INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Epidemiologic data on asthma prevalence, morbidity, and severity seem to point to a sex difference that varies with age in the risk of having asthma. Asthma and wheezing are more prevalent in boys than in girls (1, 2); during puberty the male disadvantage disappears (3, 4), and women older than 20 yr seem to have higher prevalence and morbidity rates of asthma than do males (5).

This pattern, assessed in different studies with different methods, has raised several hypotheses about the different susceptibility to asthma of men and women (6). Hormonal status has been invoked as one of the factors most likely to be able to differentially influence the risk of asthma occurrence (4); sex differences in the rate of lung growth and in airway size have been suggested as a potential alternative explanation (6).

However, prevalence and morbidity data cannot conclusively demonstrate a differential risk pattern for asthma in women and men because they are influenced by determinants of the incidence or determinants of the duration of the disease, such as the treatment and management of asthmatic patients. Indeed, it has been shown that the higher prevalence of asthma attacks in women could be partly explained by women's lesser ability to properly manage the disease (9).

Unfortunately, epidemiologic data on the pattern of asthma incidence according to age and sex are very poor, and those that do exist are limited to local communities or particular age groups (10).

The aim of the present study was to retrospectively assess the difference in asthma incidence, from birth to age 44 yr, in men and women in a wide set of developed nations, and to verify the consistency of the difference pattern among countries, using the data of a multicenter, cross-sectional, general population-based survey conducted according to a common and highly standardized protocol. An effort was also made to assess the influence of smoking habit and airway caliber on the incidence of asthma in men and women through a case-control study.

For this purpose, we used data collected in 16 countries in the European Community Respiratory Health Survey (ECRHS) (13).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Design

The data presented in this study were collected in the framework of Stage 2 of the ECRHS, from 1991 to 1993, and came from 37 centers in 16 countries (Table 1). The ECRHS is an international, cross-sectional, multicenter survey on the prevalence, determinants, and management of asthma (13). The methods used in the study have been described previously (13, 14).

Participating centers selected areas for study that were defined by preexisting administrative boundaries, had populations of at least 150,000 inhabitants and, where possible, had up-to-date sampling frames. The design of the ECRHS involved two stages. In the first stage, a screening questionnaire of respiratory symptoms was mailed to a random sample of at least 3,000 subjects (male to female ratio = 1) aged 20 to 44 yr. In the second stage, a 20% random sample of respondents to the mail questionnaire were invited to the local chest clinic, where they underwent a standardized structured interview (administered by trained interviewers), blood and skin tests, a spirometric assessment of lung function, and an airway challenge with methacholine. The study protocol had been approved by the local ethics committees, and written informed consent was obtained from all participants.

Interview Information and Lung Function

The presence of lifelong asthma and the age of the subject at first attack of asthma were identified on the basis of the answers subjects gave to the following questions: (1) Have you ever had asthma? (2) How old were you when you had your first attack of asthma?

The subjects were classified into the following, mutually exclusive groups according to their smoking habits (based on the answers to the clinical interview questions):

Nonsmokers. Subjects who had not smoked for a total of 1 yr (i.e., subjects who had not smoked more than 20 packs of cigarettes or 360 g of tobacco in their lifetime, or had smoked fewer than one cigarette per day or one cigar per week for 1 yr).

Ex-smokers. Subjects who had smoked for at least 1 yr but not during the month preceding the study.

Smokers. Subjects who had smoked for at least 1 yr and had also smoked during the month preceding the study.

For each subject, FEV₁ was measured with a water-sealed spirometer (Biomedin, Padua, Italy), according to the recommendations of the European Respiratory Society (15). The largest of at least three technically acceptable FEV₁ readings, not exceeding the next highest reading by more than 5%, was recoded.

