INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: asthma; obesity; body mass index; African American; physical activity; epidemiology; environmental tobacco smoke; passive smoking; sex differences

One approach to reducing the considerable and increasing morbidity of asthma is to identify modifiable risk factors as targets for preventive interventions. Epidemiologic investigations have consistently identified age, parental history of asthma or allergies, a history of certain respiratory viral infections at particular stages of childhood development, atopy, house dust mite or cockroach sensitization and exposure, and lower socioeconomic status as risk factors for asthma in children (1, 2). In addition, other risk factors have been identified, among them increased body weight, and skinfold thickness, an indicator of obesity (3). Although perhaps not a major risk factor, obesity has the advantage of being potentially modifiable.

Secular trends for increased asthma prevalence and for increased body mass index (BMI) have been noted in the United States (4). Because the association of obesity and asthma has previously been described primarily in children, and because childhood and adult asthma differ importantly, we tested whether chronic asthma or new asthma diagnosis during 10 yr of follow-up is associated with body weight, weight gain, or lower activity level in a community-based cohort of urban young adults who were 18 to 30 yr of age at the time of entry into the Coronary Artery Risk Development in Young Adults (CARDIA) Study. We also tested whether smoking and indices of environmental tobacco smoke (ETS) exposure were associated with asthma diagnosis.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Details of the CARDIA subject recruitment and study design have been published elsewhere (5). Analyses here included 4,547 participants who completed baseline and at least two additional examinations.

Young adults were recruited in 1985 and 1986 primarily by telephone from Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota; and Oakland, California. Protocols were approved by institutional review boards of participating centers. Follow-up examinations were conducted at 2, 5, 7, and 10 yr after baseline with participation of 90%, 86%, 81%, and 79% of surviving participants, respectively. Activity level at each examination was based on the Physical Activity History Score, eliciting frequency of participation in a range of specific heavy and moderate intensity activities in the last year. Of those with Physical Activity Scores in the range 300 to 399 exercise units, approximately 74% met the American College of Sports Medicine recommendation for developing cardiorespiratory and muscular fitness (6,7). A very high activity score is 800 or higher (8).

A nonsmoker was someone who admitted to smoking less than 100 lifetime cigarettes, and a regular smoker was one who smoked at least five cigarettes per week almost every week. At the first examination, serum cotinine level was measured to verify self-reported smoking status; concentrations less than 14 ng/ml were considered indicative of nonsmoking status, concentrations from 2 to 13 ng/ml of ETS exposure, and above 13 ng/ml of active smoking.

Asthma Case Definition

At each examination, participants were asked whether they were taking medications for asthma or other breathing problems. In most cases, this was verified by examining medications containers. At Years 0, 2, 7, and 10, subjects were asked whether a doctor or nurse had ever told them they had asthma. Wheeze and other symptoms associated with asthma were also queried at Years 0, 2, and 10. Combinations of conditions associated with asthma were considered in constructing the diagnostic algorithm. Among these combinations, the algorithm selected had the highest occurrence of repeat diagnosis. Thus, it was associated with the most consistent medication and self-report information of the algorithms evaluated in reidentifying as asthmatic individuals who had previously been identified as asthmatic. Asthma was classified as being present if (1) a subject was taking a medication typically used to treat asthma; failing this, (2) reporting at one examination taking a medication not typically used to treat asthma, but reporting at a later examination taking a medication typically used to treat asthma; failing this, (3) self-reported doctor or nurse diagnosis of asthma. Participants whose only indication of asthma was a self-report of taking an "asthma" medication that is not typically used to treat asthma were classified as not having asthma.

Data Analysis

Asthma was assumed to be present starting at the first examination it was diagnosed. Asthma incidence was computed assuming participants did not have asthma at any intervening missed examination, if they did not have asthma at their final examination. Statistical adjustment and testing were carried out for the cross-sectional study at baseline using logistic regression with asthma prevalence at baseline as the dependent variable and independent variables as indicated in the tables. Incidence rates were computed using life table rules, excluding from those at risk at a particular examination those who had asthma at any previous examination or who missed that examination and all subsequent examinations. During follow-up, asthma diagnoses could only occur after exactly 2, 5, 7, or 10 yr of follow-up. Asthma incidence between each examination was calculated, as well as cumulative incidence for asthma diagnoses occurring at any time during Years 2 to 10. Statistical adjustment and testing were carried out for the incidence study using proportional hazards regression, with asthma incidence at each follow-up examination as the dependent variable and independent variables as indicated in the tables. For simplicity, F tests for any differences among groups of participants (e.g., quintiles or centers) were performed using generalized linear models.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Asthma prevalence at baseline (7.0%) and 10-yr asthma incidence (5.6%) were lower in Birmingham than the other three centers. Oakland had the highest baseline asthma prevalence (12.6%) but had 10-yr asthma incidence (7.8%) similar to Chicago (8.4%) and Minneapolis (8.3%). The p value for any difference among centers is 0.001 for the baseline prevalence, and 0.06 for 10-yr asthma incidence. Age was not related to asthma prevalence or incidence.

