Räihä I, Kemppainen H, Kaprio J, Koskenvuo M, Sourander L. Lifestyle, Stress, and Genes in Peptic Ulcer DiseaseA Nationwide Twin Cohort Study. Arch Intern Med. 1998;158(7):698-704. doi:10.1001/archinte.158.7.698
The familial accumulation of peptic ulcer disease observed in several studies may be attributable to genetic effects, aggregation of environmental exposure (shared environment), or both. The intrafamilial spread of Helicobacter pylori infection has raised the question whether shared environment could explain the familial aggregation of peptic ulcer disease rather than genetic similarity of family members.
To examine the contribution of genetic and environmental factors to the pathogenesis of peptic ulcer disease in a nationwide population-based cohort of adult twins.
The Finnish Twin Cohort consists of all same-sexed twin pairs born before 1958 with both twins alive in 1975. The total number of twin pairs is 13888, of whom 4307 are monozygotic (MZ) and 9581 are dizygotic (DZ) twins. Questionnaire surveys of twins were carried out in 1975, 1981, and 1990, including medical and psychosocial questions. One question asked whether a physician had ever made a diagnosis of gastric or duodenal ulcer. In addition, hospital discharge data from 1972 to 1991 were linked with the twin cohort to obtain those twin individuals who had been treated for gastric or duodenal ulcer. The prevalence of and concordance for peptic ulcer disease were examined in MZ and DZ twins. Model-fitting analysis was used to specify the relative roles of genetic and environmental factors. The contribution of lifestyle factors and stress was examined prospectively in an incidence study and by comparison of discordant pairs.
The prevalence of peptic ulcer disease was 6.2% in men and 2.8% in women in 1975. There were 63 MZ and 86 DZ pairs concordant for peptic ulcer disease. Concordance for disease was significantly higher in MZ than in DZ twin pairs; the probandwise concordance rate was 23.6% (95% confidence interval [CI], 20.9%-26.3%) in MZ twins and 14.8% (95% CI, 13.3%-16.3%) in DZ twins. In the model-fitting analysis, a model with both additive genetic and unshared environmental effects had the best goodness-of-fit. Thirty-nine percent (95% CI, 32%-47%) of the liability to peptic ulcer disease was explained by genetic factors and 61% (95% CI, 53%-68%) by individual environmental factors. In the incidence study (logistic regression analysis of the entire cohort initially free of peptic ulcer disease, with subjects diagnosed as having peptic ulcer after 1975 as cases), current smoking (relative risk, 2.2; 95% CI, 1.5-3.2) and high stress levels (relative risk, 3.2; 95% CI, 1.4-7.6) in men and regular use of analgesics (relative risk, 3.3; 95% CI, 1.3-8.1) in women predicted peptic ulcer disease during the follow-up from 1976 to 1991. In the analysis of discordant pairs, smoking in men and regular use of analgesics in both sexes were predictors of peptic ulcer disease.
The questionnaire and hospital usage data on peptic ulcer disease in the population-based twin cohort suggest that the familial aggregation of the disease is modest, and attributable almost solely to genetic factors. Environmental effects not shared by family members were significant predictors of disease, and they were attributable to smoking and stress in men and the use of analgesics in women. The minor effects of shared environment to disease liability do not support the concept that the clustering of risk factors, such as H pylori infection, would explain the familial accumulation of peptic ulcer disease.
