Authors: Hyun Woo Lee, Kyungjoo Kim, Jung-Kyu Lee, Chin Kook Rhee, Kyung Hoon Min, Seong Yong Lim, Deog Kyeom Kim
Categories: Article, Asthma, Anti-bacterial agents, Drug-related side effects and adverse reactions, Symptom flare up, Mortality, Health care costs, Respiratory tract diseases, Epidemiology
Source: Scientific Reports
Authors: Hyun Woo Lee, Kyungjoo Kim, Jung-Kyu Lee, Chin Kook Rhee, Kyung Hoon Min, Seong Yong Lim, Deog Kyeom Kim
Asthma exacerbations contribute significantly to morbidity, mortality, and healthcare costs, with antibiotics commonly prescribed despite uncertain benefits and risks. This retrospective cohort study analyzed claim database between 2020 and 2022. Patients were categorized into extended (≥ 4 weeks) or limited (< 4 weeks) antibiotic exposure groups based on cumulative antibiotic use. We assessed the associations between prior antibiotic exposure and moderate-to-severe exacerbations, all-cause mortality, and total healthcare costs. We identified 549,425 adult asthma patients, classified into extended (25.2%) and limited (74.8%) antibiotic exposure groups. Extended antibiotic exposure was significantly associated with moderate-to-severe exacerbations (aOR = 1.25, 95% CI = 1.23–1.26) and nearly doubled the odds of all-cause mortality (aOR = 1.99, 95% CI = 1.92–2.07). Patients in the extended exposure group incurred significantly higher healthcare costs (median 7,433 vs. 5,471 thousand KRW, P < 0.001). A cumulative association pattern was observed, where longer durations of antibiotic use, greater cumulative doses, and exposure to multiple antibiotic classes were associated with progressively greater odds of exacerbations, mortality, and increased medical costs. Significant interactions revealed that rhinitis and sinusitis mitigated attenuated these associations, while bronchiectasis further strengthened them. Extended antibiotic use in adult asthma patients was linked to worse outcomes. However, as causality cannot be confirmed due to potential biases, further studies are needed.
The online version contains supplementary material available at 10.1038/s41598-025-12318-7.
Asthma is a prevalent chronic respiratory disorder characterized by persistent airway inflammation and bronchial hyperreactivity, leading to recurrent exacerbations that impose a substantial burden on both patients and healthcare systems. Exacerbations, defined as acute worsening events of asthma symptoms requiring additional treatment, represent a key determinant of accelerated lung function decline and are major contributors to increased morbidity, mortality, and healthcare costs in asthma^1,2^. Accordingly, early achievement of clinical remission to minimize the risk of exacerbations is increasingly being recognized as a key therapeutic goal in asthma management^3^. While clinical remission of asthma itself is important, it is also noteworthy that exacerbations are often precipitated by external factors such as respiratory infections, which are considered significant triggers of exacerbations^4^. This has led to the widespread and more frequent use of antibiotics in patients with asthma compared to those without asthma^5,6^, although clinical benefit of antibiotic treatment has not been clarified in managing acute exacerbations^7,8^.
Prior antibiotic exposure may exacerbate the pathological processes of asthma. A meta-analysis reported that antibiotic use in children doubled the risk of developing asthma^9^. Especially, higher risks were observed in those receiving multiple prescriptions or broad-spectrum antibiotics in children with asthma^10–13^. In adults, antibiotic exposure has also been associated with an increased risk of asthma development, with specific subgroups showing elevated risks following exposure to quinolones, cephalosporins, or macrolides^14^. Moreover, in adults, the cumulative dose and number of antibiotic classes used over the preceding five years were positively correlated with the risk of developing asthma, particularly with the use of penicillins, cephalosporins, fluoroquinolones, and macrolides^15^. Despite these findings, there remains a lack of data on the temporal associations between prior antibiotic use and subsequent clinical outcomes in adult asthma patients.
We aimed to evaluate the associations between previous exposure to antibiotics and clinical outcomes, including moderate-to-severe exacerbations, all-cause mortality, and total medical costs. Furthermore, we examined whether these associations varied according to the duration and type of antibiotic exposure, as well as the presence of comorbidities and different asthma treatment regimens, serving as indirect indicators of disease severity.