Sex and Race

Table 1 shows baseline and follow-up prevalence and incidence of asthma by race and sex. Statistical testing adjusts for age and center. The F test for any race-sex difference corresponded to p = 0.08 for asthma prevalence and p = 0.0001 for incidence. Baseline prevalence was greatest in black males and least in white females. However, incidence was nearly 1.5 times as high in women as in men, in both whites and blacks. A slight excess incidence in blacks compared with whites was not statistically significant. The same pattern is seen for incidence during each of the follow-up examinations: the highest incidence was seen for black women, followed closely by white women, and then black men and white men.

Educational Attainment

Among participants whose maximal educational attainment was less than high school, asthma prevalence at baseline was 1.8 times higher than those whose maximum education was high school degree or greater than high school (Table 2). Both the group with less than high school education and the group with high school education have significantly higher 10-yr asthma incidence than those with greater than high school education.

Smoking and ETS

The baseline prevalence of asthma was similar among never smokers, ex-smokers, and smokers, but asthmatics diagnosed during the study were more frequently cigarette smokers or ex-smokers than were those never diagnosed with asthma, with an adjusted hazard rate ratio of 1.38 (1.07 to 1.77) for Years 2-10 incidence (Table 3). A similar but not statistically significant relation was seen comparing ex-smokers with never-smokers. Further subclassification of smokers into those who quit at various stages of follow-up and those who continued smoking or started smoking revealed that 10-yr asthma incidence was greater among quitters than nonquitters. Stratification by sex and race showed a significant association of active smoking with asthma in females but not males, and in blacks but not whites.

Asthma at baseline and subsequent diagnosis of asthma was not associated with serum cotinine at baseline when grouped by no ETS exposure (< 2 ng/ml), ETS exposure (2-13 ng/ml), and active smoking (> 13 ng/ml), (Table 4). In this study, about 31% of subjects had serum cotinine levels indicative of active cigarette smoking at the first examination. Previous analysis of cotinine concentrations in cigarette smokers in CARDIA showed that they correlated strongly with self-reported smoking status. Among nonsmokers, higher self-report of ETS (not shown) was also not associated with subsequent asthma diagnosis.

BMI

Subjects were divided into five equal-sized groups (quintiles) according to their BMI at baseline. Asthma prevalence at baseline and incidence during 10-yr of follow-up according to these quintiles are presented in Table 5. The asthma prevalence was smallest among Group 2 (20 to 39th percentiles) and greatest among the highest, but the difference in asthma prevalence among quintiles at baseline was not significant (p = 0.74). This J-shaped pattern is even more evident for 10-yr asthma incidence. The hazard rate ratio for quintile 5 versus 2 is 1.60 (1.12 to 2.30), after adjusting for age, maximal educational attainment, race, sex, and center, whereas the corresponding hazard rate ratio for quintile 1 versus 2 is 1.39 (0.97 to 1.99). The p value is 0.007 for any difference in 10-yr asthma incidence according to baseline BMI quintiles.

A J-shaped pattern of risk of having asthma is consistently found when asthma incidence is regressed on quintiles of 10-yr BMI change (e.g, categories of weight gain during the period of follow-up), including Year 0 BMI quintiles in the same model (Table 6). A J-shape describes a pattern in which risk for the lowest quintile (lowest 10-yr BMI change) is slightly higher than for the second quintile, and where risk for the three highest quintiles is more markedly elevated. Those in quintile 5 who have the greatest BMI gain have the largest asthma incidence (11.2%, hazard rate ratio 1.63 compared with quintile 2). In the same model, the Year 0 BMI quintiles 1 and 5 have the highest risk, indicating that BMI at entry into the study is associated with risk for subsequent asthma diagnosis, even holding subsequent weight gain constant. The Year 10 BMI (Table 6) is the sum of the Year 0 BMI and BMI change during 10 yr. This summary measure has highest risk in BMI Year 10 quintile 5, with a slight increase in risk in quintile 1. The associations shown in Tables 5 and 6 persist after adjustment for physical activity (not shown).

Stratified by sex, a significant association of Year 10 BMI and change of BMI was seen in females but not in males. Stratified by race, a significant association of BMI at Year 10 with asthma was seen in blacks and whites, whereas an association of asthma with change in BMI was seen only in blacks.

