Results. Assessment of IBS. Determination of Concordance

Transcription

1 GASTROENTEROLOGY 2001;121: Irritable Bowel Syndrome in Twins: Heredity and Social Learning Both Contribute to Etiology RONA L. LEVY,* KENNETH R. JONES, WILLIAM E. WHITEHEAD, SHARA I. FELD, NICHOLAS J. TALLEY, and LINDA A. COREY *University of Washington, Seattle, Washington; University of North Carolina at Chapel Hill, Chapel Hill, North Carolina; Lakeside School, Seattle, Washington; University of Sydney, Sydney, Australia; and Virginia Commonwealth University, Richmond, Virginia Background & Aims: Heredity has been suggested to explain the finding that irritable bowel syndrome (IBS) tends to run in families. Research in this area has been limited. The aim of the present study was to assess the relative contribution of genetic and environmental (social learning) influences on the development of IBS by comparing concordance rates in monozygotic and dizygotic twins to concordance between mothers and their children. Methods: Questionnaires soliciting information on the occurrence of more than 80 health problems, including IBS, in self and other family members were sent to both members of 11,986 twin pairs. Results: Analysis is based on 10,699 respondents representing 6060 twin pairs. Concordance for IBS was significantly greater (P 0.030) in monozygotic (17.2%) than in dizygotic (8.4%) twins, supporting a genetic contribution to IBS. However, the proportion of dizygotic twins with IBS who have mothers with IBS (15.2%) was greater than the proportion of dizygotic twins with IBS who have co-twins with IBS (6.7%, P < 0.001), and logistic regression analysis showed that having a mother with IBS and having a father with IBS are independent predictors of irritable bowel status (P < 0.001); both are stronger predictors than having a twin with IBS. Addition of information about the other twin accounted for little additional predictive power. Conclusions: Heredity contributes to development of IBS, but social learning (what an individual learns from those in his or her environment) has an equal or greater influence. The irritable bowel syndrome (IBS) is a chronic functional gastrointestinal disorder characterized by abdominal discomfort or pain that begins with a change in the frequency or consistency of stool and/or is relieved by defecation (2 of 3 symptoms required), in the absence of other diseases that could explain the symptoms. 1 It has been observed that IBS tends to run in families. 2 Furthermore, Levy et al. 3 found that children of parents with IBS tend to use health care significantly more for gastrointestinal problems than children of parents who do not have IBS. Two predominant etiologic models have been proposed that could explain these familial patterns: pathophysiological mechanisms, 4 which are inherited, and psychological phenomena, such as learned responses to visceral sensations. 5 One study 6 has explored the heritability of IBS, but this work was limited by a small sample size. The present study explored the possible genetic contribution to IBS by studying the concordance rate of IBS among a large population-based group of twin pairs and their parents in the United States. The aim of the study was to assess the relative contribution of genetic and environmental (social learning) influences on the development of IBS by comparing the prevalence of IBS in the mothers of monozygotic (MZ) and dizygotic (DZ) twins with IBS to the prevalence of IBS in co-twins of MZ and DZ twins who have IBS. A higher prevalence rate among the mothers of twins with IBS than among the co-twins of DZ twins with IBS would demonstrate an environmental influence because DZ twins share approximately the same number of genes with each other as they share with their mothers. Subjects and Methods Subjects The population-based panel for this study included all twins born in the Commonwealth of Virginia between 1915 and Of 40,896 twin pairs ascertained from vital records, 19,453 could be traced, and 11,986 twin pairs ultimately agreed to participate in the registry by completing a zygosity questionnaire. Zygosity determinations were based on responses to questions concerning the degree of similarity of pair members perceived by the twins, family members, and close associates. This method of assigning zygosity has been shown to permit correct zygosity assignment with 97% accuracy. 7 Zygosity testing using blood typing or DNA fingerprinting, Abbreviations used in this paper: DZ, dizygotic; IBS, irritable bowel syndrome; MZ, monozygotic by the American Gastroenterological Association /01/$35.00 doi: /gast

