Skip to main content
NCBI home page
ERJ Open Research logoLink to ERJ Open Research
. 2025 Aug 26;11(4):00969-2024. doi: 10.1183/23120541.00969-2024

Biomarkers of type 2 inflammation as predictors of response to biologics for severe eosinophilic asthma

Duong Duc Pham 1, Ji-Hyang Lee 1, Hyouk-Soo Kwon 1, Woo-Jung Song 1, You Sook Cho 1, Hyunkyoung Kim 1, Jae-Woo Kwon 2, So-Young Park 3, Sujeong Kim 4, Gyu Young Hur 5, Byung Keun Kim 6, Young-Hee Nam 7, Min-Suk Yang 8, Mi-Yeong Kim 9, Sae-Hoon Kim 10, Byung-Jae Lee 11, Taehoon Lee 12, So Young Park 13, Min-Hye Kim 14, Young-Joo Cho 15, ChanSun Park 16, Jae-Woo Jung 17, Han Ki Park 18, Joo-Hee Kim 19, Ji-Yong Moon 20, Pankaj Bhavsar 21, Ian M Adcock 21, Kian Fan Chung 21, Tae-Bum Kim 1,✉
PMCID: PMC12378948  PMID: 40874166

Abstract

Background

The relationship between pre-treatment levels of blood eosinophil count (BEC), fractional exhaled nitric oxide (FENO) and sputum eosinophils (Sp-EOS) and treatment response to monoclonal antibodies (mAbs) in severe eosinophilic asthma (SEA) remains unclear. We evaluated pre-treatment levels of BEC, FENO, Sp-EOS and their combinations as predictors of treatment responses in patients with SEA undergoing anti-interleukin (IL)-5/IL-5Rα or anti-IL-4Rα antibody therapies.

Methods

The study included 153 adult patients with SEA (59 anti-IL-5/IL-5Rα and 94 anti-IL-4Rα users). Logistic regression models were used to evaluate the association between predictors and 12-month treatment responses and clinical remission across four domains: exacerbation rate, maintenance of oral corticosteroid dose, forced expiratory volume in 1 s (FEV1) and asthma control test (ACT) improvement.

Results

Pre-treatment BEC and Sp-EOS were not associated with treatment responses in either mAb group. For combined data from anti-IL-5/IL-5Rα and anti-IL-4Rα users, the adjusted odds ratios (95% confidence intervals) for a 1-unit increase in log-transformed FENO were 1.8 (1.21–2.74) for FEV1 response and 2.15 (1.29–3.75) for ACT response. For anti-IL-4Rα users, these values were 2.34 (1.39–4.17) and 3.6 (1.73–8.84), respectively. No significant association between FENO and treatment response was found among anti-IL-5/IL-5Rα users. Additionally, no associations were observed between BEC, Sp-EOS or FENO and clinical remission across mAb categories. Combining biomarkers did not significantly enhance predictive ability.

Conclusion

In patients with SEA treated with anti-IL-4Rα antibodies, pre-treatment FENO may be a good predictor for certain treatment response domains.

Shareable abstract

Pre-treatment BEC is not associated with response to anti-IL-5/IL-5Rα or anti-IL-4Rα in SEA patients. FENO effects vary by outcome domain and biological therapy. Combining these biomarkers does not improve prediction. https://bit.ly/3GG9kaT

Introduction

Severe asthma (SA) is a chronic inflammatory condition of the airways that requires the use of high-dose inhaled (ICSs) or oral corticosteroids (OCSs) for maintenance therapy. However, despite such treatment, patients continue to experience frequent recurrent exacerbations and have a decline in respiratory function [1]. This complex pathological condition involves numerous interconnected factors and diverse individual responses to various exposures and treatments [2]. Recent studies have focused on specific cytokines in the type 2 (T2) inflammatory pathway, identifying a T2-high phenotype and developing monoclonal antibodies (mAbs) that specifically inhibit these cytokines [3, 4]. mAbs targeting downstream cytokines, such as interleukin (IL)-5, IL-4 and IL-13, compared to placebo decrease exacerbation rates, reduce maintenance oral corticosteroid (mOCS) use and improve lung function in patients with severe eosinophilic asthma (SEA) [5–8]. However, individual responses to these agents vary widely within the population. Therefore, identifying the predictors that can forecast the therapeutic response to such treatment is essential for personalising therapy for SEA patients.

The variability in response to biological treatments may stem from inconsistencies in defining the concept of “response”. Reduction in exacerbation frequency is considered the most critical outcome in treating SEA [1]. However, recent therapeutic goals for SEA have used a more multidimensional approach due to the heterogeneous nature of asthma. The proposed recommendations on the definition of “response” focus on four domains: reduction in exacerbation frequency, decrease in long-term OCS use, improvement in asthma control and increase in forced expiratory volume in 1 s (FEV1) [9–11]. The responder rate varies greatly depending on the domains included in the criteria, and many individuals do not respond in one domain but show a significant improvement in another domain [10, 12]. Recently, there has been emerging interest in combining these response factors to achieve long-term clinical remission and complete disease elimination. A consensus framework defined clinical remission as requiring at least 12 months of assessment, with no exacerbations or mOCS, along with optimised/stabilised lung function, and the absence of significant symptoms [13]. Furthermore, estimates of the effect size of the relationship between predictors and outcomes in randomised controlled trials (RCTs) and real-world studies differ, especially in the context of the heterogeneous nature of asthma [14].

Patient characteristics before mAb administration are used to identify potential predictors of response [15]. The post hoc analyses from RCTs revealed that pre-treatment blood eosinophil count (BEC) levels in patients with SA were associated with the extent of exacerbation reduction among anti-IL-5/IL-5Rα or anti-IL-4Rα users [16–19]. Meanwhile, high baseline fractional exhaled nitric oxide (FENO) levels may be independently associated with a better response in patients treated with anti-IL-4Rα [20]. High BEC and FENO levels before mAbs administration have been recommended for optimising treatment response [1]. However, the relationship between baseline BEC and FENO levels and response outcomes is mostly based on secondary analyses from RCTs. Additionally, whether the relationship between pre-treatment T2 inflammation biomarkers and clinical treatment responses persists remains unclear for SEA patients eligible for biological add-on therapy (i.e., those with high pre-treatment T2 inflammation biomarkers). BEC and FENO are indirect markers of eosinophilic airway inflammation and can be influenced by other pathological factors [21, 22]. The induced sputum eosinophil (Sp-EOS) level is considered a direct proxy for eosinophilic airway inflammation [23], but whether combining this biomarker with currently available biomarkers can improve the prediction of response to mAb treatment is unknown.

In this study, we investigated the predictive power of three main T2 inflammatory biomarkers (BEC, Sp-EOS and FENO) in forecasting the various domains of treatment responses and clinical remission to either anti-IL-5/IL-5Rα or anti-IL-4Rα mAbs in SEA patients.