A FAMILIAL accumulation of peptic ulcer disease has been well established.1,2 Theoretically, the familial overrepresentation of disease may be attributable to the aggregation of environmental exposures (shared environment), genetic effects, or both. The concept of the inheritance of peptic ulcer disease has been commonly accepted, while the effect of environmental predisposing factors such as the use of anti-inflammatory analgesics,3 smoking,4 alcohol consumption, and mental stress5 has also been emphasized. Several characteristics with genetic influence are associated with duodenal ulcer: blood group O, nonsecretors of blood group substances, hyperpepsinogenemia I, rapid gastric emptying, and postprandial hypergastrinemia.6- 8 Hyperpepsinogenemia I is inherited as an autosomal dominant trait and it is present in 50% of patients with duodenal ulcer disease. There is, however, evidence that Helicobacter pylori infection elevates serum pepsinogen levels, and therefore the genetic transmission of hyperpepsinogemia has been questioned.9,10
The causal linking of H pylori infection and peptic ulcer disease has raised the need for reevaluation of the role of inheritance. Intrafamilial clustering of H pylori infection has been shown,11- 14 and it is also evident that different members of the same family are infected by similar strains of H pylori.15,16 Thus, person-to-person transmission of H pylori within families could explain familial clustering of peptic ulcer disease even without genetic transmission. The intrafamilial spread of H pylori infection could be attributable to the close contact of family members but also genetic liability could modify susceptibility to infection.17
The relative contribution of genes and environment to interindividual variation in disease susceptibility can be investigated in twins. Comparison of monozygotic (MZ) and dizygotic (DZ) twin pairs provides an estimation of the magnitude of genetic factors.18 If MZ twins are not fully concordant, nongenetic and environmental factors would be implicated. A higher concordance rate in MZ rather than in DZ twin pairs suggests the presence of a genetic influence. For a disease entirely determined by environmental factors, the concordance in MZ twins should be equal to that in DZ twins. Twin studies have supported the concept that a large part of the familial aggregation is attributable to genetic factors. The concordance in MZ twins has varied between 14.0% and 80.0% and that of DZ twins between 0% and 35.7%.19,20 Only the study by Jensen19 has been based on a nationwide cohort of twins, while other published studies have been based on selected materials. Restricted samples may bias the estimates of concordance rates.21
To shed more light on the roles of genetic and environmental factors in peptic ulcer disease in the H pylori era, we analyzed prospective data of the nationwide Finnish Twin Cohort, including several environmental exposures.
The Finnish Twin Cohort was compiled from the Central Population Registry of Finland using selection procedures described in detail elsewhere.22- 24 In brief, the Central Population Registry of Finland is a computerized data bank of personal information on all Finnish citizens from 1967 onward. All sets of individuals with the same birth date, the same sex, the same surname at birth, and the same local community of birth were identified. This yielded all the pairs (N=17357) of same-sex adult twins born in Finland before 1958 with both twins alive in 1967, as well as a small number of subjects who satisfied these criteria but who were not biological twins. A baseline questionnaire, administered in August through October 1975, asked whether the subjects were twins and included questions for zygosity classification (see below). The overall response rate was 89%. From local parish records, further inquiries were made about the parents of all nonresponders and individuals with conflicting responses to the item on twinship to clarify twinship.
The questions used to assign the zygosity of twins were similar to those used in questionnaires for other large twin samples.25- 28 The validity of zygosity questions was confirmed by blood typing a subsample of 104 twin pairs living in the Helsinki area.29 About 93% of all responding pairs were classified as MZ or DZ. Classified pairs had a 1.7% probability of misclassification. The algorithm for classifying zygosity left 7% of respondent pairs unclassified because of conflicting responses to the items.29 A total of 2489 pairs were left unclassified by zygosity. These also consisted of pairs with an unknown address or with nonresponse to the 1975 questionnaire. A total of 13888 pairs (4307 MZ and 9581 DZ) who were 18 years or older at baseline were identified.
