Study design

This retrospective, observational multicenter study analyzed data from patients diagnosed with severe asthma who underwent biologic therapy at 4 tertiary hospitals in Japan: Tohoku University Hospital (Miyagi), National Hospital Organization Sagamihara National Hospital (Kanagawa), Japanese Red Cross Ishinomaki Hospital (Miyagi), and South Miyagi Medical Center (Miyagi) (Figure 1A). This study was conducted in accordance with the principles outlined in the Declaration of Helsinki, and the study protocol was approved by the Institutional Review Board of Tohoku University Hospital (IRB No. 2022-1-198). Written informed consent was obtained from all participants prior to enrollment in the study. Patient enrollment was conducted between July 2020 and July 2022. The study timeline comprised 2 main phases: baseline and follow-up (Figure 1A). The cohort entry date for each participant was defined as the date of initiation of Bx. The baseline period was established as the first visit to the cohort entry date, during which pre-Bx clinical characteristics were assessed. For all patients, baseline measurements (including FeNO and spirometry) were obtained immediately before initiation of their current biologic therapy, regardless of prior biologic exposure history. This approach ensured standardized baseline assessments for both biologic-naïve patients and those switching from previous biologic treatments. Pre-Bx exacerbations were evaluated for 12 months preceding the cohort entry date. This period was designated as the pre-Bx exacerbation assessment period. The follow-up period began on the cohort entry date and extended for at least 12 months. Patients were required to have received the current biologic agent for a minimum of 12 consecutive months to be eligible for the study analysis. The post-Bx exacerbation assessment period was defined as the 12 months before patient enrollment, which coincided with the end of the follow-up period.

Display full size

Figure 1. Overview of study timeline and patient selection process. (A) Study design and timeline of the clinical research study. The study period consisted of two main phases: a baseline period before Bx initiation and a follow-up period after Bx initiation. The pre-Bx exacerbation assessment period was defined as 12 months before Bx initiation. The post-Bx exacerbation assessment period was defined as 12 months before patient enrollment. (B) Patient selection flowchart. Of 106 patients initially screened for eligibility, 53 patients were included in the final study population. Bx: biologics

Patients

Asthma was diagnosed by respiratory physicians in accordance with the Japanese guidelines or the Global Initiative for Asthma (GINA) guidelines, while severe asthma was defined according to the GINA guidelines. Eligible patients were ≥ 18 years of age, had been undergoing biologic therapy for ≥ 1 year, and had baseline assessments conducted at enrollment and upon initiation of the first biologic therapy. Biologic agents included in the present study were omalizumab, mepolizumab, benralizumab, and dupilumab, the choice of which was at the discretion of the attending physician. Patients who switched Bx upon enrollment were included, provided that the current biologic used at enrollment had been administered for ≥ 1 year. Patients with insufficient data for analysis were excluded to ensure the integrity and completeness of the dataset. Additionally, individuals with a current biologic treatment duration of less than 12 months were excluded. Patients with comorbidities requiring chronic systemic corticosteroid use were also excluded to avoid potential confounding effects on asthma-related outcomes. Non-steroidal anti-inflammatory drug-exacerbated respiratory disease, allergic bronchopulmonary mycosis, and eosinophilic granulomatosis with polyangiitis were also excluded due to their distinct pathophysiology, which could obscure the core features of severe asthma.

Data collection and evaluation

The following patient characteristics were analyzed: sex; age; body mass index (BMI); and baseline treatments, including Bx, duration of asthma, and comorbid diseases. Additionally, the following parameters were evaluated during the baseline period: peripheral blood eosinophil count (BEC); serum IgE, FeNO; pulmonary function test (PFT) results [forced vital capacity (FVC), forced expiratory volume in 1 second (FEV₁), FEV₁/FVC ratio, percent predicted FEV₁ (FEV₁ %predicted)]; and daily maintenance OCS (mOCS), expressed as prednisone equivalents (mg per day). The post-bronchodilator values for the PFT were collected. The highest BEC and FeNO levels were recorded during the baseline period. Serum IgE levels were determined on the date nearest the cohort entry date. Specific IgE positivity for ≥ 1 aero-allergen(s) was required to confirm the presence of atopy. Data regarding exacerbation history, including the frequency and severity of exacerbations in the 12 months preceding biologic initiation or the enrollment, were also collected.

