Real‐world efficacy of treatment with benralizumab, dupilumab, mepolizumab and reslizumab for severe asthma: A systematic review and meta‐analysis

Abstract Background Severe asthma is a major cause of morbidity. Some patients may benefit from biological therapies. Most evaluations of these treatments are derived from randomized controlled trials (RCTs), but few patients are eligible for these trials. Studies involving more diverse groups of participants exist, but there is a lack of precise pooled estimates. Objective This systematic review aims to evaluate the real‐world efficacy of recently and nearly licensed biological therapies for severe asthma to assess the generalizability of the RCT data. Methods Clinical outcomes including exacerbation rate, oral corticosteroid usage, forced expiratory volume in 1 second (FEV1) and fractional exhaled nitric oxide (FeNO) were examined. Studies were assessed for risk of bias using the Critical Appraisal Skills Programme checklist tool. The certainty of evidence was assessed using Grading of Recommendations, Assessment, Development and Evaluations (GRADE). Results A total of 21 studies examining biologicals in real‐world settings were identified; they mostly focused on benralizumab and mepolizumab. The introduction of biologicals reduced the annualized exacerbation rate significantly by −3.79 (95% confidence interval [CI] −4.53, −3.04), −3.17 (95% CI −3.74, −2.59) and −6.72 (95% CI −8.47, −4.97) with benralizumab, mepolizumab and reslizumab, respectively. Likewise, improvements were observed in FEV1 (0.17 L 95% CI 0.11, 0.24) and FeNO (−14.23 ppb 95% CI −19.71, −8.75) following the treatment with mepolizumab. After treatment with benralizumab, there was an increase in FEV1 (0.21 L 95% CI 0.08, 0.34). Conclusions These data demonstrate that anti‐IL5 biologicals may improve the clinical outcomes of patients with severe asthma in a clinic environment with similar effect sizes to RCTs. The data were mainly retrospective and unadjusted, so estimated effect sizes may not be reliable. More data are needed to acquire accurate effect estimates in different subpopulations of patients.


| INTRODUC TI ON
Asthma is a heterogenous non-communicable chronic disease which is characterized by wheeze, shortness of breath, cough and chest tightness. Approximately, 5%-10% of patients have severe asthma, a form associated with increased mortality, reduced quality of life and increased healthcare costs. 1,2 According to the American Thoracic Society/European Respiratory Society (ATS/ERS) guidelines, asthma can be defined as severe when it requires high doses of inhaled corticosteroids and a second controller agent (which can include oral corticosteroid [OCS]) to prevent it from becoming uncontrolled, or when it is uncontrolled despite this therapy. 3 In patients with asthma, clinical objectives include reducing the rate of exacerbation, improving symptom control and reducing the use of oral corticosteroids. 4 Recent developments in our understanding of the molecular biology of asthma has facilitated new treatment options.
Asthma can be subdivided into endotypes based on the pathophysiology observed in patient populations. In some groups, activated Type 2 T Helper cells produce Interleukin-4 (IL-4), Interleukin-5 K E Y W O R D S asthma, asthma control, benralizumab, dupilumab, exacerbations, FeNO, FEV 1 , mepolizumab, real-world studies, reslizumab

G R A P H I C A L A B S T R A C T
This study examines the evidence for the efficacy of biologics in typical patients with severe asthma. Many of these patients would not have qualified for randomized controlled trials (RCTs). We demonstrate that biologicals are effective for typical patients and that across certain outcomes including forced expiratory volume in 1 second, the gains are comparable to results from RCTs.