Our retrospective cohort study utilized data from the Korean Health Insurance Review and Assessment service (HIRA) database, which covers the general population of South Korea with a three-year follow-up period from January 2020 to December 2022. Eligible patients met an operational definition of asthma during the period from January 2020 to December 2020. Specifically, asthma was defined by meeting all of the following (1) a primary or a secondary diagnosis of asthma with ICD-10 codes J45 or J46, (2) age ≥ 15 years, and (3) evidence of using at least one asthma medication on two or more occasions per year. This operational definition combines diagnostic and treatment information to enhance specificity and reflect clinical practice^16^. While formal validation studies are limited, this operational definition has been consistently applied in prior studies using the same data source^17,18^. The medications considered included long-acting muscarinic antagonists (LAMA), long-acting beta-2 agonists (LABA), inhaled corticosteroids (ICS), ICS/LABA combinations, short-acting muscarinic antagonists (SAMA), short-acting beta-2 agonists (SABA), SAMA/SABA combinations, theophylline, leukotriene receptor antagonists, systemic corticosteroids, and systemic beta agonists.
Patients were categorized based on the total number of days of antibiotic use during the previous year. We defined extended exposure as at least 4 weeks of cumulative antibiotic use and limited exposure as less than 4 weeks. The list of specific antibiotics analyzed in this study is summarized in Supplementary Table 1.
Baseline characteristics included age, sex, insurance type, type of hospital, and previous history of pneumonia. Age was analyzed in the following 15–24, 25–34, 35–44, 45–54, 55–64, 65–74, and ≥ 75 years. Insurance types were categorized into health insurance and medical aid, and hospital types were classified as primary, secondary, or tertiary. Pneumonia history included outpatient clinic visits or hospitalizations, hospitalizations alone, and the length of stay in hospital. The operational definition of pneumonia was established based on the following (1) ICD-10 codes indicative of pneumonia, (2) diagnostic imaging codes for chest X-ray or computed tomography, and (3) prescription of antibiotics following the index date. Among the comorbidities, sinusitis (identified by ICD-10 code J32 with at least two documentations within a year), rhinitis (identified by ICD-10 codes J30 or J31 with at least two documentations within a year), and bronchiectasis (identified by ICD-10 code J47 with at least one documentation within a year) were specifically collected as comorbidities of interest in this study. Other comorbidities were assessed using the Charlson Comorbidity Index (CCI), categorized as ≤ 1, 2–3, and > 3. These comorbidities included myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular accident or transient ischemic attack, COPD, connective tissue disease, peptic ulcer disease, liver disease, diabetes mellitus, hemiplegia, chronic kidney disease, malignancy, leukemia, and lymphoma.
Inhaled treatments were grouped into SABA monotherapy, ICS monotherapy, ICS/LABA, and ICS/LABA + LAMA. Oral medications included methylxanthine or methylxanthine derivatives, leukotriene receptor antagonists, and oral beta-agonists. Systemic corticosteroids were documented as use of oral corticosteroids and cumulative oral corticosteroid dose, expressed as the prednisolone-equivalent dose. Antibiotic-related variables included type of antibiotic class (fluoroquinolones, β-lactam antibiotics, and macrolides), course of antibiotics, number of antibiotic classes, and duration of antibiotic use. The course of antibiotics refers to a distinct event in which an antibiotic was administered. The number of antibiotic classes indicates the extent of exposure to different antibiotic classes among the three specified.
Our study outcomes were moderate-to-severe exacerbations, all-cause mortality, and total medical cost during a 2-year period from January 2021 to December 2022. Moderate exacerbations were defined as outpatient department visits accompanied by additional treatment, such as antibiotics or systemic steroids. Severe exacerbations were characterized by emergency room visits or hospitalizations, also requiring additional treatment with antibiotics or systemic steroids.
The associations between antibiotic exposure and the clinical outcomes (acute exacerbation, all-cause mortality, and total medical cost) were evaluated across subgroups based on the profile of antibiotic use, including the type of antibiotic class, the course of antibiotic therapy, the number of antibiotic classes, and the duration of antibiotic use (categorized as 4–8 weeks, 8–12 weeks, 12–24 weeks, and ≥ 24 weeks). Interaction analyses were conducted according to specific comorbidities (rhinitis, sinusitis, and bronchiectasis) and inhaled treatments for asthma, serving as an indirect indicator of asthma severity. Sensitivity analyses were performed to assess whether exposure to each specific antibiotic drug had a significant association with the clinical outcomes.
