INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In recent decades asthma has become one of the most common chronic diseases in children and adolescents in industrialized countries (1, 2). It is generally believed that genetic predisposition is important in asthma, whereas environmental factors are assumed to trigger and modify the expression of the disease. The rapid increase of asthma in developed and urban environments is more likely explained by changes in environment and life style rather than by changes in gene pool (3). In this new situation it is unclear whether the significance of genetics has diminished in the development of asthma.

Twins provide a unique setting to analyze the proportion of genetic and environmental factors in the development of a disease. This can be done by comparing the concordance of a disease in genetically identical monozygotic twins (MZ) with that in dizygotic twins (DZ) who represent full sibs. The comparison is based on an assumption that the environmental exposures are not different between MZ twin pairs compared with those among DZ twin pairs. On the other hand, the mode of inheritance can be studied only in family settings that have shown that parental atopy increases the risk of asthma and other atopic diseases in offspring (4, 5). There is also some evidence that a specific pattern of allergic symptoms in parents increases the risk of the same pattern in offspring (6, 7). With respect to the statistical power, studies of twin families are from two to five times more informative than are conventional family settings (8). Furthermore, the twin-family design allows better distinctions between the influence of shared cultural inheritance and biologic inheritance than the traditional family design (9). There are no previous twin-family studies published on asthma. In population-based twin pair studies, the estimated effects of genetic factors in asthma have ranged from 35 to 70%, depending on the population and study design (10).

Like other complex disorders, asthma does not follow a simple Mendelian pattern of inheritance. Although the clinical picture of the disease is very heterogeneous and the transmission mechanisms are poorly understood, genetic mapping has been initiated to identify the genes that contribute to asthma. These studies have already revealed linkage of bronchial hyperreactivity (13) and serum immunoglobulin E levels to chromosome 5q31 (14, 15), and linkage of atopy to point mutations in the high affinity receptor of immunoglobulin E on chromosome 11q13 (16). This gene-mapping approach is completely genetic and based on the systematic comparison of inheritance patterns between affected and unaffected individuals. Thus, it may yield new information on the molecular pathogenetic mechanisms involved in asthma independently of descriptive disease studies. At present, the genes predisposing to asthma remain unknown.

The present twin-family study assesses the importance of true genetic factors in the development of asthma in adolescence and gives some new guidelines for recruitment of patients for genetic studies in the future.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Samples and Measures

The Finnish Twin Cohort Study is a project for the study of genetic and environmental determinants of chronic diseases and risk factors (17). Twin pairs born between 1958 and 1986 have been identified from Finland's Central Population Registry (17). For twins born from 1974 onwards, the number of twins identified from the Central Population Registry is nearly identical to the number of twins born annually according to birth statistics compiled by the Central Statistical Office of Finland (17).

Data were collected by questionnaires. After a pilot study, a questionnaire was mailed to twins born in 1975 through 1978 within 2 mo of their sixteenth birthday inquiring about similarity in appearance, health habits, symptoms, as well as personality and social relationships. The parents of the twins were mailed a questionnaire about the birth, development, and growth of the twins during early childhood and adolescence until 16 yr of age. Questions on chronic conditions and diseases such as asthma were included. The parents were also asked to reply separately to a questionnaire on their own health and life style, including a question on asthma. A total of 2,483 families were contacted and 2,023 questionnaires on twins' childhoods were returned. Individual response rates were 93% for girls, 87% for boys, 86% for mothers, and 82% for fathers.

Asthma in twins was defined on the basis of the parental response in the family questionnaire as to whether asthma had been diagnosed in either, neither, or both of the twins. The question was left unanswered in 213 questionnaires, and hence these twins were not used in the analyses. The mother and the father were also asked if asthma had ever been diagnosed in either one of them by a physician. Asthma status data were available for both parents in 1,648 families.