Key messages
• Observational data from typical populations can help to establish the generalizability of results observed in randomized controlled trials (RCTs).
• Treatment with biologicals is associated with significant improvements in asthma control, exacerbations and lung function.
• The improvement in results is comparable to RCTs, implying RCTs are applicable to typical populations.
(IL-5) and Interleukin-13 (IL-13) which act as the principal drivers of inflammation. 5 ILC2 cells are another important source of these inflammatory cytokines. These cells are activated by alarmins produced by epithelial cells in response to various biological molecules. 6 IL-5 is involved in the maturation of eosinophils and in the migration of eosinophils to the lungs, where they trigger inflammation and cause hyper-responsiveness of the airways. 7 IL-4 and IL-13 interact with the IL-4Ra receptor subchain and stimulate the production of immunoglobulin E and mediators of airway remodelling. 8 IL-13 also modulates nitric oxide production within the respiratory system, increases mucus production and increases smooth muscle contractility. 8 Monoclonal antibodies target these pathways, mepolizumab and reslizumab bind to IL-5, benralizumab interacts with the IL-5 receptor and dupilumab binds to the shared component of the IL-4/ IL-13 receptor. 9,10 These biological therapies disrupt the action and activities of these key molecules.
Biologicals have been shown to be effective in randomized controlled trials (RCTs). However, it has been recognized that only a minority of patients with severe asthma would also be eligible for inclusion within an RCT. 11 Moreover, in RCTs, many patients randomized to the placebo arm experience a significant improvement in asthma control and a reduction in the number of exacerbations suggesting that these apparent severe asthmatics simply require increased monitoring rather than increase drug therapy. 12 Consequently, there are questions over the generalizability of results from RCTs. It has been suggested that, in general, the conclusions derived from RCTs could be more impactful if supported by evidence of therapeutic effectiveness from clinical practice. As such, real-world studies are becoming more influential, informing decisions by healthcare regulatory bodies including the National Institute of Clinical Excellence (NICE) and Germany's Institute for Quality and Efficiency in Health Care. 13,14 However, when assessing real-world studies, their small sample size and, in some cases, the lack of a control group can limit interpretation. These can both be addressed within the context of a systematic review and meta-analysis through reference to effect measures derived from RCTs and by collating and pooling multiple studies to provide precise effect estimates. In this systematic review, we aimed to evaluate the efficacy of recently licenced biological agents in populations of participants with severe asthma reported in real-world studies to assess the generalizability of the RCT data.

| Protocol and registration
The study protocol was registered on the International Prospective Register of Systematic Reviews (PROSPERO, CRD42020207080).

Grading of Recommendations, Assessment, Development and
Evaluations (GRADE) and Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines were followed. 15,16 The review aimed to assess adults and children with a diagnosis of asthma from 6 years of age treated with any nearly licenced or licenced biological therapies. Clinical outcomes examined included exacerbation rate, asthma control and reduction in the mean daily dose of OCS. Additional outcomes examined included forced expiratory volume in one second (FEV 1 ), fractional exhaled nitric oxide (FeNO) and eosinophil count.
Benralizumab, mepolizumab, reslizumab and dupilumab were examined. Omalizumab was excluded from this analysis as it has been in routine use for over a decade, and the authors chose to focus on newer agents where healthcare professionals have less clinical experience. then selected relevant records by a two-step process. Initially, titles and abstracts of records were screened for eligibility. The full text of any eligible record was then acquired, and a detailed evaluation was performed. Any disagreements about inclusion were resolved by discussion and consensus, or by consultation with a third reviewer (G. R.). Studies were excluded if they were very small (included <20 participants in total) to minimize bias, were animal studies, qualitative research, systematic reviews, narrative reviews, RCTs, editorials, conference abstracts, studies not relating to asthma, studies not relating to specified biologics, studies not in English or studies not focused on real-life research. References were managed with Endnote Version X9 (Thomson Reuters) and Rayyan. 17

| Data extraction and risk of bias assessment
Two reviewers (D. C. and S. N. T.) independently extracted the characteristics and data from the eligible studies. These data sets were checked against each other, and any disagreements were resolved by arbitration and consensus or by consultation with a third reviewer (G. R.). Where necessary authors were contacted and requested to provide additional data, including unadjusted results. Risk of bias of included studies was evaluated with the use of the Critical Appraisal Skills Program tool for cohort studies. 18 Disagreements were solved through discussion; if agreement could not be reached, arbitration with a third reviewer (G. R.) was held.

| Meta-analysis
Descriptive tables that included information on population characteristics, interventions and key outcomes were created. Data were pooled and meta-analyzed using STATA 19 using a random effects model. As outcomes were continuous, mean differences (MD) with 95% confidence intervals (CI) were used. Where insufficient data were present to perform a meta-analysis, authors were contacted by members of the review team to request additional data.
Heterogeneity was calculated by the I 2 statistic. Heterogeneity around baseline eosinophilia was explored by meta-regression owing to the possibility of variation in real patient populations.