Categorical variables were analyzed using Pearson’s chi-square test, while continuous variables were analyzed using t-tests. We used logistic or linear regression model to estimate 2-year outcome risks because all patients had a fixed follow-up period. This approach avoids immortal time bias that could occur if baseline timing differed by exposure status. Univariable and multivariable logistic or linear regression analyses were performed to assess the risk of clinical outcomes associated with antibiotic exposure. The multivariable logistic regression models were adjusted for covariates, including age, sex, insurance type, hospital type, prior history of pneumonia, comorbidities (rhinitis, sinusitis, and bronchiectasis), CCI, and asthma treatments. Relevant adjustment variables were selected based on a directed acyclic graph (DAG), constructed from a comprehensive literature review (Supplementary Fig. 1). A P-value of < 0.05 was considered statistically significant. All statistical analyses were conducted using SAS Enterprise Guide 7.1. (SAS Institute Inc., Cary, NC, USA).
Of the 2,742,087 individuals screened for any asthma-related diagnostic or treatment codes in the HIRA database, 954,661 were identified with an ICD-10 diagnosis code for asthma (J45–J46) and documented receipt of at least two prescriptions for asthma treatment. Patients younger than 15 years of age (n = 405,236) were excluded from the analysis. The final study population consisted of 549,425 asthma patients aged ≥ 15 years who had received at least two asthma treatment prescriptions. A total of 138,586 (25.2%) patients with antibiotic use for ≥ 4 weeks were categorized into the extended exposure group, while 410,839 (74.8%) patients with < 4 weeks of use or no antibiotic exposure were classified into the limited exposure group (Supplementary Fig. 2).
Baseline characteristics of the eligible patients are summarized in Table 1. The extended exposure to antibiotics group were older and had a lower percentage of males compared to the limited exposure group. The extended exposure group had a higher prevalence of medical aid recipients and more frequent visits to tertiary hospitals. A greater proportion of patients in the extended exposure group had a history of pneumonia, both in terms of outpatient visits and hospitalizations, and they also experienced longer hospital stays. Comorbidities such as sinusitis, rhinitis, and bronchiectasis were more prevalent in the extended exposure group, and these patients had higher CCI scores.
Table 1Baseline characteristics according to exposure to antibiotics.Total(N = 549,425)Extended exposure to antibiotics (n = 138,586)Limited exposure to antibiotics (n = 410,839)P-valueAge, years, mean (± SD)60.5 ± 18.061.6 ± 17.660.1 ± 18.1< 0.001 15–24, n (%)23,448 (4.3)4,827 (3.5)18,621 (4.5)< 0.001 25–34, n (%)35,876 (6.5)7,472 (5.4)28,404 (6.9)< 0.001 35–44, n (%)51,511 (9.4)13,598 (9.8)37,913 (9.2)< 0.001 45–54, n (%)69,997 (12.7)16,329 (11.8)53,668 (13.1)< 0.001 55–64, n (%)106,511 (19.4)26,438 (19.1)80,073 (19.5)< 0.001 65–74, n (%)125,306 (22.8)33,122 (23.9)91,914 (22.4)< 0.001 ≥75, n (%)137,046 (24.9)36,800 (26.6)100,246 (24.4)< 0.001Sex, male, n (%)259,474 (47.2)64,176 (46.3)195,298 (47.5)< 0.001Insurance type Health insurance, n (%)485,599 (88.4)118,981 (85.9)366,618 (89.2)< 0.001 Medical aid^a^, n (%)63,826 (11.6)19,605 (14.1)44,221 (10.8)< 0.001 Type of hospital Primary, n (%)56,690 (10.3)10,053 (7.3)46,637 (11.3)< 0.001 Secondary, n (%)52,002 (9.5)12,633 (9.1)39,369 (9.6)< 0.001 Tertiary, n (%)440,733 (80.2)115,900 (83.6)324,833 (79.1)< 0.001History of pneumonia Outpatient clinic visit or hospitalization, n (%)15,107 (2.8)11,829 (8.5)3278 (0.8)< 0.001 Hospitalization, n (%)7,935 (1.4)6,697 (4.8)1,238 (0.3)< 0.001 Length of stay in hospital, days, median (IQR)12 (7–21)13 (8–22)8 (5–13)< 0.001Comorbidities Sinusitis, n (%)96,134 (17.5)33,745 (24.4)62,389 (15.2)< 0.001 Rhinitis, n (%)467,547 (85.1)121,696 (87.8)345,851 (84.2)< 0.001 Bronchiectasis, n (%)32,000 (5.8)11,772 (8.5)20,228 (4.9)< 0.001Charlson comorbidity index ≤1, n (%)149,336 (27.2)31,383 (22.6)117,953 (28.7)< 0.001 2 ~ 3, n (%)218,159 (39.7)53,595 (38.7)164,564 (40.1)< 0.001 >3, n (%)181,930 (33.1)53,608 (38.7)128,322 (31.2)< 0.001Other comorbidities Myocardial infarction, n (%)89,489 (16.3)25,331 (18.3)64,158 (15.6)< 0.001 Congestive heart failure, n (%)59,656 (10.9)17,482 (12.6)42,174 (10.3)< 0.001 Peripheral vascular disease, n (%)100,153 (18.2)28,635 (20.7)71,518 (17.4)< 0.001 Cerebrovascualar accident or transient ischemic attack, n (%)47,428 (8.6)13,593 (9.8)33,835 (8.2)< 0.001 COPD, n (%)156,044 (28.4)45,235 (32.6)110,809 (27.0)< 0.001 Connective tissue disease, n (%)1,280 (0.2)391 (0.3)889 (0.2)< 0.001 Peptic ulcer disease, n (%)404,671 (73.7)110,216 (79.5)294,455 (71.7)< 0.001 Liver disease, n (%)60,414 (11.0)17,066 (12.3)43,348 (10.6)< 0.001 Diabetes mellitus, n (%)176,027 (32.0)49,156 (35.5)126,871 (30.9)< 0.001 Hemiplegia, n (%)5,465 (1.0)1,774 (1.3)3,691 (0.9)< 0.001 Chronic kidney disease, n (%)21,278 (3.9)6,427 (4.6)14,851 (3.6)< 0.001 Malignancy, n (%)57,003 (10.4)17,571 (12.7)39,433 (9.6)< 0.001 Leukemia, n (%)906 (0.2)394 (0.3)512 (0.1)< 0.001 Lymphoma, n (%)1,065 (0.2)368 (0.3)697 (0.2)< 0.001^a^Medical aid serves economically vulnerable populations, including recipients of the National Basic Livelihood Security Program, individuals covered by specific legal mandates, and those experiencing homelessness.COPD, chronic obstructive pulmonary disease; IQR, interquartile range; SD, standard deviation.
In the analyses of specific antibiotics, extended exposure to antibiotics was consistently associated with higher age, greater prevalence of comorbidities, and increased use of medical aid and tertiary hospital services (Supplementary Table 2 ).
Patients with comorbidities such as rhinitis, sinusitis, and bronchiectasis had longer durations of antibiotic use across various classes (Supplementary Table 3). Bronchiectasis patients had a higher exposure to broad-spectrum antibiotics (levofloxacin, moxifloxacin, Piperacillin + β-lactamase, and third-generation and anti-pseudomonal cephalosporins) compared to those with rhinitis or sinusitis. Similarly, patients with sinusitis showed higher use of antibiotics, particularly for β-lactam antibiotics and macrolides, than those with rhinitis.
The extended antibiotic exposure group was more likely to receive inhaled treatments, such as ICS/LABA + LAMA, and oral medications, including methylxanthines, leukotriene receptor antagonists, and oral beta-agonists (Table 2**).** In contrast, ICS/LABA was more commonly used in the limited antibiotic exposure group. Additionally, the extended exposure group had a higher proportion of patients using systemic corticosteroids and received significantly greater cumulative doses of oral corticosteroids compared to the limited exposure group. These patterns were consistent across different antibiotic classes, including fluoroquinolones, β-lactam antibiotics, and macrolides (Supplementary Table 4). Patients receiving ICS/LABA + LAMA combinations had longer median durations of antibiotic use compared to those on SABA, ICS monotherapy, or ICS/LABA (Supplementary Table 5). They had higher rates of antibiotic use, including specific drugs like levofloxacin, moxifloxacin, broad-spectrum β-lactam antibiotics, and azithromycin.