2 800 LEVY ET AL. GASTROENTEROLOGY Vol. 121, No. 4 while preferable, is not feasible in surveys of this type and magnitude, particularly where questions concerning the similarity of twins have been shown to be a reasonable alternative. 8,9 The questionnaire survey was reviewed by the Virginia Commonwealth University Institutional Review Board for Human Research, and the study was reviewed and approved during each year it was sent out. The analyses on which the present report is based were conducted using an anonymized data set. Assessment of IBS Systematic questionnaire data soliciting information on the occurrence of a number of different health problems, including IBS, were sent to both members of the 11,986 twin pairs who agreed to participate and to their spouses when the twin pair member reported being married. (Data from spouses were not examined as part of this study.) Questionnaires were returned by mail. Subjects were asked, Have you or any of your close family been affected by this disease? A list of 79 health problems common to men and women plus 7 health problems specific to women and 3 health problems specific to men followed. The list included irritable bowel syndrome. If the disorder was reported for a member of the respondent s family, he or she then identified which other family members experienced it (i.e., twin, father, mother, brothers, sisters, sons, and daughters). Determination of Concordance The total proband concordance estimate was computed using the standard formula (2C 2 C 1 )/(2C 2 C 1 D 1 ), 10 where C 2 is the number of pairs in which both twins independently reported IBS (double ascertainment), C 1 is the number of twin pairs in which only 1 twin completed a questionnaire but reported both self and the co-twin to have IBS (single ascertainment), and D 1 is the number of pairs discordant for IBS, based on the reports of either both twins or only 1 twin. As a check on the reliability of including singly ascertained data in the estimate of concordance, we also calculated concordance for the subset of twins in which both twins completed questionnaires (double ascertainment only), using the formula 2C/(2C D). 10 Data Analysis The significance of differences in concordance rates between MZ and DZ twins was tested using 2 analysis. For comparisons of concordance rates in MZ and DZ twins, the unit of analysis was the twin pair (not individual respondents). To assess the relative contribution of genetic and social learning influences, we compared the probability that a twin with IBS would have a co-twin with IBS with the probability that a twin with IBS would have a mother with IBS. We restricted the comparison to 1 parent to avoid inflating the a priori probability of a match between parent and child, and we chose the mother because (1) women are more likely to have IBS than men and (2) mothers typically spend more time with their children than fathers and thus are more likely to influence the child s illness behavior through modeling. These analyses were also limited to twin pairs in which at least 1 twin had IBS. Tests of the significance of differences in proportions were used for these comparisons when samples were unrelated, 11 and McNemar s test 12 was used when proportions were compared from related samples. Step-wise logistic regression was used as a second, confirmatory method of assessing the relative contribution of heredity and social learning to IBS. The unit of analysis here was the individual twin respondent, but to avoid entering cases that were not truly independent, we entered data only for the first twin of each twin pair. The dependent variable was the twin s self-report of IBS; however, when the first twin did not complete the questionnaire, we used the second twin s report about the first twin. (Twins were defined as first and second twins a priori by the twin registry, not based on who first returned a questionnaire.) Independent variables in the initial regression analyses were (1) whether the twin s mother had IBS, (2) whether the twin s father had IBS, and (3) whether the co-twin reported IBS. A second regression analysis was run in which the independent variables were not allowed to enter freely, but the twin s report of whether the co-twin had IBS was forced to enter first. This second logistic regression was run to control for halo effects (i.e., the tendency to attribute one s own symptoms to others) as an explanation for the observation that IBS twins reported that their parents had IBS at greater than chance rates. Tests were interpreted as significant at P 0.05 and as a trend at P 0.10 but P For logistic regression analysis, we reported the Nagelkerke R 2 as a measure of the predictive relationship among variables (comparable to an R statistic in a linear regression) and the c statistic as a measure of goodness of fit. Results Demographics