Statistics

Statistical analyses were performed using Prism version 10.4.1 (GraphPad Inc., San Diego, CA, USA), JMP Pro version 17.0.0 (SAS Institute, Cary, NC, USA), or Python version 3.12. Data are expressed as median with interquartile range (IQR) for continuous variables, and numbers with percentages for categorical variables unless otherwise indicated. Changes from pre-Bx to post-Bx were analyzed using the Wilcoxon matched-pairs signed-rank test for non-normally distributed continuous variables, paired t-test for normally distributed continuous variables, and McNemar's test or Fisher’s exact test for categorical variables. Changes in lung function categories were visualized using a Sankey diagram created in Python version 3.12, and analyzed using Markov chain analysis, with a chi-squared test for statistical significance using the “scipy.stats” package on Python. JMP was used for hierarchical clustering analysis to identify distinct disease activity groups. Canonical correlation analysis was used to evaluate relationships between clinical variables and disease activity groups, and analysis of covariance (ANCOVA) was used to assess the relationship between pre-Bx and post-Bx outcomes. Differences in clinical outcomes across groups identified through the clustering analysis were assessed using the Kruskal-Wallis test, followed by Dunn’s multiple comparison test for continuous variables and negative binomial regression, followed by Dunnett’s multiple comparison test for exacerbation frequencies. Multinomial logistic regression analysis was performed to identify predictors of post-Bx control status. Odds ratios were calculated as exp(β), where β represents the regression coefficient for each variable. Covariates for the multinomial logistic regression model were selected based on clinical relevance, previous literature, and statistical considerations. We initially included variables with p < 0.10 in univariable analyses. Additionally, we incorporated variables deemed important for group discrimination based on canonical correlation analysis, even if they did not meet the p < 0.10 threshold in univariable comparisons. To address multicollinearity, we employed variance inflation factor analysis. For highly correlated variables, we selected representative variables to include in the model, avoiding the simultaneous inclusion of strongly correlated predictors. We then performed backward stepwise selection to retain variables with p < 0.05 in the final multivariable model, balancing statistical significance with clinical relevance. Receiver operating characteristic (ROC) curves were constructed using GraphPad Prism by plotting sensitivity versus (vs.) 1-specificity across all thresholds, with optimal cutoff values determined via the Youden index to maximize diagnostic accuracy. Differences with p-value < 0.05 were considered to be statistically significant in all analyses.

Clinical outcomes and biomarker changes after biologic therapy in patients diagnosed with severe asthma

Data from 106 patients with severe asthma were initially screened for eligibility during the enrollment period (Figure 1). The final study population consisted of 53 patients, all of whom had severe asthma classified as GINA step 4 or 5 prior to initiating biologic therapy. The clinical characteristics of the study population at baseline are summarized in Table 1. A comprehensive analysis comparing clinical parameters pre-Bx vs. post-Bx was used to determine the clinical impact of biologic therapies on patients diagnosed with severe asthma. The median duration of overall biologic use was 42.0 months (range: 12.9–138.5 months). For the currently administered biologic agent, the median duration of use was 34.5 months (range: 12.9–138.5 months). Findings revealed significant improvements in multiple clinical outcomes following biologic treatment. The annual exacerbation frequency decreased significantly from a median of 3.0 (IQR 2.0–6.0) pre-Bx to 0.0 (0.0–0.0) post-Bx (p < 0.0001) (Figure 2A). The proportion of patients experiencing zero exacerbations increased dramatically, whereas that of patients with frequent exacerbations decreased substantially (p < 0.0001) (Figure 2B), demonstrating the efficacy of Bx in reducing exacerbation rates. Use of mOCS was significantly reduced after biologic therapy (p < 0.001) (Figure 2C). Concurrently, the number of patients requiring mOCS decreased significantly (p < 0.05) (Figure 2D), reflecting the steroid-sparing effect of biologic therapies. Lung function parameters also demonstrated a notable improvement. Median FEV₁ %predicted values increased significantly from pre-Bx to post-Bx (Figure 2E). Markov chain analysis revealed a significant shift toward improved lung function categories after biologic therapy (p < 0.0001) (Figure 2F). The distribution of absolute change in FEV₁ (ΔFEV₁) post-Bx is illustrated in Figure 2G, with a threshold of 200 mL improvement, demonstrating the positive impact of Bx on airway obstruction. BEC decreased significantly, from a median of 590/μL (IQR 265–1,206/μL) pre-Bx to 180/μL (IQR 0–415/μL) post-Bx (p < 0.0001) (Figure 2H). Similarly, median FeNO levels were markedly reduced from 55 ppb (IQR 31–87 ppb) to 26 ppb (IQR 16–46 ppb) (p < 0.001) (Figure 2I). Serum IgE levels remained unchanged after treatment (Figure 2J).