Analysis of Lifetime Asthma Incidence

For analysis of lifetime asthma incidence, the onset of asthma was defined as the age at which the first attack of asthma was reported. Incidence rates were computed by dividing the number of events (the occurrence of the first attack) by the total number of person-years at risk, that is, the time during which a subject was at risk for developing asthma (i.e., the time from birth to age at the first attack of asthma if the subject was asthmatic, or the time from birth to the time the subject was interviewed if the subject was not asthmatic). Events and person-years were computed by 5-yr classes for men and women in the age range of 0 to 44 yr. In this way, subjects reporting that they had developed asthma at the age of 18 yr, each contributed 5 person-years to the first three age-classes, and 3 yr and 1 event to the 15- to 19-yr age class. The rate ratios for women with respect to men were computed in each age class for each country; homogeneity of rate ratios among countries was tested with the Mantel-Haenszel test for heterogeneity, and the pooled rate ratio was expressed according to the Mantel- Haenszel estimator (16).

Since the rate ratios for women versus men were computed with retrospective data, our estimates could have been biased by two sources of error: (1) recall bias could have differentially influenced the recall of the age at the first asthma attack in men and women; and (2) a potential cohort effect could have differentially influenced asthma incidence in men and women. In order to control for these two sources of error in the estimates of age-specific rate ratios, we adjusted the age-sex specific incidence of asthma for age at the time of interview (17), which is directly related to the birth-cohort effect and is a function of the time of recall (given the age of onset, the recall time varies with the age at interview).

For this purpose, the following Poisson regression model was fitted to the data (18):

where  is the logarithm of the incidence of asthma; _1', ₂, and ₁₂ are parameters that estimate the joint effect of sex and age at first attack (onset) on asthma incidence; ₃ and ₂₃ estimate the effect of age at interview (age_int) on the age-specific incidence of asthma (17); and ₁₃ is a parameter testing whether the effect of sex varies according to the age at interview (differential recall/cohort-effect bias). In this model, age at the first attack of asthma was considered a categorical variable (5-yr age classes), and age at interview was considered a continuous variable.

The analysis was performed using the Stata software system (Stata, Inc., College Station, TX) (19).

Case-Control Study to Assess the Influence of Sex, Airway Caliber, and Smoking Habit on Asthma Incidence

The aim of the case-control study was to verify whether differences in smoking habit and basal FEV₁ (an indicator of airway caliber) could explain differences in asthma susceptibility in women and men.

Because information on smoking habit and lung function was available at the time of the interview, this information could not be related to lifetime asthma incidence, but only to the incidence of asthma in a short period before the interview. For this purpose we considered all new cases of asthma (first attack of asthma) that occurred during the 3 yr before the interview and for which complete information was available on smoking habit, FEV₁, and height.

Each case was matched with eight controls (subjects who did not report asthma in their lifetime) who were randomly sampled from the population at the same study center as the subject and from the same age group (5-yr class) at the time of the study.

The subjects' smoking habits at the start of the risk period (3 yr before the survey) were reconstructed from the answers to the following questions: "How old were you when you started smoking" and "How old were you when you stopped or cut down smoking". Two of 343 smokers at the interview were classified as nonsmokers at the beginning of the risk period because they reported having started smoking during the 3 yr period before the interview (one case and one control); for a similar reason, 58 of 186 ex-smokers were classified as smokers at the beginning of the risk period (seven cases and 51 controls). The FEV₁ at the start of the risk period was computed on the basis of to two assumptions: (1) No decline in FEV₁ occurred during the 3-yr period before the interview (FEV₁ at interview equals FEV₁ 3 yr earlier). (2) The FEV₁ at the beginning of the 3-yr risk period could be computed on the basis of the yearly decline in FEV₁ according to the sex of the subject and the presence of asthma derived by the estimates of Peat and colleagues (20).

