Browse the corpus

Severe asthma affects approximately 5% of the global asthma population [], yet it imposes a disproportionate burden in terms of exacerbations requiring hospitalizations, impaired Quality of Life (QoL) [, ], and socioeconomic costs [, ]. Although most patients achieve adequate control with high‐dose inhaled corticosteroids (ICS) and additional controller therapies, a subset remains uncontrolled despite maximal pharmacotherapy []. For these patients, targeted biologics as add‐on treatments can reduce exacerbation frequency and oral corticosteroid (OCS) use while improving lung function and symptom control [].

Biologics herald an era of personalised medicine for severe asthma, but they are expensive [] and have associated side effects []. Pertinently, not all patients benefit equally, with non‐response rates ranging from 15% to 17% [, , ] and partial response rates from 43% to 69% [, ]. This can be attributed to factors such as heterogeneity at the pathophysiological and population level, and differences in treatment strategy []. Therefore, in lieu of a ‘one‐size‐fits all’ approach, it is necessary to identify the right biologic for the right patient i.e., prediction of response. Inflammatory biomarkers, such as raised fractional exhaled nitric oxide (FeNO) and high blood eosinophil counts (BEC) have been established as predictors of biologic response using trial registry data, and subsequently informed regulatory approval criteria and treatment guidelines [, , ]. However, these guidelines do not always align with real life, partly due to the homogenous populations included in the underpinning randomised controlled trials (RCTs). Additionally, composite predictors of biologic response have been explored [, ], including a tool to predict omalizumab response based on asthma control [], though this has not been widely implemented as it lacks physiological markers crucial for gauging treatment benefit.

The pursuit of biologic response predictors is complicated by the heterogeneity in definitions of response []. These range from measurement of improvement in individual outcomes (e.g., exacerbations) to composites including multiple outcomes, cutoffs, and disparate nomenclature (e.g., non‐response, response, super‐response) [, ]. Inconsistency in the timepoints of evaluation further hampers the quest for response predictors, with some studies investigating potential predictors using baseline data collected pre‐biologic, while others use data from an early assessment several months after initiating therapy. A meta‐analysis established that baseline BEC and total serum IgE levels can serve as reliable predictors of omalizumab response in patients with allergic asthma []. However, a comparable synthesis for other biologics is lacking. Therefore, this systematic review aimed to identify predictors of response to biologics (except omalizumab) for severe asthma.

A search strategy was developed on EMBASE (OVID) and subsequently adapted for the following databases: MEDLINE (OVID), CINAHL (EBSCOhost, Cumulative Index to Nursing and Allied Health Literature), and ISI Web of Science (Thomson Web of Knowledge) (see Data ). Databases were searched from 1990 to 10 January 2024. ClinicalTrials.gov ( CT.gov ) and International Trials Registry Platform (ICTRP) were also searched. The reference lists of identified studies, systematic reviews, reviews, and guidelines were searched for additional articles. Experts within the 3TR Respiratory Work Package were also consulted to identify additional relevant articles. The search strategy included terms for omalizumab as it was developed prior to the identification of the systematic review on predictors of omalizumab response []. As detailed in the subsequent section, records focused on omalizumab were excluded during the citation screening process.

References were pooled and de‐duplicated in Endnote version X9 (Thomson Reuters, Philadelphia, PA ), and subsequently uploaded to Rayyan [] ( rayyan.qcri.org ). Record titles, abstracts, and full texts were screened independently by two reviewers ( AR , PD ) according to the predefined eligibility criteria. Data about study design, sample size, participant characteristics, intervention and comparator (where appropriate), administration regime, predictor variable(s), response outcome(s) and associated effect estimates were extracted into a pilot‐tested form by three reviewers ( AR , PD , DC ). The final extraction was cross‐checked. Any disagreements were resolved through discussion or involvement of the third reviewer ( GR )

The methodological quality and certainty of evidence was evaluated in a two‐step process by two independent reviewers (AR, PD). Any discrepancies were resolved through discussion, or arbitration by the third reviewer (GR). First, the Risk of Bias (RoB) was assessed using the Clinical Appraisal Skills Programme (CASP) checklist for clinical prediction rules (CPR) [], which was modified to appraise studies investigating predictors of treatment response. Each study was assessed and assigned a rating of ‘yes’, ‘no’, or ‘unclear’ for selected domains. The RoB assessment was conducted by two independent reviewers (AR, PD). Any discrepancies were resolved through discussion, or arbitration by the third reviewer (GR). Second, the certainty of evidence was assessed using the modified Grading of Recommendations, Assessment, Development and Evaluation (GRADE) framework [, , ]. The quality of evidence was graded as ‘high’, ‘moderate’, ‘low’, or ‘very low’. The criteria specific to this review are provided in Table .

