Although there was considerable heterogeneity among the few available studies, treatment with CS appeared to have a significant impact on the makeup of the microbiome.
The overall trend for microbiome diversity following treatment with CS was an increase: two studies showed a significant increase both in α-diversity after inhalation of FP (25, 28). One of these studies included patients with stable COPD who received 12 months of FP, the other young adult asthmatics who received FP for 6 weeks. Perhaps this can be interpreted as an indication of an effect independent of the underlying disease, which starts relatively soon and extends for a prolonged period. An increased diversity appears to be beneficial in COPD and protective against asthma (30, 31). The only study with an opposite, albeit non-significant, trend for a decreased diversity was the only study which used oral CS for 14 days (26). Unfortunately, due to the limited number of studies it is impossible to conclude whether this difference was related to the route of CS application.
Most studies detected significant shifts in the composition of the airway microbiome (25, 27, 28). However, there was no clearly shared direction between these shifts. In the respiratory microbiome, Proteobacteria and Firmicutes appeared to be inversely correlated (30). Additionally, elevated levels of the Proteobacteria phylum in COPD appeared to be associated with exacerbations (pre-treatment) (26, 32). Contoli et al. found the Proteobacteria phylum to be significantly reduced in the group treated with FP for 12 months, perhaps indicating a beneficial effect on microbiome composition (25). Contrarily, Durack et al. found Neisseria and Moraxella (both belonging to the Proteobacteria phylum, see Table 2) to be increased in their cohort of steroid-responders (27). Only one study found a significant effect of steroid treatment on the burden of the microbiome in COPD patients (25). This specific endpoint was measured using culture (not a culture-independent method), suggesting a possible methods effect.
Baseline assessments of steroid-naïve asthma patients by Durack et al. tended to have a higher phylogenetic diversity (Faith index) compared to healthy controls (p = .06) (27). A further study comparing steroid-naïve asthma patients to healthy controls found no difference in α-diversity, but detected a significant composition difference on the taxonomical level (33). A clear differentiation between the influence of the treatment and that of the underlying condition is difficult to achieve, particularly in such small cohorts.
Garcia-Nunez et al. found a significant correlation between lung function and bacterial diversity in sputum in COPD patients (30), however, whether these differences also stem from the underlying disease or the treatments prescribed is not known. In COPD, inhaled steroids are an important component of the available armamentarium and significantly slow the decline in quality of life and lower the exacerbation rate (34). Inhaled CS treatment is likewise a pillar of asthma therapy recommended by national and international guidelines (35, 36). The role of the respiratory microbiome in mediating these effects is only being uncovered gradually. Significant differences in the composition of the microbiome were detected between ICS-responders and non-responders (27, 37), potentially allowing for better prediction of treatment response or the development of new treatment options.
Additionally, it can be presumed that the route of application (inhaled or systemic) and substance choice might also play a significant role in how CS affect the microbiome. Due to the current paucity of studies, our review unfortunately cannot adequately assess this question. Three of the included studies investigated the effect of FP on the respiratory microbiome (25, 27, 28). FP might, however, be an outlier among inhaled steroid therapies, as it has been associated with an increased risk of pneumonia in adults as discussed in an overview of systematic reviews comparing FP with budenoside by Janson et al. (38). The authors listed differences in pharmacokinetics and immunosuppressive efficacy as potential reasons for this difference in pneumonia incidence (38). Other pharmacological treatments may also influence the airway microbiome. For example, Durack et al. proposed that the compositional shifts observed in the placebo cohort of their study may be caused by the lactose contained in the placebo medication (27).
Eosinophilia appears to further influence the lung microbiome parameters in COPD and asthma (25, 26, 39). Additional factors such as diet and probiotics also affect the makeup of the respiratory microbiome via the gut-lung axis (13, 15, 40). However, these factors were not adequately assessed or controlled for in the studies identified by and included in our systematic review.
The main limitation of this systematic review is the small number of studies fulfilling our inclusion criteria. Three studies covering the topic of interest only analysed samples from one time point (37, 41, 42). The teams of two further studies responded to our requests for further information but did not include data from non-steroid treated controls (32, 43). The resulting small number of selected studies and their recency is certainly related to the relative novelty of this field of microbiome research. As evidence of ongoing exploration, we identified and contacted the authors of four registered clinical trials and one conference abstract without a full published report investigating this topic. However, only one team had already finalised data analysis by that point. Thus, an update of this systematic review including the data from these studies would be interesting once these ongoing studies become available.
The second apparent limitation is the great heterogeneity between the five included studies regarding analytic methods, such as sequencing techniques and platform, study populations, tested CS agents, dosages, applications and duration of CS treatment. The control populations varied between healthy controls (28, 29), patients with the same disease receiving placebo (27) or SOC (25, 26). Each of these factors potentially affects the outcomes of interest, complicating the interpretation and comparison between the studies. Consequently, the discrepancies in the reported outcomes may also in part be due to the different methods used, particularly between studies using culture vs. non culture-based methods. Therefore, it is challenging to disentangle the true CS effect from these or other unidentified confounding factors on microbiome diversity, burden and composition.
The strengths of this review include the methodologically precise execution and the maximisation of the scope of the search. This allowed an accurate documentation of the current status quo of research on this question and the considerable heterogeneity of methodology. Thus, our results show the limits of the current understanding of this important topic and are informative for the planning of future studies in this field.