Display full size

Figure 2. Clinical outcomes and biomarkers before and after Bx treatment in patients with severe asthma. (A–D) Exacerbations and maintenance oral corticosteroid (mOCS) use. (A) Number of exacerbations during pre- and post-biologics (Bx) periods (Wilcoxon matched-pairs signed-rank test, ^****p < 0.0001). (B) Patient distribution by annual exacerbation frequency (Fisher’s exact test, ^****p < 0.0001). (C) mOCS dose pre- and post-Bx (Wilcoxon matched-pairs signed-rank test, ^***p < 0.001). (D) Patients requiring mOCS (McNemar’s test with continuity correction, ^*p < 0.05). (E–G) Lung function parameters. (E) Percent predicted forced expiratory volume in 1 second (FEV₁ %predicted values) pre- and post-Bx (mean ± SD; paired t-test, ^**p < 0.01). (F) Sankey diagram illustrating transitions in lung function status based on FEV₁ %predicted values (> 80%, 50–80%, and < 50% categories) pre- and post-Bx (Markov chain analysis with chi-square test, ^****p < 0.0001). (G) Patient distribution by absolute change in FEV₁ (ΔFEV₁) post-Bx (threshold: 200 mL). (H–J) Type 2 inflammation biomarkers. Pre- and post-Bx levels of (H) blood eosinophils (/μL), (I) fractional exhaled nitric oxide (FeNO, ppb), and (J) serum immunoglobulin E (IgE, IU/mL). Wilcoxon matched-pairs signed-rank test: ^****p < 0.0001, ^***p < 0.001. ns: not significant; FEV₁: forced expiratory volume in 1 second. Box plots show median, interquartile range, and whiskers (minimum to maximum)

Comparative clinical profiles and treatment outcomes of anti-IgE, anti-IL-5/IL-5R, and anti-IL-4R therapies in patients with severe asthma

Among the 53 patients analyzed, 19, 15, and 19 patients currently underwent anti-IgE, anti-IL-5/IL-5R, and anti-IL-4R therapy, respectively. Of the 15 patients receiving anti-IL-5/IL-5R therapy, 2 had previously been treated with anti-IgE. Similarly, among the 19 patients receiving anti-IL-4R therapy, 5 had a history of treatment with anti-IgE alone, while 2 had previously been treated with anti-IL-5/IL-5R. Additionally, 2 of the anti-IL-4R users had a history of treatment with both anti-IgE and anti-IL-5/IL-5R Bx. Baseline characteristics of patients receiving anti-IgE, anti-IL-5/IL-5R, or anti-IL-4R therapies were compared (Table 2). It should be noted that for patients with previous biologic exposure (n = 9), baseline clinical parameters represent measurements obtained immediately before starting their current biologic agent, ensuring comparable baseline assessments across all study participants. Patients in the anti-IL-5/IL-5R group had significantly higher pre-Bx BEC compared to both the anti-IgE and anti-IL-4R groups (median: 846 vs. 470 and 420 cells/μL, respectively; p < 0.05 and p < 0.01). Additionally, FeNO levels were significantly elevated in the anti-IL-5/IL-5R group compared to the anti-IgE group (median: 76.5 vs. 24.0 ppb, p = 0.0084). A higher proportion of patients in the anti-IgE and anti-IL-4R groups demonstrated the positivity of skin prick test (SPT) or the presence of specific IgE compared to the anti-IL-5/IL-5R group.

The clinical outcomes before and after each biologic therapy are shown in Figure 3. All three Bx significantly reduced annual exacerbation rates compared to baseline (Figure 3A). Among patients who were receiving mOCS prior to the initiation of Bx, mOCS dose also decreased (Figure 3B). mOCS dose between before and after anti-IgE therapy was not statistically evaluated because only 2 patients underwent mOCS in this group. Reductions in mOCS dose were observed even among patients who had switched Bx (red lines, Figure 3B). FEV₁ %predicted values improved significantly only in the anti-IL-4R group (Figure 3C). We conducted a sensitivity analysis by excluding patients who had received prior biologic treatments (n = 44). The clinical outcomes in this subgroup were very similar to those in the full cohort, suggesting that previous biologic use did not have a major impact on the response patterns observed (Table 3).

