The study population included 348 asthmatics and their family members and 416 control subjects. Probands (asthmatics and controls) aged 7 to 65 were ascertained in five French cities (Paris, Lyons, Marseilles, Montpellier, and Grenoble). All pro-bands and their two parents were born in France. In roughly 10,000 self-completed questionnaires (6,400 for potential cases and 3,600 for potential control subjects), 1,589 probands fulfilled inclusion criteria regarding place of birth and family structure (either proband with a spouse and at least one child, or proband with at least one sibling and two parents). A positive answer to four questions (Table 1) identified those with asthma. A total of 278 probands fulfilled this criterion. Seventy more who only answered positively to two or three questions were also included after examination of their medical record. Subjects with asthma comprised 213 adults and 135 children. Children control subjects were recruited in surgery hospitals. Two thirds of adult control subjects were population based (electoral rolls), and one-third were from surgery hospitals or check-up centers. Overall, half of the controls were population based. Cases, relatives, spouses, and control subjects totaled 1,854 (Table 2).

Subjects answered a detailed questionnaire on upper and lower airway symptoms, allergic symptoms, childhood events including infections, medical history, and environmental factors, using British Medical Research Council/European Coal and Steel Community, American Thoracic Society, and European Community Respiratory Health Survey (ECRHS) questionnaires with additional questions when needed. Environmental data included active smoking (age at start, amount) passive smoking (in utero, during childhood, at present, at home and at work), presence of domestic animals during childhood and at present, pollen exposure (countryside house, month of birth), area and month of survey, indoor moisture, home characteristics related to dust-mite exposures (carpet, curtains, etc.), type of heating and cooking fuel, occupational exposures known to be related to respiratory diseases, and occupation and industry for the last two jobs.

For subjects with FEV₁ > 80% predicted, a methacholine bronchial challenge test was performed (maximum dose, 4 mg). For others, a bronchodilator test was performed. Skin-prick tests to 11 allergens, total immunoglobulin E (IgE) levels, multiRAST Phadiatop test, leukocyte count, and standard differential count were performed. Bronchial testing and allergen skin-prick tests were performed according to the European Respiratory Health Survey protocol (16).

Criteria chosen to define asthma were as simple and standardized as possible. The phenotype considered was based on the answers to four questions (Table 1), all from international standardized questionnaires, with the idea of having undisputable asthma according to questionnaire data (favoring specificity over sensitivity). A self-completed questionnaire was used to avoid discordances among medical practices.

A procedure was set up to include true asthmatics who may not answer positively to the four questions due to specific aspects related to hospital ascertainment (no to “attack in the last year” in those who recover after a very severe attack the previous year, no answer to “doctor's diagnosis of asthma” in new patients). The clinical center was asked to report any pertinent element of the medical record to substantiate the diagnosis, information which was then anonymously submitted by mail to all centers, asking for a conclusion about asthma. Five successive sets of patients were assessed. Clinical assessment by seven centers was obtained (except for one set, for which only five assessments were available) for each potential case. The inclusion decision was finally based on the analysis of the seven clinical assessments by an epidemiologist (F.K.). Answers from the centers were separated into “yes,” “no,” and all others (“maybe” or any comment). When most centers said no, the patient was excluded without further examination; when all or all but one said yes, the patient was included. In the other cases, the responses were analyzed, taking into account the question(s) for which a positive answer was not obtained. In five series, 126 subjects were evaluated and 108 of them included (86%). Out of the 348 examined cases, 70 subjects (20%) entered the study. The reproducibility of the procedure, assessed on a subset of 33 patients chosen at random (Table 3) was poor when considering separately the evaluations of each center as shown by the κ values (Table 3). The κ values were similar when considering patient evaluation based on three classes (no, maybe, yes) or based on two (yes versus [no + maybe], or [yes + maybe] versus no). However, there was good reproducibility of the decision-making procedure, with κ = 0.71.