| GRADE assessment
This study assessed the strength of outcome as per the GRADE guidelines. 16 As this study dealt with observational studies, all studies were initially given two of four points. This was adjusted by deducting points if there was risk of bias, inconsistency, publication bias, imprecision or indirectness. Points were added as per the quality assessment criteria, for example if large effects were seen or if there was evidence of a dose response. Based on these, we assigned outcomes to four categories of certainty based on the overall GRADE score for each comparison. Outcomes were categorized as high (at least 4 points), moderate (3 points), low (2 points) and very low certainty (1 point or less).

| Comparison to RCTs
Studies examined by a recent systematic review 20 were selected and data on the effect size for key measures: FEV 1 , exacerbation and control, amongst the active treatment group were extracted. Where required authors were contacted, and data were obtained. In cases where this was not possible data points were estimated from figures using Graph Grabber 2.0 (Quintessa).

| Search results
The search strategy retrieved 1392 records. After duplicates were excluded, 839 unique records were identified. A total of 117 records were appraized in full. A total of 51 records were found to relate to omalizumab and so were considered beyond the scope of this review.
Additionally, 45 further records were excluded due to differences in the population, study designs, outcome, dosage or route, inclusion of too few participants (<20 total participants), publication type or language.

| Characteristics of included studies
The main characteristics of the studies are included in Tables S1-S4. The real-world trials included 1512 participants who were followed for up to 2 years. All studies recruited adults. Mean ages for participants ranged between 52 and 60 years. All studies were conducted between 2018 and 2020. The percentage of women in the studies ranged between 78.6% and 34%. Of the studies that reported the number of smokers, the percentage ranged between 0% and 41%. Most studies were retrospective studies apart from three studies examining mepolizumab (Table S16). Approximately, 47.6% of the studies use the ERS/ATS definition of severe asthma, 9.5% of studies used the Global Initiative of Asthma guideline definition whilst in 42.9% of studies, severe asthma was physician diagnosed.

| Risk of bias of included studies and publication bias
Four studies were found to have a low risk of bias. 24,31,35,38 One study was found to have a high risk of bias. 36 The remaining studies were deemed to be at moderate risk of bias (Tables S5-S7). Studies were grouped according to biological agent and clinical outcome and assessed using funnel plots (Figures S17-S21). Publication bias was incorporated into the GRADE assessment (Tables S5-S7).

| Asthma control
Asthma control was evaluated in studies using the Asthma Control Test (ACT) and the Asthma Control Questionnaire (ACQ-6). An improvement was seen in asthma control with mepolizumab and benralizumab, with low and very low certainty of evidence, respectively. Eight studies assessed the effect of mepolizumab on ACT. 27 Significant heterogeneity was seen in the effect sizes for studies where participants were treated with both mepolizumab and benralizumab. The heterogeneity was explored and found to be unrelated to baseline eosinophilia in the case of mepolizumab (p = .544) but related in the case of benralizumab (p < .001) (Tables 1-3; Table S9).

| Changes in oral steroids
Studies analyzed the change in mean daily steroid doses following the treatment with mepolizumab, benralizumab and reslizumab.
Although a decrease was found with all three agents, the quality of evidence was found to be very low for mepolizumab and benralizumab and low for reslizumab. Five trials examined treatment with mepolizumab, 30 (Tables 1-3; Table S9).

| Change in eosinophils
Blood eosinophilia was assessed after treatment with three biologicals. Each demonstrated a decrease, but the quality of evidence was found to be very low in the case of mepolizumab, benralizumab and reslizumab. Eight trials examined mepolizumab, 27,28,31,32,33,34,35,41 five trials examined benralizumab 22,23,24,25,41 and two trials examined reslizumab. 39 The minimal clinically important difference (MCID) for FEV 1 is 0.2 L as per previous studies. 19 The MCID for ACT is 3 as per previous studies. 42 Pooled effects represent mean difference [95% confidence interval] from a random effects meta-analysis model after therapy with mepolizumab in real-world studies.
Abbreviations: ACT, Asthma Control Test (higher scores indicating better control); CI, confidence interval; FeNO, fractional exhaled nitric oxide; FEV1, forced expiratory volume in 1 s. Pooled effects represent mean difference [95% confidence interval] from a random effects meta-analysis model after therapy with Reslizumab in real-world studies.