To assess the effect of sex, FEV₁, and smoking habit on the occurrence of asthma, a conditional logistic regression model was fitted to the data (18, 21), using the dichotomous case-control indicator as the dependent variable and the potential risk factors as the predictors. FEV₁ was adjusted for height by dividing its value (in liters) by the square of the subject's height in meters (22), and was recoded in tertiles, using the case distribution. The matched-sets indicator (105 sets of one case and eight controls) was used as a stratification factor. The effect of each factor was expressed as an odds ratio (OR) with 95% confidence intervals (CIs) and associated p values. Unadjusted and adjusted ORs were calculated via maximum likelihood methods, using EGRET software (Statistics and Epidemiology Research Corporation, Seattle, WA) (23).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Lifetime Asthma Incidence

Of 37,349 subjects selected for the second stage of the ECRHS, 18,659 participated in our survey (50.0%).The participation rate varied from 91% in Gardabaer to 12% in Montpellier (Table 1). The age of the subjects was 33.8 ± 7.1 yr (mean ± SD) and 52% were women (9,797). Three hundred twenty-four subjects did not answer the questions on asthma and were excluded from the analysis. The total person-years amounted to 593,978, and 1,578 subjects reported lifetime asthma.

The rate ratios for women versus men (Table 2) in each age-class were constant in all 16 countries studied, as shown by the results of the Mantel-Haenszel heterogeneity test (Table 2, Column 4 ) which were never statistically significant. When the influence of age at interview was tested according to the Poisson model (Table 3), the results showed that this variable significantly influenced the incidence of asthma (p < 0.001), but the rate ratios for women versus men were constant across different ages at the interview (the interaction term between sex and age at time of interview was not significant, at p = 0.38). Figure 1 shows the age-specific rate ratios for women versus men adjusted by age at the time of interview, obtained with the Poisson model. These estimates were almost equal to those reported in Table 1. The relative risk (RR) of women developing asthma with respect to men increased with age. In particular, women had a significantly lower risk than men of developing asthma during infancy: the RRs for girls versus boys were 0.74 (95% CI: 0.61 to 0.91) and 0.56 (95% CI: 0.45 to 0.73) in the 0- to 5-yr and 5- to 10-yr age groups, respectively; around puberty, the risk was almost equal in both sexes (0.84, [95% CI: 0.66 to 1.12]); and after puberty the risk in women was significantly higher than in men, ranging from 1.38 (95% CI: 1.02 to 1.88) to 5.91 (95% CI: 2.31 to 15.26).

View larger version (6K): [in this window] [in a new window]   Figure 1.   Age-specific rate ratios (RR) of asthma for women versus men, adjusted by age at interview.

The age-specific incidence rates of asthma by age at time of interview are shown in Figure 2, top panel for men and in Figure 2, bottom panel for women. As can be seen, the age-specific incidence of asthma decreased as age at interview increased. This could reflect either an increase in asthma incidence in the most recent generations (cohort effect), a decrease in the recall of past events (recall bias) in the older age groups, or both. In any case, it can be appreciated that, independently of age at the time of interview, the incidence of asthma showed a different age pattern according to sex. In men, the risk of developing asthma was at a peak in the first years of life (0 to 5 yr) and decreased rapidly until puberty: from then on, it was almost constant. In women the risk decreased slowly from infancy to puberty, and started to increase after that.

View larger version (12K): [in this window] [in a new window]   View larger version (12K): [in this window] [in a new window]   Figure 2.   Age-specific incidence of asthma by age at interview (1991 to 1993) for men (top panel ) and women (bottom panel ). Lines were obtained by cubic spline interpolation of observed incidences (dots).

Case-Control Study

Of 142 subjects reporting their first attack of asthma during the 3 yr before the survey, 105 had complete information on smoking habit and FEV₁. The latter were enrolled as cases. Of these 105 subjects, 26.7%, 35.2%, and 38.1% reported their first attack of asthma as occurring 1, 2, and 3 yr, respectively, before the survey. Table 4 shows the distribution and unadjusted ORs of the risk factors studied, in both cases and in the randomly matched controls. Women had a twofold greater risk of having asthma than did men (OR: 2.04; 95% CI: 1.32 to 3.15); the risk of developing asthma was inversely associated with airway caliber: in particular, subjects in the highest tertiles of height-adjusted FEV₁ had a 63% lower risk than did subjects in the lowest tertiles. Smokers showed a 42% lower risk than did non-smokers (OR: 0.58; 95% CI: 0.36 to 0.92).