The present study examines data from a questionnaire mailed to 32,240 known twins included in 19,453 twin pairs between 1987 and Before mailing, addresses were updated by cross-referencing motor vehicle records; however, 10% were returned with incorrect addresses. Respondents included 10,699 twins from 6060 twin pairs enrolled in the Virginia Twin Registry. There were 281 twin pairs in which 1 or both twins reported IBS status and at least 1 member reported having IBS. These 281 twin pairs included 117 MZ and 164 DZ twin pairs. The mean age of the twins at the time of completing the questionnaire was 29.9 years for the MZ twins and 30.3 years for DZ twins (Table 1). The young age of the respondents is representative of twin cohorts and is attributable to the fact that the rate of twin births has risen dramatically since the 1920s. In addition, the method of

3 October 2001 IRRITABLE BOWEL SYNDROME IN TWINS 801 Table 1. Demographics of Study Subjects MZ DZ All Female Female Male Male All Female Female Male Male Female Male Male Female n Mean age (yr) verifying current addresses using motor vehicle records more readily identified younger women who had not yet changed their maiden names. The MZ twin pairs included 84 female and 33 male twin pairs. The DZ twin pairs included 62 female, 19 male, and 83 mixed-sex pairs (Table 1). In this sample, women reported IBS at a rate more than twice that of men, as is often seen in Western samples. 1 In terms of socioeconomic status, 13% of respondents reported incomes more than $40,000 per year compared with 3% of the whole population of the state based on census figures for Virginia. The registry is 85.9% white and 14.1% African American. This compares well with census figures for Virginia (19.5% are African Americans). Twin Twin Concordance The concordance for the MZ twins was 17.2%, compared with 8.4% for DZ twins ( , P 0.030). This analysis included 62 MZ and 106 DZ twin pairs in which only 1 twin reported for the pair and 55 MZ and 58 DZ twin pairs in which both twins reported IBS status. When the analysis was restricted to doubly ascertained twin pairs, the results were essentially the same: concordance estimates were 16.7% for MZ and 6.7% for DZ twins ( , P 0.09). The genetic liability for a disorder can be estimated by subtracting the DZ concordance from the MZ concordance. The liability for an MZ twin having IBS when the other twin reports IBS is estimated to be 8.8%. Figure 1 shows the prevalence of IBS in the total Figure 1. Evidence for parental influence in IBS: prevalence of IBS in subgroups. sample for twin respondents (2.6%) and the prevalence of IBS among mothers of all respondents (2.2%); these are the base rates and reflect the a priori probability of reporting IBS for this sample of subjects. Figure 1 also shows the proportion of twins with IBS who also have co-twins with IBS (a simple approximation of the concordance rate, but calculated somewhat differently) and the proportion of twins with IBS who report that their mothers have IBS. Each twin pair was only counted once; when both twins completed questionnaires, the first twin to report IBS was the index twin. These data show that compared with the base rate of IBS in all twins (2.6%), the probability of having IBS is significantly greater in DZ twins whose co-twins have IBS (6.7%, P 0.001), and the probability of having IBS is significantly greater if the co-twin with IBS is an MZ (13.7%) rather than a DZ (6.7%, P 0.003) twin. The data also show that compared with the base rate of IBS among all mothers (2.2%), the probability of having a mother with IBS is greater for DZ (15.2%, P 0.001) and MZ (17.1%, P 0.001) twins with IBS. However, the probability of having a mother with IBS is similar for DZ (15.2%) and MZ (17.1%) twins. These data are consistent with a genetic contribution to IBS. Contribution of Social Learning Figure 1 also compares the probability of having a mother with IBS with the probability of having a co-twin with IBS for both MZ and DZ twin pairs. The probability that a DZ twin with IBS would have a mother with IBS (15.2%) was significantly greater (P 0.001) than the probability that a DZ twin with IBS would have a co-twin with IBS (6.7%). The probability that an MZ twin with IBS would have a mother with IBS (17.1%) was also significantly greater (P 0.001) than the probability that a DZ twin with IBS would have a co-twin with IBS (6.7%). This finding supports a social learning contribution to IBS because mothers share approximately the same number of genes with their children as DZ twins share with each other, yet the proportion of IBS twins whose mothers had IBS was significantly greater than the proportion of DZ twins with IBS whose co-twins had IBS.