The definition of asthma was chosen to be more specific in relatives of probands, in order to build a sample of affected sibling pairs for linkage studies (Table 1). Initially, we considered BHR a prerequisite. However, as subjects with FEV₁ ⩽ 80% predicted did not have the bronchial challenge, and because of the effect of therapy on the degree of BHR, it was not possible to get a sufficient number of families. Therefore, treatment for asthma, hospitalization, or a positive test to bronchodilation were also included in the criteria (Table 1). Seventy-four families among the 348 studied fulfilled the criteria described in Table 1. In the whole data set among 138 adult relatives who fulfilled the criteria of the phenotype with BHR, bronchodilatation, hospitalization or asthma therapy, the great majority fulfilled criterion for asthma therapy, eight entered only through the bronchial hyperresponsiveness criterion, one through the hospitalization criteria, and none only through the bronchodilator criteria. Among the 72 child relatives fulfilling the criteria, only five entered through the bronchial hyperresponsiveness criterion alone, and none through the bronchodilator criterion. Thus, the most frequent criteria were asthma diagnosis and treatment for asthma.

For association studies, a panel of subjects without asthma was selected. Ninety adult subjects were nonasthmatic and strictly nonatopic (total IgE < 40 IU/ml, no wheal > 0 to any of 11 allergens tested by skin-prick tests, and younger than 50). Negative answers to the four questions were used to exclude asthma (Table 1). The cutoff point for IgE was chosen to get in the adult controls the same prevalence of low IgE as lack of atopy. Instead of adjusting IgE level or skin-test response for age, these criteria were defined in subjects < 50 yr in order to have very simple inclusion criteria.

Phenotypes other than usual asthma need to be considered. Phenotypic heterogeneity may help to disentangle etiologic heterogeneity. Severe asthma represents a clinical subphenotype of interest, and studying severe asthma (extreme phenotype) may increase the power to detect linkage. Clinically-based ascertainment likely selects a more severe group of asthmatics than population-based ascertainment. It would be useful to define asthma phenotypes unencumbered by the activity of the disease, which may depend heavily on nongenetic factors, and free of gene-trigger interactions. Besides the activity, environmental factors and inadequate treatment can modify the severity of the asthma phenotype. Consideration of treatment as a marker of severity implies that the relevant phenotype for etiological research is masked by the treatment. Figures from the EGEA study or other clinically and population-based studies should help to address quantitative markers of severity.

Most asthmatics recruited for genome-wide search have childhood-onset asthma (2, 5). Data from the EGEA study will allow us to analyze both childhood- and adult-onset asthma, which will be interesting because asthma may be a developmental disease. Smokers have often been excluded (2, 5) yet we need to understand smoking's effects on the inflammatory pattern of asthma (9). Intrinsic, nocturnal, aspirin-sensitive asthma, and other clinical forms may be relevant subphenotypes based on clinical classifications.

Intermediate phenotypes are important with respect to pleiotropy and etiological heterogeneity (Figure 4). Bronchial hyperresponsiveness, total IgE, allergen skin-test response, and to a lesser extent, eosinophilia are usually considered intermediate phenotypes for asthma (1-3). Definitions of BHR vary among studies. A PD₂₀ = 2 mg (maximal dose in the ECRHS) was considered in an Australian study (18), whereas BHR was defined by a PD₂₀ < 25 mg/ml in the CSGA (5) and PD₂₀ < 32 mg/ml in the Dutch study (2). In EGEA, most of the asthmatic cases with BHR had severe BHR, as three quarters of those with BHR decreased their FEV₁ by 20% at a dose ⩽ 0.25 mg. The way of handling missing data in BHR is particularly critical. For asthmatics, etiologic risk factors of those with BHR and those with low FEV₁ are likely similar, but that hypothesis is unlikely to be true for other subjects. Epidemiological reports have introduced quantitative variables for BHR (19, 20), and the slope-response to the agonist is now used in the epidemiological and genetic literature (3). Distinction of sensitivity from reactivity was proposed 20 years ago (21), and factors related to reactivity and sensitivity may be different (22). Work is planned in the EGEA study to model the dose-response curve of the bronchial provocation test in order to better define various dimensions.

Fig. 4. Intermediate phenotypes (P) under the hypotheses of pleiotropy and etiologic heterogeneity.