The questionnaire surveys carried out in 1975, 1981, and 1990 contained 97 to 103 multiple-choice demographic, medical, and psychosocial questions. In each survey, twins were asked whether a physician had ever made a diagnosis of gastric or duodenal ulcer (yes or no). Smoking (none, occasionally, former smoker, or current smoking), alcohol consumption (reported use of beer, wine, and hard liquors was converted to alcohol grams and classified as: abstainer, 1-399 g/mo, 400-899 g/mo, or ≥900 g/mo), and use of analgesic drugs and antacids (none, <10, 10-59, 60-180, or >180 days yearly) were also determined. Experienced stress was measured in the 1975 questionnaire, which included 4 items of self-report of stress of daily activities presented as 4-point scales. The items of stress of daily activities were the following: (1) In general I am unusually tense and nervous; (2) there is great deal of stress connected with my daily activities; (3) at the end of day I am completely exhausted mentally and physically; and (4) my daily activities are extremely trying and stressful. There were 4 alternatives to answer: the statement describes me "very well," "well," "not very well," or "not at all." The sum scores were divided into 5 categories (none, low, some, moderate, or high). Correlation coefficients of scale items and total score distributions have been described.24 Stress of daily activities score has been found to predict morbidity from mental disorders in a 6-year follow-up among 31116 Finnish adults.30
Hospital discharge data were also used to identify twins with peptic ulcer disease. The National Agency for Welfare and Health has kept a national registry of hospital discharges since 1969. This registry covers all discharges of inpatients from all hospitals in Finland. The discharge diagnoses have been assigned by those physicians who treated the patient using the International Classification of Diseases, Ninth Revision (ICD-9).31 Up to 4 different diagnoses per patient could be listed at each discharge. Data from 1972 to 1991 were linked with the Twin Cohort, using the unique social security identification number assigned to each Finnish citizen, to obtain the twin individuals who had gastric or duodenal ulcer disease as a discharge diagnosis (ICD-9 codes 531-534 or their equivalents in ICD-8). By combining the questionnaire and hospital usage data, a total of 1692 twin individuals with peptic ulcer disease were identified.
The prevalence of peptic ulcer disease at baseline was calculated for all individuals (twins and nontwins) who responded to the 1975 questionnaire. For the analysis of predictors of peptic ulcer disease, the cohort of individuals initially free of ulcer disease was defined as those subjects who replied also to both of the follow-up questionnaires and who at baseline reported that they did not have a history of or hospitalization for peptic ulcer disease and they did not use antacids. Incident cases of disease were those individuals who reported a physician-diagnosed peptic ulcer disease in 1981 or 1990, or who had been hospitalized for peptic ulcer in 1976 or later. Unconditional logistic regression analysis32 was used to estimate the odds ratios of disease by age (5 groups), smoking status (4 categories), analgesic use (5 categories), use of alcohol (4 categories), and stress of daily activities score (5 categories).
Twin similarity for peptic ulcer disease can be summarized using estimates of concordance. Concordance can be assessed using 2 concordance rates (termed pairwise and probandwise), each calculated separately for MZ and DZ pairs. Pairwise concordance rate gives the proportion of affected pairs that are concordant and they are descriptive statistics. The probandwise concordance rate is the proportion of all probands (primarily ascertained affected individuals) that belong to concordant pairs. This gives information on the recurrence risk of disease (corresponding to cumulative risk) associated with the degree of relationship of the pair, and this can then be compared with the risk of disease in the population at large, and with different types of relatives. The 95% confidence intervals (CIs) for concordance and discordance of peptic ulcer disease were computed. The age-adjusted relative risks (RRs) in MZ and DZ twins when 1 twin had peptic ulcer disease, compared with the population risk, were also calculated.
For the disease, which has both a genetic liability and multiple known environmental causes, the multifactorial model was considered the most appropriate to estimate the contribution of genetic factors to the underlying liability of the disease. The model assumes that there is normally distributed liability to disease, and when the threshold of liability is reached, disease becomes manifest. Both genes and environmental factors are assumed to contribute to the liability and they result from the joint effects of many genes with small effects and a multitude of environmental effects. These assumptions were considered reasonable for this analysis as most MZ pairs were discordant for disease, and there are several known environmental agents increasing risk of disease. The proportion of individuals exceeding the threshold in the population is equivalent to the population risk.