4 802 LEVY ET AL. GASTROENTEROLOGY Vol. 121, No. 4 Confirmatory Regression Analysis Logistic regression analysis provided an alternative way of assessing the relative contribution to IBS of social learning and heredity. For these analyses, possible predictors of IBS status included reports of whether the father had IBS as well as whether the mother and the co-twin had IBS. Logistic regression analysis (Appendix) revealed that reporting the father as having IBS was the best single predictor of self-reporting of IBS (Nagelkerke R , c 0.559, P 0.014), and reporting the mother as having IBS loaded next (Nagelkerke R , c 0.635, P 0.001, when combined with reporting father as having IBS). The co-twin s report of IBS loaded last (Nagelkerke R , c 0.683, P when combined with the first 2 variables). To control for a possible halo effect in reporting of IBS in others by twins with IBS, the regression analysis was repeated with the index twin s report of the co-twin s IBS status forced to enter the logistic regression model before the IBS status of either mother or father. This allowed us to study the effects of the mother s and father s IBS status after excluding any variance attributable to the co-twin s status. The co-twin s IBS status yielded a c value of (P 0.001), but the IBS status of the father and the mother each independently contributed significantly to the model (P for both) and, when combined, increased the c value to Thus, the predictive ability of parents IBS status was preserved when potential halo effects were removed. Discussion This study used a novel method to compare genetic influences to a specific type of environmental influence, namely the influence of illness behavior in parents. It is also the largest twin study on the heritability of IBS, involving data from 6060 twin pairs. The findings suggest that genetic effects contribute to IBS but that social learning (what an individual learns from those in his or her environment through such mechanisms as modeling and positive reinforcement) plays an equal or perhaps larger role. Heritability The gold standard for the study of heritability of any disorder is to follow MZ and DZ twins separated at birth. 13 If adoption is unrelated to zygosity or to the presence of the disorder, differences between MZ and DZ twins in the concordance for the disorder can be attributed solely to heredity. Unfortunately, it is often difficult to follow this ideal standard, primarily because of insufficient numbers of twins reared apart. Therefore, studies often follow a compromise methodology in which the concordance for a disorder is compared in MZ and DZ twins, most of whom were reared together. In this case, concordance reflects both genetic and shared environmental influences. This methodology was included in our study and also is the methodology followed by Morris- Yates et al., 6 as discussed below. Our data show that the concordance for IBS is twice as great in MZ twin pairs (17.2%) as in DZ twin pairs (8.4%; P 0.030). Moreover, compared with the base rate of IBS among twin respondents (2.6%), twins were significantly more likely to report IBS if they had a DZ co-twin with IBS (6.7%; P 0.001), and the probability of reporting IBS was significantly greater for co-twins of MZ twins with IBS (13.7%) than for co-twins of DZ twins with IBS (P 0.003; Figure 1). Morris-Yates et al. 6 previously reported on genetic factors in IBS by examining data from the Australian Twin Study. Using structured linear modeling, they concluded that genetics accounted for approximately 57% of the variance in the reporting of symptoms of a functional bowel disorder. Table 2 compares the Australian study with the U.S. study. The criteria for case inclusion were different: in the Australian study, endorsement of 2 or more bowel symptoms defined a subject as having functional bowel disease, whereas subjects in the U.S. study had to understand what IBS meant and to report that they had IBS. Sample sizes were also different: the Australian study included data for 462 families, whereas the U.S. study represented 6060 twin pairs from different families. The Australian study identified only 33 twin pairs in which at least 1 twin reported symptoms consistent with a functional bowel disorder, whereas the present study included 281 twin pairs (including 113 doubly ascertained twin pairs) in which 1 or both twins reported that they had IBS. The 2.6% prevalence rate of IBS in our total sample is consistent with the 2.9% estimate of the prevalence of the diagnosis of IBS in surveys in the U.S. 14 Table 2. Comparison of Australian and U.S. Studies Australian study U.S. study No. of families No. of twins ,699 Case definition FBD symptoms 1 Report of IBS diagnosis Twin pairs with FBD Concordance for MZ 33.3% 17.2% twins Concordance for DZ twins 13.3% 8.4% FBD, Functional bowel disorder.

