Discussion
This is the first real-world study to compare the parallel biologic naïve characteristics of OMA and MEPO treated subjects against a common comparator that remained biologic naïve (SNB) within the same cohort. A key finding was that despite potentially overlapping clinical indications for these drugs, OMA and MEPO treated patients showed distinctive asthma phenotypes. Regardless of that, both biologic response rates were comparable to RCTs and other reports (19–26). In addition, for OMA, even though 43.3% (45/104) patients were potentially dual-eligible for MEPO, the OMA response rate was 88.5%, reiterating that OMA was efficacious in the phenotype that received it. Indeed, both OMA and MEPO conferred substantial multidimensional clinical benefit to the real-world populations which received them. This testifies to the success of both biologics in targeting treatable traits in difficult asthma patients. Nevertheless, non-responders emerged to both drugs, typically characterised by worse baseline disease and psychological comorbidity.
Of those with highest disease burden in our cohort (OMA, MEPO, SNB), 51.6% did not receive biologics. Both biologic receiving groups had hallmarks of greater disease severity with significantly worse lung function (Clinic FEV1%), worse airway inflammation (FENO) and greater mOCS dependence compared to SNB. The SNB group are noteworthy as their characterisation derives from the timepoint of WATCH enrolment, when 42.3% of them were new clinic referrals, and thus at an early phase of conventional treatment optimisation. Their lack of subsequent biologic need infers a responsiveness to that conventional optimisation. Conversely, biologics treated subjects had already undergone substantial treatment optimisation, without sufficient response, prior to their characterisation in this study.
Although our three high burden groups were similar with regards to ACQ6, exacerbations and AHE, they were phenotypically distinct. OMA patients had a younger, early-onset, atopic phenotype, with high proportions of co-morbid ABPA and rhinitis. Conversely, MEPO patients had an older, male, late-onset, eosinophilic but less atopic phenotype, associated with higher prevalence of nasal polyposis but less dysfunctional breathing. This adds further insight to preliminary suggestions of typical patient features for these biologics outlined by GINA (27). Additionally, stratification of biologic use by our recently described age-of-onset/sex phenotypes (15) further corroborate these findings, as we found significantly different biologic use across the four phenotypes. Notably, the female/early-onset phenotype showed highest prevalence of OMA and lowest prevalence of MEPO use, while the male/adult-onset phenotype showed highest prevalence of MEPO use. This may part explain a disparity in biologic effect, whereby OMA but not MEPO, significantly improved Clinic FEV1%, mirroring findings of another real-world UK study (26). This observed difference could be partly explained by the higher representation of MEPO patients among the male/adult-onset phenotype which had poorest baseline lung function (15) in our cohort. Other reports have also confirmed that such patients have more severe, persistent airflow limitation (28), potentially explaining their limited lung function improvement. Another disparity was in steroid-sparing effect, whereby MEPO but not OMA, significantly reduced mOCS dependency. This may reflect different trial durations. Indeed, other studies which evaluated OMA beyond 16-weeks, found that it reduced the proportion of patients on mOCS (16,21,22,29,30). This may also reflect clinical practice during the single biologic phase, whereby a more conservative mOCS weaning approach may have been adopted, given the lack of alternatives.
Overall, there is limited knowledge on clinical predictors of OMA and MEPO response. A pooled analysis of seven clinical trials found that baseline characteristics were unable to reliably predict OMA benefit (23). For MEPO, post-hoc analyses of RCT data suggested that baseline PBE could be a useful predictor of response (31), but this was not consistently observed in real-world studies (24,26), including ours. Instead, our data suggests that patients with the most severe and poorly controlled baseline disease were poorest responders. Thus, for MEPO, better baseline asthma control was independently associated with response and super-response. This mirrored the findings of Kavanagh et al. (26), where in their cohort, poor disease control at baseline was independently associated with MEPO non-response. Similarly, in OMA, more ‘severe’ exacerbations, AHE, at baseline were associated with non-response, while being on mOCS, was associated with non-super-response. However, while AHE may represent more ‘severe’ asthma exacerbations, they may also reflect impact of multiple influences beyond just airways disease. Indeed, Burke et al. identified that those with repeated AHE were a subgroup of difficult asthma patients with multiple aggravating comorbidities including obesity, Gastrointestinal reflux disease, dysfunctional breathing and psychological morbidity(32). Such complex multifactorial health events may be less responsive to a simple biologic approach. It is notable that by adopting a holistic, asthma MDT approach, they reduced AHE significantly(32). Collectively these findings emphasise the importance of comprehensive, up-front characterisation of difficult asthma patients, focused on addressing all treatable traits to maximise biologic outcome.
