Business & Finance

A Multidimensional Twin Study of Mental Health in Women

A Multidimensional Twin Study of Mental Health in Women
of 8
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  506 Am J Psychiatry 157:4, April 2000  Regular Articles A Multidimensional Twin Studyof Mental Health in Women Kenneth S. Kendler, M.D., John M. Myers, M.S., andMichael C. Neale, Ph.D. Objective: While researchers have increasing insight into the role of genetic and envi-ronmental factors in the etiology of psychiatric and substance use disorders, they knowmuch less about how such factors influence the dimensions of healthy psychologicalfunctioning. Method: In a population-based sample of 794 female-female twin pairs, theauthors examined, by using multivariate structural equation modeling, six dimensions ofmental health: perceived physical health, nonconflictual interpersonal relationships, anx-ious-depressive symptoms, substance use, social support, and self-esteem. Results: Thebest-fit model was complex and constituted five common factors (two genetic, one familyenvironmental, and two unique environmental); variable-specific genetic effects for physi-cal health, substance use, and social support; and variable-specific family environmentaleffects for interpersonal relationships and substance use. Genetic effects were seen for allsix dimensions; total heritabilities ranged from 16% to 49%. Family environment was animportant influence on interpersonal relationships, substance use, and social support. Conclusions: Mental health is a complex phenotype that is influenced by a diverse arrayof genetic and environmental factors. While genetic factors appear to be of moderate etio-logic importance in all major dimensions of mental health, the family environment is an im-portant influence on only interpersonal relations, social support, and substance use. (Am J Psychiatry 2000; 157:506–513) A n increasing number of twin, family, and adoptionstudies are clarifying the role of genetic and environ-mental factors in the etiology of a wide range of psy-chiatric and substance use disorders (1). By contrast,we understand much less about how these factors in-fluence overall mental health. While twin studies haveexamined individual variables such as happiness (2),self-esteem (3), and social support (4, 5), these con-structs reflect only a part of what is generally consid-ered to be mental health.In this study, we begin to address this disparity byexamining, in a large sample of female-female twinpairs ascertained from a population-based registry, sixvariables chosen to reflect at least part of the diversityof functions that constitute mental health. We attemptto address the following specific questions about thesevariables:1. What is the structure of the genetic factors that in-fluence these dimensions of mental health? Is there aset of mental health genes, or is mental health more ge-netically complex, with distinct genetic factors that in-fluence certain dimensions?2. While family environment appears to play littlerole in the etiology of major psychiatric disorders (1,6), is the same pattern true of dimensions of healthy  Received Jan. 19, 1999; revision received June 21, 1999;accepted July 12, 1999. From the Department of Psychiatry andthe Department of Human Genetics, Virginia Institute for Psychiat-ric and Behavioral Genetics, Medical College of Virginia, VirginiaCommonwealth University. Address reprint requests to Dr. Kendler,Department of Psychiatry, Box 980126, Medical College of Virginia,Richmond, VA 23298-0126; (e-mail).Supported by NIMH and National Institute on Alcohol Abuse andAlcoholism grant MH/AA-49492, NIMH grants MH-40828 and MH-54150, and a NIMH Research Scientist Award (MH-01277) to Dr.Kendler. The Virginia Twin Registry, established by W. Nance,M.D., Ph.D., and maintained by L. Corey, Ph.D., is supported byNational Institute of Child Health and Human Development grantHD-26746 and National Institute of Neurological Disorders andStroke grant NS-31564.The authors thank Carol Prescott, Ph.D., for her comments on anearlier version of the article.  Am J Psychiatry 157:4, April 2000  507 KENDLER, MYERS, AND NEALE psychological functioning? If family environment isimportant in mental health, does it affect all dimen-sions or only some? Is there a single factor that reflectsa health-inducing family environment, or are there sev-eral dimensions of family experience that affect mentalhealth?3. While individual environmental experiences cer-tainly influence mental health, to what extent do theseenvironmental effects influence all dimensions or se-lected dimensions of psychological functioning? METHOD Participants  The subjects examined in this study were participants in a longi-tudinal investigation of the risk factors for common psychiatric dis-orders in women. They were members of Caucasian female same-sextwin pairs from the Virginia Twin Registry (6), a population-basedregistry