When the three foregoing factors were studied simultaneously (Table 5), the OR for women decreased from 2.04 to 1.47 (95% CI: 0.89 to 2.44), losing its statistical significance. This decrease was entirely due to the introduction of the height-adjusted FEV₁ into the model, suggesting that the greater susceptibility to asthma in women than in men is explained, at least in part, by their smaller airway caliber. The effect of smoking habit was almost unchanged. There was no statistically significant interaction between the three variables studied. When the analysis was repeated, assuming a different decline for cases and controls during the risk period (20), the results substantially confirmed the role of FEV₁ in at least partly explaining the different susceptibility of the sexes to asthma.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main result of our analysis of the incidence of asthma from 0 to 44 yr of age in a wide range of countries was the finding of a reversal in sex-related incidence around the time of puberty, due to an increase in reported asthma in women and a decrease in men. The consistency of this pattern in all the countries examined in our study strongly suggests that during puberty, something happens that causes a greater susceptibility to asthma in women than in men. This pattern, also found in a recent study of the prevalence of asthma in British adolescents (4), has been attributed to hormonal changes, a major factor contributing to the sex differentiation that occurs during puberty. Furthermore, other studies, of pregnancy, oral contraceptive use, and hormonal replacement therapy, suggest that hormonal factors could play a role in the exacerbation and severity of asthma and allergy (24).

One feature of our analysis of gender differences in asthma incidence according to age is that the onset of asthma was estimated retrospectively, asking randomly selected subjects, aged 20 to 44 yr and identified at a specific point in time, whether they had had an attack of asthma and when they had had their first attack. Two kinds of bias could arise when this retrospective approach is used. The first comes from the selection only of subjects who are survivors of a potential cohort identified at birth. Subjects could have been lost because they died from asthma or because they emigrated after onset of the disease. This bias, however, could have influenced our estimates only to a minor extent, since mortality from asthma is very rare and emigration because of asthma is unlikely. The second, and more important, source of bias is related to the cohort effect (the incidence pattern could have changed over generations) or to recall bias (subjects of different sex and age could recall past events in different ways). In particular, a recall bias could have distorted the age-specific incidence rates of asthma (17), since people tend to attribute the onset of an illness to an age closer to the time of interview than to the true age of onset (telescoping bias). As a consequence, incidences in younger age groups should be estimated as lower than they occur, whereas those in older age groups should be estimated as higher. However, if the extent of the bias is the same in women and men for the different age groups, the ratios of the age-specific incidence of asthma for women versus men are only affected to a minor extent, and are likely to reflect the true RRs. In our data it was impossible to disentangle the recall and cohort effects. For this reason, we controlled for the age at interview, which is, to different degrees, related to both of the previously described effects. Our analysis showed that the age at interview was a strong determinant of asthma incidence: younger subjects reported a higher incidence than did older subjects, whereas the age pattern of the incidence was very similar in younger and older subjects both for women and for men. The age at interview (and accordingly the potential recall or cohort bias) had the same effect in women and in men, as is also documented by the absence of a statistically significant interaction between sex and age at the time of interview (Table 3).

Hence, our age-specific rate ratios strongly document that women's risk of developing asthma was significantly lower than that of men before puberty (rate ratios ranging from 0.74 [95% CI: 0.61 to 0.91] to 0.84 [95% CI: 0.66 to 1.12]) and significantly higher after puberty (rate ratios ranging from 1.38 [95% CI: 1.02 to 1.88] to 5.91 [95% CI: 2.31 to 15.26]). A similar result was also reported in one study done in the United States, where the incidence of asthma was determined from individual medical records which are not influenced by recall bias (12).