Results were summarized using narrative synthesis and descriptive tables. Predictor variables with heterogeneous definitions across included studies (e.g., different thresholds for high BEC) were grouped for the narrative summary but were downgraded during the appraisal as per the criteria in Table . Pooled data from selected studies were included in a meta‐analysis (MA) using a random‐effects model, conducted in the STATA statistical package (STATA/IC 16.0 for Windows). Only studies similar in terms of predictor and response definitions, intervention, and comparator were included. Eligible response definitions included the COMSA outcomes [, ] (exacerbations, FEV 1 , maintenance OCS use, asthma control, and QoL). All response definition types were considered for example, partial response, super‐response, clinical response, and non‐response. Discrepancies were addressed by grouping similar definitions together, such as combining studies reporting clinical response based on exacerbations and asthma control. The MA identified the biologic response predictive power of the investigated variables. Where insufficient data were available for the MA, study authors were contacted to request additional data. It was not possible to assess publication bias due to the limited number of eligible studies.

The study selection process is summarized in Figure . The search strategy yielded a total of 5853 records, with three additional records identified from the review of citations. Following the deletion of duplicate records, 4203 records were screened by title and abstracts and excluded if they did not meet the eligibility criteria. The full text of 508 records was assessed for eligibility; predominant reasons for exclusion were reporting of pooled data across mild‐to‐severe asthma, absence of data on predictors of biologic response, and lack of published data for clinical trial records. Ultimately, 13 studies were identified for mepolizumab [, , , , , , , , , , , ], 6 for benralizumab [, , , , , ], 3 for dupilumab [, , ], and 1 for tezepelumab []. The characteristics of included studies are presented in Table . Most studies were real‐world studies conducted in multinational settings. The sample sizes of the included studies ranged from 20 to 822 participants, with participant ages spanning from 28 to 82 years. None of the identified studies investigated response predictors for pediatric severe asthma, or biologics targeting the non‐T2 pathway.

The ability of inflammatory biomarkers to predict response was investigated by 17 studies, including BEC (11 studies) [, , , , , , , , , , ], FeNO (6 studies) [, , , , , ], BEC and FeNO combined (1 study) [], T2 Phenotype (1 study) [], Th17 cells (1 study) [], Group 3 innate lymphoid cells (ILC3) (1 study) [], periostin levels (1 study) [], peripheral neutrophils (1 study) [], and atopic status (4 studies) [, , , ]. Six studies assessed lung function parameters as predictors of response [, , , , , ]. 12 studies investigated clinical parameters for their predictive value, including exacerbation history (5 studies) [, , , ], OCS use/dose (10 studies) [, , , , , , , ], ICS dose (1 study) [], prior biologic use (1 study) [], number of respiratory medicines (1 study) [], asthma control (7 studies) [, , , , , , ], QoL (1 study) [], age of asthma onset (6 studies) [, , , , , ]. The presence of co‐morbidities as predictors of response was assessed in six studies, including allergic disease (1 study) [], chronic rhinosinusitis (1 study) [], non‐allergic rhinitis with eosinophilia syndrome (NARES) (1 study) [], Nasal Polyps (NP) (5 studies) [, , , , ], bronchiectasis (2 studies) [, ], Gastroesophageal Reflux Disease (GERD) (2 studies) [, ], obesity (1 study) [], osteoporosis (1 study) [], dysfunctional breathing (1 study) [], and mental health conditions (1 study) []. Eleven studies assessed socio‐demographic variables as predictors of response, including age (9 studies) [, , , , , , , , ], gender (6 studies) [, , , , , ], BMI (10 studies) [, , , , , , , , , ], smoking status (5 studies) [, , , , ] and geographic region (1 study) [].

A summary of the risk of bias assessment for included studies is provided in Table . While most studies clearly defined the predictors of biologic response under investigation, about half lacked a diverse sample with varied clinical and demographic characteristics, leading to concerns about external validity. The primary reasons for negative ratings were a single‐centre real‐world design, small sample sizes ( n < 30), or uneven gender distribution [, , , , , , , ]. None of the studies validated the response predictors identified in a training set of patients in a separate, validation patient set, limiting the generalizability of the findings to the broader severe asthma population. Only two studies [, ] evaluated the response predictors and outcomes in a blinded manner, putting the majority at risk of compromised internal validity. Blinding was not possible in 10 studies due to their retrospective real‐world [, , , , , , ] or post hoc design [, , ]. Although the original RCTs were conducted with blinding, post hoc analyses are particularly susceptible to publication bias, with significant results more likely to be reported. Notably, seven studies reported imprecise estimates of treatment effect, as indicated by wide confidence intervals [, , , , , , ], which limits the conclusiveness of their findings. This assessment did not consider reporting of p ‐values due to their dependence on sample size.