| Comparison to RCT data
In this study, precise effect estimates were derived from real-world studies. These estimates were comparable to figures derived from the active treatment groups of RCT studies. The increase in FEV 1 observed in real-world studies examining mepolizumab of 170 ml was comparable to the range of effects observed in RCTs (range 111-183 ml). [43][44][45] In real-world studies, a 210 ml increase was seen in FEV 1 with benralizumab, and this was similar to the RCT results (range 239-330 ml) 46 Figure 2). It was not possible to make a direct comparison for asthma control between real-world studies and RCT data owing to differences in measurement methodology (Tables 1-3; Table S10).

| Effect of dupilumab
One study examined the effect of dupilumab on key outcomes. 40 Owing to the limited number of studies, no meta-analysis was possible, and results were tabulated ( Table 4).

| Summary of main findings
This systematic review provides evidence about the real-world effects of anti-IL5 biologicals in asthma. There was moderate certainty of evidence that three biologicals, mepolizumab, benralizumab and reslizumab, decrease the annualized rate of exacerbation and moderate evidence for a decrease in FeNO with mepolizumab. There was low certainty of evidence that the use of mepolizumab and benralizumab in a real-world environment was associated with an improvement in asthma control as measured by ACT. There was low certainty of evidence for improvement in FEV 1 with mepolizumab and benralizumab. However, the data suggest that benralizumab has no effect on FeNO. In this study, evidence of a decrease in blood eosinophilia for mepolizumab, Note: Evidence examined using the GRADE Criteria where High Confidence indicates that the true effect lies close to that of the estimate of the effect, Moderate Confidence indicates the true effect is likely to lie close to the estimate of the effect, but there is a possibility that it is substantially different, Low Confidence indicates the true effect may be substantially different from the estimate of the effect, Very Low Confidence indicates the true effect is likely to be substantially different from the estimate of effect.
The minimal clinically important difference (MCID) for FEV 1 is 0.2 L as per previous studies. 19 The MCID for ACT is 3 as per previous studies. 42 Pooled effects represent mean difference [95% confidence interval] from a random effects meta-analysis model after therapy with benralizumab in real-world studies.
Abbreviations: ACT, asthma control test (higher scores indicating better control); CI, confidence interval; FEV 1 , forced expiratory volume in one second; FeNO, fractional exhaled nitric oxide. benralizumab and reslizumab was observed but with very low certainty.
In this study, benralizumab and mepolizumab were demonstrated to have differing effects on clinical parameters including FENO, FEV 1 and exacerbations. These should be considered alongside the data from previous studies. 20,49 As in previous RCT studies, the data provided in this meta-analysis suggest that mepolizumab, benralizumab and reslizumab have the capacity to reduce the number of exacerbations. 43,44,46,50,51,52,53,54 Likewise, these have shown that FEV 1 is improved by treatment with biologicals. 43,44,46,50,51,52,54 However, unlike previous work, real-world trials appear to demonstrate that benralizumab has an effect on FEV 1 which is above the MCID whilst mepolizumab causes a statistically significant change which is below the MCID. 20 Interestingly, despite the effect on FEV 1 , in this study, benralizumab exerts no clear effect on FeNO. This effect has not been replicated in the context of systematic reviews of RCTs. 20 In patients with severe eosinophilic asthma with a FeNOhigh phenotype, the use of benralizumab has been associated with a fall in FeNO. 55 However, this does not occur in individuals with a FeNO-low phenotype. 56 Further studies are required to assess the different phenotypes in context of treatment with these monoclonals. This may reflect subtle differences in the mechanism of action F I G U R E 2 Comparison between meta-analysis of real-world studies and active group randomized controlled trial (RCT) data for changes with biological therapy. Active group only data extracted from biological RCTs in EACCI systematic reviews (black squares) is compared against the meta-analysis of biological RWS from this review (red squares). Markers placed at 1× and 2× the minimally clinically important difference (MCID). Studies clustered by therapeutic agent.  (Table S17). The methodological variation may have introduced a small amount of heterogeneity. These differences should be taken into consideration when forming conclusions. However, the data suggest that real-world studies can help to provide a more comprehensive understanding of biologics in real clinical practice.