Table 2Asthma treatments according to antibiotic exposure.Total(N = 549,425)Extended exposure to antibiotics (n = 138,586)Limited exposure to antibiotics (n = 410,839)P-valueInhaled treatments SABA monotherapy, n (%)52,854 (9.6)13,178 (9.5)39,676 (9.7)0.105 ICS monotherapy, n (%)18,574 (3.4)5,016 (3.6)13,558 (3.3)< 0.001 ICS/LABA, n (%)364,041 (66.3)79,445 (57.3)284,596 (69.3)< 0.001 ICS/LABA + LAMA, n (%)28,927 (5.3)9,323 (6.7)19,604 (4.8)< 0.001Oral medications Methylxanthine or methylxanthine derivative, n (%)215,087 (39.2)58,582 (42.3)156,505 (38.1)< 0.001 Leukotriene receptor antagonist, n (%)304,154 (55.4)82,664 (59.7)221,490 (53.9)< 0.001 Oral beta-agonist, n (%)137,264 (25.0)40,017 (28.9)97,247 (23.7)< 0.001Systemic corticosteroids Use of oral corticosteroid, n (%)305,541 (55.6)102,385 (73.9)203,156 (49.5)< 0.001 Cumulative oral corticosteroid^a^, mg, mean (± SD)383 ± 937499 ± 1104325 ± 834< 0.001^a^Only individuals who received oral corticosteroids were included. The cumulative dose of oral corticosteroids was calculated as prednisolone-equivalent doses (mg) over the past year.ICS, inhaled corticosteroids; LABA, long-acting beta-2 agonists; LAMA, long-acting muscarinic antagonists; SABA, short-acting beta-2 agonists; SD, standard deviation.
During the two-year observation period, 149,036 asthma patients (27.1%) experienced at least one moderate-to-severe exacerbation. The extended exposure group had significantly higher rates of moderate, severe, and overall moderate-to-severe exacerbations compared to the limited exposure group (P-value < 0.001, Table 3). This trend was observed consistently across individual antibiotic classes (Supplementary Table 6).
Table 3Clinical outcomes according to antibiotic exposure.Extended exposure to antibiotics (n = 138,586)Limited exposure to antibiotics (n = 410,839)P-valueAcute exacerbation Moderate, n (%)36,995 (26.7)93,434 (22.7)< 0.001 Severe, n (%)11,396 (8.2)20,917 (5.1)< 0.001 Moderate-to-severe, n (%)43,194 (31.2)105,842 (25.8)< 0.001All-cause mortality, n (%)5,757 (4.2)7,658 (1.9)< 0.001Medical cost Total cost, 1000 Korean Won, median [IQR]7,433 (3,308–18,794)5,471 (2,442–13,401)< 0.001 Outpatient cost, 1000 Korean Won, median [IQR]2,596 (1,397–4,586)1,917 (990–3,453)< 0.001 Inpatient cost, 1000 Korean Won, median [IQR]4,837 (1,911–14,208)3,554 (1,452–9,948)< 0.001 Medication cost, 1000 Korean Won, median [IQR]4,428 (1,921–10,976)2,783 (1,209–6,728)< 0.001IQR, interquartile range.
In the multivariable logistic regression model, extended antibiotic exposure was associated with an increased adjusted odds ratio (aOR) for moderate-to-severe exacerbations in asthma patients (aOR = 1.25 [95% CI = 1.23–1.26], Fig. 1). Among the antibiotic classes, fluoroquinolones had the highest aOR (1.32 [95% CI = 1.28–1.37]), followed by β-lactam antibiotics and macrolides. The odds of exacerbations increased with the number of antibiotic courses and classes prescribed, with the highest odds observed in patients receiving four or more courses (aOR = 1.30 [95% CI = 1.28–1.32]) or those exposed to multiple antibiotic classes (aOR = 1.49 [95% CI = 1.44–1.54]). Longer durations of antibiotic exposure, particularly ≥ 24 weeks, were also significantly associated with higher odds of exacerbations (aOR = 1.51 [95% CI = 1.46–1.56]).
Fig. 1Adjusted risk of moderate-to-severe exacerbations according to antibiotic exposure. Odds ratio for moderate-to-severe exacerbations was evaluated using a multivariable logistic regression model, adjusting for covariates, including age, sex, insurance type, hospital type, prior history of pneumonia, comorbidities (rhinitis, sinusitis, bronchiectasis), Charlson Comorbidity Index, and asthma treatments. aOR, adjusted odds ratio.