Reinforcing this, our data uniquely showed that psychological co-morbidities may be associated with biologic non-response, an unexplored aspect by other real-world biologic studies. Anxiety was independently associated with OMA non-response while depression was independently associated with OMA non-super-response and was associated with MEPO non-response. Psychopathologies have been associated with biologic non-response in other diseases. Analysis of the British Society for Rheumatology Biologics registry showed that depression reduced the odds of biologic response (33). The impact of psychopathology on biologic outcome could be secondary to the well documented interplay between psychological disease and SA (34). Psychopathologies have been associated with worse asthma control, more exacerbations and more AHE (35–37). Furthermore, studies have shown that proinflammatory cytokines associated with asthma are raised in depression and anxiety (38,39), which may dampen biologic effect. Brown et al showed in a RCT that 12-week continuous escitalopram therapy for SA patients with co-morbid major depression significantly reduced OCS use and asthma control (40). As such, our findings encourage proactive management of psychological comorbidity alongside consideration of asthma biologics.
Analysis of our pooled biologic data allowed us to describe an overall biologic unresponsive group. They had early-onset asthma, were predominantly female yet had comparable exacerbations, mOCS dependence, FENO, lung function and asthma ICU admissions to responders. However, they were characterised by more AHE, a larger proportion of multiple AHE, significantly worse baseline asthma control, alongside greater proportions of anxiety and depression. We postulate their biologic unresponsiveness may have been augmented by their high burden of psychopathologies as although their objective disease markers and clinical co-morbidities were equivalent, their subjective markers of disease were not. Our recent work has shown that this group of early-onset, female patients have the highest prevalence of psychological co-morbidities, yet also have the highest frequency of biologic use (15). This reiterates the importance of holistically addressing treatable traits, through addressing psychopathologies before biologic therapy.
Head-to-head comparisons between OMA and MEPO response rates were not appropriate in our data, given their different phenotypic traits, and the different response tools employed. However, notably,10/15 MEPO ‘non-responders’ displayed responses in domains outside NICE criteria. Particularly noteworthy were those who sustained an improvement in AHE status, including one who had both ACQ and AHE status response. However, despite improvements in disease control and healthcare utilisation, important markers of economic and patient-centred efficacy, these MEPO patients were not classified as responders according to NICE criteria(14). A post-hoc analysis of two MEPO RCTs found that ACQ was unreliable in predicting MEPO response (41). Though important, their findings were based on RCT data which may have limited transferability to real-world patients (7–9). Additionally, ACQ is used to gauge MEPO response in the Australian Mepolizumab Registry, and shown to correlate with improvements in objective measures(25). Conversely, few MEPO responders did not sustain an ACQ or AQLQ response. This could be because improvement in NICE defined domains may not equate to the patient’s perception of better asthma control, or quality-of-life. Thus, in those who are borderline responders, consideration might be made to measure MEPO response more holistically, perhaps by taking into account a wider range of measures. However, the economic implications of any such move need careful deliberation.
Our study had limitations. Inherent to real-world observational studies, we had some missing data. However, real-world data capture is representative of clinical populations receiving these treatments. Our report is also limited by the small numbers in the MEPO group, which prevented us from uncovering whether the different age-of-onset/sex phenotypes had differing response predictors. Therefore, future studies are needed to clarify these findings and further explore age-of-onset/sex related signals. Our study had several strengths. We are the first to report detailed real-world clinical outcomes on both OMA and MEPO in parallel, against a non-biologic comparator in a difficult asthma cohort. Additionally, our cohort represents an extensively characterised difficult asthma population from a wide geographical catchment, enhancing generalisability of findings. This allowed mapping of previously described clinical clusters onto our data, consolidating our observations. We also undertook pooled analysis of the non-responder group and explored other definitions of response in MEPO, compared to OMA.