Caffeine Intake and Asthma Symptoms Joel Schwartz,
PhD, and Scott T. Weiss, MD, MS
Methylxanthines in theform of coffee and tea may be dietary factors that function as phawnacologic bronchodilators. To examine this hypothesis, we analyzed data from the Second National Health and Nutrition Examination Survey (NHANES II). NHANES II was conducted on a sampk representative of the civilian noninstitutionulired US population, which included white and black adults aged 30 years or older. We examined the relationship of usual coffee and tea consumption: from a medical history questionnaire to the presence of asthma and wheezing symptoms. Subjects who drank coffee on a regular basis had a 29% reduction in the odds of having current asthma symproms (odds ratio, 0.71; 95% conjidence interval, 0.55 to 0.93) when compared with non-coffee drinkers. The effect exhibited a significant dose-response relationship, with the number of cups of coffee consumed per day being inversely related to asthma prevalence. This relationship was independent of age, gender, and cigarette smoking. Tea consumption was not significantly related to asthma prevalence in this cohort. Ann Epidemiol J992;2:627-635. KEY WORDS:
Asthma,
coffee, nutrition.
INTRODUCTION Recently,
interest
has been
generated
in the
hypothesis
that
dietary
factors
may
influence the risk of chronic obstructive pulmonary (COPD) disease ( 1, 2). Postulated mechanisms for the association of diet and lung disease include an influence on oxidant and antioxidant
capacities
airway inflammation
of lung parenchyma
(1). Laboratory
and clinical
or modification
of the host response
data that have been presented
an effect of vitamin C on respiratory symptoms (l), pulmonary airway responsiveness (4); of serum copper on pulmonary function symptoms (1); and of dietary sodium and potassium on respiratory
to
suggest
function (3), and (5) and respiratory symptoms (1) and
airway responsiveness (6, 7). However, relatively little attention has been devoted to dietary factors functioning as pharmacologic agents, specifically as bronchodilators (8). Methylxanthines, in the form of coffee and tea, are the oldest known bronchodilators, the time of Ramazzini to treat and prevent bronchospasm. We used data quantitate relationship
MATERIALS
AND
from
the prevalence
the National of coffee
to the respiratory
Health
and Nutrition
and tea consumption
symptoms
having
of asthma
been
Examination
by American
used since Survey
adults
to
and its
and wheeze.
METHODS
The Second National Health and Nutrition Examination Survey (NHANES II) was conducted from February 1976 through February 1980. The sample was selected by the National Center for Health Statistics to be representative of the civilian noninstitutionFrom the US Environmental Protection Agency, Washington, DC (J.S.); and the Channing Laboratory, Brigham and Women’s Hospital; Pulmonary Division, Beth Israel Hospital; and Harvard Medical School, Boston, MA (S.T.W.). Address reprint requests to: Scott T. Weiss, MD, MS, Channing Laboratory, 180 Longwood Avenue, Boston, MA 02115. Received July 24, 1991; revised January 16, 1991. 0 1992Elsev~erScience PublishingCo., Inc.
1047-2797/92/$05.00
628
AEP Vol. 2, No. 5 September 1992: 627-635
Schwartz and Weiss CAFFEINE INTAKE AND ASTHMA SYMPTOMS
alized US population
aged 6 months
examined,
12,102
including
to 74 years. A total of 20,322
white and nonwhite
subjects
persons were
aged 30 years or older.
Details of the complex survey design, the examination procedures, and the laboratory measurements were published previously (9). Before medical examinations were performed, trained interviewers visited the subjects in their homes to administer socioeconomic and medical history questionnaires, of the National tionnaire
Heart,
on chronic
and examination used to maintain
which included questions similar to those
Lung, and Blood Institute-American
Thoracic
respiratory disease. A manual of operations,
protocols, and specifically trained interviewers standardized conduct of the survey (10).