Threshold models with genetic sources of variation, ie, additive and dominance (nonadditive effects of alleles at a given locus summed over all relevant loci); and environmental sources of variation, ie, shared (family influences common to both twins) environmental or environmental sources of variation unique to the individual (ie, not shared with twin) in the underlying liability were estimated.33 These models can be fitted to the 2×2 contingency tables (no disease or disease in the first twin vs no disease or disease in the second twin).34 Both for men and women the contingency tables are set out separately for MZ and DZ pairs. These models were estimated using the Mx software34 as programmed for twin analyses. These models as such do not permit gene-environment analysis.
Goodness-of-fit statistics were used to assess to what degree the model specified by the investigators adequately corresponds to the data; a small goodness-of-fit square value and high P value indicates good correspondence between model and data. Alternative models that specify different components of variance can be compared by assessing the change in χ2 relative to changes in the df between models.35
Because the subjects in this study were twins, pairs discordant for incident peptic ulcer disease were identified to examine whether the risk factors of individuals with peptic ulcer disease differ from the risk factor distribution among their age-matched siblings (either MZ and DZ twins) without peptic ulcer disease. These twins represent individuals who have generally shared the same childhood environment as the exposed subject, and share part of or all their genes in descent in common. An estimate of the RR comes from the ratio of the number of pairs in which the exposed twin had peptic ulcer disease, but the unexposed twin had not, to the number of pairs in which the opposite has occurred; a multivariate extension is conditional regression analysis. The significance of a variable in the analysis was assessed by a likelihood ratio test.
The number of individuals with peptic ulcer disease is shown in Table 1. All individuals who responded to the questionnaire in 1975 were included. The overall prevalence of peptic ulcer disease was 6.2% in men and 2.8% in women. The prevalence increased with age, being 14.4% in men and 6.7% in women older than 65 years. Of twin individuals who responded to all 3 questionnaires, 784 (7.3%) of 10807 reported peptic ulcer and 98 (0.9%) of them also had a discharge diagnosis of peptic ulcer disease. Seventeen subjects did not report peptic ulcer disease, although they had a discharge diagnosis.
Of twin individuals who at baseline had no history of peptic ulcer disease, no hospitalization before 1976, nor any use of antacids, 197 men (4.7%) and 171 women (3.2%) reported peptic ulcer disease during the 15-year follow-up. Common risk factors of peptic ulcer disease were compared in these cases and in those who remained free of peptic ulcer disease. In the logistic regression analysis (including age group, smoking, use of analgesics, use of alcohol, and stress of daily activities score), current smoking in men and long-term use of analgesics in women and high stress score of daily activities in both sexes predicted peptic ulcer disease (Table 2).
Pairs in which both twins responded to the questionnaire in 1975, 1981, or 1990 are included in the analyses. Cases are subjects who reported a history of peptic ulcer disease in a questionnaire or who had been hospitalized. The prevalences did not differ between MZ and DZ twin individuals. The pairwise and probandwise concordance rates are also shown in Table 3. There were 63 MZ and 86 DZ pairs concordant for peptic ulcer disease, giving the pairwise concordance rates of 13.4% (95% CI, 11.8%-16.5%) and 8.0% (95% CI, 6.4%-9.6%) and the probandwise concordance rates of 23.6% (95% CI, 20.9%-26.3%) and 14.8% (95% CI, 13.3%-16.3%), respectively. The differences in the concordance rates between MZ and DZ twins were significant. The age-adjusted RR of peptic ulcer disease, when the other twin was affected, was higher in MZ men (RR, 2.7; 95% CI, 2.0-3.6), MZ women (RR, 2.7; 95% CI, 1.6-4.5), and DZ men (RR, 1.7; 95% CI, 1.4-2.2) compared with the population risk calculated from individual data of twins, while it did not differ in DZ women (RR, 1.2; 95% CI, 0.7-1.9) (Table 4). The concordance rates did not differ in twin pairs living together (260 men and 220 women) compared with those pairs who did not live together in 1981.
