In 1875, Galton [12] pointed out the value of twins for studying the relative importance of heredity and environment. Subsequently, twin studies have proved to be a valuable method for investigating the genetic component of traits. The basic tenet is that if a trait (for example, H. pylori infection) has no genetic component, the trait is displayed equally by both members of monozygotic twin pairs and by both members of dizygotic twin pairs. Greater similarities within monozygotic twin pairs suggest the involvement of genetic effects.

One of the most powerful study designs to examine the relative importance of genetic and environmental variation for individual differences in susceptibility to infection is to compare the results of identical twins separated at an early age with those of twins who were reared together. By comparing twins reared together with twins reared apart, it is possible to estimate the importance of rearing environment on twin similarity.

We investigated the relative importance of genetic and environmental factors on the acquisition of H. pylori infection by studying twins separated at an early age and reared apart and twins reared together [13].

Methods

Top Methods Statistical Analysis Results Discussion Author & Article Info References

Sample

We studied twins from a subregistry of the Swedish Twin Registry, which includes entries for about 25 000 twin pairs born in Sweden between 1886 and 1958. The subregistry consists of a number of twins who indicated that they were separated before the age of 11 years (and were reared apart) and of a sample of twins who were reared together; all these twins were matched on the basis of sex, date, and county of birth [13]. This sample is used in the Swedish Adoption-Twin Study of Aging. When this aging study was initiated in 1984 with a mail-out questionnaire, both twins were alive from 591 twin pairs reared apart and from 627 twin pairs reared together. Two years later, a subset of twin pairs reared together and reared apart in which both members responded to the first questionnaire was invited to participate in an examination that included health evaluation and administration of cognitive tests. Both members of 291 pairs who were 50 years of age or older and 12 pairs younger than 50 years participated in this in-person testing phase. This represents 60% of all pairs 50 years or older in which both responded to the first questionnaire. If one twin in a pair declined participation in the in-person testing phase, the co-twin was not invited to participate (resulting in a lower pairwise participation rate). A follow-up of twins declining participation indicated that cognitive decline and poor health were the major reasons for unwillingness to participate. The average age at the time of testing was 65.6 years (SD, 8.4); 60% of the twins were women.

The present analyses include 36 pairs of monozygotic twins reared apart, 64 pairs of monozygotic twins reared together, 88 dizygotic pairs reared apart, and 81 dizygotic pairs reared together for whom serum samples were available. Zygosity diagnoses were first made using physical similarities and were confirmed using the serologic assay. One pair was excluded from the analyses because adequate certainty of zygosity (>98%) could not be obtained. The distribution of age at separation is highly skewed: 52% of the twins reared apart were separated before their first birthday, 69% by their second birthday, 82% by the age of 5 years, and all by the age of 11 years. Reasons for separation varied; most were separated because of the death of one or both parents or economic hardship (or both). Further details of the procedures, sample, and design of the Swedish Adoption-Twin Study of Aging are described by Pedersen and colleagues [14].

Helicobacter pylori Assay

The presence of anti-H. pylori IgG was assessed using the commercially available high-molecular-weight cell-associated protein H. pylori immunoassay (Enteric Products Inc, Westbury, New York). The test was scored positive when the optical density was 32.0. The test has a sensitivity greater than 98% and a specificity greater than 95% [15]. Helicobacter pylori infection is chronic and even lifelong. After successful antimicrobial agent therapy, the antibody titer is not detectable within a year; thus, the presence of serum antibody is a reliable indication of H. pylori status.

Statistical Analysis

Top Methods Statistical Analysis Results Discussion Author & Article Info References

As a preliminary indication of the importance of genetic effects for H. pylori infection, probandwise and pairwise concordances were calculated for monozygotic and dizygotic twins, reared apart and reared together. A greater concordance for monozygotic pairs than for dizygotic pairs suggests the importance of genetic effects. Pairwise concordances were calculated as the percentage of pairs concordant for expressing H. pylori among the total number of pairs in which at least one member was infected with H. pylori. Probandwise concordances represent the number of affected twins in concordant pairs divided by the number of proband patients infected with H. pylori. The latter form of concordance is more genetically meaningful because it can be compared with risk figures for other family groups.