Society
ques-
special interview
and examiners
were
Definitions Current
asthma was defined as an affirmative response to these two questions:
“Has a
doctor ever told you that you had asthma?” “Do you still have it?” Wheeze was defined as an affirmative response to the question: “In the past 12 months, apart from colds and flu, have you frequently had trouble with wheezing?” The medical history questionnaire
elicited
information
on usual daily intake of
coffee and tea. A coffee drinker was defined as a subject who responded affirmatively to this question: “Do you drink coffee ?” Consumption was quantitated in cups per day as reported by subjects in response to the question: “On average, how many cups or glasses per day do you drink?” Tea drinkers and tea consumption
in cups per day were
defined as for coffee. Smoking
history was obtained
by questionnaire.
Current
number
of cigarettes
smoked per day, maximal number of cigarettes smoked per day, and duration of smoking were ascertained. Cumulative smoking was indexed by the product of the maximal number of cigarettes smoked per day with the duration (in years) of smoking, to maintain comparability between current and former smokers. Body weight (kg) was measured in a mobile medical examination center.
Statistical
Methodology
The probability that a subject had current asthma or frequent wheeze was modeled by logistic regression. All regressions controlled for age, race, gender, and cigarette-years of smoking exposure. Initial models tested whether the risk of asthma or wheezing increased linearly or nonlinearly with cumulative smoking. Once the appropriate relationship
was identified,
it was used in all subsequent
analyses.
Initial analyses of coffee consumption merely compared coffee drinkers to noncoffee drinkers. To see whether a dose-response relationship existed, coffee drinkers were then divided into three categories: one cup per day; two cups per day; and three or more cups per day. To examine whether there was any interaction between gender and coffee consumption and smoking and coffee consumption, two interaction terms (cups/kg of body weight by gender and pack-years by cups/kg of body weight) were added to the basic logistic regression models described above. The effects of increasing cigarette smoking exposure on asthma occurrence stratified by various levels of coffee consumption and age and controlling for gender and race by regression analysis are presented later. For the analysis, subjects were also stratified by age less than 60 years and 60 to 74 years, to look at differential effects of age. If caffeine is really acting like theophylline in these subjects,
values in cups per
kilogram of body weight should be a better index of dose than data in cups per day. To test this hypothesis, the analyses were repeated using this alternative measure. All the analyses above were repeated for tea consumption instead of coffee con-
Schwartzand Weiss CAFFEINE INTAKE AND ASTHMA SYMPTOMS
AEP Vol. 2, No. 5 September 1992: 627435
sumption.
Finally,
both coffee and tea were considered
simultaneously
629
in regression
models. The NHANES
II is a stratified clustered random sample of the US population.
This complex design raises some issues for data analysis, whose resolution depends on the specific type of analysis performed. Clustering
results in the following statistical
issues. People who live in the same
town are often more alike than average. If the risk factors they share in common
are
not all controlled for in the means model, a correlation between subjects in the same town will occur in the residuals of the regression model. This results in off-diagonal terms in the covariance
matrix, an issue that has also arisen in clinical trials performed
at multiple
centers.
The
RTILOGIT
(1 l), a logistic regression procedure in Statistical
use of a random site effect addresses this issue. We used Analysis Systems (SAS,
Cary, NC) to account for this random site effect in our analysis. Alternative to the use of RTILOGIT exist. The site locations for the NHANES
procedures
II were chosen to result in an oversampling
of blacks, the elderly, and low-income families. Hence, the sample as drawn is not exactly representative of the US population. This situation does not differ from the vast majority representative
of epidemiologic
literature,
of the US population.
The
where the samples are almost never
census bureau,
however,
has compiled
estimates of each subject’s probability of being selected given the stratification. That probability can be used to construct weights, so that weighted means, rates, and so on are representative in calculating,
of the US population.
for example,
While such weights are entirely appropriate
the mean diastolic blood pressure in the United States,
to
use the weights in a regression of blood pressure on weight is to insist that some subjects in the sample contribute more to that correlation than others. Although use of the weights has been advocated for epidemiologic analysis (12), we do not believe this would be the optimal statistical approach. In the first case, each subject can be thought of as “representing”
a number of people in the US population.