| Areas for further development
In this review, one real-world study on dupilumab was identified precluding further efforts to perform a meta-analysis. Further assessment of this agent in real-world studies is suggested to provide precise effect estimates.

| Strengths and limitations
This systematic review and meta-analysis of data regarding recently licenced biologicals in a real-world environment provides some of the first estimates of the effect of newer biologicals in real clinic patients. A key strength of the data presented in this study is that there is consistency in effect across primary studies, and that many of the effects observed were significantly above the no-effect line. Importantly, these outcomes were some of the most widely used in clinical practice. FEV 1 , for example, has been linked to respiratory hospitalization, healthcare costs and mortality amongst asthmatics. [57][58][59] Nevertheless, there are some limitations to this systematic review which should be considered when interpreting the results.
One limitation was the heterogeneity found in the analysis of some outcomes. Although many studies had narrow confidence intervals, heterogeneity was identified between studies. However, there was a consistent direction of effect found amongst studies, and in most, the effects were far above the no-effect line. Some of the heterogeneity surrounding the effect size between studies was explained by the baseline level of blood eosinophilia. Differences in study populations and endotypes of asthma may have some effect, but there were insufficient analyses to explore this, which prevented us from presenting effect sizes for different endotypes. Methodological variation may have also contributed to the heterogeneity seen. One critical feature is that real-world studies reflect real clinical environments, and these environments may be different to one another. As such, heterogeneity is to be expected. Conversely, the limited data available for agents like reslizumab artificially reduced some of the observed heterogeneity. However, through synthesis, broad consensus can be obtained.
Many of the included studies are observational and as expected they are at higher risk of bias. This is reflected in the assessment of risk of bias for the included studies. Some of this risk of bias is derived from the methodology used. In this meta-analysis, most of the studies were retrospective. These studies are an efficient way of exploring the effects of biologicals on severe asthma. However, the presence of these studies has the potential to introduce bias into the analysis particularly via recall bias. This needs to be considered when

| Implications for practice and research
We demonstrate that effect estimates from real-world studies of benralizumab, mepolizumab and reslizumab are similar to those derived from RCTs confirming that these therapeutics are also effective in typical clinical patients with severe asthma. Effect estimates generated in this study can be used to optimize patient services. However, there are areas where further targeted work could be impactful. For instance, given that reducing OCS use is a major factor in the use of biologicals in clinical practice, gaining further insight into this parameter represents a priority for future studies.
To allow future meta-analysis to create more precise estimates, we suggest that studies report the following: (1)  The biologicals reviewed in this study showed a significant improvement in several key clinical parameters in a real-life environment.
We argue that to produce a comprehensive profile for biologicals, it is essential for data from real-world studies, RCTs and registries to be combined and analyzed. Additionally, we need more studies examining the impact of biologicals in different asthma endotypes. Furthermore, in both real-world studies and RCTs, there are very limited data for the use of these newer biologicals in the paediatric populations, and this remains as a key area for future research. 20

| CON CLUS ION
This meta-analysis shows that in real-world studies mepolizumab, benralizumab and reslizumab are effective treatments for asthma when looking at key clinical parameters. The effects observed in real-world trials are similar to those seen in the active group of equivalent RCTs. Further research is required to provide precise effect estimates for dupilumab in a real-world setting.

ACK N OWLED G EM ENTS
D.C was responsible for study design, execution, data gathering, analysis, drafting manuscripts and critical discussion. J.S. and S.N.T are joint second authors and were involved in data gathering, execution and analysis. S.N.T contributed to editing the manuscript and figure creation. A.R, J.S, S.N.T and E.K participated in critical discussion. G.R was responsible for study design, analysis methodology, drafting manuscripts and critical discussion. We thank authors of studies for responding to enquiries and providing data. We would also like to thank Paula Sands, Engagement Librarian, University of Southampton for assistance with search strategies.

CO N FLI C T O F I NTE R E S T
The authors know of no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.