The primary aim of the analyses was to describe individual differences in the acquisition of H. pylori infection by partitioning these differences (total variance) into genetic and environmental variance components. Quantitative genetic theory posits that total phenotypic variance is the sum of genetic variance plus environmental variance and is twice the covariation of genetic and environmental effects. Genetic variance can be further partitioned into additive genetic variance, reflecting the effect of many segregating genes of equal effect, and into nonadditive genetic variance because of interactions within and among genetic loci. Environmental variance can be partitioned into environmental effects shared by family members (causing familial similarity) and into nonshared environmental effects (causing differences among family members). An assumption of the present analyses, that no genotype-environment covariance exists, has been supported for most behavioral and biomedical phenotypes [16]. Heritability is defined as the proportion of total variance attributable to genetic variance. In these models, all variance that is not genetic is considered environmental. Thus, the heritability statistic provides an estimate of the relative importance of genetic differences for individual differences in susceptibility to H. pylori infection in the sample of twins (relative to environmental effects). This statistic refers to the sample and does not describe risk to an individual.

Intraclass Correlations

Several methods exist to estimate heritability [16]. Nonadditive genetic effects are indicated when monozygotic correlations are more than twice the magnitude of dizygotic correlations. The intraclass correlation of monozygotic twins reared apart provides a direct and unbiased estimate of heritability. The influence of shared rearing environment can be estimated by comparing correlations of twins reared apart with those of twins reared together. The effect of correlated environments in an adoption-twin study refers to all similarities that cannot be explained by genetic factors or shared rearing environments. The effect of correlated environment may reflect prenatal influences and similarities in adult life experience (such as degree of contact when the twins are adults) and is estimated by subtracting the estimate of heritability from the intraclass correlation for monozygotic twins. Residual variance that is not explained by heredity, shared rearing environment, or correlated environments is attributed to nonshared environmental effects unique to the individual.

Model-Fitting Analysis

The variance comparisons of correlations and their interpretations are summarized in Table 1. Examination of intraclass correlation provides valuable insights about the relative importance of genetic and environmental factors. However, it is difficult to estimate the importance of several variables (for example, heritability, nonshared environment, shared rearing environment, and other forms of correlated environment) by separate comparisons of pairs of correlations. Model-fitting analyses are more powerful because the data from all four groups identified by zygosity status and rearing status (rearing-by-zygosity groups) are weighted appropriately and because tests of alternative models are permitted. The models for fitting are based on quantitative genetic expectations of factors contributing to twin similarities and differences.

The model used in the present study Table 1 has been described in detail by Neale and Cardon [17]. Because the phenotype (presence of H. pylori infection) is categorical (0 or 1), tetrachoric intraclass correlations were calculated using the PRELIS program (Scientific Software, Inc., Chicago, Illinois) for each of the four rearing-by-zygosity groups. Expected correlations based on the model for the four rearing-by-zygosity groups were fitted to the observed tetrachoric correlation matrices using a maximum likelihood estimation procedure in Lisrel VII (Scientific Software, Inc., Chicago, Illinois). Heritability, nonshared environment, shared rearing environment, and age were included in the model.

Results

Top Methods Statistical Analysis Results Discussion Author & Article Info References

We tested the differences in concordance rates between monozygotic and dizygotic twin pairs reared apart and reared together and between the combined sample regardless of the rearing status (Table 2). A difference was noted in the pairwise and probandwise concordance rates for monozygotic compared with dizygotic twin pairs, pooled over rearing status (67% for monozygotic twin pairs compared with 46% for dizygotic twin pairs, P = 0.004 for pairwise concordance; 81% for monozygotic twin pairs compared with 63% for dizygotic twin pairs, P = 0.001 for probandwise concordance). For those who were reared apart, the concordance rates were also greater for monozygotic twins than for dizygotic twins (69% for monozygotic twins compared with 38% for dizygotic twins, P = 0.006 for pairwise concordance; 82% for monozygotic twins compared with 66% for dizygotic twins, P = 0.003 for probandwise concordance) (Table 2).

Intraclass Correlations

The intraclass correlations for the four rearing-by-zygosity groups are listed in Table 3. The pattern of correlations shown in Table 3 suggests a statistically significant role for genetic effects on the acquisition of H. pylori infection. The intraclass correlation for monozygotic pairs reared apart is considered the best single estimate of broad-sense heritability, assuming that other types of correlated environment are not important. In the sample of twins in the Swedish Adoption-Twin Study of Aging, the interclass correlation coefficient was 0.66 for monozygotic twins who were reared apart.