The available evidence for inflammatory markers as predictors of biologic response is summarized in Table (detailed in Table ). For mepolizumab, ‘moderate’ quality evidence suggested that high BEC levels (≥ 300 and ≥ 500 cells/μL) are associated with greater improvement in asthma control defined by ACQ‐5 score [, ], compared to lower BEC levels (≥ 150 cells/μL), but not with improvement in the SGRQ score []. Similarly, ‘moderate’ quality indicated that high peak BEC (0.70 cells × 10 9 /L) in the year prior to benralizumab treatment is associated with a higher likelihood of super‐response [], though evidence for high BEC levels predicting clinical response was inconsistent [, ]. The evidence was downgraded primarily due to heterogeneity in the measurement of predictor variables and response definitions across studies. For dupilumab and tezepelumab, ‘high’ quality evidence showed that high BEC levels (≥ 150 and ≥ 300 cells/μL) are associated with significantly greater reductions in exacerbations [, ], improvement in lung function [, , ] and a higher likelihood of OCS dose reduction [, ] compared to lower BEC levels (< 150 and < 300 cells/μL). Despite the limited number of studies, the evidence is based on large RCTs, lending greater confidence to the findings.

Some studies suggested that high FeNO levels (> 40 ppb) were associated with clinical response [] and super‐response [] to benralizumab, while others reported no significant predictive value []. The meta‐analysis revealed that elevated FeNO levels in the range of 40–44 ppb have good discriminatory ability for identifying clinical responders versus non‐responders to benralizumab (Figure ). In contrast, evidence for FeNO as a predictor of mepolizumab response was limited and of ‘very low’ quality; a small observational study suggested that elevated FeNO levels were not associated with clinical non‐response but did not provide comparisons between different cut‐off values []. FeNO levels between 25 and 50 ppb were associated with a greater reduction in exacerbation risk and OCS dose following dupilumab, compared to both higher (> 50 ppb) and lower (< 25 ppb) levels []. FeNO < 25 ppb was associated with lung function improvement, though post‐BD FEV 1 improvements did not reach the Minimal Important Difference (MID) of 0.2L [, ]. A post hoc analysis of a large RCT showed that combining high BEC (≥ 150 cells/μL) with low FeNO levels (< 25 ppb) predicts clinically significant improvements in pre‐ and post‐BD FEV 1 , surpassing the MID []. However, the evidence quality was downgraded to ‘moderate’ due to imprecision indicated by wide confidence intervals.

There was predominantly ‘low’ quality evidence that biomarkers such as high periostin levels, peripheral neutrophils [], Th17 cells, and ILC3 [] are associated with anti‐IL5/5Rα response. This evidence was downgraded due to imprecision, arising from the small sample size of the included observational studies.

Lung function and airflow obstruction parameters (such as FEV 1 < 60% and FEV 1 /FVC < LLN) were not significantly associated with a reduction in exacerbations or clinical (non‐) response following mepolizumab [, ] and benralizumab treatment [, , ] (Table ; detailed synthesis in Table ). The quality of evidence was rated ‘very low’ due to a serious risk of bias, stemming from heterogeneity in response definitions and the lack of comparison between different thresholds for the investigated lung function parameters. The small sample sizes of the included real‐world studies and imprecise effect estimates further reduced the certainty of evidence.

The quality of evidence for the biologic response prediction value of clinical parameters is presented in (Table ; detailed synthesis in Table ). For Mepolizumab, a high baseline exacerbation rate was a significant predictor of super‐response [] but the evidence quality was rated ‘low’ due to risk of bias and imprecision. There was ‘high’ quality evidence that lack of OCS use or low OCS dose (< 10 mg/day) at baseline is associated with a reduced risk of exacerbations [], while baseline OCS use reduced the likelihood of achieving OCS‐free status []. For dupilumab, a low OCS dose (< 10 mg/day) was significantly associated with a greater reduction in exacerbation rates, greater improvement in pre‐BD FEV 1 , and a higher likelihood of achieving steroid‐free status compared to high OCS dose (≥ 10 mg/day) []. This evidence was upgraded to ‘high’ quality due to the clinical significance of the effect estimates and a clear dose–response gradient. Biologic naïve patients were more likely to achieve benralizumab response [], with the evidence upgraded to ‘high’ certainty due to the large magnitude of effect. Adult‐onset asthma was associated with greater asthma control improvement [] and an increased likelihood of achieving OCS‐free status [] following mepolizumab, though evidence for the latter was rated ‘very low’ due to lack of comparative data on age cut‐offs. On the other hand, there was ‘moderate’ quality evidence that childhood‐onset asthma is associated with benralizumab response [].