A further source of bias influencing the observed incidence pattern of asthma could be the improvement in diagnostic criteria over recent decades (27). This could partly explain the increased asthma incidence in younger generations. However, changes in diagnostic habits are likely to have affected the incidence of asthma to the same degree in both sexes. In other words, the age-specific RRs estimated for women versus men could hardly be biased by changes in diagnostic trends.

In order to understand whether other factors related to sex could explain the greater susceptibility to asthma of young adult women, we undertook a case-control study, using all the incident cases of asthma in the 3 yr before our survey, which were identified in the ECRHS international database. The results of our analysis showed that airway caliber at least partly explained the greater incidence of asthma in young women. In fact, in young women (17 to 44 yr old), the relative risk of developing asthma (OR: 2.04; 95% CI: 1.32 to 3.15) decreased significantly with control for height-adjusted FEV₁ (OR: 1.47; 95% CI: 0.89 to 2.44).

Since airway size is on average smaller in males than in females in infancy, and becomes larger in boys than in girls during and after puberty (28), it is likely that even the different susceptibility to asthma before the age of 17 yr could be explained by differences in airway size. This hypothesis is also supported by previous findings (8, 29) that the higher bronchial reactivity of women is almost totally due to their smaller airway caliber. The negative association between airway size and bronchial hyperresponsiveness has been explained by physics, since flow through the airway depends on the fourth power of the airway radius (30). Therefore, a given stimulus that shortens the airway smooth muscle will cause a greater obstruction to flow in relatively small airways, even if the airways are normal. The observed association between height-adjusted FEV₁ and asthma incidence is less obvious, but one could speculate that the effect of the inflammatory processes sustaining the occurrence of asthma is amplified by a small or reduced airway caliber, which is in turn related to sex, age, and atopy (29). However, in interpreting the previous association, it should be remembered that the risk period considered in our analysis was quite short ( 3 yr), and that the assessment of lung function at the beginning of the risk period depended on the airway caliber measured at the time of the interview, when the first attack of asthma had already occurred. Our correction, based on the assumption of a different decline in FEV₁ for asthmatic and non asthmatic individuals during the risk period, could only partly overcome the problem of the intrinsic cross-sectional nature of our data. This implies that the observed association between height-adjusted FEV₁ and asthma incidence could result from the physiopathologic process of the disease, entailing a reduction in bronchial airway caliber, that has caused or will cause a first attack of asthma. Only longitudinal data can give a conclusive answer to this problem. The ongoing follow-up of the subjects enrolled in the ECRHS from 1991 to 1993 will address the question of whether a lower level of lung function increases the susceptibility to asthma or is simply a step in the pathologic process of the disease.

Previous studies, based on asthma prevalence and morbidity, suggested an association between smoking and asthma prevalence (31, 32), but as far as we know, only two studies addressed this relationship with incident cases. In a Swedish cohort (11), smoking increased the incidence of physician-diagnosed asthma in women but not in men, whereas in a British cohort (33), a strong positive association was found between active smoking and wheezing. Our result, suggesting that the risk of developing asthma is significantly lower for smokers (OR: 0.58, 95% CI: 0.36 to 0.93) than for nonsmokers, is partly in contrast with these two previous studies, but mainly with the British one, which, however, used a less specific definition of asthma than did our study (34). A possible nonconflicting explanation for our result could be the presence of a strong "healthy-smoking effect" (35) in our young population (i.e., smokers were initially healthier on average than otherwise comparable people who never started to smoke). For example, persons with respiratory diseases during early life are more likely to be lifetime nonsmokers than are initially healthy people, and this effect, like the well-known "healthy worker effect" (36), is more pronounced in younger than in older people. However, these last results have to be interpreted cautiously because of the relatively small number of cases analyzed.

In conclusion, our analysis of the incidence of asthma using the ECRHS international database strongly confirms a gender reversal during puberty caused by an increase in reported asthma in women and a decrease in men, and suggests that factors other than hormonal ones, and particularly the size of the airway, could play an important role in explaining the different pattern of asthma incidence in men and women.