficult. Finally, the relatively small number of participants included in the CIT study (due to the reluctance of some study participants to undergo CIT) may have led to non-significant findings in terms of inflammatory phenotype changes. Despite the latter two limitations, the trends observed (i.e. increased EA with suboptimised treatment, reduced eosinophils, improvement in asthma control and increased NEA with optimised treatment) were generally similar to those previously described in studies in which ICS was withdrawn or added to asthma therapy [10, 11].