Methods

Participants

The Precision Medicine Intervention in Severe Asthma (PRISM) study is a nationwide ongoing multicentre (23 centres) prospective cohort study since May 2020 [24]. Adults with SA were enrolled based on diagnoses by respiratory experts as either “controlled” with high-dose continuous ICS or a second anti-asthma drug, or “uncontrolled” despite active treatment. Patients who had used any biologics for asthma treatment in the past 3 months, had prolonged OCS use in the past 4 weeks (exceeding a total of 90 mg prednisolone), or had hypereosinophilic syndrome, allergic bronchopulmonary aspergillosis, eosinophilic granulomatosis with polyangiitis or any other severe respiratory diseases were excluded. Biologics were indicated for SEA patients according to the Global Initiative for Asthma recommendations [1]. In this analysis, we evaluated patients treated with anti-IL-5/IL-5Rα (mepolizumab, reslizumab or benralizumab) or anti-IL-4Rα (dupilumab), who completed a 12-month follow-up. The criteria for administration, dosage and usage protocol of biologics are presented in supplementary table 1S. In brief, mepolizumab was administered to patients with a baseline BEC of ≥150 cells·μL−1 or ≥300 cells·μL−1 in at least one test within the previous 12 months. Reslizumab was indicated for patients with a BEC >400 cells·μL−1, and benralizumab for patients with a baseline BEC >150 cells·μL−1. Dupilumab was indicated for patients with a baseline BEC >150 cells·μL−1 or FENO >25 ppb. A total of 153 SEA patients met the inclusion criteria, comprising 59 anti-IL-5/IL-5Rα users (22 patients on mepolizumab, 28 patients on reslizumab and nine patients on benralizumab) and 94 anti-IL-4Rα users. The number of patients per biologics group at each hospital is listed in supplementary table 2S. The Institutional Review Board at each study site approved the study protocol and registered it on ClinicalTrials.gov (registration number NCT05164939) [25]. All participants provided informed written consent.

Measurements

Baseline characteristics collected before treatment included BEC, Sp-EOS, FENO, asthma control test (ACT) score, pre-bronchodilator FEV1, age, sex, age at diagnosis, body mass index (BMI), number of exacerbations in the previous year, ICS use (in budesonide equivalent dose; Bude_eq), OCS use (in prednisolone equivalent dose; Pred_eq) at baseline and history of comorbidities, such as allergic rhinitis, gastroesophageal reflux and nasal polyps. BEC values, measured in cells per microlitre, were derived from standard complete blood cell counts. FENO measurements were performed using an airway inflammation monitor (Niox Vero, SE-754, Uppsala, Sweden), adhering to the American Thoracic Society/European Respiratory Society guidelines and technical standards. Sp-EOS percentages were obtained from induced sputum samples collected after nebulised hypertonic saline inhalation. Supplementary table 3S shows detailed instructions for calculating Bude_eq and Pred_eq.

Definition of outcome

The definition of “responder” in this analysis was based on four consensus domains: reduction in exacerbation rate, reduction in mOCS, improvement in FEV1 and improvement in the ACT score [10]. An “EXA-OCS-Res” referred to patients who, after 12 months of treatment, met one of the following conditions: 1) a ≥50% reduction in the annual exacerbation rate; 2) a ≥50% reduction in mOCS use; or 3) no history of exacerbations in the year before treatment, maintaining this status throughout the 12-month follow-up. An “FEV1-Res” was defined as a patient who, after 12 months of treatment, had a predicted pre-bronchodilator FEV1 of ≥80%, or that increased by at least 100 mL and by ≥10% at the 12-month visit compared with pre-treatment. An “ACT-Res” was defined as a patient who, after 12 months of treatment, had an ACT score of ≥20, or that increased by at least 3 points at the 12-month visit compared with pre-treatment. Clinical remission (Clin-Rem) was defined as patients who had no exacerbations and did not use mOCS throughout the follow-up period, while also fulfilling the criteria for FEV1-Res and ACT-Res at the 12-month checkup.

Statistical analysis

Statistical analyses were conducted using R version 4.3.1, with statistical significance set at p<0.05. Descriptive analysis compared the baseline patients’ characteristics among biologics groups and between responders and non-responders in each domain. T-tests and rank sum tests were employed for normally and non-normally distributed continuous variables, respectively. For categorical variables, Fisher's exact tests were applied. Logistic regression examined the association between predictors and their combination with the outcome. Seven models were performed for each response domain with predictors as (M1) BEC, (M2) Sp-EOS, (M3) FENO, (M4) BEC+Sp-EOS, (M5) BEC+FENO, (M6) Sp-EOS+FENO and (M7) all three biomarkers. The models were adjusted for covariates including age and sex. The predictors (BEC, Sp-EOS and FENO) were log-transformed owing to their skewed distribution. The area under the curve (AUC) of the receiver operating characteristic (ROC) was calculated for each model and compared with the AUC of other models using the DeLong test [26]. The sensitivity–specificity equality criterion (SSEC) was used to estimate the optimal cut-off and corresponding sensitivity and specificity for baseline BEC, Sp-EOS and FENO in predicting each domain outcome [27]. The same analyses were performed separately on overall data, anti-IL-5/IL-5Rα users and anti-IL-4Rα users. Because the “responder” and “remission” labels were quite skewed, logistic regression models were more likely to favour the majority labels. To address this issue, we applied the random over-sampling examples (ROSE) technique to adjust the data with the ratio of outcomes as 1:1 to avoid overfitting [28]. The analysis of the data after applying the ROSE technique was reported as a sensitivity analysis.

Results

Patients

At baseline, patients had a mean age of 54 years and a high mean BMI (25 kg·m−2), with nearly 90% having a BEC level >150 cells·μL−1, 75% having an FEV1 <80%, and 75% having an ACT score <20 points. Anti-IL-5/IL-5Rα and anti-IL-4Rα users did not differ regarding demographics, lung function, history of exacerbations and ICS/OCS doses used. Anti-IL-5/IL-5Rα users had a higher baseline BEC, with a notably higher proportion of BEC >300 cells·μL−1 (88% versus 64%, p=0.001). They also had higher baseline Sp-EOS and FENO compared to anti-IL-4Rα users (all p<0.05; table 1).

TABLE 1.

Baseline characteristics of anti-interleukin (IL)-5/IL-5Rα and anti-IL-4Rα users

Variable Total n anti-IL-5/IL-5Rα anti-IL-4Rα p-value
Male, n (%)+ 66 (43.1) 59/94 20 (33.9) 46 (48.9) 0.093
Age, y# 53.8±10.3 59/95 52.4±10.6 54.8±10.1 0.177
Agediag, y# 42±13.3 57/89 43.0±18.0 42.0±20.0 0.756
BMI, kg·m−2# 24.9±3.8 59/94 24.3±3.6 25.3±3.9 0.098
FEV1, %# 63.9±19.6 59/94 63.1±19.2 64.4±19.9 0.670
FEV1u80, n (%)+ 116 (75.8) 59/94 47 (79.7) 69 (73.4) 0.441
BEC, cells·μL−1 538 (581.5) 59/94 674.0 (645.5) 428.5 (379.0) <0.001
BEC150, n (%)+ 137 (89.5) 59/94 56 (94.9) 81 (86.2) 0.107
BEC300, n (%)+ 112 (73.2) 59/94 52 (88.1) 60 (63.8) 0.001
Sp-EOS, % 7 (27) 47/78 15.0 (36.0) 4.5 (21.8) 0.010
Sp-EOS2%, n (%)+ 80 (64) 47/78 35 (74.5) 45 (57.7) 0.083
FENO, ppb 53.5 (60.2) 58/94 62.0 (69.0) 45.5 (70.2) 0.038
FENOa25, n (%)+ 121 (79.6) 58/94 52 (89.7) 69 (73.4) 0.022
ACT, points# 16.1±4.9 59/94 15.3±5.2 16.6±4.6 0.141
ACTu20, n (%)+ 114 (74.5) 59/94 46 (78.0) 68 (72.3) 0.568
EXAhis, events 1 (10.1) 59/94 2.0 (5.0) 0.0 (3.8) 0.321
Bude_eq, μg·day−1 640 (465.3) 59/94 640.0 (400.0) 640.0 (306.0) 0.244
Pred_eq, mg·day−1 0 (5.6) 59/94 0.00 (2.50) 0.00 (5.00) 0.069
AR, n (%)+ 113 (73.9) 59/94 44 (74.6) 69 (73.4) 0.999
GR, n (%)+ 57 (37.3) 59/94 15 (25.4) 42 (44.7) 0.017
NP, n (%)+ 38 (24.8) 59/94 14 (23.7) 24 (25.5) 0.850
Open in a new tab