Since the number of
people represented by each subject varies, they must contribute varying amounts to the means to make them representative. To apply the same framework to within sample epidemiologic
correlation
seems inappropriate.
Each subject is a sample of one, and
inherently has no more or less relativity than the others. The issue with oversampling is generalizability, not internal validity. If the entire sample were blacks, the results might be valid for blacks but not necessarily whites. In the NHANES
II data, where blacks are over-represented,
the intercept will be effected
if blacks have a different underlying rate of asthma than whites. The coefficient of coffee consumption will be effected only if the slopes are different between whites and blacks.
We tested this possibility with an interaction
term in our regression models
and found no difference. The other stratification variables where oversampling occurs are in the elderly and low-income populations. Our analysis was stratified by age, and income was not significantly related to our outcomes. By directly using the stratification variables in our analysis, we address any issue of bias, without weakening by assigning low weights to part of the sample.
our power
RESULTS Table 1 presents the crude rates per thousand subjects for asthma and wheeze stratified by cigarette smoking status, age, and coffee-drinking status. The crude rates suggest that wheeze is more prevalent than asthma and both symptoms are lower in coffee
630
Schwartz and Weiss CAFFEINE INTAKE AND ASTHMA SYMPTOMS
AEP Vol. 2, No. 5 1992: 627-635
September
TABLE 1 Number of subjects and crude rates of asthma and wheezing (per thousand) by age, smoking status, and coffee categories No. of subjects
Category
12,095 6005 6090 3928 8167 2518 2024 2532 5021
Overall 2 60 y
< 60 y Smoker Nonsmoker No coffee 1 cup/d 2 cups/d 2 3 cups/ d
drinkers than non-coffee
Wheezing
Asthma
84.5 96.4 72.8 123.1 65.9 84.7 90.4 73.6 87.5
31.7 33.3 30.1 33.4 30.9 37.8 33.1 27.3 30.3
drinkers, although a clear dose-response
relationship
is not
evident. Older age and cigarette smoking are associated with increased rates of symptom prevalence
and may also be associated with coffee drinking. Therefore,
adjustment
for
these variables seemed warranted. In models for both current asthma and wheeze, a linear term (pack-years) for cumulative smoking had greater explanatory power than the quadratic and logarithmic transforms, which increase faster and less fast than the linear transform, respectively. Hence, pack-years were used to index cumulative smoking exposure in the rest of the analysis. Subjects who drank coffee regularly had a significantly lower risk of current asthma (odds ratio [OR], 0.71; 950/o confidence interval [CI], 0.55 to 0.93) and an insignificantly lower risk of wheeze (OR, 0.87; 95% CI, 0.73 to 1.03) than did subjects who did not drink coffee. Tea drinkers had insignificantly lower risks of current asthma (OR, (OR,
0.85; 0.81;
95% CI, 0.63 to 1.16) but significantly lower risks of frequent 95% CI, 0.66 to 0.98) than non-tea drinkers.