Download Figure | Download Powerpoint

[More] [Minimize]

To dichotomize skin-prick test response, various thresholds have also been proposed. Such variability in part reflects the difference in population prevalences (high in clinical settings, low in population-based studies) for studies evaluating their predictive value for asthma or allergic diseases. More than 90% of adult cases had atopy or high IgE or eosinophilia. Because, in this descriptive phase of the study, IgE and eosinophilia were not age-adjusted, figures only allow estimates of intrinsic asthma in adults, although it is likely to be very rare in children. Few reports (23, 24) have tried to build quantitative scores that consider the size of the responses from the battery of skin tests. Following the hypothesis of a general atopy phenotype, a composite score taking into account IgE, skin-prick test, and eosinophilia has been proposed (3). Besides quantitative or composite scores, phenotypes may be built to delineate various aspects of allergic markers. There is increasing evidence that the associations of skin-test responses, total IgE, and eosinophilia with environmental and clinical factors are different (25). Some studies attempting to delineate basal IgE level from specific response to allergens (by adjusting for or excluding specific IgE response) suggest that different genes control specific and nonspecific IgE production (1, 6). The existence of intrinsic asthma has been challenged by Burrows and colleagues (26), who showed that any asthma was related to high age-adjusted IgE. Asthma has even been called eosinophilic bronchitis. All so-called intermediate phenotypes do not play the same role. The role of the biologic allergic markers is hypothesized to be intermediate in a causal pathway for some or all asthmas, whereas BHR is considered a hallmark of asthma, i.e., a manifestation of the disease, or even a definition for asthma (27).

Considering environmental factors and potential interactions with genetic factors may increase our ability to detect genetic factors in multifactorial diseases such as asthma. Suitable strategies need both good design and good analysis. Analyses may be conducted on adjusted phenotype scores, by including environmental factors as covariates in genetic models, or by excluding subjects exposed to environmental factors such as smoking. This has been done in several studies (3, 5); conversely the analysis could be restricted to exposed subjects. Comparing subjects who have asthma in the absence of exposure with healthy subjects in the presence of exposure may be relevant for association studies, under the hypothesis of additive effects of genetic and environmental factors (28). Under the hypothesis of strong genetic-environment interaction, searching for genetic factors among exposed subjects with and without asthma may increase the study's power. This approach could be taken in EGEA.

Statistical modeling such as classification procedures, or principal component analysis might suggest various phenotypic dimensions (13), but is most relevant when a priori hypotheses regarding heterogeneity are not easy to formulate within related homogeneous variables. Previous segregation analyses may also help in the choice of subphenotypes to favor those showing higher familial aggregation or those most likely to be determined by a major gene effect (5).

Phenotype constructs that consider phenotype heterogeneity may be built under the hypothesis of pleiotropy, i.e., of one latent common cause, or under the hypothesis of etiological heterogeneity. Hypotheses of various phenotype constructs need to be more explicit. Whereas both lumping and splitting approaches have advantages, it seems that there has not been enough work on the splitting hypothesis for asthma. Standardizing primary data and subphenotypes is a prerequisite for comparing different studies and dissecting the complex mechanisms underlying asthma.

Respiratory epidemiology. I. Annesi, F. Kauffmann (coordinator), M. P. Oryszczyn (INSERM U169, Villejuif); F. Neukirch, M. Korobaeff (INSERM U408, Paris).

Genetics. M. H. Dizier, J. Feingold (INSERM U155, Paris); F. Demenais (INSERM U358, Paris); M. Lathrop (INSERM U358, now at the Wellcome Trust Center of Human Genetics, Oxford).

Clinical centers. Grenoble—I. Pin, C. Pison; Lyon—D. Ecochard (deceased), F. Gormand, Y. Pacheco; Marseille—D. Charpin, D. Vervloet; Montpellier—J. Bousquet; Paris Cochin—A. Lockhart, R. Matran (now in Lille); Paris Necker—E. Paty, P. Scheinmann; Paris Trousseau—A. Grimfeld.

Data management. J. Hochez (INSERM U155); N. Le Moual (INSERM U169).