formed from a systematic review of all birth certificates inthe Commonwealth of Virginia. Twins were eligible to participate if they were born during 1934–1971 and both members had previ-ously responded to a mailed questionnaire. In our first interview, weassessed 92% of the eligible individuals (N=2,163)—90% face-to-face, the rest by telephone. Zygosity was determined blindly by stan-dard questions (7), photographs, and—when necessary—DNA (6,8). We recently conducted a validation study that involved perform-ing polymerase chain reaction zygosity tests on an additional 269pairs of twins and oversampling those for which our prior zygosityassignment was uncertain. On the basis of these tests (where themean number of markers tested per pair was 17.5, SD=8.4), zygositywas changed in 12 pairs (4.5%).We performed three additional waves of telephone interviewswith this sample, but here we use data only from the third interview,which was completed by 1,898 individuals—87.7% of the individ-ual twins from the srcinal sample. The mean number of months be-tween the first interview and the third interview was 61.3 (SD=5.1).The mean age of the participating twins in the third interview was34.6 years (SD=7.5, range=22–59). From all twins who participatedin interview wave 1, pairs in which both members participated in thethird wave were similar in age (t=1.33, df=2161, p=0.18) to pairs inwhich one or both twins did not cooperate in the third wave, but theformer group had, on average, approximately 0.6 more years of ed-ucation (t=5.56, df=2161, p < 0.0001). All interviews were conductedby interviewers who were blind to information about the co-twin.Written informed consent was obtained before face-to-face inter-views, and oral assent was obtained before telephone interviews. Measures  We chose, a priori, six measures from the first and third inter-views that reflected broad aspects of functioning that relate to men-tal health. For simplicity, we scaled all dimensions so that increasinglevels reflected increasing mental health.Self-perceived physical health was assessed by two items in thewave 1 interview: 1) satisfaction with health and 2) number of daysspent sick in bed in the last 12 months. This variable was includedbecause of the strong evidence for a positive association betweenphysical and emotional health (9). The mean scores for these twoitems were calculated after each was rescaled to contribute equalweight. Cronbach’s alpha (10), a standard measure of internal scalereliability, for these two items was 0.41.Nonconflictual interpersonal relationships were assessed by sepa-rate questions from a standard inventory of stressful life events (11),and participants were asked at both the wave 1 and wave 3 inter-views about any serious problems getting along during the preceding12 months with 1) parents, 2) twin, 3) other siblings, 4) other rela-tives, 5) close friends, 6) neighbors, or 7) in-laws. We examined themean number of problems across the two interview waves. Cron-bach’s alpha for these items was 0.61.Levels of anxiety and depression were assessed by the response to18 questions asked at both interviews about the experience, for atleast 5 days over the last 12 months, of 16 disaggregated symptomsof major depression, as outlined by DSM-III-R (e.g., separate itemsfor weight gain and loss and psychomotor retardation and agita-tion), and two screening questions for generalized anxiety disorder(“feeling anxious, nervous, or worried” and “your muscles felttense, or you felt jumpy or shaky inside”). Positive responses to theseitems, which were adapted from the Structured Clinical Interviewfor DSM-III-R (12), were not counted if the symptoms were judgedto be the result of medications or physical illness. We examined themean number of symptoms reported across the two waves. Cron-bach’s alpha for these items was 0.89.Levels of substance use were measured with three variables thatassessed 1) alcohol consumption at both waves, measured as theproduct of the number of days per month when an alcoholic drinkwas consumed and the number of drinks consumed on an averageday when drinking occurred, 2) the frequency of use of “medicinefor your nerves or sleeping medicines,” assessed at the wave 1 inter-view, and 3) the average number of cigarettes smoked per day in thelast year, assessed at the wave 3 interview. We examined the firstprincipal component derived from these measures. Cronbach’s alphafor these three items was 0.39.Social support was assessed as the first principal component of a15-item social interaction scale that was previously developed andused at the Institute for Social Research (13) and with which ratingswere obtained at both interviews. These items reflected the fre-quency and quality of contact with friends and relatives, the fre-quency of attendance at clubs or other organizations, and the num-ber of friends and confidants. The highest negative loadings on