Data are presented as mean±sd or median (interquartile range) unless indicated otherwise. n: the number of patients among anti-IL-5/IL-5Rα and anti-IL-4Rα users, respectively, who had data for the corresponding variable. Agediag: age at asthma diagnosis; BMI: body mass index; FEV1: predicted pre-bronchodilator forced expiratory volume in 1 s; FEV1u80: patients with baseline FEV1 <80%; BEC: blood eosinophil count; BEC150 and BEC300: patients with baseline BEC ≥150 and ≥300 cells·μL−1, respectively; Sp-EOS: sputum eosinophils as a percentage; Sp-EOS2%: patients with baseline Sp-EOS ≥2%; FENO: fractional exhaled nitric oxide; FENOa25: patients with baseline FENO >25 ppb; ACT: asthma control test score; ACTu20: patients with baseline ACT <20%; EXAhis: recall exacerbation number during the previous year; Bude_eq and Pred_eq: budesonide and prednisolone equivalent doses, respectively; AR: allergic rhinitis; GR: gastroesophageal reflux; NP: nasal polyp. p-values were determined using: #: ANOVA for continuous variables; : rank sum test for non-parametric variables; +: Fisher's exact test for categorical variables. p-values <0.05 are indicated in bold.

Rate and baseline characteristics of responders

Figure 1 illustrates the response rate by each response domain and the remission rate. After 12 months of treatment, >85% of patients met the criteria for EXA-OCS-Res and ACT-Res, and >75% fulfilled the criteria for FEV1-Res. Only ∼40% of them achieved clinical remission according to the combined criteria of all four domains. Across all response domains and clinical remission, the response and clinical remission rates in the anti-IL-5/IL-5Rα group were slightly higher compared with the anti-IL-4Rα group.

FIGURE 1.

FIGURE 1

Open in a new tab

Responder rate. Bars represent the responder rate in the overall data, anti-interleukin (IL)-5/IL-5Rα users, and anti-IL-4Rα users for each response domain and remission. EXA-OCS-Res: responder is defined as a reduction by ≥50% in the annual exacerbation rate, or a reduction by ≥50% in the maintenance oral corticosteroid (mOCS) dose, or remaining without exacerbations; FEV1-Res: FEV1 responder is defined as having a 12-month visit FEV1 ≥80%, or an FEV1 increase by ≥100 mL and ≥10%; ACT-Res: ACT responder is defined as having a 12-month visit ACT score ≥20, or an ACT score increase by ≥3 points; Clin-Rem: clinical remission is defined as having no mOCS use and no exacerbations during the 12-month follow-up, plus meeting the criteria for FEV1-Res and ACT-Res.

At baseline, EXA-OCS responders had a higher proportion of females (both in the overall dataset and among anti-IL-5/IL-5Rα users) and experienced more exacerbation events in the previous year (in the overall dataset and among anti-IL-4Rα users) compared with non-responders (table 2, supplementary tables 4S and 5S). For FEV1-Res and ACT-Res, responders had higher FEV1 values and elevated BEC levels (both absolute counts and the proportion of patients with >300 cells·µL−1, all p<0.05), as well as higher FENO levels (both absolute values and the proportion of patients with FENO >25 ppb, all p<0.05) in the overall dataset and among anti-IL-4Rα users. Among anti-IL-5/IL-5Rα users, no group differences were found for FEV1-Res; however, for ACT-Res, responders had higher Sp-EOS levels (table 2, supplementary tables 4S and 5S). Patients who met the criteria for Clin-Rem had elevated BEC levels (both absolute counts and the proportion of patients with >300 cells·µL−1, all p<0.05) in the overall dataset and among anti-IL-4Rα users, but not among anti-IL-5/IL-5Rα users. These patients also required higher doses of mOCS (in the overall dataset and among anti-IL-5/IL-5Rα users). Additionally, anti-IL-4Rα users who achieved Clin-Rem after 12 months of treatment were younger and had higher baseline FEV1 and FENO levels (table 2, supplementary tables 4S and 5S).

TABLE 2.

Baseline characteristics stratified by response groups for each outcome domain on overall data

Variable EXA-OCS Res FEV1-Res ACT-Res Clin-Rem
Non Responder Non Responder Non Responder Non Remission
Male, n (%)+ 11 (68.8) 55 (40.1)* 20 (57.1) 46 (40.4) 10 (47.6) 56 (42.7) 45 (48.4) 21 (35.6)
Age, y# 54.4±11.4 53.8±10.2 55.5±8.2 53.6±10.9 58.8±10.9 53.0±10.0* 54.4±10.1 53.0±10.7
Agediag, y# 42.5±24.5 42.0±18.0 39.0±18.0 42.0±21.0 50.0±21.0 42.0±20.0 40.5±19.0 44.0±20.0
BMI, kg·m−2# 26.9±5.3 24.7±3.5 25.1±4.5 24.8±3.6 24.0±3.2 25.1±3.9 25.4±4.1 24.2±3.3
FEV1, %# 60.8±23.9 64.3±19.1 56.2±17.8 66.4±19.4* 54.5±25.0 65.4±18.3 62.9±18.9 65.2±20.7
BEC, cells·μL−1¶ 533.0 (326.0) 538.0 (450.0) 338.0 (484.5) 550.0 (442.5)* 295.0 (304.0) 550.0 (430.0)* 480.0 (488.0) 567.0 (377.0)*
BEC150, n (%)+ 15 (93.8) 122 (89.1) 26 (74.3) 107 (93.9)* 17 (81.0) 119 (90.8) 81 (87.1) 55 (93.2)
BEC300, n (%)+ 12 (75.0) 100 (73.0) 20 (57.1) 88 (77.2)* 10 (47.6) 101 (77.1)* 62 (66.7) 49 (83.1)*
Sp-EOS, % 4.5 (13.8) 7.0 (32.0) 3.0 (21.5) 7.5 (32.0) 3.5 (34.5) 7.5 (31.2) 5.5 (25.2) 11.0 (34.5)
Sp-EOS2%, n (%)+ 9 (64.3) 71 (64.0) 16 (59.3) 63 (65.6) 9 (56.2) 70 (64.8) 47 (63.5) 33 (66.0)
FENO, ppb 42.0 (31.0) 56.0 (69.5) 35.0 (72.5) 56.0 (61.5)* 36.5 (35.8) 56.0 (70.5)* 45.5 (71.2) 58.0 (58.5)
ACT, points# 15.8 (4.9) 16.1 (4.9) 15.3 (4.5) 16.6 (4.8) 15.0 (3.7) 16.3 (5.0) 15.6 (4.7) 16.9 (5.0)
EXAhis, events 0.0 (1.0) 1.0 (5.0)* 0.0 (5.5) 1.0 (3.0) 4.0 (13.0) 0.0 (3.0)* 1.0 (5.0) 0.0 (3.0)
Bude_eq, μg·day−1¶ 400.0 (290.0) 640.0 (250.0) 400.0 (297.0) 640.0 (240.0) 640.0 (352.0) 640.0 (240.0) 640.0 (280.0) 640.0 (240.0)
Pred_eq, mg·day−1¶ 0.0 (0.0) 0.0 (5.0) 0.0 (3.7) 0.0 (5.0) 0.0 (7.5) 0.0 (4.4) 2.0 (5.0) 0.0 (0.0)*
AR, n (%)+ 11 (68.8) 102 (74.5) 29 (82.9) 81 (71.1) 15 (71.4) 97 (74.0) 69 (74.2) 44 (74.6)
GR, n (%)+ 7 (43.8) 50 (36.5) 16 (45.7) 39 (34.2) 9 (42.9) 47 (35.9) 39 (41.9) 18 (30.5)
NP, n (%)+ 4 (25.0) 34 (24.8) 8 (22.9) 29 (25.4) 3 (14.3) 35 (26.7) 23 (24.7) 15 (25.4)
Open in a new tab