The relationship
between coffee consumption
wheeze
and current asthma showed some
signs of a dose-response pattern, particularly for subjects aged 60 years and over. Table 2 shows the odds ratios and 95% confidence limits for all ages and for the two age subgroups by number of cups per day usually drunk. For comparison, Pagan0 and colleagues (8) are summarized. A dose-response relationship
TABLE
2
Odds Ratios (95% confidence intervals) for current asthma by coffee
consumption category Survey (INHS) (8)
and age”: NHANES
Coffee consumption All ages (cups/d) 0
1 2 23
the results of was indicated
1.00 0.84 (0.60-1.17) 0.66 (0.47-0.92) 0.70 (0.53-0.93)
30-59
II 1976-1980,
and Italian National
NHANES
II
y
60-74
1.00 0.79 (0.47-1.33) 0.56 (0.33-0.90) 0.75 (0.50-1.11)
Health
INHS y
1.00 0.87 (0.56-1.35) 0.73 (0.47~1.12) 0.62 (0.41-0.97)
45-64 1.00 1.00 0.64 0.78
y
> 64 y 1.00 0.90 0.88 0.68
AEP Vol. 2, No. 5 September1992: 627435
Schwartz and Weiss CAFFEINE INTAKE AND ASTHMA SYMmOMS
o.4]
O.ool -.03
.osol -.0549
631
W.050
Cups of Coffee by kg of Body Weight
0
O.OOl-.Os
.OstJl-.054a
.-.Oxi
Cups of Coffee by kg of Body Weight
FIGURE 1 Odds ratios for the relationship between cups of coffee consumed per kilogram of body weight and the presence of current asthma (A) and frequent wheeze (B). Subjects 30 to 59 years old are depicted by open circles and subjects 60 to 74 years old are depicted by stars.
for wheezing as well (results not shown). Coffee consumption was reported by 85% of subjects under age 60 years and by 87% of subjects 60 years or older. For tea consumption, the respective rates were 74% and 73%. Figure 1 shows the odds ratios for current asthma and frequent wheeze according to quantities of coffee consumed as defined by cups per kilogram of body weight. Results are shown for the two age subgroups. Analyses stratified by age showed signs of a doseresponse relationship. Tea consumption, in contrast, showed little evidence of a doseresponse relationship (data not shown).
632
Schwartz and Weiss
AEP Vol. 2, No. 5 September 1992: 627-635
CAFFEINE INTAKE AND ASTHMA SYMPTOMS
TABLE 3 Odds ratios (95% confidence intervals) for an additional 10 pack-years of smoking exposure by age and coffee consumption patterns: NHANES II, 1976-1980 Coffee consumption pattern (cups/kg of body weight) 0
o-0.03 0.030-0.055 2 0.055 0
o-0.03 0.030-0.055 2 0.055
Age (Y)
Current
Wheeze
asthma
< < <
PhD, and Scott T. Weiss, MD, MS
Methylxanthines in theform of coffee and tea may be dietary factors that function as phawnacologic bronchodilators. To examine this hypothesis, we analyzed data from the Second National Health and Nutrition Examination Survey (NHANES II). NHANES II was conducted on a sampk representative of the civilian noninstitutionulired US population, which included white and black adults aged 30 years or older. We examined the relationship of usual coffee and tea consumption: from a medical history questionnaire to the presence of asthma and wheezing symptoms. Subjects who drank coffee on a regular basis had a 29% reduction in the odds of having current asthma symproms (odds ratio, 0.71; 95% conjidence interval, 0.55 to 0.93) when compared with non-coffee drinkers. The effect exhibited a significant dose-response relationship, with the number of cups of coffee consumed per day being inversely related to asthma prevalence. This relationship was independent of age, gender, and cigarette smoking. Tea consumption was not significantly related to asthma prevalence in this cohort. Ann Epidemiol J992;2:627-635. KEY WORDS:
Asthma,
coffee, nutrition.
INTRODUCTION Recently,
interest
has been
generated
in the
hypothesis
that
dietary
factors
may
influence the risk of chronic obstructive pulmonary (COPD) disease ( 1, 2). Postulated mechanisms for the association of diet and lung disease include an influence on oxidant and antioxidant
capacities
airway inflammation
of lung parenchyma
(1). Laboratory
and clinical
or modification
of the host response
data that have been presented
an effect of vitamin C on respiratory symptoms (l), pulmonary airway responsiveness (4); of serum copper on pulmonary function symptoms (1); and of dietary sodium and potassium on respiratory
to
suggest
function (3), and (5) and respiratory symptoms (1) and
airway responsiveness (6, 7). However, relatively little attention has been devoted to dietary factors functioning as pharmacologic agents, specifically as bronchodilators (8). Methylxanthines, in the form of coffee and tea, are the oldest known bronchodilators, the time of Ramazzini to treat and prevent bronchospasm. We used data quantitate relationship
MATERIALS
AND
from
the prevalence
the National of coffee
to the respiratory
Health
and Nutrition
and tea consumption
symptoms
having
of asthma
been
Examination
by American
used since Survey
adults
to
and its
and wheeze.
