INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The genetics of asthma has received interest since the beginning of the 20th century. As early as 1909, Drinkwater, using the example of a three-generation pedigree, suggested that asthma is a dominant Mendelian trait (1), whereas a subsequent large study of the genetic transmission of asthma by Cooke and van der Veer in 1916 (2) was not able to show such a simple mode of inheritance (3). Although there was speculation in the early 1950s that the observed familial aggregation of asthma could be due to an infective agent, several studies have shown that a genetic component is more likely. Tips, in 1954, was the first to suggest a polygenetic mode of inheritance of asthma (4). This theory was further developed in 1967 by Leigh and Morley, who suggested that asthma is a multifactorial trait in which multiple gene loci interact with one or more external factors (3, 5).

Studies of inbred populations, twin registries, family analyses, and migration and adoption studies (6) are difficult to interpret because of the different designs they employ. Furthermore, they have been criticized for the absence of proper controls, nonsystematic sampling methods, and lack of standardized diagnostic procedures. Although the first large-scale twin study of the Swedish registry observed only a 0.19 concordance rate of asthma in 2,434 monozygotic twins (9), a reanalysis showed a heritability of 0.65 (7), a finding similar to that in at least 10 later twin studies. Although previous family studies have shown an increased frequency in first-degree relatives, asthma seems not to be inherited in a simple Mendelian fashion (10).

The major problem in all studies is the lack of a clear-cut definition of asthma. Most studies use a self-stated diagnosis to define asthma. Recurrent wheezing and breathlessness are the main symptoms of asthma, but these symptoms can vary over time and sometimes vanish in adulthood. Although physicians may agree in diagnosing severe asthma, the diagnostic threshold for minor symptomatic cases varies considerably. Also, the interaction with concomitant allergic disease or environmental exposure to allergens, viruses, and pollutants has been very difficult to analyze. Although genetic and environmental factors as well as their interactions are important, their precise roles in the pathogenesis of asthma are not yet understood.

The analysis described here used the data from the European Community Respiratory Health Survey (ECRHS), which is at the moment one of the world's largest epidemiologic studies of airway disease. Since 1990 the ECRHS has been collecting information about the prevalence of asthma and suspected risk factors for this disease. The study involves 34 centers in 11 countries in the European Community, seven centers in five Cooperation in Science and Technical Research (COST) states in Europe, and 15 centers in seven other countries (13).

The goal of the present analysis was: (1) to provide data for the counseling of parents on the risk of asthma in their offspring; (2) to show possible geographic variation between/ within different countries in the genetic influence on asthma; (3) to identify potential subgroups with primarily genetic transmission of asthma; and (4) to investigate the mode of inheritance of asthma.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Area, Subjects, and Study Design

The study was performed from 1990 to 1992 in Australia, Belgium, France, Germany, Greece, India, Italy, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, and the United States. Cities or regions qualified as study areas if they had: (1) preexisting administrative boundaries; (2) a total population of at least 150,000 people; and (3) up-to-date sampling frames that could be used to sample adults in the required age range. Additionally, it was intended to collect data from at least three areas in each region in order to reduce the confounding effects of countries and languages.

Random samples of subjects born between 1946 and 1971 were drawn in most of the centers from lists of census bureaus. Three thousand subjects in the age range of 20 to 48 yr were to be included in Stage I data collection and 600 in Stage II at each center. The complete study design has been published elsewhere (14). Briefly, this was a cross-sectional study using a two-step approach: In Stage I, a screening questionnaire standardized within the EC Respiratory Health Survey was mailed. Stage II comprised a detailed questionnaire, spirometric measurements, methacholine or bronchodilator inhalation tests, skin testing, and determination of total and specific IgE. For Stage II, a random sample of all individuals in Stage I was selected. The study protocol had been approved by the local ethics committees, and written informed consent was obtained from all participants.

Questionnaire

For the analysis described in this report, only answers of the index individual and what they reported about their siblings and parents were used. The 71 items of the standardized interview were developed from preexisting questionnaires (14, 15). The questionnaire used for the analysis had been translated from the English version into the various languages of the study centers, and was backtranslated into English after its pilot use with 30 volunteers to check for comprehensibility. All interviewers were trained to ask exactly the same questions.