Data are presented as mean±sd or median (interquartile range) unless indicated otherwise. EXA-OCS Res, FEV1 Res, ACT Res: response based on exacerbation and oral corticosteroid use, FEV1 and ACT, respectively; Clin-Rem: clinical remission; Agediag: age at asthma diagnosis; BMI: body mass index; FEV1: predicted pre-bronchodilator forced expiratory volume in 1 s; BEC: blood eosinophil count; BEC150 and BEC300: patients with baseline BEC ≥150 and ≥300 cells·μL−1, respectively; Sp-EOS: sputum eosinophils as a percentage; Sp-EOS2%: patients with baseline Sp-EOS ≥2%; FENO: fractional exhaled nitric oxide; ACT: asthma control test score; EXAhis: recall exacerbation number during the previous year; Bude_eq and Pred_eq: budesonide and prednisolone equivalent doses, respectively; AR: allergic rhinitis; GR: gastroesophageal reflux; NP: nasal polyp. p-values were determined using: #: ANOVA for continuous variables; : rank sum test for non-parametric variables; +: Fisher's exact test for categorical variables. *: p-values <0.05.

Association with treatment response and remission

Multivariate regression analysis adjusted for age and sex showed that baseline FENO was positively correlated with EXA-OCS-Res, FEV1-Res and ACT-Res in the overall dataset (odds ratio (OR) 1.54 (95% CI 0.92–2.64), 1.8 (1.21–2.74) and 2.15 (1.29–3.75), respectively). Meanwhile, baseline FENO was positively correlated with FEV1-Res and ACT-Res among anti-IL-4Rα users (OR 2.34 (95% CI 1.39–4.17) and 3.6 (1.73–8.84), respectively). Baseline Sp-EOS levels showed no association with any outcome in any dataset. Clin-Rem also did not show any significant association with any of the three T2-inflammation biomarkers, either in the overall data or when stratified by biologic groups (table 3).

TABLE 3.

Multivariate regression and optimal cut-off for BEC, Sp-EOS and FENO in predicting treatment response

Outcome/predictor All biologics Anti-IL-5/IL-5Rα users Anti-IL-4Rα users
OR (95% CI) p-value Cut-off Se Sp OR (95% CI) p-value Cut-off Se Sp OR (95% CI) p-value Cut-off Se Sp
EXA-OCS-Res
 BEC 0.96 (0.55–1.58) 0.864 ≤538 0.50 0.50 0.66 (0.18–1.6) 0.438 ≥638 0.56 0.60 1.05 (0.55–1.92) 0.885 ≤470 0.54 0.55
 Sp-EOS 1.19 (0.68–2.16) 0.543 ≥6 0.58 0.50 1.31 (0.43–4.21) 0.620 ≥16 0.51 0.75 1.1 (0.56–2.23) 0.787 ≥6 0.51 0.60
 FENO 1.54 (0.92–2.64) 0.106 ≥45 0.59 0.69 2.4 (0.89–8.19) 0.109 ≥51 0.66 0.60 1.37 (0.72–2.64) 0.335 ≥44 0.53 0.55
FEV1-Res
 BEC 1.47 (1–2.2) 0.055 ≥480 0.61 0.60 1.76 (0.89–3.94) 0.121 ≥638 0.59 0.70 1.25 (0.78–2.03) 0.343 ≥350 0.59 0.60
 Sp-EOS 1.29 (0.83–2.02) 0.263 ≥7 0.53 0.56 0.94 (0.38–2.17) 0.879 ≤24 0.55 0.57 1.31 (0.77–2.31) 0.338 ≥5 0.53 0.60
 FENO 1.8 (1.21–2.74) 0.004 ≥45 0.63 0.66 0.85 (0.4–1.74) 0.657 ≤70 0.61 0.60 2.34 (1.39–4.17) 0.002 ≥43 0.69 0.68
ACT-Res
 BEC 1.44 (0.89–2.3) 0.13 ≥427 0.68 0.67 1.03 (0.33–2.81) 0.960 ≤674 0.53 0.50 1.75 (0.96–3.26) 0.069 ≥312 0.67 0.67
 Sp-EOS 1.15 (0.67–2.02) 0.622 ≥7 0.52 0.56 0.03 (0–0.47) 0.075 ≤36 0.79 0.75 1.83 (0.88–4.34) 0.127 ≥4 0.58 0.58
 FENO 2.15 (1.29–3.75) 0.005 ≥44 0.62 0.65 0.55 (0.18–1.45) 0.241 ≤73 0.64 0.60 3.6 (1.73–8.84) 0.002 ≥39 0.72 0.73
Clin-Rem
 BEC 1.27 (0.89–1.85) 0.194 ≥550 0.53 0.58 1.19 (0.66–2.33) 0.580 ≥700 0.54 0.59 1.31 (0.85–2.1) 0.236 ≥459 0.60 0.59
 Sp-EOS 1.27 (0.88–1.84) 0.210 ≥9 0.56 0.57 0.81 (0.43–1.48) 0.490 ≤19 0.52 0.48 1.57 (0.98–2.58) 0.065 ≥7 0.55 0.61
 FENO 1.38 (0.98–1.97) 0.071 ≥56 0.59 0.59 0.71 (0.4–1.24) 0.239 ≤63 0.58 0.55 1.95 (1.23–3.28) 0.007 ≥52 0.66 0.66
Open in a new tab

p-values <0.05 are indicated in bold. BEC: blood eosinophil count; Sp-EOS: sputum eosinophils as a percentage; FENO: fractional exhaled nitric oxide; OR (95% CI): odds ratios and 95% confidence intervals adjusted for age and sex for a 1-unit increase in the log-transformed predictor; cut-off: optimal cut point of raw values of the predictor estimated by the sensitivity–specificity equality criterion; Se and Sp: corresponding sensitivity and specificity of the identified cut-off; IL: interleukin; EXA-OCS-Res, FEV1-Res, ACT-Res: response based on exacerbation and oral corticosteroid use, predicted pre-bronchodilator forced expiratory volume in 1 s (FEV1) and asthma control test score (ACT), respectively; Clin-Rem: clinical remission.