Better asthma control at baseline was associated with a higher likelihood of achieving OCS‐free status [], as well as clinical response and super‐response [, ] following mepolizumab treatment, with ‘high’ and ‘moderate’ quality evidence, respectively. The forest plot in Figure shows that better asthma control at baseline, defined as a low Asthma Control Questionnaire score, may be able to distinguish responders versus non‐responders to mepolizumab. However, the high heterogeneity between the included studies suggests that this relationship may not be consistent across populations. While no association was found between baseline asthma control and response to benralizumab, a post hoc analysis of a large real‐world study showed that better asthma control at 3 months can predict long‐term clinical response [].

The quality of evidence for co‐morbidities as predictors of response to biologics for severe asthma is summarised in Table (detailed in Table ). NP was significantly associated with clinical response and super‐response to mepolizumab [], but the evidence quality was rated ‘very low’ owing to heterogeneity in response definitions and the small sample size of the included studies. NP in patients treated with benralizumab was associated with a reduced likelihood of response, but the ‘low’ quality evidence was based on a small real‐world cohort []. Evidence for the predictive value of other co‐morbidities, including allergy [], chronic rhinosinusitis [], NARES [], Bronchiectasis [, ], GERD [, ], obesity [], osteoporosis [], dysfunctional breathing, and mental health disorders [], was inconclusive and of ‘low’ to ‘very low’ quality. This was primarily due to the lack of comparative data for patients with and without these co‐morbidities, as well as imprecise effect estimates indicated by wide confidence intervals.

A summary of the quality of evidence for socio‐demographic characteristics as predictors of biologic response is presented in Table (detailed synthesis in Table ). There was ‘moderate’ to ‘low’ quality evidence that a higher BMI (> 30 kg/m 2 ) is associated with a higher likelihood of experiencing more exacerbations [], reduced asthma control improvement [], and a lower likelihood of achieving steroid‐free status [] following mepolizumab treatment. Regarding gender, a single large observational study provided ‘low’ quality evidence that male participants are less likely to benefit from mepolizumab in terms of asthma control improvement []. Other investigated variables, such as age and smoking status, were not associated with anti‐IL5/5Rα response, with several studies lacking comparative data on clearly defined age cut‐offs [, , , ] and smokers versus non‐smokers [, ].

This systematic review is the first comprehensive synthesis of evidence on predictors of response to anti‐IL5/5Rα, −4Rα, and anti‐TSLP therapies for severe asthma. We found ‘moderate’ to ‘high’ quality evidence that high BEC (≥ 300 cells/μL) reliably predicts response to mepolizumab, dupilumab, and tezepelumab. Elevated FeNO levels (> 40 ppb) predicted benralizumab response, while FeNO levels between 25 and 50 ppb were linked to reduced exacerbations and OCS dose after dupilumab treatment. ‘High’ quality evidence showed that being OCS naïve or lower baseline OCS dose (< 10 mg/day) is associated with better outcomes following mepolizumab and dupilumab. Additionally, better asthma control at baseline and post 3 months were significant predictors of mepolizumab and benralizumab response, respectively. However, evidence for other variables was limited, downgraded primarily due to the heterogeneous measurement and reporting of response definitions. This highlights the urgent need for standardized, universally accepted definitions of response in studies of targeted therapies for severe asthma. Additionally, no evidence was found for biologics targeting the non‐T2 pathway or the pediatric severe asthma population.

Our finding that elevated BEC levels can predict response to anti‐IL‐5/5R, anti‐IL‐4, and anti‐TSLP therapies aligns with GINA guidance, which recommends using raised BEC levels for biologic initiation assessment []. Several large RCTs and post hoc analyses that informed the regulatory approval of T2 biologics [, , , , , , ] and reported on the response predictive value of raised BEC levels [, ] were not eligible for this review due to their inclusion of patients with moderate‐to‐severe asthma or lack of subgroup analysis. Consequently, it was not possible to conduct a meta‐analysis to quantitatively delineate the BEC cut‐offs for predicting response in the true severe asthma population. Nonetheless, the real‐world studies comprising most of our evidence base represented a heterogeneous clinic population, where treatment response can be influenced by unique non‐pathophysiologic factors that are typically absent in RCTs. However, our conclusions were limited by the small sample size of the eligible studies, highlighting the need for larger real‐world studies to better understand the traits influencing biologic response in the severe asthma subgroup [, ].