In interaction analyses, the presence of rhinitis or sinusitis attenuated the association between extended antibiotic exposure and moderate-to-severe exacerbations, whereas bronchiectasis strengthened this association (Supplementary Table7). Conversely, the adjusted odds of moderate-to-severe exacerbations associated with extended antibiotic exposure were significantly higher in patients receiving inhaled asthma treatments, with the greatest odds observed in those on ICS/LABA + LAMA therapy.
In sensitivity analyses, all specific antibiotics within the three classes were consistently associated with higher odds of moderate-to-severe exacerbations (Supplementary Table 8). Among these, moxifloxacin showed the strongest association (aOR = 1.25 [95% CI = 1.19–1.31]), whereas clarithromycin had the weakest association (aOR = 1.03 [95% CI = 1.01–1.04]).
Over the two-year follow-up, patients in the extended antibiotic exposure group had significantly higher all-cause mortality rates compared to those in the limited exposure group (4.2% vs. 1.9%, P-value < 0.001, Table 3). Similar results were found in the patients exposed to fluoroquinolones and β-lactam antibiotics, but not macrolides (Supplementary Table 6).
In the multivariable analyses, extended antibiotic exposure was associated with nearly double the odds of all-cause mortality (aOR = 1.99 [95% CI = 1.92–2.07], Fig. 2). Among antibiotic classes, β-lactam antibiotics showed the highest aOR (2.09 [95% CI = 2.00–2.18]), followed by fluoroquinolones and macrolides. The odds of mortality increased with the number of antibiotic courses and classes, with the greatest odds observed in patients receiving four or more courses (aOR = 2.25 [95% CI = 2.15–2.37]) or exposed to multiple antibiotic classes (aOR = 3.16 [95% CI = 2.92–3.41]). Longer durations of antibiotic use, particularly ≥ 24 weeks, were associated with higher odds of mortality (aOR = 3.31 [95% CI = 3.09–3.55]).
Fig. 2Adjusted risk of all-cause mortality according to antibiotic exposure. Odds ratio for all-cause mortality was evaluated using a multivariable logistic regression model, adjusting for covariates, including age, sex, insurance type, hospital type, prior history of pneumonia, comorbidities (rhinitis, sinusitis, bronchiectasis), Charlson Comorbidity Index, and asthma treatments. aOR, adjusted odds ratio.
Interaction analyses showed that rhinitis or sinusitis attenuated the association between extended antibiotic exposure and all-cause mortality linked to extended antibiotic exposure (Supplementary Table 9). Conversely, the adjusted odds of all-cause mortality associated with extended antibiotic exposure were higher in patients receiving inhaled asthma treatments, with the highest odds observed in those on ICS monotherapy.
In sensitivity analyses, extended exposure to specific antibiotics was associated with varying odds of all-cause mortality (Supplementary Table 10). Among fluoroquinolones, moxifloxacin showed the strongest association. Piperacillin + β-lactamase had the strongest association among β-lactam antibiotics. For macrolides, azithromycin was associated with higher odds of mortality, while clarithromycin was associated with to lower odds.
For two years, patients in the extended exposure group incurred significantly higher total medical costs compared to those in the limited exposure group (median 7,433 vs. 5,471 thousand Korean Won, P-value < 0.001, Table 3). This cost disparity was consistent across inpatient, outpatient, and medication costs, with significantly higher expenditures in the extended exposure group. This trend was found consistently across different antibiotic classes (Supplementary Table 6).
In the multivariable linear regression model, extended antibiotic exposure was associated with a significant increase in total medical costs (adjusted β = 2,202 [95% CI = 2,109–2,294], Fig. 3). Among antibiotic classes, fluoroquinolones had the highest associated cost (adjusted β = 3,461 [95% CI = 3,227–3,695]), followed by β-lactam antibiotics. Macrolides, in contrast, were linked to a slight reduction in costs. The number of antibiotic courses and classes was associated with increased medical costs, with the highest costs observed in patients receiving four or more courses (adjusted β = 2,310 [95% CI = 2,199–2,420]) or exposed to multiple antibiotic classes (adjusted β = 3,470 [95% CI = 3,240–3,701]). Patients exposed to antibiotics for ≥ 24 weeks had the most significant increase in costs (adjusted β = 5,931 [95% CI = 5,706–6,156]).