In an analysis of data after over-sampling using the ROSE technique, baseline BEC was inversely correlated with EXA-OCS-Res in the overall dataset (OR 0.55, 95% CI 0.37–0.8) and positively correlated with EXA-OCS-Res (OR 1.67, 95% CI 1.05–2.85) and ACT-Res (OR 1.65, 95% CI 1.03–2.76) among anti-IL-4Rα users. Baseline Sp-EOS was inversely correlated with ACT-Res and Clin-Rem among anti-IL-5/IL-5Rα users (OR 0.24 (95% CI 0.06–0.61) and 0.42 (0.19–0.86), respectively) and positively correlated with Clin-Rem among anti-IL-4Rα users (OR 1.92, 95% CI 1.19–3.26). Baseline FENO was positively correlated with EXA-OCS-Res, FEV1-Res, ACT-Res and Clin-Rem in the overall dataset (OR 1.46 (95% CI 1.02–2.11), 2.08 (1.43–3.15), 2.45 (1.65–3.84) and 1.44 (1.03–2.04), respectively). Among anti-IL-4Rα users, FENO was positively correlated with FEV1-Res, ACT-Res and Clin-Rem (OR 1.73 (95% CI 1.11–2.79), 6.03 (2.98–14.39) and 2.25 (1.39–3.89), respectively), whereas among anti-IL-5/IL-5Rα users, FENO was positively correlated only with EXA-OCS-Res (OR 1.94, 95% CI 1.05–3.89) (supplementary table 6S).

Discriminative power

Figure 2 illustrates the AUC of the seven models adjusted for age and sex for each treatment response domain and clinical remission. The analysis was conducted on the overall dataset, anti-IL-5/IL-5Rα users and anti-IL-4Rα users. For the treatment response domains (figure 2a–i), there were no differences in AUC between the three T2-inflammation factors across EXA-OCS-Res (all datasets: figure 2a–c), FEV1-Res for the overall dataset and among anti-IL-5/IL-5Rα users (figure 2d, f), and ACT-Res for the overall dataset and among anti-IL-4Rα users (figure 2g, i). For FEV1-Res among anti-IL-4Rα users, the FENO model (M3) had a higher AUC compared to the BEC model (M1) (AUC: 0.75 versus 0.62). Adding BEC or Sp-EOS into M3 did not improve the model performance (figure 2f). For ACT-Res among anti-IL-5/IL-5Rα users, Sp-EOS outperformed BEC and FENO in predicting ACT-Res (AUC: 0.91, 0.4 and 0.65, respectively, with all p-values from the DeLong test <0.05). However, adding BEC or FENO to the Sp-EOS model did not increase the AUC (figure 2h). Overall, the combination of the three factors did not significantly increase the AUC compared to models using individual factors. An exception was observed among anti-IL-5/IL-5Rα users and FEV1-Res, where the combination of three factors (M7) improved the AUC of the Sp-EOS (M2) and FENO (M3) models (AUC: 0.92 versus 0.69 and 0.72, respectively) (figure 2e). For Clin-Rem, there were no differences in AUC across all models, whether individual or combined, in any dataset (figure 2j–l).

FIGURE 2.

FIGURE 2

Open in a new tab

Area under the curve (AUC) and 95% confidence interval of the models predicting each response outcome. Datasets: overall data, anti-interleukin (IL)-5/IL-5Rα users and anti-IL-4Rα users. Model 1 (M1): outcome—BEC+covariates; Model 2 (M2): outcome—Sp-EOS+covariates; Model 3 (M3): outcome—FENO+covariates; Model 4 (M4): outcome—BEC+Sp-EOS+covariates; Model 5 (M5): outcome—BEC+FENO+covariates; Model 6 (M6): outcome—FENO+Sp-EOS+covariates; Model 7 (M7): outcome—BEC+Sp-EOS+FENO+covariates. Covariates were age and sex. Data are presented as the mean and 95% confidence intervals of the AUC of the receiver operating characteristic (ROC) curves. The vertical dashed grey line is the reference line for the AUC value of 0.5. Pairwise comparisons of AUC by DeLong test with p-value <0.05: a: M1 versus M2; b: M1 versus M3; c: M1 versus M4; d: M1 versus M5; e: M1 versus M6; f: M1 versus M7; g: M2 versus M3; h: M2 versus M4; i: M2 versus M5; j: M2 versus M6; k: M2 versus M7; l: M3 versus M4; m: M3 versus M5; n: M3 versus M6; o: M3 versus M7; p: M4 versus M5; q: M4 versus M6; r: M4 versus M7; s: M5 versus M6; t: M5 versus M7; u: M6 versus M7. EXA-OCS-Res: responder is defined as a reduction by ≥50% in the annual exacerbation rate, or a reduction by ≥50% in the maintenance oral corticosteroid (mOCS) dose, or remaining without exacerbations; FEV1-Res: FEV1 responder is defined as having a 12-month visit FEV1 ≥80%, or an FEV1 increase by ≥100 mL and ≥10%; ACT-Res: ACT responder is defined as having a 12-month visit ACT score ≥20, or an ACT score increase by ≥3 points; Clin-Rem: clinical remission is defined as having no mOCS use and no exacerbations during the 12-month follow-up, plus meeting the criteria for FEV1-Res and ACT-Res. BEC: blood eosinophil count; Sp-EOS: sputum eosinophils; FENO: fractional exhaled nitric oxide.