5 October 2001 IRRITABLE BOWEL SYNDROME IN TWINS 803 The concordance for the Australian study was nearly double that of the current study, probably because of differences in case definition. However, the pattern of differences in concordance rates between MZ and DZ twins were similar (33% vs. 13.3% in Australia; 17.2% vs. 8.4% in the U.S.). Thus, the 2 studies are in agreement in showing differences between MZ and DZ concordance rates that are consistent with a genetic effect. Social Learning The present study also differed from the Australian study in its use of data on the reported incidence of IBS among the mothers and fathers of these twins. These data allowed us to estimate nongenetic or environmental influences on the reporting of IBS. To assess the relative contribution of social learning and genetics to IBS, we compared the probability of having a mother with IBS with the probability of having a co-twin with IBS for MZ twins with IBS (Figure 1). From a purely genetic perspective, the probability that the mother of an MZ twin with IBS will have IBS should be similar to the probability that the DZ co-twin of a twin with IBS will also have IBS because the numbers of shared genes are similar. However, we found that the probability that an MZ twin with IBS will have a mother with IBS is similar to the probability of having IBS in the co-twin of an MZ twin with IBS, and it is significantly greater than the probability of having IBS in the co-twin of a DZ twin with IBS. Thus, having a mother with IBS accounts for as much variance as having an identical set of genes with a co-twin who has IBS, suggesting that the contribution of social learning to IBS is at least as great as the contribution of heredity. In the regression analysis, having a father with IBS was a stronger predictor of IBS than having a mother with IBS, although only 8.2% of IBS twins had fathers with IBS, compared with the 14.9% of IBS twins who had mothers with IBS. The explanation for this discrepancy is that IBS was more common among the mothers (1.5%) than among the fathers (0.6%) of non-ibs twins. Reports that family members have the same illness as oneself could be influenced by a phenomenon that has been referred to as a halo effect, i.e., a tendency to attribute to others the symptoms one has because one has become sensitized to notice them. To determine whether this halo effect explained the concordance between respondents and their parents for IBS, we performed a second regression analysis in which we controlled for the halo effect by removing the variance in self-report of IBS that is associated with reporting that a twin has IBS. (We assumed that the halo effect would have a greater impact on the tendency to attribute one s own symptoms to a twin than on the tendency to attribute these symptoms to parents.) We still found that the combined effect of reporting of IBS in the father and mother accounted for more variance (6.2%) in reporting IBS in self than did reporting IBS in a co-twin (5.6%). These findings are consistent with previous work from our laboratory 3,15 and others, 16,17 suggesting that parental modeling and reinforcement of illness behavior contribute to the cause of IBS. This research has shown similar patterns of health care use among parents with IBS and their children. 3,16 Levy et al. 3 found that children of parents in a health maintenance organization who had received an IBS diagnosis made significantly more health care visits overall, as well as significantly more health care visits for gastrointestinal problems, than children of control subjects. Parental modeling was hypothesized to be a major contributor to this observation. Studies of childhood learning have also shown an influence of parental teaching through reinforcement or reward learning. 15,18 In one study of nursing students, Whitehead et al. 15 found that childhood social learning could account for up to 25% of the variance in symptom reports, 16% of the variance in disability days, and 5% of the variance in clinic visits. Moreover, there was evidence for specificity of learning; encouragement and modeling of illness behavior for menstrual symptoms during childhood was more highly correlated with adult reports of menstrual symptoms, clinic visits for gynecologic problems, and disability days for menstrual symptoms than with reports of symptoms, clinic visits, and disability days for nongynecologic symptoms. Study Limitations Neither this study nor the Australian one used consensus criteria (such as the Rome Criteria 1 ) to assess IBS. In addition, both studies could have been improved by including expert physician assessment of patients to confirm diagnosis. Twin studies of the concordance for specific traits are vulnerable to ascertainment bias, i.e., the tendency for individuals with the disorder to be more likely to participate in the research. Ascertainment bias is unlikely to be a significant factor in the present study because IBS was only 1 of a list of more than 80 health problems included in the survey, and no emphasis was given to gastrointestinal complaints. Conclusion and Future Directions This study is complementary to the Australian study in suggesting that there is a hereditary component to IBS. However, this study goes beyond the Australian study in demonstrating that social learning has an effect