METHODS
The Second National Health and Nutrition Examination Survey (NHANES II) was conducted from February 1976 through February 1980. The sample was selected by the National Center for Health Statistics to be representative of the civilian noninstitutionFrom the US Environmental Protection Agency, Washington, DC (J.S.); and the Channing Laboratory, Brigham and Women’s Hospital; Pulmonary Division, Beth Israel Hospital; and Harvard Medical School, Boston, MA (S.T.W.). Address reprint requests to: Scott T. Weiss, MD, MS, Channing Laboratory, 180 Longwood Avenue, Boston, MA 02115. Received July 24, 1991; revised January 16, 1991. 0 1992Elsev~erScience PublishingCo., Inc.
1047-2797/92/$05.00
628
AEP Vol. 2, No. 5 September 1992: 627-635
Schwartz and Weiss CAFFEINE INTAKE AND ASTHMA SYMPTOMS
alized US population
aged 6 months
examined,
12,102
including
to 74 years. A total of 20,322
white and nonwhite
subjects
persons were
aged 30 years or older.
Details of the complex survey design, the examination procedures, and the laboratory measurements were published previously (9). Before medical examinations were performed, trained interviewers visited the subjects in their homes to administer socioeconomic and medical history questionnaires, of the National tionnaire
Heart,
on chronic
and examination used to maintain
which included questions similar to those
Lung, and Blood Institute-American
Thoracic
respiratory disease. A manual of operations,
protocols, and specifically trained interviewers standardized conduct of the survey (10).
Society
ques-
special interview
and examiners
were
Definitions Current
asthma was defined as an affirmative response to these two questions:
“Has a
doctor ever told you that you had asthma?” “Do you still have it?” Wheeze was defined as an affirmative response to the question: “In the past 12 months, apart from colds and flu, have you frequently had trouble with wheezing?” The medical history questionnaire
elicited
information
on usual daily intake of
coffee and tea. A coffee drinker was defined as a subject who responded affirmatively to this question: “Do you drink coffee ?” Consumption was quantitated in cups per day as reported by subjects in response to the question: “On average, how many cups or glasses per day do you drink?” Tea drinkers and tea consumption
in cups per day were
defined as for coffee. Smoking
history was obtained
by questionnaire.
Current
number
of cigarettes
smoked per day, maximal number of cigarettes smoked per day, and duration of smoking were ascertained. Cumulative smoking was indexed by the product of the maximal number of cigarettes smoked per day with the duration (in years) of smoking, to maintain comparability between current and former smokers. Body weight (kg) was measured in a mobile medical examination center.
Statistical
Methodology
The probability that a subject had current asthma or frequent wheeze was modeled by logistic regression. All regressions controlled for age, race, gender, and cigarette-years of smoking exposure. Initial models tested whether the risk of asthma or wheezing increased linearly or nonlinearly with cumulative smoking. Once the appropriate relationship
was identified,
it was used in all subsequent
analyses.
Initial analyses of coffee consumption merely compared coffee drinkers to noncoffee drinkers. To see whether a dose-response relationship existed, coffee drinkers were then divided into three categories: one cup per day; two cups per day; and three or more cups per day. To examine whether there was any interaction between gender and coffee consumption and smoking and coffee consumption, two interaction terms (cups/kg of body weight by gender and pack-years by cups/kg of body weight) were added to the basic logistic regression models described above. The effects of increasing cigarette smoking exposure on asthma occurrence stratified by various levels of coffee consumption and age and controlling for gender and race by regression analysis are presented later. For the analysis, subjects were also stratified by age less than 60 years and 60 to 74 years, to look at differential effects of age. If caffeine is really acting like theophylline in these subjects,
values in cups per
kilogram of body weight should be a better index of dose than data in cups per day. To test this hypothesis, the analyses were repeated using this alternative measure. All the analyses above were repeated for tea consumption instead of coffee con-
Schwartzand Weiss CAFFEINE INTAKE AND ASTHMA SYMPTOMS
AEP Vol. 2, No. 5 September 1992: 627435
sumption.