Results are summarized in Table 5. The first model (E) specifies that the observed pattern of distribution of disease in the twin pairs can be accounted for solely by factors uncorrelated between twins; this model fit the data poorly and could be rejected (P<.001). The next model (C and E) fit the data significantly better (P<.001 for change in χ2), but the C and E model overall fit was still poor (P<.001) and failed to account adequately for the data. Addition of the additive genetic term to the model again improved model fit, and the overall model fit (P=.32) was adequate, reflecting the greater similarity of MZ pairs compared with DZ pairs for peptic ulcer occurrence. The estimate for shared environmental effects was zero, and was dropped; the more parsimonious model (A and E; A is additive genetic effects) was sufficient to account for the data (P=.41). This is also the best model of all those tested based on the Akaike information criterion,19 a statistic summarizing the goodness-of-fit and model simplicity,35 indicating that 39% (95% CI, 32%-47%) of the liability to peptic ulcer disease is explained by genetic factors and 61% (95% CI, 53%-68%) by individual environmental factors. Inclusion of nonadditive genetic effects did not improve the model. In all the above model-fitting analyses, different thresholds were used for men and women. This reflects the different rates of peptic ulcer disease by sex, while the model specifies that the relative roles of genetic and environmental factors in disease liability are equal in men and women. An A and E model with equal thresholds for men and women gave a significantly worse fit to the data than the A and E model in Table 5 (χ2 difference, 127; df=1; P=.001).
A total of 83 (20 MZ) male pairs and 87 (25 MZ) female pairs initially without a history of peptic ulcer disease or any use of antacids became discordant for peptic ulcer disease during the follow-up and had full data on the risk factors of interest. The distribution of potential risk factors differed significantly between cases and the controls (their age-matched twins). In a conditional logistic regression analysis with all variables, smoking (P=.03) and use of analgesics (P=.02) were associated with peptic ulcer in men, while stress of daily activities and use of alcohol were not. The RR associated with current smoking was 3.8 (95% CI, 1.1-13.3). Among women, only the use of analgesics was significantly associated with the disease (P=.03). Because of small numbers, zygosity-specific analyses are not reported.
The purpose of this nationwide twin study was to examine the roles of genetic and environmental factors in the origin of peptic ulcer disease. The higher concordance rate for peptic ulcer disease among MZ pairs compared with DZ pairs and the population risk indicates a genetic liability. In the model-fitting analysis, the model with both additive genetic and unshared (unique) environmental effects had the best goodness-of-fit value. Approximately 40% of the variability of susceptibility to peptic ulcer disease is attributable to genetic factors, while the remainder is attributable to individual environmental effects or measurement error. Despite the higher prevalence of peptic ulcer disease in men than in women, the relative roles of genetic and environmental factors in disease liability seem to be equal.
In the Swedish twin study,17 the same rearing environment contributed to familial similarity for acquired H pylori infection, but the extent of shared environmental effect was only 20%, while the extent of the genetic effect was 66%. In our study the model with shared environmental effects fit the data poorly, suggesting that the intrafamilial transmission of H pylori infection cannot explain the familial aggregation of peptic ulcer disease without genetic traits. The shared environmental effects measured come mostly, but not necessarily, from the childhood family. In our study, there were also 240 twin pairs who had lived together even in adulthood, and the comparison of these pairs with the pairs who did not live together showed no difference. This also confirms the finding that the shared environmental effect did not exist. The reason why the shared environmental effect was not seen in peptic ulcer disease, although it was present in acquisition of H pylori infection, may be attributable to the multifactorial origin of peptic ulcer disease; H pylori infection is just one major risk factor for the disease, and the majority of people with H pylori infection do not develop peptic ulcer disease in their lifetime. Another explanation would be that H pylori is so prevalent in the population studied that its effect on variation in risk of peptic ulcer disease cannot be detected. Our study does not contradict the theory that the inheritance of peptic ulcer disease could be mediated by the genetic susceptibility to acquisition of H pylori infection.17
The direct influence of H pylori infection on disease liability was not possible to examine in this study. On the other hand, detailed data on other environmental risk factors such as smoking, use of alcohol, use of analgesics, and mental stress were available. The incidence study and the analysis of discordant twin pairs showed that smoking and mental stress in men and the use of analgesics in women predicted peptic ulcer disease. The growing use of nonsteroidal anti-inflammatory drugs has risen as the major cause of peptic ulcer disease aside from H pylori infection.36,37 On the one hand, with the exception of acute stress ulcer found most often in children and critically ill patients, findings have been conflicting regarding the role of stress in the development of peptic ulcer disease. In a cohort study by Anda et al,38 self-perceived stress predicted peptic ulcer disease during the 13-year follow-up (RR, 1.8). On the other hand, none of the psychological indicators were associated with an increased risk of peptic ulcer disease in a large Norwegian population-based survey.39 In the discordant analysis of the Danish twin study,19 mentally distressing family and occupational conditions were associated with an increased risk of peptic ulcer disease, but the influence of smoking and alcohol was not studied.