In the ROSE-adjusted analysis, differences emerged across analyses for treatment response domains. In the overall dataset, the BEC model (M1) demonstrated a higher AUC compared to the Sp-EOS model (M2) for EXA-OCS-Res (0.75 versus 0.61, p<0.05). However, adding Sp-EOS (M4) or FENO (M5) did not enhance the AUC of the model (supplementary figure 1Sa). The FENO model showed superior performance compared to the Sp-EOS model for both FEV1-Res (AUC: 0.73 versus 0.56, p<0.05) and ACT-Res (AUC: 0.75 versus 0.60, p<0.05). Combining BEC or Sp-EOS with FENO did not improve the model's predictive performance (supplementary figure 1Sd and g). For anti-IL-5/IL-5Rα users, similar to the pre-over-sampling analysis, the Sp-EOS model had a superior AUC for predicting ACT-Res compared to the BEC and FENO models (AUC: 0.83 versus 0.59 and 0.59, respectively). However, adding BEC or FENO to the Sp-EOS model did not enhance the predictive power (supplementary figure 1Sh). There were no differences in predictive power among the three predictors in the analysis of EXA-OCS-Res or FEV1-Res (supplementary figure 1Sb, e). For EXA-OCS-Res, combining all three factors in one model (M7) significantly increased the AUC compared to models with a single predictor (M1, M2, M3) (AUC: 0.96 versus 0.83, 0.76 and 0.82, respectively) (supplementary figure 1Sb). For FEV1-Res, the model incorporating all three factors (M7, AUC=0.9) outperformed the FENO model (M2, AUC: 0.64) but was not superior to the BEC (M1) or Sp-EOS (M3) models (supplementary figure 1Se). For anti-IL-4Rα users, the FENO model (M3) was superior to the BEC and Sp-EOS models for ACT-Res (AUC: 0.90 versus 0.77 and 0.71). However, combining other factors did not improve the performance of the FENO model (Figure 1Si). No differences in performance were observed between the seven models for EXA-OCS-Res (supplementary figure 1Sc). The FENO model (M3) outperformed the BEC model (M1) for FEV1-Res (AUC: 0.73 versus 0.65). Combining BEC or Sp-EOS with FENO did not enhance the performance of the model (supplementary figure 1Sf). For the Clin-Rem outcome, there were no differences in predictive power between the BEC, Sp-EOS and FENO models in the overall datasets and among anti-IL-5/IL-5Rα users (supplementary figure 1Sj, k), whereas FENO showed a higher AUC than BEC (AUC 0.76 versus 0.65, supplementary figure 1Sl) in the anti-IL-4Rα group. The model combining Sp-EOS and FENO (M6) increased the predictive power of the model with FENO alone (M3) in the anti-IL-5/IL-5Rα group (AUC 0.89 versus 0.69, supplementary figure 1Sk). However, combining predictors did not improve the single predictor model in the anti-IL-5/IL-5Rα group or across the entire dataset.

Discussion

Predicting the response to biologics and remission in SEA treatment will enhance effectiveness and cost-efficiency in personalised medicine. Previous discoveries of predictors of response were mostly based on secondary data analyses from RCTs, which selected patients for mAb therapy according to strict criteria [16–19]. However, asthma pathogenesis is highly heterogeneous, and few studies have investigated the predictors of treatment response and remission to biologics in real-life settings [12, 29–33]. Our study showed that the 1-year response rate to biologics in patients with SEA was relatively high and varied depending on outcome domains. Meanwhile, only one-third of patients with SEA fulfilled the requirements for Clin-Rem. The discriminative power of the predictors among the three T2 inflammation biomarkers also differed depending on outcome domains and mAbs used. T2 inflammation markers were less likely to predict outcomes in exacerbation and mOCS domains than in lung function and asthma control outcomes. Higher pre-treatment BEC was observed in individuals who responded to biologics, particularly among anti-IL-4Rα users and for outcomes such as ACT-Res and Clin-Rem. However, when adjusted for covariates such as age and sex, this relationship no longer persisted. FENO was a predictor of treatment response in ACT and FEV1 domains for anti-IL-4Rα users. Analysis of the over-sampled data highlighted the relationship between predictors and treatment response. However, the association between T2 inflammation markers and Clin-Rem was weak in all analyses across groups, regardless of the type of biologics used. The combination of predictors did not significantly improve their predictive ability.

The concept of response to biologics treatment should be evaluated from multiple aspects [9–11]. Studies on mAb effects mostly relied on the reduction in exacerbation rate as the primary outcome [5]. However, asthma is heterogeneous, with many patients showing a response in one domain but not in others [12]. The frequency of exacerbations is directly proportional to the decline in lung function, which occurs at a faster rate in individuals aged <40 years [34]. For these individuals, rapid improvement in lung function to prevent irreversible lung remodeling should be a critical outcome. Recently, data from the International Severe Asthma Registry revealed that the rate of treatment response using the exacerbation domain was the highest (>80%), while response rates in other domains were <50%. Another real-world study reported response rates of <50% across all four response domains [35]. The current study showed the response rate to biological therapy using two domains, exacerbation and mOCS, as well as one domain (either FEV1 or ACT), all >75%. This is because we also considered patients who had good pre-treatment conditions (no exacerbations, or high FEV1 or ACT) and maintained these conditions throughout treatment as responders, rather than solely defining a response as an improvement in pre-treatment impairment. Given that conditions of patients with SEA often deteriorate gradually and irreversibly, maintaining any outcome domain in a good state is clinically significant. The rates of achieving clinical remission have also been reported inconsistently. Although there has been consensus on the four criteria in the definition of Clin-Rem [13], specific thresholds for lung function and asthma control status vary between studies. Recent real-world studies have shown that remission rates in patients using biologics range from 19% to 32.1%, depending on the type of biologics [31, 33]. Our data indicate a relatively high Clin-Rem rate (ranging from 37% to 41%), suggesting that Clin-Rem is a feasible target for a substantial proportion of patients with SEA.

BEC is widely recognised as a predictor of biologics treatment efficacy. This understanding stems from post hoc analyses of RCTs comparing biologics to placebo across various BEC levels [15–18]. These analyses concluded that higher baseline BEC correlates with improved response. Consequently, Global Initiative for Asthma (GINA) recommends anti-IL-5/IL-5Rα and anti-IL-4α therapies for patients with high BEC (typically >150 cells·μL−1) [1]. However, it remains uncertain whether this association persists in real-life patients with SEA who already present with elevated BEC before treatment. Some real-world studies have confirmed the association between higher BEC and better treatment response [12, 32] or clinical remission [31, 33] for biologic users. However, in our study, the association between BEC and treatment response was not as strong. While BEC was associated with ACT-Res and Clin-Rem in univariate regression, this association disappeared after adjusting for covariates. The discrepancy between our findings and prior studies may be attributed to several factors. First, the small sample size may have significantly impacted the estimates, especially in the presence of imbalanced outcome variables. When we analysed the over-sampled data, BEC was associated with EXA-OCS-Res and ACT-Res in anti-IL-4α users. BEC is a skewed variable, and the correlation between BEC and eosinophilic airway inflammation is stronger at particularly high levels of BEC. Several previous studies have found a correlation between BEC and biological treatment responses, particularly when comparing groups with high BEC (>300 cells·μL−1) to those with lower BEC levels [31, 32] or examining BEC doubling concentration [33]. In contrast, our analysis used log-transformed BEC as a continuous variable, which may explain the differing results.

Post hoc analyses of RCT data showed higher baseline FENO levels predicted better treatment response to dupilumab [20]. Real-life data indicated that patients with SEA with high FENO who did not respond to anti-IL-5/IL-5Rα therapy achieved remission after transitioning to anti-IL-4Rα [36]. In our study, FENO was associated with responses in FEV1-Res and ACT-Res among anti-IL-4α users, consistently observed in both original and over-sampled data. This highlights the importance of FENO as a predictor in anti-IL-4 therapy.

Sp-EOS, as a direct marker of eosinophilic airway inflammation, is less commonly used owing to its invasive nature and technical requirements. Two RCT analyses found Sp-EOS less predictive than BEC for anti-IL-5/IL-5Rα treatment [37, 38]. In our study, higher Sp-EOS levels were observed in anti-IL-5/IL-5Rα users with improved ACT scores, but multivariate analyses found no significant associations with treatment responses or remission, likely because of modest sample size and incomplete Sp-EOS data.