Finally,
both coffee and tea were considered
simultaneously
629
in regression
models. The NHANES
II is a stratified clustered random sample of the US population.
This complex design raises some issues for data analysis, whose resolution depends on the specific type of analysis performed. Clustering
results in the following statistical
issues. People who live in the same
town are often more alike than average. If the risk factors they share in common
are
not all controlled for in the means model, a correlation between subjects in the same town will occur in the residuals of the regression model. This results in off-diagonal terms in the covariance
matrix, an issue that has also arisen in clinical trials performed
at multiple
centers.
The
RTILOGIT
(1 l), a logistic regression procedure in Statistical
use of a random site effect addresses this issue. We used Analysis Systems (SAS,
Cary, NC) to account for this random site effect in our analysis. Alternative to the use of RTILOGIT exist. The site locations for the NHANES
procedures
II were chosen to result in an oversampling
of blacks, the elderly, and low-income families. Hence, the sample as drawn is not exactly representative of the US population. This situation does not differ from the vast majority representative
of epidemiologic
literature,
of the US population.
The
where the samples are almost never
census bureau,
however,
has compiled
estimates of each subject’s probability of being selected given the stratification. That probability can be used to construct weights, so that weighted means, rates, and so on are representative in calculating,
of the US population.
for example,
While such weights are entirely appropriate
the mean diastolic blood pressure in the United States,
to
use the weights in a regression of blood pressure on weight is to insist that some subjects in the sample contribute more to that correlation than others. Although use of the weights has been advocated for epidemiologic analysis (12), we do not believe this would be the optimal statistical approach. In the first case, each subject can be thought of as “representing”
a number of people in the US population.
Since the number of
people represented by each subject varies, they must contribute varying amounts to the means to make them representative. To apply the same framework to within sample epidemiologic
correlation
seems inappropriate.
Each subject is a sample of one, and
inherently has no more or less relativity than the others. The issue with oversampling is generalizability, not internal validity. If the entire sample were blacks, the results might be valid for blacks but not necessarily whites. In the NHANES
II data, where blacks are over-represented,
the intercept will be effected
if blacks have a different underlying rate of asthma than whites. The coefficient of coffee consumption will be effected only if the slopes are different between whites and blacks.
We tested this possibility with an interaction
term in our regression models
and found no difference. The other stratification variables where oversampling occurs are in the elderly and low-income populations. Our analysis was stratified by age, and income was not significantly related to our outcomes. By directly using the stratification variables in our analysis, we address any issue of bias, without weakening by assigning low weights to part of the sample.