Display full size

Figure 3. Clinical outcomes before and after anti-IgE, anti-IL-5/IL-5R, or anti-IL-4R therapy in patients with severe asthma. (A) Number of exacerbations per year during pre- and post-anti-IgE (left), anti-IL-5/IL-5R (middle), and anti-IL-4R (right) therapies (Wilcoxon matched-pairs signed-rank test, ^****p < 0.0001, ^***p < 0.001). (B) Daily maintenance oral corticosteroid (mOCS) dose in patients treated with anti-IgE (left, n = 2), anti-IL-5/IL-5R (middle, n = 6), or anti-IL-4R (right, n = 5) therapies (paired t-test, ^**p < 0.01, ^*p < 0.05). Analysis was limited to patients receiving mOCS prior to biologic therapy. mOCS dose between before and after anti-IgE therapy was not statistically tested because only 2 patients underwent mOCS in this group. Points and lines in red represent patients who had previously received Bx different from the current therapy. (C) FEV₁ %predicted pre- and post-treatment of anti-IgE (left), anti-IL-5/IL-5R (middle), and anti-IL-4R therapies (paired t-test, ^*p < 0.05). ns: not significant; IgE: immunoglobulin E; IL: interleukin; IL-5R: IL-5 receptor; Bx: biologics; FEV₁ %predicted: percent predicted forced expiratory volume in 1 second

Identification of distinct patterns of clinical status and disease activity of patients treated with biologic therapy

A hierarchical clustering analysis was used to evaluate the heterogeneity in clinical status and disease activity of patients treated with Bx, based on 3 key outcome measures during the follow-up period: exacerbation frequency; mOCS dose; and FEV₁ %predicted values. We analyzed the clinical status and disease activity irrespective of the specific biologic agent, because we sought to identify common response patterns across different type2-targeted Bx. This clustering analysis revealed that patients were classified into three distinct control groups using the following criteria: well-controlled patients (n = 23) experienced fewer than 3 exacerbations, required no mOCS use, and achieved post-biologic FEV₁ more than 80% predicted; moderately controlled patients (n = 22) experienced fewer than 3 exacerbations and required no mOCS use, but did not achieve post-biologic FEV₁ more than 80% predicted; poorly controlled patients (n = 8) experienced 3 or more exacerbations or required daily mOCS during biologic therapy. Among the eight subjects with the poorly controlled cluster, two met only the exacerbation criterion, five met only the mOCS criterion, and one met both criteria (Figure 4A). Individual components of the disease control status were analyzed (Figure 4B–D). Exacerbation rates varied significantly between groups, with poorly controlled patients experiencing substantially higher exacerbation rates than well-controlled and moderately controlled patients (Figure 4B). The use of mOCS was markedly higher in the poorly controlled patients, indicating a greater dependence on systemic corticosteroids for their disease control (Figure 4C). Well-controlled patients demonstrated superior lung function, while moderately and poorly controlled patients exhibited similar levels of airflow limitation, both of which were significantly lower than those of well-controlled patients (Figure 4D). An analysis of lung function transition from pre-Bx to post-Bx revealed that 26.1% of well-controlled patients (n = 23) improved to achieve FEV₁ ≥ 80% predicted post-Bx (Figure 4E). The moderately controlled group (n = 22) exhibited heterogenous responses.

Display full size

Figure 4. Hierarchical clustering analysis of clinical responses in Bx-treated patients with severe asthma. (A) Hierarchical clustering heatmap stratifying patients into well-controlled, moderately controlled, and poorly controlled patients based on post-biologics (Bx) treatment outcomes. Color intensity represents the magnitude of three key clinical parameters: FEV₁ %predicted values (FEV₁; blue to red: 120% to 30%), maintenance oral corticosteroid (mOCS) dose (prednisolone equivalent, blue to red: 0 to 10 mg/day), and annual exacerbation frequency (Ex; blue to red: 0 to 6 events/year). Dendrogram shows the hierarchical relationship between patient clusters. (B–D) Post-Bx clinical characteristics across groups. Bar graphs showing (B) annual exacerbation frequency, (C) mOCS dose (prednisolone equivalent, mg per day), and (D) FEV₁ %predicted values post-Bx in well-controlled (W; red), moderately controlled (M; blue), and poorly controlled (P; green) patients. Data are presented as median with interquartile range. Statistical analysis was performed using negative binomial regression followed by Dunnett’s multiple comparison test for annual exacerbation frequency, and by Kruskal-Wallis test followed by Dunn’s multiple comparison test for mOCS dose and FEV₁ %predicted values. ^****p < 0.0001, ^***p < 0.001, ^*p < 0.05. ns: not significant. (E) Sankey diagrams illustrating transitions in lung function status based on FEV₁ %predicted values (> 80%, 50–80%, and < 50% categories) pre- and post-Bx in well-controlled, moderately controlled, and poorly controlled patients. FEV₁ %predicted: percent predicted forced expiratory volume in 1 second; FEV₁: forced expiratory volume in 1 second