Selection bias in twin studies can influence estimated heritabilities for chronic diseases and cause overestimation of the concordance rates in geographically or otherwise restricted samples.21 In our study, the sample was based on an unselected nationwide twin population. Since MZ twins are more similar than DZ twins with respect to smoking, food habits, physical activity, and many other behavioral characteristics, the heritability can also be overestimated.21 Our findings are consistent with previous twin studies19,20 regarding the genetic contribution in peptic ulcer disease. In our study, moreover, the detailed data from the 1975 survey also made it possible to specify the effects of genetic and several environmental factors.
The information on peptic ulcer disease was based on self-reports by subjects susceptible to errors because of problems in recall or validity of the questionnaire. Diagnosis was not ascertained by review of medical records. Of the subjects who responded to all 3 questionnaires, 17 had a discharge diagnosis of peptic ulcer disease without reporting it. While there are no endoscopically controlled studies on the validity of the peptic ulcer disease data collected by the questionnaire, the comparison of questionnaire and medical record data is available. In a Danish cohort study,40 the questionnaire data controlled by the reviewing of medical records revealed 4.3% false-positive and 3.1% false-negative statements of ulcer occurrence. In Finland, peptic ulcer disease represents a well-defined entity, and it is generally distinguishable from dyspepsia. We suggest that this simple question offers considerably valid information without a meaningful possibility for misunderstanding. In fact, the prevalence of peptic ulcer disease in our study was comparable with a Finnish population study,41 in which the prevalence of endoscopically verified peptic ulcer disease or ulcer scar was 5.9%. In the study by Anda et al,38 the lifetime ulcer prevalence was also consistent with the prevalence in our study, being 7.7% in men and 3.6% in women (6.2% and 2.8% in our study), with a men-women ratio of 2.1:1 (2.2:1 in our study). Also the age distribution of ulcer prevalence was similar to that in the present study. The similar prevalence figures also suggest that our questionnaire presumably provided a reasonably valid estimation of the prevalence of peptic ulcer disease.
A limitation of our study was that gastric and duodenal ulcer disease were not asked about separately in the questionnaire. Different pathogenetic mechanisms are probably involved in the susceptibility of these 2 entities. The contribution of a genetic transmission and H pylori infection has been proposed especially in duodenal ulcer disease, while the use of nonsteroidal anti-inflammatory agents has been considered a major risk factor in gastric ulcer disease.
In conclusion, the questionnaire and hospital usage data on peptic ulcer disease in the population-based twin cohort suggest that the familial aggregation of the disease is modest, and attributable almost solely to genetic factors. Environmental effects were not attributable to factors shared by family members, and they were related to smoking and stress in men and the use of analgesics in women.
Accepted for publication September 11, 1997.
The study was supported by the Finnish Foundation for Gastroenterological Research and the Emil Aaltonen Foundation, Helsinki, Finland.
We thank Kauko Heikkilä for statistical assistance and Pirjo Piekka for secretarial assistance.
Reprints: Ismo Räihä, MD, Department of Geriatrics, University of Turku, Kunnallissairaalantie 20, FIN-20700 Turku, Finland.