Several attempts have been made to combine potential predictors to improve predictive models for treatment response to biologics in patients with SEA [19, 39]. Combining the two criteria of BEC >150 and FENO >25 ppb could enhance the likelihood of response to treatment in patients with SA using anti-IL-4Rα [19]. However, no real-world study has evaluated the combination of all three T2 inflammation markers. In this study, combining the markers in pairs or using all three markers together did not significantly improve the predictive power compared to using only a single marker with the highest predictive power. This result only considered pre-treatment conditions and did not account for the variability of these markers during follow-up, possibly influencing the treatment response.

The findings should be interpreted within the study's strengths and limitations. This real-world study comprehensively examined multiple domains of treatment response using three T2 markers and their combinations. However, small sample sizes limited separate analyses for specific drugs among anti-IL-5/IL-5Rα users. High response rates in certain domains (FEV1, ACT) may bias prediction models, though this was addressed with the ROSE technique. The follow-up was limited to only 12 months, which may not fully capture the long-term effects of biological therapy. The predictive models employed only three markers and their combinations, without considering interactions with other confounding factors. The results should not be extended to all biological treatments because the study did not include other mAbs, such as anti-IgE or anti-TSLP. Potential bias from site differences in this multicentre cohort was minimised through standardized protocols and rigorous training.

We conclude that in this study, pre-treatment BEC and Sp-EOS do not have a clear association with the treatment response to T2-biologics in SEA patients. Pre-treatment FENO better predicts specific responses, particularly with anti-IL-4 therapy. Combining T2 biomarkers does not significantly improve predictive power compared with the best single marker.

Footnotes

Provenance: Submitted article, peer reviewed.

This study is registered a www.clinicaltrials.gov as NCT05164939.

Ethics statement: The Institutional Review Board at each study site approved the study protocol and registered it on ClinicalTrials.gov (registration number NCT05164939).

Conflict of interest: W-J. Song serves as the Chief Editor of this journal. P. Bhavsar has received grants from GSK and MSD for investigator-initiated studies, as well as honoraria from GSK and AstraZeneca; and is an associate editor of this journal. I. Adcock has received grants from GSK, Sanofi, UKRI and EPSRC for investigator-initiated studies, along with honoraria from GSK and AstraZeneca. K.F. Chung has received grants from MRC, GSK, Merck, NIEHS and EPSRC for investigator-initiated studies, and honoraria from GSK, Novartis and AstraZeneca. All other authors declare no conflicts of interest.

Support statement: This research was supported by a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number: RS-2024-00403700). Funding information for this article has been deposited with the Open Funder Registry.

Supplementary material

Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.

Supplementary material

00969-2024.SUPPLEMENT.pdf (541.7KB, pdf)
DOI: 10.1183/23120541.00969-2024.Supp1
Open in a new tab