Fig. 3Adjusted risk for increased total medical costs according to antibiotic exposure. β-coefficient for total medical cost was evaluated using a multivariable linear regression model, adjusting for covariates, including age, sex, insurance type, hospital type, prior history of pneumonia, comorbidities (rhinitis, sinusitis, bronchiectasis), Charlson Comorbidity Index, and asthma treatments.
In interaction analyses, rhinitis or sinusitis alleviated the increased medical costs associated with extended antibiotic exposure, whereas bronchiectasis worsened them (Supplementary Table 11). In contrast, the adjusted β for total medical costs due to extended antibiotic exposure was higher in patients receiving inhaled asthma treatments, with the highest adjusted β observed in those with ICS/LABA + LAMA.
In sensitivity analyses, the highest medical costs were associated with extended exposure to piperacillin + β-lactamase (adjusted β = 8,990 [95% CI = 8,745–9,235]) and anti-pseudomonal cephalosporins, while macrolides, such as clarithromycin and roxithromycin, were associated with reduced costs **(**Supplementary Table 12).
Our study showed that extended antibiotic exposure in adult asthma patients is associated with significantly worse clinical outcomes, including higher odds of moderate-to-severe exacerbations, greater odds of mortality, and elevated total medical costs. Our analysis identified a cumulative association pattern, where longer durations of antibiotic use, greater cumulative doses, and exposure to multiple antibiotic classes were associated with progressively higher odds, mortality, and increased medical costs. In asthma patients, the impact of extended antibiotic exposure on outcomes was influenced by comorbidities, with coexisting bronchiectasis associated with greater odds of exacerbations and higher total medical costs. Furthermore, extended antibiotic exposure interacted with inhaled therapies, likely reflecting both disease severity and the protective effects of these treatments, with varying associations for exacerbation, mortality, and total medical costs depending on the specific therapy. Among specific antibiotics, moxifloxacin was most strongly associated with moderate-to-severe exacerbations, while piperacillin combined with β-lactamase inhibitors was most strongly associated with all-cause mortality. Although macrolides were associated with increased odds of moderate-to-severe exacerbations, they were associated with lower total medical costs. Notably, azithromycin was associated with higher odds of all-cause mortality, whereas clarithromycin was associated with lower odds. These findings underscore the need for a more cautious approach to antibiotic use in asthma management, particularly when prolonged, repeated, or combination therapies are considered.
The impact of prior exposure to antibiotics on the clinical course of asthma patients following exacerbation events remains poorly understood. A systematic review and meta-analysis suggested that antibiotics may improve symptom scores and peak expiratory flow, but significant heterogeneity across studies limits the generalizability of these findings^7^. In infective asthma exacerbations, a 7-day course of amoxicillin was associated with a reduction in repeated antibiotic prescriptions for lower respiratory tract infections within 14 days, though long-term outcomes were not examined^19^. In a randomized controlled trial examining procalcitonin-guided antibiotic use during asthma exacerbations, there was no reduction in hospital length of stay, nor was there a difference in the incidence of severe exacerbations during 12 months of follow-up between those who received antibiotics and those who did not^20^. Additionally, two observational studies on pediatric asthma patients found no significant association between early-life antibiotic exposure and the risk of asthma exacerbations. However, a noteworthy finding in the PACMAN database was a trend toward increased asthma exacerbations with antibiotic use for more than 30 days, although this did not reach statistical significance^21^. In our study, we set the threshold for antibiotic exposure at four weeks, targeting a more intensively exposed adult asthma population compared to prior studies. Notably, we found a significant association between increased moderate-to-severe exacerbations and longer durations, more frequent courses, and exposure to multiple antibiotic classes, a pattern similar to the association between antibiotic exposure and asthma development^15^. These findings suggest that unnecessary antibiotic use in adult asthma patients may be harmful in the long term, potentially worsening asthma control.