The practical application of biologic response predictors requires careful consideration of feasibility. This review found that raised FeNO levels (> 40 ppb) can identify responders to benralizumab, and levels between 25 and 50 ppb are linked to greater benefit with dupilumab. However, FeNO remains less commonly used in practice compared to biomarkers like BEC, primarily due to the ease of measuring BEC, while FeNO testing may be limited by equipment availability in certain clinical settings. Feasibility becomes even more critical when considering exploratory predictors, such as genetic polymorphisms in the IL‐4, 5, and 13 pathways []. While genetic testing holds potential for biologic response prediction, it introduces challenges related to cost, integration into clinical workflows, and real‐time data access to guide treatment decisions. Addressing these feasibility concerns is key to ensuring the rapid adoption of emerging treatment response predictors into practice.

The ‘high’ quality evidence supporting OCS naïveté and low OCS dose as predictors of mepolizumab and dupilumab response aligns with the literature, which shows that OCS use is associated with poor response to anti‐IL5 [], while OCS naïveté or low baseline OCS dose predicts remission []. However, the predominantly ‘low’ quality evidence for the response predictive value of other variables across clinical, lung function, sociodemographic, and co‐morbidity domains contrasts with the wider literature. For example, although our review found limited ‘very low’ quality evidence for NP as a predictor of mepolizumab response, post hoc analyses [] suggest that NP is associated with greater reduction in exacerbations, potentially serving as a surrogate T2 marker. This disparity stemmed from studies such as the latter recruiting moderate‐to‐severe asthma patients, with those included in our review downgraded due to inconsistent definitions and measurements of predictors (e.g., patient‐reported vs. clinically diagnosed NP) and lack of comparison data for patients with versus without the predictor (e.g., effect size reported for NP group only), or above versus below predictor thresholds. Standardizing the definition, measurement, and reporting of predictor variables in future studies is critical to enable data pooling for meta‐analysis and drawing more definitive conclusions.

Heterogeneity in response definitions, particularly among studies using clinical response criteria, was a major factor in downgrading the quality of evidence and hindering data pooling for meta‐analysis. Response criteria varied across studies in terms of measurement (e.g., exacerbation reduction measured by hospitalisations vs. OCS bursts), definition (e.g., > 50% vs. 80% exacerbation reduction), and outcome combinations (e.g., clinical, inflammatory, and patient‐reported vs. clinical plus patient‐reported only). This inconsistency led to cases where predictors of response in one study did not hold true in others with differing definitions. Additionally, bias was introduced when the predictor variable was part of a clinical response definition, thus increasing the likelihood of meeting that definition. There was also variability in the directionality of effect, with some studies assessing a variable's ability to predict response, while others focused on its role in predicting non‐response. To address these issues, there is a clear need to implement a universal response definition, such as CONFIRM [], and further research to unravel the relationships between predictors and biologic response, alongside the development of multivariate prediction tools [].

We found limited evidence for the response predictive value of variables such as periostin, yet T2 biomarkers are often used in practice to define eligibility for biological therapy. This underscores the need for robust studies to identify reliable predictors beyond traditional markers that influence individual patient response. The 3TR Asthma Biologics Cohort study [] is addressing this gap by using advanced multi‐omics methods and repeated bio‐sampling from multiple compartments to identify biomarkers that can predict (non‐) response. To facilitate a step change, funding bodies such as the EU/IHI and NIH should prioritize and support more controlled interventional trials focused on emerging biomarkers.

Although six studies in this review indicated eligibility for patients with severe asthma aged ≥ 12 years [, , , , , ], most participants were adults. A previous review [] reported on clinically available biomarkers and their ability to predict T2 biologic response in children, but it primarily focused on moderate‐to‐severe asthma, highlighting a gap in research specifically targeting the paediatric severe asthma population. Additionally, our review identified a lack of data for non‐T2 populations. Given the observed benefits of anti‐TSLP therapy in patients with non‐T2 characteristics [, , , ], further investigation into response predictors for the non‐atopic population is urgently needed. This study excluded predictors of omalizumab response to prevent repetition, as a comprehensive review has already been conducted and published on this []. That analysis identified baseline BEC and total serum IgE levels as reliable predictors of omalizumab response in patients with allergic asthma [].