The questions used in the analysis included: Have you ever had asthma? Do you have any nasal allergies, including "hay fever"? Have you ever had eczema or any kind of skin allergies? How many brothers (separate question: sisters) do or did you have? How many of your brothers (separate question: sisters) ever had asthma? How many of your other brothers (separate question: sisters) ever had eczema, skin or nasal allergy or "hay fever"? Did your mother (next question: father) ever have asthma? Did your mother (separate question: father) ever have eczema, skin or nasal allergy or "hay fever"? Further questions related to personal characteristics of the probands and smoking history in the family. Unfortunately, skin-prick test results were not available for all family members. Therefore, extrinsic asthma was defined as asthma plus any atopic diseases (eczema, skin, or nasal allergy as defined earlier), and intrinsic asthma was defined as asthma without any of these diseases.

Analysis

Data from 30 centers were available in June 1995, of which details were reported earlier (15). Of 15,449 randomly selected single data sets, only those with complete information on first-degree family members were used. This yielded data from 13,963 interviews concerning 75,392 individual persons. The number of persons included from each center is listed in Table 1, together with the response rates. The regional category (N = northern Europe, C = central Europe, M = Mediterranean region, O = non-European countries) is shown next to each center. The risk of having asthma is given as an odds ratio (OR) with the corresponding 95% confidence interval (95% CI). Additionally, the ORs for different subgroups of asthma probands compared with healthy subjects were investigated. These analyses were done with the SAS 6.11 software system (SAS Institute, Cary, NC) (16).

Beginning with the pedigrees of the probands, a segregation analysis was performed with the regressive logistic approach (17) implemented in REGD of the Statistical Analysis for Genetic Epidemiology (SAGE) computer programs (Case-Western Reserve University, Cleveland, OH) (18). This method represents an extension of the conventional logistic regression for family data, and allows genetic components to be incorporated under a simple Markovian dependence structure where the risk for an individual is written as a function of his or her own observed covariates and the phenotypes of his or her relatives (17). Since the index persons were randomly sampled, no ascertainment correction was necessary. Because of a size restriction in the data array of SAGE, only families with up to nine family members could be included, resulting in the omission of a further 593 families with 6,796 family members.

In the simplest regression models (Class A), common parentage alone accounts for sib-sib correlation. The risk of an individual being affected with asthma is a function of a baseline risk (), the risk based on observational covariates (X) for the individual, and the additional (residual or regressive) risk due to the individual's having an affected father, mother, or spouse (Z_F, Z_M, and Z_S are scores for the father's, mother's and spouse's phenotype, respectively, and are equal to 1 if the respective individual is affected and equal to zero if the individual is unaffected). Assuming an additive model, the log odds of asthma () for a person can be written as

(1)

where  represents the person's baseline risk, _F, _M, and _S represent the residual risk of having an affected father, mother, or spouse, respectively, that is not due to g, and  is the relative risk associated with each observational covariate in the vector X. Each parameter is on the logit scale and hence interpretable as a log-odds ratio. The penetrance is the probability of an individual's being affected, conditional on type, sex, the phenotypes of parents and spouse, and the covariates entered in the model. For different genetic and nongenetic dispositions the penetrance can be calculated from the estimated parameters using the formula

(2)

where  is given by Equation 1.

The baseline parameter  may depend on gender and an unobservable qualitative (genetic or nongenetic) factor, g, with a population frequency _g. The parameter g is an individual's "type," which allows for a major effect: either a major gene locus or a random major environmental effect. For a single autosomal two-allele (A,B) locus, g can take one of three possible genotypes AA, AB, or BB, and, assuming Hardy-Weinberg equilibrium _AA = q_A², _AB = 2 q_A(1  q_A), and _BB = (1  q_A)², where q_A denotes the frequency of allele A.

The transmission parameter _g is defined as the probability that a parent of type g transmits factor A, and is involved in the calculation of the likelihood of an individual's having asthma. The transmission parameter is generally allowed to range from 0 to 1, but is restricted to _AA = 1, _AB = 0.5, and _BB = 0 for Mendelian inheritance. Consequently, with different restrictions on these parameters (type frequency, _g; transmission probabilities, _g; baseline parameters, (g, sex); and residual familial effects, _F, _M, and _S), it is possible to generate various genetic and nongenetic models. For the general model, the type frequency (_g) must sum to one, with boundaries ranging from zero to one. The general model estimates the parameters without restrictions, and hence provides the standard against which the other models are compared. The environmental model of asthma constrains the three transmission probability or  values to be equal to one another (i.e., transmission independent of type). The "cultural" transmission model is defined by _AA = _AB = 1, and _BB = 0. The non-major-gene model postulates no genetic transmission of asthma, and the parameters q_A and  are therefore ignored in this model. In contrast, the genetic Mendelian models restrict the transmission probabilities to _AB = 1, _AB = 0.5, and _AB = 0, but allow the baseline risks to vary freely between zero and one. Dominant inheritance is modeled by requiring that the baseline risk for (AA) = (AB), recessive inheritance by requiring that (BB) = (AB), and codominant inheritance by requiring that (AB) = 0.5 ([AA] + [BB]). The last model may also be viewed as an "allele-dosage-model."