Twin zygosity was determined by examining the responses of both members of each twin pair to two questions on the similarity of appearance at primary school age. By using a set of decision rules described earlier (18), the twin pairs were classified as MZ, DZ, or undetermined zygosity. This method has been validated among twins 18 yr of age or older by using 11 genetic markers (18). Twins with undetermined zygosity were classified by discriminant analyses using the data transformations and an analytic approach described earlier (19). Additional information such as that available from photographs, fingerprints, and DNA-marker studies was used for zygosity determination when needed. Uncertainty about zygosity remained for 96 twin pairs, who were consequently excluded from the analyses. The final series was 1,713 twin pairs.

Statistical Methods

Standard univariate twin analyses were used to describe patterns of familial aggregation of asthma. As to descriptive statistics, probandwise concordance rates were computed, and tests of homogeneity of prevalence rates across twin types were performed. Twin similarity was assessed using the probandwise concordance rate (20, 21), and tetrachoric correlations were computed from pairwise contingency tables. Additionally, the parents' pairwise similarity for asthma was analyzed. The risk of asthma in the offspring was assessed after stratifying the twin data on parental asthma status. The effect of the parental asthma status on the occurrence of asthma in their offspring was examined for each parent separately and independently of the other parent.

Model-fitting was carried out to estimate the components of variance in the liability to asthma. The model assumes that the genetic component is polygenic, i.e., the result of many independent Mendelian genes acting additively to control liability. The multifactorial threshold model further implies that the unobserved liability is normally distributed, with the observed disease status determined by an unobserved threshold selected to match the overall prevalence of the disease (20). The disease becomes manifest when a certain level or threshold of liability is reached. These assumptions were considered reasonable for two reasons. Firstly, the intrapair correlation for asthma liability was considerably less than unity in MZ twins and less than double in DZ twins, indicating that genetic effects are probably additive. Secondly, no single major gene controlling the expression of asthma has been identified. The modeling was first carried out on unstratified twin-data and then on twin-data stratified according to parental asthma status.

From data on twins reared together it is possible to model four separate parameters: an additive genetic component (A), effects caused by dominance (D), common environmental components (C), and unique environmental components (E). Common environmental effects are those shared by both twins or all family members, whereas unique environmental effects are uncorrelated between family members and represent the exposures of each individual. Unique environmental effects include also measurement errors, for example, in the classification of asthma. One can fit models based on the different combinations of these parameters: AE, ACE, ADE, and CE, but effects caused by dominance and common environmental effects cannot be simultaneously modeled (20). The variance components estimates are calculated using the contingency table in different zygosity groups. Chi-square (²) goodness-of-fit statistics were used to assess how well a model fits the data. The goodness-of-fit of alternative, hierarchically nested models can be compared by the difference in ² values of those models. The difference in two such ², goodness-of-fit statistics is itself distributed as an ² statistic with degrees of freedom equal to the difference in the degrees of freedom of the two models being compared. Analyses were carried using Mx, a statistical program for genetically informative data sets (22).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Descriptive Statistics

The cumulative incidence of asthma at the age of 16 was higher among boys (4.7%) than among girls (3.1%), whereas there was no difference in incidences by twin type (Table 1). Monozygotic pairs in both sexes were more concordant than dizygotic pairs. Within zygosity groups, male pairs were more concordant than female pairs, as indicated both by the probandwise concordance rates and the tetrachoric correlations of liability. The probandwise concordance rate for MZ pairs (42%) indicated that the monozygotic cotwin of a twin with asthma had an eleven-fold higher risk of asthma compared with the population at large, whereas for a dizygotic cotwin the risk was fivefold higher. Asthma concordance in opposite sex DZ pairs was not different from the concordance in like-sexed DZ pairs (Table 1)

Asthma was reported by 4.6% of mothers (n = 76) and 3.8% of fathers (n = 62). In only eight families did both parents have asthma, and for spouse pairs the tetrachoric correlation of asthma was 0.27 (95% confidence interval = 0.08 to 0.46). The parents were somewhat more similar for asthma than would have been expected on the basis of random association alone. The effect of the parental asthma status on the occurrence of asthma in their offspring was examined for each parent separately and independently of the other parent (Table 2). Because there was no significant difference in the risk of asthma between the boys and the girls, the data for both sexes were pooled. Asthma was reported for 11% of the children of asthmatic mothers and 10% of children of asthmatic fathers, whereas only 3% of the children of nonasthmatic parents had asthma. The risk of a child to be asthmatic was four times greater if either the mother or the father had asthma than if the child had been born to nonasthmatic parents.