Although the intraclass correlation coefficient for monozygotic twins reared apart provides the single best estimate of heritability, comparisons between twins reared together and twins reared apart permit investigation of the role of shared rearing environment. Average reared-together correlations were greater than reared-apart correlations, indicating the importance of shared rearing environments (Table 3). When monozygotic correlations are not greater than dizygotic correlations, twin similarity may reflect correlated environments rather than genetic similarity. However, this did not seem to occur in our study, and thus correlated environment was not included in subsequent analyses. Further, because the average monozygotic correlations were not greater than twice the magnitude of dizygotic correlations, nonadditive genetic effects were not indicated and were not included in model-fitting analyses.

Model-Fitting Results

A satisfactory fit of the model to the data, indicated by a chi-square with a P value > 0.05, was obtained (chi-square = 11.09, P = 0.921, df = 19). The most relevant information pertains to the amount of variation that can be accounted for by each of the genetic and environmental effects. The heritability estimate of 0.57 from the model-fitting analysis is similar to the 0.66 estimate based on the correlation obtained from analyzing monozygotic twins reared apart. Shared rearing environment accounts for 20% of the variance. The remaining variance (23%) is attributable to nonshared environmental factors that contribute to differences, not similarities, among family members. The statistical significance of the genetic effects was evaluated by eliminating the genetic variable from the model because the fit in the model was now worse; this indicated that genetic effects were statistically significant.

Discussion

Top Methods Statistical Analysis Results Discussion Author & Article Info References

Many seroepidemiologic studies [1, 18-24] have shown that H. pylori is associated with histologic gastritis and peptic ulcer disease; these studies used sensitive and specific assays to detect anti-H. pylori antibody. Helicobacter pylori is now accepted as the most common cause of gastritis, and the previous associations with gastritis are now attributed to H. pylori infection [1]. The role of genetic susceptibility to H. pylori infection has been difficult to study. The rate of acquisition appears to be higher in developing countries than in developed countries [23, 24]. Even within developed countries, ethnic and racial variation has been noted [19, 25-30]. For example, in the United States, the rate of acquisition of H. pylori among black and Hispanic persons is twice that among white persons, and these differences remained after adjustments were made for age and socioeconomic factors [19]. Such variation suggests that genetic and environmental factors are important for determining the susceptibility to H. pylori infection. Helicobacter pylori infection clusters within families with children [9, 10, 31, 32], and an increased frequency of H. pylori infection occurs among spouses and children living in the same household with a positive index case; these observations suggest person-to-person transmission, a common source of infection, or both. The genetically unrelated spouse of an infected index case is more likely to be infected than is the spouse of an uninfected index case, suggesting that environmental factors are also important [10].

Investigators have been interested in finding a genetic component for H. pylori-related diseases, such as peptic ulcer disease, gastritis, and gastric cancer. Although it is impossible to entirely separate environmental factors from genetic influences, several family studies suggest a genetic component for these diseases [33-35]. Initially, the strongest evidence of a genetic relation came from studies showing an increase risk for duodenal ulcer in persons with hyperpepsinogenemia [34, 36, 37]; however, increased levels of serum pepsinogen I also occur in persons with H. pylori infection [18, 37-39].

Our study provides strong evidence of a genetic influence on the acquisition of H. pylori infection. Further, sharing the same rearing environment also contributes to a familial tendency for acquiring H. pylori infection. The results extend our understanding of the cause of H. pylori infection in two important ways. First, genetic factors account for most (57%) of the variation we observed in this sample of Swedish twins for H. pylori infection. These results most likely reflect a genetic influence for susceptibility to the infection. Second, environmental factors are also important, as shown by correlations for twins who were reared together that were greater than correlations for twins who were reared apart. Thus, sharing the same rearing environment contributes to familial tendency for H. pylori infection. Although the literature on twins with duodenal ulcer is limited, concordance rates for peptic ulcer have been reported in a follow-up study [33] of twins from the Danish Twin Registry. A monozygotic concordance rate of 52.6% and a like-sexed dizygotic concordance rate of 35.7% suggest that peptic ulcer may also be influenced by genetic effects. Because we have established the importance of genetic factors for susceptibility to H. pylori infection, we can now analyze whether the relation between H. pylori and diseases such as peptic ulcer, chronic gastritis, and gastric cancer may be mediated by a common genetic mechanism or whether peptic ulcer disease arises from the causal influence of acquiring H. pylori infection; we intend to study these questions using this data set.