Smoking in a family's past was included as a covariate, since the symptoms of chronic bronchitis from smoking may interfere with asthma. Dividing Europe into three regions and aggregating areas outside Europe into a further region, we tried to adjust for the different ethnic subgroups with some common mode of inheritance. Correspondingly, in the segregation models, three dummy variables were added to account for the study region. To compare the different hypotheses, twice the difference in the log likelihood of the data under a specific hypothesis of interest (e.g., dominant model) is compared with the unrestricted general model using a chi-square test in which the degrees of freedom is the difference in the number of estimated parameters not maximized at boundary values. Additionally, Akaike's information criterion (AIC) was used to compare the models. AIC is given by

(3)

where lower values of AIC of a model correspond to a better fit to the data. The estimation of the parameters was done with a maximum likelihood iterative method, using the direct search method as implemented in the MAXFUN routine of the SAGE package (18).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Table 1 lists the main results of the analysis by study center. There was a wide range in the participation rates, ranging from 12.2% in Montpellier to 90.3% in Umea. Also, in some centers the number of participants exceeded the required sample size. More than half of the study centers had a participation rate higher than 65%. Owing to the restrictive sampling scheme, the age distribution was very similar in the centers. Women were slightly overrepresented (52.5%, as compared with 47.5% for men), a ratio that could be observed in almost all centers.

A total of 13,963 families were evaluable, resulting in 75,392 individual persons. The overall reported asthma prevalence was 6.9% (95% CI: 6.5% to 7.3%), with a range from 2.1% in Erfurt to 16.2% in Montpellier. The Mediterranean countries had slightly lower prevalence rates than did central Europe, whereas the northern and the non-European countries had higher asthma prevalances. Interestingly, the prevalence rates within countries also varied widely.

The overall reported asthma prevalence of 6.1% in the parental generation was 0.8% lower than that in the index generation. Taking into account additional late-onset cases experienced only in the parent generation, this difference would have been much greater if we had been able to compare age-adjusted prevalences.

Unexpectedly, the risk of a subject's developing asthma if a parent had asthma varied greatly by region, with an OR of 0.9 to 9.1. The risk if a mother had asthma (3.2; 2.6 to 3.9) was slightly higher than if a father had asthma (2.9; 2.4 to 3.5). The risk if at least one parent had asthma was lower in the Mediterranean countries (1.8; 1.2 to 2.9) and higher in the non- European countries (4.2; 2.9 to 6.2) than in central and northern Europe. The risk was more than doubled (7.0; 3.9 to 12.7) if both parents were affected by asthma. Although the risk estimates if both parents had asthma were very similar within Europe (northern Europe 5.7 [2.3 to 14.1], central Europe 7.7 [1.9 to 30.8], Mediterranean 6.2 [1.7 to 22.9], the risk estimates in non-European countries were about threefold greater with an OR of 14.1 [2.6 to 77.5]).

Because additional information on parental asthma was very limited, we could examine only an intrinsic and an extrinsic subgroup (Table 2). Extrinsic asthma in the parent appeared to be a much stronger risk factor for extrinsic (4.9; 3.9 to 6.0) than for intrinsic asthma in the offspring (1.5; 0.8 to 2.6). In contrast, intrinsic asthma in any parent was a risk for intrinsic asthma in the offspring, but to a lesser extent than for extrinsic asthma. A genetic component is therefore more likely in atopic asthma, but is also possible in nonatopic asthma.