Model-fitting

Alternative models fitting additive genetic (A) and common (C) and unique (E) environmental sources of variation to the data on the twins are shown in Table 3. The ACE and the AE models explained the data statistically equally well. In the more parsimonious AE model (i.e., the one that accounts for the data with fewer parameters), the genetic effect accounted for 79% and unique environment for 21% of the variance in the development of asthma. In the ACE model, the proportion of genetic effect was 65%, whereas the effects of unique and common environments were 23 and 12%, respectively.

Model-fitting on twin data stratified according to parental asthma is shown in Tables 4 and 5. In the most parsimonious model (AE), among the twins with at least one asthmatic parent, the genetic effects accounted for 87% and unique environmental effect for 13% of the variance in liability. In the less parsimonious ACE model, the estimate for common environment was zero; thus, the proportions of genetic and unique environmental effects were equal to those in the AE model. We were able to exclude both models, including only environmental effects (CE and E). In the twin pairs with nonasthmatic parents (Table 5), the proportion of variance caused by genetic effects was lower than that in the above-mentioned twins with at least one asthmatic parent, whereas environmental factors accounted for a higher proportion of variance in susceptibility to asthma. The estimates for the genetic effects in the ACE and AE models were 52 and 76%, respectively. In contrast to the first group, the CE model including only environmental effects explained the twin data equally well. Given the low occurrence of asthma among the offspring of nonasthmatic parents, we had insufficient statistical power to reliably distinguish between the AE, ACE, and CE models. Only the model comprising unique environmental effects (E) alone was rejected.

To test whether the proportions of variance caused by genetic effects were significantly different in the families with and without affected parents, a chi-square test of heterogeneity was computed by fitting the data from both types of families to an AE model constraining variance components and thresholds to be the same over family type. Subtracting from the combined chi-square, the chi-squares for the AE models in Table 4 and 5 yielded a test of homogeneity (² = 34.91, degrees of freedom = 2, p < 0.0001), indicating that a model with the same genetic and environmental effects cannot account for the twin data in both family types.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We present for the first time combined twin-family data on the inheritance of asthma. Additionally, the twin design includes several improvements over earlier studies in this area (10, 12, 23). Instead of including several age groups, the twins were all studied at the same chronological age of 16. At this age, nearly all were living together with their parents in the same household, and they were not yet occupationally active. Thus, the differences in environmental exposures within the twin pairs were probably much smaller compared with that in adults. In the adult population, one can also expect that the effect of heredity is lower because of longer exposures to different environmental risk factors provoking asthma. In Finland the diagnostic criteria of asthma are uniform, being based on the definition of asthma by the American Thoracic Society (24). Recent Finnish epidemiological studies have shown an increasing, but still fairly low, prevalence of asthma among teenagers compared with many other Western countries (25, 26). Because the recruitment of twin families was nationwide with a high response rate and because the incidences of asthma in twins and in Finnish teenagers who were not twins were equal, it can be assumed that in respect to asthma the twins studied here are representative of 16 yr-old adolescents in Finland in general.

Previous epidemiological studies have shown that questionnaire-based studies are able to give reliable data on asthma morbidity in large population-based studies (27, 28). In Finland, the older part of the twin cohort showed that self-reported, physician-diagnosed asthma predicts asthma mortality with a sensitivity of 92% and with a specificity of 98% (29). Recently a questionnaire-based study on asthma and allergy showed that all cases of self-reported asthma among Finnish university students were verifiable by clinical studies (Marita Kilpeläinen, personal communication). Even though the symptoms of asthma are well known both by Finnish physicians and by patients, it is still possible that mild asthma is underdiagnosed. Obviously unreported cases of asthma are missed when an approach like ours is used. In this study design, as the proportion of true negatives (healthy subjects) is much larger than the proportion of true positives (asthmatic subjects), the effect of false negatives remains of minor importance in the results of model-fittings. On the other hand, more detailed phenotyping (serum IgE level and skin prick testing, bronchial hyperreactivity) would have provided additional evidence for or against asthma while allowing at the same time the analysis of phenotypes associated with asthma. However, these quantitative traits most probably also represent very heterogeneous entities and have a limited value in predicting the clinical symptoms of asthma (30). Possible random classification errors will increase the unique environmental variance component in genetic modeling of the disease. The symptoms of asthma may be also recognized more easily in the families with already one affected family member. Although this possible confounding factor could increase concordance and aggregation of asthma into families, it should be equal for MZ and DZ twins, reducing the difference between the twin groups and thus be seen as a common (family) environmental effect. This ascertainment bias should also increase the incidence of asthma in the study population compared with the normal population.