Physical Activity

The asthma prevalence at baseline and 10-yr asthma incidence were similar across the baseline physical activity quintiles (Table 7). There was also no significant difference of asthma prevalence at baseline and incidence during 10-year follow-up across the 10-yr change of physical activity quintiles and Year 10 physical activity quintiles (Table 8). Findings were unchanged when only vigorous or only moderate-intensity activities were considered. Analyses for asthma rate ratios adjusting BMI for physical activity and physical activity for BMI barely changed odds ratio and hazard rate ratios. The absence of an association of asthma with physical activity was not altered by adjustment for BMI. Thus, the association of BMI increase with asthma incidence is independent of reported level of physical activity.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The association of asthma diagnosis with higher weight or BMI was originally described in children but has also been found in adults. Camargo and coworkers found an association of increased BMI with incident asthma diagnosis in adult female registered nurses 26 to 46 yr of age (9); Chen and coworkers found an association of obesity with asthma diagnosis in women but not men in a cross-sectional study of the Canadian National Population Health Survey (10). In a longitudinal study of subjects older than 40 yr of age, Guerra and coworkers found that baseline BMI  28 was associated with subsequent asthma diagnosis only in women (11).

Schachter and coworkers found in a pooling of three cross-sectional studies that severe obesity was a significant risk factor for doctor diagnosis of asthma, history of wheeze, and medication use after adjusting for atopy, sex, smoking history, and family history. Although their method for dividing subjects into five BMI groups differed from ours, Schachter and coworkers also found increased asthma in the lowest BMI group ("underweight, BMI < 18.5") compared with the second group, and highest asthma in the highest BMI group ("severe obesity, BMI  35.0"). In addition, they found that obesity was not associated with increased airway reactivity on methacholine challenge testing or with reduced FEV₁/FVC ratio (12). This suggested that asthma diagnosis, wheeze, and medication use in obese subjects is not due to typical asthma, which is associated with methacholine hyperresponsiveness. However, in a prospective study of nonasthmatic aging males, Litonjua and coworkers found that 29 middle-aged to older men who had new onset methacholine airway hyperresponsiveness (although not necessarily asthma) had significantly higher baseline BMI (before developing airway hyperresponsiveness) and greater increase in BMI between the first and second examinations, after controlling for age, smoking, IgE concentration, baseline FEV₁, and height (13).

Platts-Mills and others have hypothesized that a secular trend of increasing asthma prevalence among school-aged children is in part causally associated with decreased daily physical activity and weight gain (14). A previous analysis of physical activity level over the first 7 yr follow-up of the CARDIA study showed decreasing activity levels with the passage of time in this age group. In addition, analysis showed that some of the decline in activity levels was independent of birth cohort or age, thus that there was a time-related or secular trend toward diminished activity level in young adults (15). Our results in young adults starting at ages 18 to 30 demonstrate little association of asthma with BMI at baseline, but the longitudinal design demonstrates an association of higher BMI at baseline and increase in BMI with subsequent asthma incidence during the period of follow-up. No such association of asthma incidence is seen with level or change in physical activity. This does not support a causal sequence of asthma incidence leading to lower physical activity resulting in weight gain, but rather suggests that weight gain precedes and thus could be causally associated with the diagnosis of asthma. Our finding of an association of asthma with higher BMI in women but not in men cannot be attributed to sex-related differences in amount of weight gain, physical activity levels, cigarette smoking, or exposure to ETS in these young adults.

In the CARDIA subjects, lung function (spirometry) has previously been analyzed for the first examination, and lower sex-specific and race-specific lung function was significantly and independently associated with cigarette smoking, higher BMI, skinfold thickness (an index of body fat), and lower time to stopping a graded treadmill test (15). Thus, new cases of asthma associated with weight gain may have occurred in those whose height-, race-, and sex-specific lung function was already lower in association with higher BMI or body fat, and lower exercise tolerance. Lower lung function in this circumstance may be a predisposing factor for later asthma diagnosis.

An important strength of this study is the availability of both asthma prevalence and incidence, because many relationships emerge in the incidence data that are not present in the prevalence data. The incidence data are free of some of the limitations inherent in prevalence data. For example, criteria in an incidence study are uniform, whereas prevalence data combine cases occurring for a wide variety of reasons. Our study supports this argument: the association with BMI and BMI change was clearer in longitudinal than in baseline analyses.

In summary, among a large group of urban young adults, asthma diagnosis at the start of study was associated with center of study, race, sex, lower maximal education attainment, and active cigarette smoking, but not with cotinine levels indicative of ETS exposure in nonsmokers. With follow-up for 10 yr, new or incident diagnoses of asthma showed a J-shaped pattern of association with BMI, with significant association limited to females, and greatest risk associated with highest BMI at start and at the end of 10 yr, and with gain in BMI (weight gain) over the period of observation. This association of asthma incidence with weight gain was not explained by lower physical activity.