The observed associations between extended antibiotic exposure and worse clinical outcomes in asthma may be explained by several potential biological mechanisms. One prominent hypothesis involves the disruption of the gut and airway microbiome, particularly by broad-spectrum antibiotics^22^. Gastrointestinal adverse events are commonly observed in patients receiving prolonged antibiotics, which may reflect alterations in the gut microbiome as a key underlying mechanism^23,24^. Antibiotic-induced alterations in microbial diversity can lead to dysbiosis, impairing the body’s immune regulation and promoting a T2 inflammatory environment, which may exacerbate airway inflammation and increase susceptibility to exacerbations^25^. In the prospective CHILD birth cohort, infants who developed asthma showed decreased gut α-diversity, and early-life exposure to antibiotics was associated with reduced gut α-diversity, suggesting that antibiotic use during infancy may contribute to the development of asthma by altering the gut microbial community^26^. However, the gut microbiome did not significantly influence the transcriptomic profile of peripheral blood mononuclear cells in asthma patients, highlighting the need for further investigation into the role of the gut microbiome in asthma pathogenesis^27^. Another consideration is the potential for antibiotics to facilitate secondary infections, as repeated or extended exposure can lead to antibiotic resistance, necessitating more aggressive treatment and potentially contributing to more severe exacerbations^28^. This mechanism aligns with the cumulative associations observed in our study. Therefore, it is speculated that extended antibiotic use may contribute to adverse clinical outcomes by exacerbating T2 inflammation due to gut microbial dysbiosis and by increasing antimicrobial resistance.
An intriguing and paradoxical finding in our study was that while macrolides were associated with increased odds of moderate-to-severe exacerbations, they were linked to a reduction in total medical costs. These findings could be attributed to the relatively lower cost of macrolides compared to other antibiotic classes and their frequent use in outpatient settings, where hospitalization and more intensive treatments may be avoided. Furthermore, exposure to azithromycin was associated with higher odds of all-cause mortality, whereas exposure to clarithromycin was associated with lower odds. This difference may be explained by the distinct clinical uses of these macrolides. Azithromycin is used long-term in patients with poorly controlled chronic airway diseases, including severe asthma, due to its immunomodulatory properties, whereas clarithromycin is primarily used for pulmonary infections, with less frequent long-term use in chronic airway diseases. Thus, the use of azithromycin may indicate higher baseline severity of asthma compared to clarithromycin. Additionally, prolonged macrolide use may contribute to antibiotic resistance and adverse clinical events, potentially explaining the higher mortality observed in association with azithromycin exposure.
This study has several limitations. First, as this study was a retrospective analysis using claims data, we cannot establish a causal relationship between antibiotic exposure and clinical outcomes. In particular, we did not perform matching on key confounding variables, such as demographic factors, comorbidities, or baseline treatment characteristics. Although we adjusted for multiple confounders based on a DAG, we did not perform matching on key variables such as demographics, comorbidities, or baseline treatments, which may have introduced selection bias and unmeasured confounding. Additionally, residual factors such as asthma severity and physician prescribing patterns could have contributed to indication bias, particularly in comparisons between antibiotic classes. Therefore, our findings should be interpreted as associative rather than causal, and with caution. Second, the use of claims data lacks detailed clinical information, such as microbiological and inflammatory markers, potentially leading to misclassification of asthma severity and comorbidities. In addition, the accuracy of claims-based definitions for asthma and exacerbations has not been formally validated in the HIRA database. To address this, we adopted operational definitions that have been widely used in prior population-level studies. While this approach enhances the practical utility of large-scale data, we acknowledge that some degree of misclassification may persist, which could have influenced our findings. Third, the study’s focus on a single-country population may restrict generalizability to other settings with different healthcare practices. Fourth, we did not explore the microbiome or antibiotic resistance, which could help clarify the biological mechanisms underlying the observed associations. Finally, our study did not apply a time-to-event model such as the Cox proportional hazards model. We used logistic and linear regression models to evaluate outcomes over a two-year period because all patients shared a fixed one-year exposure window and a uniform two-year follow-up period. This design ensured temporal consistency and comparability between groups, allowing us to avoid potential biases such as immortal time bias.
In conclusion, our study indicates that extended antibiotic exposure in adult asthma patients was associated with higher odds of exacerbations, increased mortality, and greater healthcare costs. The cumulative effect of longer antibiotic duration, exposure to multiple antibiotic classes, and frequent courses was associated with progressively greater odds of adverse clinical and economic outcomes. However, given the inherent limitations of our observational design, including residual confounding and potential indication bias, a definitive causal relationship cannot be established. These findings underscore the need for careful antibiotic prescribing while emphasizing the necessity of further prospective studies to validate these associations.
Below is the link to the electronic supplementary material.
Supplementary Material 1
Supplementary Material 2