thisfactor ( < –0.40) were for items that reflected interpersonal tensions(e.g., “Do your relatives criticize you? Do your friends make toomany demands on you?”), whereas the highest positive loadings( > 0.40) were on items that reflected warm relationships (e.g., “Doyour relatives express interest in how you are doing? Do your rela-tives make you feel that they care about you?”). Cronbach’s alphafor these 15 items was 0.61.Self-esteem was assessed by the 10-item Rosenberg Self-EsteemScale (14), which was completed at the wave 3 interview. We exam-ined the first principal component derived from these items. Cron-bach’s alpha for this scale was 0.89.The distributional properties of these variables, even after analysisof principal components, were often highly nonnormal. Therefore,before analysis, we polychotomized all variables into three approxi-mately equal classes that reflected low, intermediate, and highscores. We then fitted to these resulting contingency tables a multi-ple-threshold model, which assumes a normally distributed underly-ing latent distribution. We tested the goodness-of-fit of this modelseparately in monozygotic and dizygotic twins. Of the 12 tests, themodel failed at the 5% level only twice, a result not different fromchance expectations (15); no variable failed in both zygosities. Wealso examined the relationship between these six variables and age.The highest correlation was with nonconflictual interpersonal rela-tionships (0.13); all other correlations were under 0.05. We did not,therefore, include age in our model. Statistical Analysis  Our approach to the analysis of twin data has been outlined in de-tail elsewhere (16, 17) and consists of inferring the action of geneticand environmental risk factors from the pattern of resemblance inmonozygotic and dizygotic twin pairs. For this article, we performeda multivariate genetic analysis of our six putative dimensions of mental health. While the goal of univariate genetic analysis is the de-composition of the variance of a trait into its genetic and environ-mental components, in multivariate genetic analysis, the focus shiftsto decomposing sources of covariance among traits. To illustrate thedifference between univariate and multivariate twin analysis, theconcept of latent or unobserved factors is introduced. In a tradi-tional or phenotypic factor analysis, latent factors are postulated tocause the resemblance (or, more technically, covariation) among  508 Am J Psychiatry 157:4, April 2000  MENTAL HEALTH IN WOMEN items. The goal of factor analysis is to explain the correlationsamong a large number of variables as a result of the effects of a smallnumber of latent factors. Multivariate genetic analysis is also amethod of explaining correlations among multiple items. However,it goes beyond traditional factor analysis in that it provides insightinto the causes of resemblance among variables.As with all our twin modeling, we here assume that the variationand covariation in liability to mental health can be ascribed to threepotential sets of factors: 1) additive genetic variables, which con-tribute twice as much to the correlations between monozygotic twinsas they do to the correlations between dizygotic twins (becausemonozygotic twins share identical genes by descent, whereas dizy-gotic twins share, on average, half their genes), 2) common or familyenvironment (those familial factors such as parental attitudes thatare shared by members of a twin pair), which contribute equally tothe correlation between monozygotic and dizygotic twins, and 3) in-dividual-specific environmental factors, which reflect environmentalexperiences not shared by both members of a twin pair and thereforecontribute to differences between them in their reported levels of mental health.In univariate analysis, information regarding the causes of varia-tion is obtained by comparing the resemblance of monozygotic anddizygotic twin pairs for a single variable. In multivariate analysis, thecorrelation between two or more variables is the primary unit of analysis. By comparing the cross-twin, cross-variable correlation inmonozygotic and dizygotic twins, and contrasting that to the cross-twin, within-variable and within-twin, cross-variable correlations,the covariation of two or more variables can be partitioned into itsgenetic and environmental components.Two alternative models were tested to describe how genetic andenvironmental factors may influence covariation. In the commonpathway model (17–19), genetic and environmental factors influ-ence covariation through a single pathway. Such a model contains noseparate genetic and environmental latent factors. Rather, geneticand environmental variables act conjointly through one or more la-tent phenotypes. By contrast, in the independent pathway model,genes and the environment can contribute to covariation throughseparate genetic and environmental latent factors. Because we hadonly six