Canonical correlation analysis was performed to identify predictors of the 3 distinct control groups using hierarchical clustering, and to compare baseline clinical parameters. This analysis clearly separated the 3 groups: lung function, BEC, and BMI pre-Bx distinguished well-controlled from moderately controlled patients, while baseline mOCS use was important in differentiating moderately from poorly controlled patients (Figure 5A). Examination of baseline characteristics revealed that the poorly controlled group exhibited higher baseline mOCS use and lower FEV₁ %predicted values than the other groups (Table 4). FEV₁ %predicted values, mOCS use, BEC, and BMI at baseline were identified as potential predictors for group classification. Multinomial logistic regression analysis using these variables as covariates revealed that FEV₁ %predicted values and BEC were independent predictors distinguishing well-controlled from moderately controlled patients (Table 5). We excluded FVC %predicted values and FEV₁/FVC to avoid the multicollinearity between these lung function parameters and FEV₁ %predicted values. While FEV₁ %predicted values demonstrated good discriminatory power (AUC = 0.73, p = 0.0095), BEC showed limited standalone predictive utility (AUC = 0.58, p = 0.3460) (Figure 5B). Using a cut-off of ≥ 90.0% FEV₁ %predicted values, well-controlled and moderately controlled patients were differentiated with a sensitivity of 59.1% and a specificity of 86.4%. ANCOVA was performed to determine the relationship between clinical outcomes and control group classification, adjusting for baseline values (Figure 5C). For exacerbation frequency, there was a significant effect of group differences (p = 0.0008), independent of pre-Bx exacerbation rates (p = 0.0548). Regarding mOCS dose, both groups (p = 0.0009) and pre-Bx mOCS dose (p < 0.0001) had significant effects, showing that post-Bx mOCS use was independently influenced by both pre-Bx dose and group classification. Similarly, for FEV₁ %predicted values, ANCOVA revealed significant effects for both group differences (p < 0.0001) and pre-Bx FEV₁ %predicted values (p < 0.0001). This suggests that post-Bx lung function was affected by baseline values, but also differed independently among the 3 control groups.

Display full size

Figure 5. Canonical correlation analysis in Bx-treated patients with severe asthma. (A) Canonical correlation analysis of clinical parameters and treatment response. The biplot displays the relationship between clinical variables and patient clusters, with vectors indicating the direction and strength of variable correlations. Ellipses represent 95% confidence regions for well-controlled (red), moderately controlled (blue), and poorly controlled (green) patients. Variables analyzed included blood eosinophil count (BEC), fractional exhaled nitric oxide (FeNO), body mass index (BMI), oral corticosteroid (OCS), and FEV₁ %predicted values (ppFEV₁). Statistical significance was determined by Wilks’ lambda test (λ = 0.14, p < 0.0001). (B) Receiver operating characteristic (ROC) curves comparing the predictive performance of FEV₁ %predicted (blue line) and BEC (red line). (C) Relationship between pre- and post-biologics (Bx) outcomes across groups stratified by hierarchical clustering analysis. Analysis of covariance (ANCOVA) was performed to evaluate exacerbation frequency (left), mOCS dose (middle), and FEV₁ %predicted values (right). Each model was adjusted for the corresponding pre-Bx values. The analysis demonstrated independent effects of group differences (exacerbation: p = 0.0008; mOCS: p = 0.0009; FEV₁: p < 0.0001) with significant effects of pre-Bx values for mOCS (p < 0.0001) and FEV₁ (p < 0.0001), but not for exacerbations (p = 0.0548) after Bx therapy. Regression lines show predicted relationships for well-controlled (red), moderately controlled (blue), and poorly controlled (green) patients, with parallel trajectories indicating consistent group differences across pre-Bx therapy values. Black dots represent individual patient data. FEV₁ %predicted: percent predicted forced expiratory volume in 1 second; mOCS: maintenance OCS; FEV₁: forced expiratory volume in 1 second; IgE: immunoglobulin E