There was no marked difference between males and females in asthma risk for paternal versus maternal extrinsic asthma. However, the genetic risk for female offspring was slightly higher than for male offspring. Extrinsic asthma in any parent was associated with a more frequent onset of asthma during the ages of 3 to 15 yr, whereas intrinsic asthma did not seem to be associated with a particular age of onset. There was no strong association between asthma and smoking in the family. However, this association is difficult to interpret, since respiratory symptoms in a family member may result in other family members quitting smoking.

Table 3 shows the results of the segregation analysis with a series of different models. Most of the specified segregation models can be rejected. Neither solely environmental nor cultural transmission of asthma is likely if we compare the AIC of these restricted models with that of the general model. The codominant type of Mendelian transmission with the allele frequencies in Hardy-Weinberg equilibrium, however, cannot be rejected whereas a clear dominant or recessive transmission is rejected. Hence, the pattern of asthma in the families considered is consistent with a codominant inheritance, but not with a dominant or recessive major gene. This is further supported by the codominant model's yielding the lowest AIC, and the similar baseline risks for males and females.

The overall gene frequency in the codominant model, calculated from Table 3, is 24.2%. Using Equations 1 and 2, we could estimate the penetrance given different genetic predispositions and covariate effects. It was low, with estimated values between 4.3% and 13.4%, in heterozygous gene carriers, but was considerably increased in homozygous gene carriers, with values between 32.6% and 70.0%. The strongest coeffect on penetrance was residence in a non-European country (OR = 2.6), and to a lesser extent residence in northern Europe (OR = 1.5).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have shown in this pooled analysis of 13,963 families from 30 centers of the ECRHS that the reported asthma prevalence was 6.9% among 20- to 48-yr-old persons and 6.1% in the parental generation. The risk of asthma for an offspring if a parent had asthma was more than threefold greater, and increased to sevenfold greater if both parents were affected. This risk was even larger in the non-European countries. Predominantly extrinsic asthma seems to have a genetic component. However, in intrinsic asthma the involvement of genes may also be possible. As shown in the segregation analysis, a major gene seems to have a high prevalence but low penetrance, and is hardly influenced by gender.

Several limitations of the study should be discussed. In asthma, as in other complex diseases, the clinical diagnosis does not have well-defined criteria, but depends more on individual experience and local diagnostic practice. Therefore, a transcultural comparison of asthma prevalence rates is difficult. This leads to the first main assumption of this report, that asthma is a disease category that is applied in a similar way in different areas, has a low genetic heterogeneity, and is frequently diagnosed. Further analysis of objective pulmonary function data from the ECRHS data will show if this assumption is correct.

The second assumption is that the selection of the study population and the disease reporting of the participants did not introduce any major bias. The large variation in the response rate, especially in low-responder areas where more severely diseased individuals may have been selected, could have resulted in a shift toward more severe asthma. On the other hand, a slight shift toward severe asthma makes it more likely that a gene effect will be detected if any exists. Because of the population-based sampling strategy, no further correction for any ascertainment bias was necessary. Temporarily excluding families in which the proband had asthma also did not significantly alter the risk estimates.

There are several possible explanations for the variation in risk for asthma within single countries. These include well-known methodologic problems, as discussed earlier (15). A genetic heterogeneity of asthma within countries is also possible. Further explanations include ethnic differences in allele frequencies or variation in exposure to common environmental factors such as air pollution, allergen load, or viral epidemics. In addition, reporting of diagnoses of asthma in parents may be more likely by index individuals who have asthma than by index cases who are not asthmatic. Attempts to understand these mechanisms, however, are at the moment speculative.

The risk for asthma in the present analysis was a bit higher than in previous reports in which the risk if any parent had asthma was reported as an OR of 1.5 (95% CI: 0.7 to 3.2) (recalculated data from a previous study [19]), 2.2 (1.4 to 3.3) (recalculated data from a previous study [20]), 2.2 and 95% CI: (1.2 to 4.0) [12] and 2.1 (1.4 to 3.1) [21].

The higher risk if a parent had extrinsic asthma also presents some interesting questions. Atopy, which may be transmitted independently of asthma, will confound any analysis of asthma genetics. To evaluate the two possibilities together will limit the solution to a polygenic mode of transmission. Excluding allergic disease, however, may also have disadvantages, since misclassification between intrinsic asthma and other obstructive diseases is also likely. The lower genetic risk for intrinsic asthma may partially originate from this diluted category. The classification of asthma as extrinsic on the basis only of a concomitant allergic disease is a less sensitive classification than the usual criterion of sensitization assessed by skin-prick testing. Therefore, it is uncertain whether intrinsic asthma is really a separate pathophysiologic disorder, as suggested by the risk distribution in Table 2.