Differences in study designs make it very difficult to compare even the results of population-based twin studies to each other (10, 12, 23). Even though concordance rate is dependent on the prevalence of the disease in each study population, the proportions of MZ concordant pairs compared with DZ concordant pairs have been approximately two in all studies, suggesting the importance of inherited factors. A previous twin study of the adult Finnish population has shown a 44% genetic effect in liability to asthma, which was considerably lower compared with the genetic effect of 79% in our analysis of adolescent asthmatics (parental asthma status ignored) (11, 12). These variance components remain comparable despite the difference in incidences of asthma in the younger (3.9%) and in the older (1.9%) twin cohort. In Australia, where asthma is more frequent than in Finland, the genetic factors have been shown to account for 74% in men and 58% in women of the liability to asthma and wheezing illness (10). The importance of genetic factors has been shown also in a recent study of Norwegian twins (18 to 25 years of age), where 75% of the variation in asthma was explained by genetic effects and the remaining variation was due to nonshared environmental influences (31). It is obvious that despite the increase of asthma, genetic factors are still of great importance in the development of asthma in adolescence. However, the relative influence of genetic and environmental factors may differ in other populations and at other ages.

Although the genetic modeling showed a strong impact of genes on asthma morbidity in adolescence, parental asthma was found to be a good indicator of a subtype of asthma predominantly genetic in origin. In the families with asthma in successive generations, genetic factors alone explained as much as 87% of the development of asthma in the offspring. These families differed in several aspects from the families in which the parents were unaffected. The cumulative incidence of asthma in twins with affected parents (14.7%) was fourfold compared with the incidence in twins without affected parents (3.3%). Also, the genetic models fitted differently in the two types of families, with a somewhat higher relative proportion of variance attributed to genetic factors and no evidence of shared factors in the affected parent families. In the families with unaffected parents, environmental factors alone were sufficient to explain the pattern of twin resemblance, even though a model with genetic effects also adequately fitted the data. The proportion of genetic control in susceptibility to multifactorial diseases can be estimated only by comparing the morbidity in two groups of patients whose proportions of genome identical by descent are different and known. The results of genetic modeling showing differences in morbidity among the twin pairs in the above-mentioned families suggest that asthma is accumulating in families more likely because of shared genes than of shared environmental risk factors. Compared with other Europeans, the Finns are culturally and genetically an exceptionally homogeneous population, which might also facilitate to some extent the identification of these two subtypes of asthma (32).

Systematic gene mapping and positional cloning have proven to be powerful tools to discover the genetic defects in classic Mendelian diseases (33). In the future, genetic studies might also be capable of producing valuable information on molecular pathogenetic mechanisms in such complex diseases as asthma, where the interactions between genes and the environment are extremely complex (33). In our study, the risk of an MZ twin of developing asthma if the cotwin is already affected at the age of 16 was much higher than in the general population. But on the other hand, only one third of the MZ pairs were concordant, even though they share the whole genome and exceptionally many environmental factors during their foetal development and early childhood. The presymptomatic prediction of the disease will remain probabilistic in the future even if the relevant genes are identified (34). Our results indicate that when patients are selected for genetic studies, a positive family history is of great importance. Early onset of asthma might also be a beneficial criteria. The ability to analyze genetic architecture in these particular families may enhance the chances of revealing the pathogenetic mechanisms involved.