Data were obtained from the National Health and Nutrition Examination Survey (NHANES), a cross-sectional survey of a representative US population, which is led by the American National Centers for Health Statistics (NCHS) and the Centers for Disease Control and Prevention (CDC), and has been conducted every 2 years since 1999 (16). NHANES researchers use a complex multi-stage sampling design to obtain a representative, non-institutionalized sample of the population of the United States (17). The survey comprised face-to-face in-home interviews and physical examinations. Further details of the survey design and data files were available on NHANES website (16, 17).

We included participants from 10 survey cycles of NHANES, between 1999 and 2018, and excluded those who were aged < 20 years (n = 44,342), did not complete the prescription medications questionnaire (n = 109) and the medical conditions questionnaire regarding asthma status (n = 3). In addition, female participants who were pregnant (n = 1,216) were also excluded. A total of 51,124 individuals were included in the study, with 25,170 (49.23%) being male and 25,954 (50.77%) being female. The detailed enrolment procedure is shown in Figure 1.

The protocols of NHANES were approved by NCHS Ethics Review Board, CDC (18), and written informed consent was obtained from all participants.

During the in-home interviews, interviewers asked participants if they had taken any prescription medication during the preceding 30 days and exactly how they had administered it. If the participant replied in the affirmative, the interviewer asked them to provide the containers for the prescription medicines, and the names of the antibiotics, the durations of their use, and the reasons for each use were recorded. This information was reported by the participant when the containers were not available. The reported drug names were converted to standard generic names and encoded using the Multum Lexicon database (19). The drugs were placed into one of three categories according to the Multum Lexicon Plus therapeutic classification system, which is used to assign a therapeutic classification to each drug and its ingredients. In the present study, we focused on drugs with an antibiotic activity, which included those in the first-level category “anti-infective,” and the second-level category, which includes penicillin, cephalosporins, macrolide derivatives, quinolones, sulfonamides, tetracyclines, lincomycin derivatives, urological anti-infectives, and other antibiotics listed in the Multum Lexicon database. Because some of the drugs were less commonly used, the final four of the groups listed above were classified as “miscellaneous antibiotics” for the analysis. Non-antibiotics such as antiviral, antifungal, antituberculosis, anti-leprosy drugs, and topical antibiotics including those in ear and eye preparations, were excluded. Using penicillin, cephalosporins, macrolide derivatives, quinolones, sulfonamides, miscellaneous antibiotics during the preceding 30 days and those did not use antibiotics were regarded as multi-classified variable.

The information regarding asthma and related symptoms collected by NHANES through a self-administered questionnaires, which were completed either during the in-home interview or clinic visit. Asthma was defined by respondents giving a positive response to either of three following questions: “Has a doctor or other health professional ever told you that you have asthma?,” “In the past 12 months have you had wheezing or whistling in your chest?” and “During the past 12 months, have you had an episode of asthma or an asthma attack?” (20).

Data regarding covariates were obtained through interviews conducted at home and physical examinations performed at the MECs. Age, sex, ethnicity, household socioeconomic status, educational level, health insurance coverage status, smoking status, alcohol consumption status, history of diseases (including infection, chronic bronchitis and emphysema), and family history of asthma were self-reported by the participants during the interview. Participants were categorized according to age as adult (age 20–40 years), middle-aged (aged 40–60 years), and older (aged > 60 years). Ethnicity was categorized as Non-Hispanic white people, Non-Hispanic black people, Hispanic people, and other. Household socioeconomic status was assessed using the family income-to-poverty ratio (family PIR), which was calculated by dividing household income by the poverty threshold for the survey year (21). This result varied according to family size and geographic location (22). The PIR was used to define three categories of socioeconomic status: low (PIR ≤ 1.3), medium (1.3 < PIR ≤ 3.5), and high (PIR > 3.5). Educational level was classified as pre-high school, high school graduate or general educational development (GED), and college or above. A participant was defined as “Covered by Health Insurance” if they had health insurance or another type of healthcare plan, such as health insurance obtained through their employment or purchased directly, or government programs, such as Medicare and Medicaid, which provide medical care or help with medical bills. The participants were also categorized as being current smokers, former smokers, or never having been a smoker of cigarettes. The alcohol consumption of a participant was judged by their response to the question “Do you have at least 12 alcoholic beverages per year?.” A standard drink was considered to be 12 fl. oz. of beer, 5 fl. oz. of wine, or 1.5 fl. oz. of liquor (23), and a response in the affirmative led to the participant being recorded as “Having a history of alcohol consumption.” A history of infection was identified using the participant’s self-reported current health status, and individuals were recorded as “having a history of infection” if they had experienced influenza, pneumonia, or ear infection that started during the preceding 30 days. Body mass index (BMI) was calculated as body mass (kg)/height (m) squared. Height and body mass were measured at the MECs according to a standard operating procedure. Individuals with BMI ≥ 25 kg/m² were recorded as being “Overweight or Obese” and those with BMI ≥ 18.5 kg/m² and < 25 kg/m² were recorded as being of “Normal weight.”

