Evaluation of real-world mepolizumab use in severe asthma across Europe: the SHARP experience with privacy-preserving federated analysis

Discussion

In this study, SHARP succeeded in linking existing patient-level data from 10 different national registries for severe asthma to evaluate the use of mepolizumab therapy across Europe. With the use of a federated analysis approach, the study shows that treatment with mepolizumab significantly reduces severe exacerbations, as well as maintenance OCS use in our sample of patients with severe asthma. In addition, the results show substantial heterogeneity among patients initiating mepolizumab, and rates of switching and discontinuing mepolizumab treatments across European countries. The registry data used for this study were pre-existing, and turned out to be far from complete for current purposes. This fact, in combination with strict requirement for defined entry and outcome data to be available to enable the federated analysis to be undertaken, resulted in a high analysis dropout rate. Thus, the results of this study, while in line with those of RCTs, should be interpreted with these limitations in mind.

The present real-world study supports and complements the results from RCTs and previous real-world studies on the effectiveness of mepolizumab therapy [5, 15–18, 26–29]. While the RCTs involve large numbers of patients from around the world, most real-world studies were conducted with data from a single institution or national registry. In addition, these studies often included small numbers of patients [30]. We are aware of only one real-world study that used data from patients with severe asthma from different countries [19]. In contrast to our study, patients in this study were prospectively followed over time, capturing data from clinical practice and recorded in a standardised way. Our study, which made use of data already collected by clinicians in 10 different European countries, not only complements the findings from the previously published studies, but also reflects the effectiveness of mepolizumab in daily clinical practice. Registries have been recognised as an important tool to provide real-world evidence [8]. By connecting several registries, our study further seeks to generate real-world evidence in large populations.

In our study, we used a relatively new approach to link privacy-sensitive data from clinical disease registries. That process was not without its challenges. The complexity and labour-intensive nature of harmonising privacy-sensitive data from different sources has been extensively described by Biedermann et al. [23]. In their study, three registries of pulmonary hypertension patients were linked and the data analysed in a federated manner. The harmonisation process in our study was even more complex, as all registries had a different data model. This makes our study the first to have harmonised and used nonstandardised real-world disease registries to obtain real-world evidence. The harmonisation process has been completed and the present proof-of-principle study has demonstrated that this federated approach can produce valid results. The platform can now be further used to obtain real-world evidence to help guide better treatment and care to the many thousands of patients with severe asthma in Europe.

In addition to the exceptional method by which we have linked registries in a privacy-protective way, our research is unique in several respects. Successfully analysing patient-data that have been collected in clinical practice and entered into local registries by clinicians from 10 different countries without the involvement or monitoring of contract research organisations including pharmaceutical companies is unprecedented to date. This is probably the best method to get as close as possible to daily clinical practice and to compare treatment practices across countries. Another strength of our study is that our large-scale approach has allowed us to analyse data from the time periods before and during the pandemic separately, so that we could avoid bias due to the alterations in circulating pathogens and changed circumstances of care for severe asthma patients.

Nevertheless, this study has its limitations. First, there are the limitations inherent to conducting real-world studies, for example lack of a control group. Due to the lack of a control group, it is not certain that the observed effectiveness of mepolizumab is caused solely by mepolizumab treatment or by other contributing factors. Second, many patients’ data could not be included in the analysis because data were missing or incomplete, or because the moment of data collection was not recorded. In addition, some patients had not given informed consent to use their data for research outside the institution in which they were treated. This feature of real-world data might have led to a selection bias between countries and may limit the generalisability of our study results. Nevertheless, we observe that the health outcomes of well-designed clinical trials can also be detected in a highly imperfect real-world setting. Third, treatment outcomes could only be evaluated for patients still on mepolizumab after 1 year of follow-up, which may have led to a selection bias, potentially overestimating our results and limiting the generalisability of our results. Furthermore, a potential limitation of the analysis is that patients considered at baseline overlapped with, but were not exactly the same as those considered at follow-up. Where paired data was available, the correlation was appropriately accounted for in the analysis. Due to reasons of statistical power, the GEE analyses were not adjusted for possible confounders. Future larger studies might be able to repeat the analyses and include covariates. Finally, there was the effect of the COVID-19 pandemic. This led to many hurdles in data collection and forced us to split the analysis into two periods. Although this reduced the statistical power of the study, we still were able to demonstrate the effect of mepolizumab in reducing exacerbations in both the pre-pandemic and pandemic time periods.

Our study shows that mepolizumab treatment leads to a reduction in the number of exacerbations and in the use of maintenance OCS in a real-world population. While this outcome was expected, it was reassuring to observe this finding in a setting alternative to RCTs or prospective observational studies, as patients included in an RCT are strictly selected, and may differ significantly from real-world populations. Our results imply that physicians should not be concerned that the effect of mepolizumab in their patients with severe eosinophilic asthma will be below expectations, even if they are less rigorously selected in terms of factors that were exclusion criteria in the RCTs. Of note was the consistent reduction in exacerbations across the countries despite the heterogeneity in the patient populations, with respect to baseline demographics.

The characteristics of real-world patients not only differ from those in the RCTs, but our study also found that patients prescribed mepolizumab differed considerably between the different European countries. This heterogeneity may be due to differences in reimbursement practices for biological therapies, to recommendations in national guidelines and therefore the eligibility criteria for biologics, or to preferences and choices of individual physicians. The differences in patient characteristics treated with mepolizumab before and during the pandemic may have influenced the extent to which some patients could or could not cope with remote care [22]. The heterogeneity of our population also illustrates that the definition of “severe asthma” does not appear to be used unambiguously. A revision of the international ERS/ATS definition of severe asthma may therefore be required. Interestingly, we also observed differences in OCS tapering strategies, and this important medical practice also requires greater attention and harmonisation. Fortunately, an important first step in the right direction was recently taken with the PONENTE study [31, 32].

A remarkable finding of our study is the difference between countries in the proportion of patients who continue or discontinue mepolizumab, or switch to another biological treatment. These switchers are likely to be patients who have partially responded to mepolizumab treatment, but have not yet shown full normalisation of all outcome parameters [33]. The fact that such a switch has not been observed in all countries probably reflects different availability of alternative biologics and contrasting practices and levels of acceptance of what is a beneficial outcome. These findings define an area in which greater insight is needed. Until now the best treatment with biologics could only be found based on physician knowledge and experience. It would be less burdensome for all parties if switching of biologics were not necessary and we had good predictors of response. By continuing to use the federated analysis approach and by further optimising national databases and enriching them with biological samples, finding reliable predictors will very likely become possible in the future. More so, if we apply artificial intelligence, machine learning and federated learning to clinical outcome data. The SHARP federated analysis platform is optimally suited for this.

In this study, SHARP demonstrated the real-world effectiveness of mepolizumab in patients with severe asthma from 10 different European countries. Mepolizumab reduced asthma exacerbations and OCS use consistent with evidence generated by RCTs. We observed heterogeneity in characteristics of patients receiving mepolizumab and in rates of switching or discontinuing treatment across countries, signalling the need for further alignment of asthma management across European countries. Our study can be seen as a successful proof of principle as to whether a federated analysis approach can be used to link privacy-sensitive data from different sources. It can thus serve as an example for other clinical research collaborations with a similar ambition. While there is still some room for improvement regarding completeness and quality of data, the SHARP federated analysis platform has great potential for future pan-European real-world severe asthma studies using patient-level data in a privacy-protected way.