6 804 LEVY ET AL. GASTROENTEROLOGY Vol. 121, No. 4 on the development of IBS that is at least as large as that of heredity. Further research should address the limitations in both this study and the study of Morris-Yates et al. 6 by using the Rome criteria 1 and/or physician assessment to confirm the diagnosis of IBS in respondents and family members. Further investigations of how family patterns of social learning affect children s responses to physiologic events would also significantly add to our knowledge of the contributions of these factors to the causes of IBS, as well as other conditions. Moreover, because 2 studies now support a contribution of heredity to the development of IBS, exploration of the mechanisms for this genetic transmission may help identify biologic markers for the disorder. Better understanding of the relative contributions of social learning (environment) and of genetics to IBS could guide the development of prevention and treatment protocols for IBS. Appendix: Results of Stepwise Logistic Regression Analyses Model 1. Variable included: father s IBS status (B 3.66; odds ratio [OR] 38.84, Wald statistic 120.6, P 0.001). Model 2. Variables included: father s IBS status (B 3.03, OR 20.71, Wald statistic 133.7, P 0.001); mother s IBS status (B 3.74, OR 42.1, Wald statistic 118.6; P 0.001). Model 3. Variables included: father s IBS status (B 3.65, OR 20.7, Wald statistic 106.1, P 0.001); mother s IBS status (B 2.54, OR 20.7, Wald statistic 76.41, P 0.001); twin s IBS status (B 1.87, OR 42.1, Wald statistic 46.7, P 0.001). References 1. Thompson WG, Longstreth G, Drossman DA, Heaton K, Irvine EJ, Muller-Lissner S. Functional bowel disorders and functional abdominal pain. In: Drossman DA, Corazziari E, Talley NJ, Thompson WG, Whitehead WE (eds). The functional gastrointestinal disorders, 2nd ed. McLean VA: Degnon Associates, 2000: Locke GR III, Zinsmeister AR, Talley NJ, Fett SL, Melton LJ III. Familial association in adults with functional gastrointestinal disorders. Mayo Clin Proc 2000;75: Levy RL, Whitehead WW, Von Korff M, Feld AD. Intergenerational transmission of gastrointestinal illness behavior. Am J Gastroenterol 2000:95: Mayer EA, Chang L, Lembo T. Brain-gut interactions: implications for newer therapy. Eur J Surg Suppl 1998;582: Whitehead WE, Crowell MD, Davidoff AL, Palsson OS, Schuster MM. Pain from rectal distention in women with irritable bowel syndrome: relationship to sexual abuse. Dig Dis Sci 1997;42: Morris-Yates A, Talley NJ, Boyce PM, Nandurkar S, Andrews G. Evidence of a genetic contribution to functional bowel disorder. Am J Gastroenterol 1998:93: Magnus P, Berg K, Nance WE. Predicting zygosity in Norwegian twin pairs born Clin Genet 1983;24: Cederlof R, Friberg L, Lundman T. Studies on similarity diagnosis in twins with the aid of mailed questionnaire. Acta Genet Stats Med 1961;11: Martin NG, Martin PG. The inheritance of scholastic abilities in a sample of twins. I. Ascertainment of the sample and diagnosis of zygosity. Ann Hum Genet 1975;39: Kendler KS, Eaves LJ. The estimation of probandwise concordance in twins: the effect of unequal ascertainment. Acta Genet Med Gemellol (Roma) 1989;38: Bruning, JL, Kintz BL. Computation handbook of statistics. Glenview, IL: Scott, Foreman, Rosner B. Fundamentals of biostatistics. Boston, MA: Prindle, Weber & Schmidt, Wright L. Twins: and what they tell us about who we are. New York, NY: Wiley, Sandler RS. Epidemiology of irritable bowel syndrome in the United States. Gastroenterology 1990;99: Whitehead WE, Busch CM, Heller B, Costa PT Jr. Social learning influences on menstrual symptoms and illness behavior. Health Psychol 1986;5: Walker LS, Garber J, Greene JW. (1991). Somatization symptoms in pediatric abdominal pain patients: relation to chronicity of abdominal pain and parent somatization. J Abnorm Child Psychol 1991;19: Lowman BC, Drossman DA, Cramer EM, McKee DC. Recollection of childhood events in adults with irritable bowel syndrome. J Clin Gastroenterol 1987;3: Whitehead WE, Crowell MD, Heller BR., Robinson JC, Schuster MM Horn S. Modeling and reinforcement of the sick role during childhood predicts adult illness behavior. Psychosomat Med 1994;56: Received October 12, Accepted May 30, Address requests for reprints to: Rona L. Levy, Ph.D., Mailstop , University of Washington, Seattle, Washington rlevy@u.washington.edu; fax: (206) Supported by grant RO1 HD36069 from the National Institute of Child Health and Human Development.