Statistical analyses were performed using the 20-year sampling weights associated with the MECs, according to the NCHS Analytic Guidelines, unless otherwise specified (24). The sampling weights associated with the MECs were calculated using the inequality of the sampling probabilities for selection in the present study and the non-response rates of participants in the NHANES as a whole and at the MECs specifically. Geometric means and 95% confidence intervals (CIs) are used to summarize continuous data, and numbers and percentages are used to summarize categorical data. The demographic and physical characteristics of the asthma and non-asthma groups were compared using the Rao-Scott χ² test.

Weighted logistic regression models were used to assess the relationship between asthma and the use of prescription antibiotics. Three models were created: model 1 was not adjusted; model 2 adjusted for age, sex, ethnicity, educational level, socioeconomic status, health insurance status, and BMI; and model 3 further adjusted for smoking status, history of diseases (including infection, bronchitis and emphysema), and family history of asthma. Odds ratios (ORs) and their 95% CIs were used to quantify the relationship between asthma and the type of antibiotic being used. In addition, subgroup analyses according to sex, ethnicity, and age were performed. When creating the model stratified according to sex, all the covariates adjusted for in model 3 were used, with the exception of sex. Analogous processes were used for the subgroup analyses according to age and ethnicity.

All the analyses were two-sided, and a p value of < 0.05 was considered to represent statistical significance. SAS v.9.4 (SAS Institute, Cary, NC, United States) and R v.4.1.3 (R Project for Statistical Computing, Vienna, Austria) were used for the statistical analyses.

As demonstrated in Table 1, the prevalence of asthma was 18.83% (9,627/51,124) and the overall prevalence of antibiotic use was 3.81% (1,946/51,124). The weighted mean age of the participants with asthma was 46.2 years, which was significantly lower than the weighted mean age of those without asthma (47.5 years, p < 0.001). Out of the 51,124 participants, 16,788 (32.84%) were aged between 20 and 40 years, 16,542 (32.36%) were aged between 40 and 60 years, and 17,794 (34.81%) were aged > 60 years. Non-Hispanic white people were the most common ethnicity (approximately 43.93%) while “other ethnicities,” including Asian, was the least common. Significant differences were found between the asthma and non-asthma groups with respect to most of the parameters assessed. Among the six categories of antibiotics included in the analysis, penicillin had the highest weighted prevalence of use (1.13%) and sulfonamide antibiotics had the lowest (0.31%).

The model 1 indicates that, compared to those not using antibiotics, individuals who used macrolide derivatives, penicillin, or quinolones had a 2.557 (95% CI: 1.811, 3.612), 1.547 (95% CI: 1.190, 2.011) and 2.053 (95% CI: 1.344, 3.137) times greater risk of asthma, respectively. However, after adjusting for demographic covariates and asthma-related factors, but only macrolides derivatives remained significantly associated with asthma (Table 2, model 3). No significant association was found between the use of the other three types of antibiotics and asthma, either before or after controlling for covariates (Table 2). The results also revealed that female participants had a 14.5% higher risk of asthma than male participants (OR: 1.145, 95% CI: 1.072, 1.223). Recent infection (OR:1.891, 95%CI: 1.648, 2.169), chronic bronchitis (OR:5.622, 95%CI: 4.954, 6.380) and emphysema (OR:4.102, 3.273, 5.141) were found to increase the risk of asthma. In addition, Hispanic participants, non-Hispanic black people and other ethnicities were all at a lower risk of asthma than non-Hispanic white people. Moreover, smoking was found to be a risk factor for asthma, with former and current smokers were at 24.3 and 78.6%, respectively, higher risks of asthma than never-smokers (former smoker, OR: 1.243, 95% CI: 1.146, 1.348; current smoker, OR: 1.786, 95% CI: 1.626, 1.961). Higher socioeconomic status, greater BMI, younger age and family history of asthma were all found to be risk factors for asthma (Supplementary Table S1).

Subgroup analyses were conducted based on demographic characteristics including sex, age, and ethnicity (Figure 2). The macrolide derivatives were closely associated with asthma in adults aged between 20 and 40 and between 40 and 60, but not among those aged > 60. Conversely, the use of quinolones was significantly associated with an increased risk of asthma in individuals aged over 60 years (Figure 2A). In addition, the use of macrolide derivatives or quinolones was significantly associated with asthma in women, while only cephalosporins had significant association with asthma in men (Figure 2B). There was a significant association between macrolide derivatives and asthma in all three races, while other antibiotics did not (Figure 2C). Interaction analyses between sex and age were also performed, and it was found that quinolone was associated with a lower risk of asthma in men between the ages of 20 and 40, but not in men of other age groups. Additionally, the association between quinolones and asthma was statistically significant in women aged between 20 and 40 and over 60, while the use of macrolides was associated with a high correlation of asthma in middle-aged women aged 40–60 (Supplementary Table S2).