our power
RESULTS Table 1 presents the crude rates per thousand subjects for asthma and wheeze stratified by cigarette smoking status, age, and coffee-drinking status. The crude rates suggest that wheeze is more prevalent than asthma and both symptoms are lower in coffee
630
Schwartz and Weiss CAFFEINE INTAKE AND ASTHMA SYMPTOMS
AEP Vol. 2, No. 5 1992: 627-635
September
TABLE 1 Number of subjects and crude rates of asthma and wheezing (per thousand) by age, smoking status, and coffee categories No. of subjects
Category
12,095 6005 6090 3928 8167 2518 2024 2532 5021
Overall 2 60 y
< 60 y Smoker Nonsmoker No coffee 1 cup/d 2 cups/d 2 3 cups/ d
drinkers than non-coffee
Wheezing
Asthma
84.5 96.4 72.8 123.1 65.9 84.7 90.4 73.6 87.5
31.7 33.3 30.1 33.4 30.9 37.8 33.1 27.3 30.3
drinkers, although a clear dose-response
relationship
is not
evident. Older age and cigarette smoking are associated with increased rates of symptom prevalence
and may also be associated with coffee drinking. Therefore,
adjustment
for
these variables seemed warranted. In models for both current asthma and wheeze, a linear term (pack-years) for cumulative smoking had greater explanatory power than the quadratic and logarithmic transforms, which increase faster and less fast than the linear transform, respectively. Hence, pack-years were used to index cumulative smoking exposure in the rest of the analysis. Subjects who drank coffee regularly had a significantly lower risk of current asthma (odds ratio [OR], 0.71; 950/o confidence interval [CI], 0.55 to 0.93) and an insignificantly lower risk of wheeze (OR, 0.87; 95% CI, 0.73 to 1.03) than did subjects who did not drink coffee. Tea drinkers had insignificantly lower risks of current asthma (OR, (OR,
0.85; 0.81;
95% CI, 0.63 to 1.16) but significantly lower risks of frequent 95% CI, 0.66 to 0.98) than non-tea drinkers.
The relationship
between coffee consumption
wheeze
and current asthma showed some
signs of a dose-response pattern, particularly for subjects aged 60 years and over. Table 2 shows the odds ratios and 95% confidence limits for all ages and for the two age subgroups by number of cups per day usually drunk. For comparison, Pagan0 and colleagues (8) are summarized. A dose-response relationship
TABLE
2
Odds Ratios (95% confidence intervals) for current asthma by coffee
consumption category Survey (INHS) (8)
and age”: NHANES
Coffee consumption All ages (cups/d) 0
1 2 23
the results of was indicated
1.00 0.84 (0.60-1.17) 0.66 (0.47-0.92) 0.70 (0.53-0.93)
30-59
II 1976-1980,
and Italian National
NHANES
II
y
60-74
1.00 0.79 (0.47-1.33) 0.56 (0.33-0.90) 0.75 (0.50-1.11)
Health
INHS y
1.00 0.87 (0.56-1.35) 0.73 (0.47~1.12) 0.62 (0.41-0.97)
45-64 1.00 1.00 0.64 0.78
y
> 64 y 1.00 0.90 0.88 0.68
AEP Vol. 2, No. 5 September1992: 627435
Schwartz and Weiss CAFFEINE INTAKE AND ASTHMA SYMmOMS
o.4]
O.ool -.03
.osol -.0549
631
W.050
Cups of Coffee by kg of Body Weight
0
O.OOl-.Os
.OstJl-.054a
.-.Oxi
Cups of Coffee by kg of Body Weight
FIGURE 1 Odds ratios for the relationship between cups of coffee consumed per kilogram of body weight and the presence of current asthma (A) and frequent wheeze (B). Subjects 30 to 59 years old are depicted by open circles and subjects 60 to 74 years old are depicted by stars.
for wheezing as well (results not shown). Coffee consumption was reported by 85% of subjects under age 60 years and by 87% of subjects 60 years or older. For tea consumption, the respective rates were 74% and 73%. Figure 1 shows the odds ratios for current asthma and frequent wheeze according to quantities of coffee consumed as defined by cups per kilogram of body weight. Results are shown for the two age subgroups. Analyses stratified by age showed signs of a doseresponse relationship. Tea consumption, in contrast, showed little evidence of a doseresponse relationship (data not shown).
632
Schwartz and Weiss
AEP Vol. 2, No. 5 September 1992: 627-635
CAFFEINE INTAKE AND ASTHMA SYMPTOMS
TABLE 3 Odds ratios (95% confidence intervals) for an additional 10 pack-years of smoking exposure by age and coffee consumption patterns: NHANES II, 1976-1980 Coffee consumption pattern (cups/kg of body weight) 0
o-0.03 0.030-0.055 2 0.055 0
o-0.03 0.030-0.055 2 0.055
Age (Y)
Current
Wheeze
asthma
< < <