00969-2024.SUPPLEMENT

References

  • 1.Global Initiative for Asthma . Global Strategy for Asthma Management and Prevention, 2023. Date last accessed: 5 June 2024. Date last updated: July 2023. https://ginasthma.org/wp-content/uploads/2023/07/GINA-2023-Full-report-23_07_06-WMS.pdf
  • 2.Kuruvilla ME, Lee FE, Lee GB. Understanding asthma phenotypes, endotypes, and mechanisms of disease. Clin Rev Allergy Immunol 2019; 56: 219–233. doi: 10.1007/s12016-018-8712-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Busse WW, Viswanathan R. What has been learned by cytokine targeting of asthma? J Allergy Clin Immunol 2022; 150: 235–249. doi: 10.1016/j.jaci.2022.06.010 [DOI] [PubMed] [Google Scholar]
  • 4.Hammad H, Lambrecht BN. The basic immunology of asthma. Cell 2021; 184: 2521–2522. doi: 10.1016/j.cell.2021.04.019 [DOI] [PubMed] [Google Scholar]
  • 5.Agache I, Beltran J, Akdis C, et al. Efficacy and safety of treatment with biologicals (benralizumab, dupilumab, mepolizumab, omalizumab and reslizumab) for severe eosinophilic asthma. A systematic review for the EAACI Guidelines – recommendations on the use of biologicals in severe asthma. Allergy 2020; 75: 1023–1042. doi: 10.1111/all.14221 [DOI] [PubMed] [Google Scholar]
  • 6.Farne HA, Wilson A, Milan S, et al. Anti-IL-5 therapies for asthma. Cochrane Database Syst Rev 2022; 7: CD010834. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Charles D, Shanley J, Temple SN, et al. Real-world efficacy of treatment with benralizumab, dupilumab, mepolizumab and reslizumab for severe asthma: a systematic review and meta-analysis. Clin Exp Allergy 2022; 52: 616–627. doi: 10.1111/cea.14112 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Castro M, Rabe KF, Corren J, et al. Dupilumab improves lung function in patients with uncontrolled, moderate-to-severe asthma. ERJ Open Res 2020; 6: 00204-201. doi: 10.1183/23120541.00204-2019 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Papaioannou AI, Fouka E, Bartziokas K, et al. Defining response to therapy with biologics in severe asthma: from global evaluation to super response and remission. Expert Rev Respir Med 2023; 17: 481–493. doi: 10.1080/17476348.2023.2226392 [DOI] [PubMed] [Google Scholar]
  • 10.Perez de Llano L, Davila I, Martinez-Moragon E, et al. Development of a tool to measure the clinical response to biologic therapy in uncontrolled severe asthma: the FEV(1), exacerbations, oral corticosteroids, symptoms score. J Allergy Clin Immunol Pract 2021; 9: 2725–2731. doi: 10.1016/j.jaip.2021.01.033 [DOI] [PubMed] [Google Scholar]
  • 11.Tiotiu A, Bikov A, Gonzalez-Barcala FJ, et al. Criteria to evaluate efficacy of biologics in asthma: a Global Asthma Association survey. Expert Rev Respir Med 2023; 17: 507–516. doi: 10.1080/17476348.2023.2223986 [DOI] [PubMed] [Google Scholar]
  • 12.Scelo G, Tran TN, Le TT, et al. Exploring definitions and predictors of response to biologics for severe asthma. J Allergy Clin Immunol Pract 2024; 12: 2347–2361. doi: 10.1016/j.jaip.2024.05.016 [DOI] [PubMed] [Google Scholar]
  • 13.Menzies-Gow A, Bafadhel M, Busse W, et al. An expert consensus framework for asthma remission as a treatment goal. J Allergy Clin Immunol Pract 2020; 145: 757–765. doi: 10.1016/j.jaci.2019.12.006 [DOI] [PubMed] [Google Scholar]
  • 14.Saturni S, Bellini F, Braido F, et al. Randomized Controlled Trials and real life studies. Approaches and methodologies: a clinical point of view. Pulm Pharmacol Ther 2014; 27: 129–138. doi: 10.1016/j.pupt.2014.01.005 [DOI] [PubMed] [Google Scholar]
  • 15.Kroes JA, Zielhuis SW, van Roon EN, et al. Prediction of response to biological treatment with monoclonal antibodies in severe asthma. Biochem Pharmacol 2020; 179: 113978. doi: 10.1016/j.bcp.2020.113978 [DOI] [PubMed] [Google Scholar]
  • 16.Ortega HG, Yancey SW, Mayer B, et al. Severe eosinophilic asthma treated with mepolizumab stratified by baseline eosinophil thresholds: a secondary analysis of the DREAM and MENSA studies. Lancet Respir Med 2016; 4: 549–556. doi: 10.1016/S2213-2600(16)30031-5 [DOI] [PubMed] [Google Scholar]
  • 17.Corren J, Weinstein S, Janka L, et al. Phase 3 study of reslizumab in patients with poorly controlled asthma effects across a broad range of eosinophil counts. Chest 2016; 150: 799–810. doi: 10.1016/j.chest.2016.03.018 [DOI] [PubMed] [Google Scholar]
  • 18.FitzGerald JM, Bleecker ER, Menzies-Gow A, et al. Predictors of enhanced response with benralizumab for patients with severe asthma: pooled analysis of the SIROCCO and CALIMA studies. Lancet Resp Med 2018; 6: 51–64. doi: 10.1016/S2213-2600(17)30344-2 [DOI] [PubMed] [Google Scholar]
  • 19.Rabe KF, Nair P, Brusselle G, et al. Efficacy and safety of dupilumab in glucocorticoid-dependent severe asthma. N Engl J Med 2018; 378: 2475–2485. doi: 10.1056/NEJMoa1804093 [DOI] [PubMed] [Google Scholar]
  • 20.Pavord ID, Deniz Y, Corren J, et al. Baseline FENO independently predicts the dupilumab response in patients with moderate-to-severe asthma. J Aller Clin Immunol Pract 2023; 11: 1213–1220.e2. [DOI] [PubMed] [Google Scholar]
  • 21.Dweik RA, Boggs PB, Erzurum SC, et al. An official ATS clinical practice guideline: interpretation of exhaled nitric oxide levels (FE) for clinical applications. Am J Resp Crit Care 2011; 184: 602–615. doi: 10.1164/rccm.9120-11ST [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Yancey SW, Keene ON, Albers FC, et al. Biomarkers for severe eosinophilic asthma. J Allergy Clin Immunol 2017; 140: 1509–1518. doi: 10.1016/j.jaci.2017.10.005 [DOI] [PubMed] [Google Scholar]
  • 23.Mukherjee M, Nair P. Blood or sputum eosinophils to guide asthma therapy? Lancet Respir Med 2015; 3: 824–825. doi: 10.1016/S2213-2600(15)00419-1 [DOI] [PubMed] [Google Scholar]
  • 24.Lee JH DP, Bhavsar P, Raby K, et al. Precision Medicine Intervention in Severe Asthma (PRISM) study: molecular phenotyping of patients with severe asthma and response to biologics. ERJ Open Res 2023; 9: 00485-2022. doi: 10.1183/23120541.00485-2022 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.N CT05164939. Precision Medicine Intervention in Severe Asthma (PRISM) Study . First received: 21 December 2021. https://clinicaltrials.gov/study/NCT05164939
  • 26.Delong ER, Delong DM, Clarkepearson DI. Comparing the areas under 2 or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988; 44: 837–845. doi: 10.2307/2531595 [DOI] [PubMed] [Google Scholar]
  • 27.Lopez-Raton M, Cadarso-Suarez C, Rodriguez-Alvarez MX, et al. OptimalCutpoints: An R Package for Selecting Optimal Cutpoints in Diagnostic Tests. J Stat Softw 2014; 61: 1–36. doi: 10.18637/jss.v061.i08 [DOI] [Google Scholar]
  • 28.Menardi G, Torelli N. Training and assessing classification rules with imbalanced data. Data Min Knowl Disc 2014; 28: 92–122. doi: 10.1007/s10618-012-0295-5 [DOI] [Google Scholar]
  • 29.Drick N, Seeliger B, Welte T, et al. Anti-IL-5 therapy in patients with severe eosinophilic asthma: clinical efficacy and possible criteria for treatment response. BMC Pulm Med 2018; 18: 119. doi: 10.1186/s12890-018-0689-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Perez-de-Llano L, Scelo G, Canonica GW, et al. Impact of pre-biologic impairment on meeting domain-specific biologic responder definitions in patients with severe asthma. Ann Allergy Asthma Immunol 2024; 132: 610–622. doi: 10.1016/j.anai.2023.12.023 [DOI] [PubMed] [Google Scholar]
  • 31.Milger K, Suhling H, Skowasch D, et al. Response to Biologics and Clinical Remission in the Adult German Asthma Net Severe Asthma Registry Cohort. J Allergy Clin Immunol Pract 2023; 11: 2701–2712.e2. doi: 10.1016/j.jaip.2023.05.047 [DOI] [PubMed] [Google Scholar]
  • 32.Soendergaard MB, Hansen S, Bjerrum AS, et al. Complete response to anti-interleukin-5 biologics in a real-life setting: results from the nationwide Danish Severe Asthma Register. ERJ Open Res 2022; 8: 00238-2022. doi: 10.1183/23120541.00238-2022 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Hansen S, Baastrup Søndergaard M, von Bülow A, et al. Clinical response and remission in patients with severe asthma treated with biologic therapies. Chest 2024; 165: 253–266. doi: 10.1016/j.chest.2023.10.046 [DOI] [PubMed] [Google Scholar]
  • 34.Soremekun S, Heaney LG, Skinner D, et al. Asthma exacerbations are associated with a decline in lung function: a longitudinal population-based study. Thorax 2023; 78: 643–652. doi: 10.1136/thorax-2021-217032 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Denton E, Hew M, Peters MJ, et al. Real-world biologics response and super-response in the International Severe Asthma Registry cohort. Allergy 2024; 79: 2700–2716. doi: 10.1111/all.16178 [DOI] [PubMed] [Google Scholar]
  • 36.Gates J, Hearn A, Mason T, et al. Long-term effectiveness of anti-IL-4R therapy following suboptimal response to anti-IL-5/5R therapy in severe eosinophilic asthma. J Allergy Clin Immunol Pract 2024; 12: 1794–1800. doi: 10.1016/j.jaip.2024.03.049 [DOI] [PubMed] [Google Scholar]
  • 37.Pavord ID, Korn S, Howarth P, et al. Mepolizumab for severe eosinophilic asthma (DREAM): a multicentre, double-blind, placebo-controlled trial. Lancet 2012; 380: 651–659. doi: 10.1016/S0140-6736(12)60988-X [DOI] [PubMed] [Google Scholar]
  • 38.Ortega H, Katz L, Gunsoy N, et al. Blood eosinophil counts predict treatment response in patients with severe eosinophilic asthma. J Allergy Clin Immunol 2015; 136: 825–826. doi: 10.1016/j.jaci.2015.05.039 [DOI] [PubMed] [Google Scholar]
  • 39.Bateman ED, Djukanovic R, Castro M, et al. Predicting responders to reslizumab after 16 weeks of treatment using an algorithm derived from clinical studies of patients with severe eosinophilic asthma. Am J Resp Crit Care 2019; 199: 489–495. doi: 10.1164/rccm.201708-1668OC [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.

Supplementary material

00969-2024.SUPPLEMENT.pdf (541.7KB, pdf)
DOI: 10.1183/23120541.00969-2024.Supp1
Open in a new tab

00969-2024.SUPPLEMENT

Articles from ERJ Open Research are provided here courtesy of European Respiratory Society

RESOURCES