Segregation analysis has been been used successfully in various diseases, such as cleft lip and/or palate (22) and hyperactivity (23). In cases of an excessive high IgE response, a major codominant gene controlling IgE in humans has been reported (24), in contrast to earlier reports in which a recessive gene was assumed (25, 26). To our knowledge there are only two reports on the segregation of asthma (27, and Jenkins, personal communication at a workshop in Paris, 1994), and one on that of bronchial hyperreactivity (BHR) (10). Although the latter analysis of 83 families indicated a clear familial component but excluded a single autosomal locus, the studies done by Holberg and colleagues rejected all specified models in 673, respective 906 families as reported subsequently (27, 28). Holberg and colleagues concluded that the inheritance of asthma is unlikely to be the result of a single major two-allele locus, whereas total serum IgE is controlled by a major autosomal codominant gene. Main restrictions of the study were its low predictive power, the selection of the families through a health maintenance organization, and an age range of the offspring that may be too low for a definite diagnosis of asthma. Jenkins presented pedigree data for 7,394 families, in which 15.9% of the index individuals had asthma. In this analysis, the general major gene model had the best fit, however, as in our analysis the AIC for the codominant model achieved the lowest value. The authors concluded that asthma is not fully described by a single-gene model. Interestingly, a major gene model with codominant transmission of the pulmonary function parameter FEV₁ has been reported as the best-fitting model in patients with chronic obstructive pulmonary disease (COPD) (29).

Any segregation analysis has constraints leading to the third major assumption about the results presented here. As in all segregation analysis, failure to reject the best-fitting model does not automatically prove the alternative model. Therefore, the results in such a case should be viewed with caution. Despite taking into consideration confounding factors, not all possible interactions can be controlled in the analysis. Unfortunately, the type of asthma (extrinsic/intrinsic) in siblings was not known in our study. For this reason, a subgroup segregation analysis of extrinsic or intrinsic asthma could not be performed. Other subgroup analyses, mainly by study center, were also not successful because of the low number of subjects in each subgroup.

Furthermore, there are some inconsistencies in the results presented in Table 3, since the  values in the general model are different from those in the Mendelian hypothesis. Obtained after many sets of initial estimates in the general model, however, these  values gave the highest likelihood of the transmission of asthma to an individual, and the chi-square statistic shows that this likelihood is not significantly different from the Mendelian likelihood. The general model also fitted the data significantly better than did a nongenetic model, in which no parent-offspring transmission and no major gene or sex effect is assumed (results not shown). Furthermore, the general model fits the data significantly better than the environmental or the cultural model. This leads to the conclusion that a genetic model is compatible with the data indicating that codominant inheritance might be a mode of inheritance.

The study of genetic factors in asthma shares many common features with that of genetic factors in other complex diseases such as schizophrenia. As recently summarized (30), there are problems with the definition of the phenotype. Although a single major gene effect could not be excluded by the present analysis, a "liability/threshold" model is supported by several other facts. For example, the risk of asthma increases with the number of relatives affected, and the number of affected relatives increases with the disease severity of affected index individuals (data not shown). It may be useful in further studies of asthma to apply a statistical model of inheritance, taking into account a multilocus model and various environmental effects. The currently available models assume an additive interaction between loci, whereas epistatic effects may be more appropriate for asthma.

The most exciting advance will be the understanding of the molecular-genetic basis of asthma by means of new molecular technologies and computer resources. This is currently the goal of at least six studies designed for an approach called positional cloning, which has already been successfully used for the high IgE phenotype (31). Any exception of the independent assortment of two genes at meiosis makes it likely that the genes are closely aligned on a chromosome. Examining this linkage of structural variants of DNA (microsatellites) and BHR or asthmatic phenotype may reveal potential chromosomal regions associated with asthma. The first preliminary results of such an approach in humans have been reported in The Netherlands (32) and in Australia (33), and recently in two genome-wide approaches (34, 35). Genes identified through the proxy of the linked microsatellites can then be searched for disease-causing mutations (36, 37). The results of the present study suggest that at least one major locus exists that may affect the risk of asthma. Despite many methodologic problems with the definition of a complex and multifactorial disease, pursuing this concept will be a promising lead toward a better understanding of the pathophysiology of asthma.