dimensions of mental health, no model with more than twocommon factors could be identified. In addition to these commonfactors, the model estimated additive genetic (A), common or familyenvironmental (C), and/or individual-specific or unique environmen-tal (E) factors that are specific to individual dimensions of mentalhealth. The common and specific factors are identified by subscriptsso that A C1  and E S  represent, respectively, the first common geneticfactor and a unique environmental factor specific to a particularvariable.The form of data for our multivariate genetic analysis is two12 × 12 polychoric correlation matrices calculated by PRELIS (20),which gave the tetrachoric correlations within and across twins forthe six dimensions of mental health and separately for monozygoticand dizygotic twins. To best describe how genes and the environ-ment influence resemblance among the dimensions of mental health,a series of multivariate models was fitted to these matrices by usingMx software (21) by the method of asymptotic-weighted least TABLE 1. Within- and Between-Twin Correlations for Indices of Mental Health in 794 Pairs of Female Twins a Twin and IndexTetrachoricTwin 1PhysicalHealthNon-conflictualRelationshipsLowAnxiety/ DepressionLowSubstanceUseSocialSupportSelf-EsteemTwin 1Physical health0. relationships0.250.420.170.300.04Low anxiety/depression0.300.530.390.320.17Low substance use0.180.230.360.22–0.02Social support0.110.410.280.070.20Self-esteem0. 2Physical health0.09 b relationships–0.030.25 b anxiety/depression– b  – substance use0. b 0.020.02Social support0. b 0.03Self-esteem–0.01–0.050.01–0.12–0.040.20 ba Shaded areas above the diagonal are monozygotic twins; clear areas below the diagonal are dizygotic twins. TABLE 2. Results of Model Fitting to Determine Influence of Genetic and Environmental Factors on Mental Health Indices in 794Pairs of Female Twins ModelNumber of FactorsAnalysisAkaike’sInformationCriterionAdditive GeneticFamilyEnvironmentalIndividual-Specific Environmental  χ 2 dfpSaturated66683.0750.25–67.0Common pathway111184.3111  < 0.001–37.7Independent pathway1111137.31020.01–66.72211105.0970.27–89.03121112.6970.13–81.44112114.6970.11–79.45221100.4920.26–83.6621293.7920.43–90.3 a 7202114.2980.13–81.8822290.7870.37–83.3 a Best-fit model according to Akaike’s information criterion (22).  Am J Psychiatry 157:4, April 2000  509 KENDLER, MYERS, AND NEALE squares. The model, which best combined the features of parsimonyand goodness-of-fit, was selected by Akaike’s information criterion(22), one of the best performing of such indices in a thorough simu-lation study (23). To identify uniquely second common factors, theloading of these factors on the first variable (physical health) was setto zero. RESULTS Complete data were available from both members of 794 twin pairs, of whom 471 were classified as mono-zygotic twins and 323 as dizygotic twins. Correlation Matrix  The complete within- and between-twin correlationmatrix for the six variables is shown in table1. Resultsfor monozygotic twins are above the diagonal (the up-per-right shaded portion), and those for dizygotictwins are below the diagonal (the lower-left unshadedportion). For both monozygotic and dizygotic twins,the matrix has three parts: correlations within twin 1,correlations between twins 1 and 2, and correlationswithin twin 2. The following patterns in this completematrix are noteworthy:1. The within-twin correlations for the six traits aregenerally positive and modest and range largely from0.15–0.30. The strongest correlations are seen betweeninterpersonal relationships and anxiety/depression, in-terpersonal relationships and social support, and anxi-ety/depression and substance use.2. The within-trait, cross-twin correlations (intable1) are higher for all six variables in monozygotictwins than in dizygotic twins. These results suggestthat genetic factors are of etiologic importance for allof these traits. However, the magnitude of twin re-semblance for these traits varied widely: for physicalhealth, the traits correlated to 0.09 in dizygotic and0.28 in monozygotic twins; for substance use, theycorrelated to 0.47 in dizygotic and 0.70 in monozy-gotic twins.3. For any two traits, there were two cross-twin,cross-trait correlations (i.e., trait 1 in twin 1 correlatedwith trait 2 in twin 2, and trait 2 in twin 1 correlatedwith trait 1 in twin 2). With few exceptions, thesecross-twin, cross-trait correlations were higher formonozygotic twins than for dizygotic twins. For exam-ple, the correlations between interpersonal relation-ships and anxiety/depression were considerably higherin monozygotic twins (0.25 and 0.35, respectively)than in dizygotic twins (0.17 and 0.10). This pattern of results was consistent with the hypothesis that geneticfactors are in part responsible for the covariance of these traits in the population. Model Fitting  The results of model fitting are outlined in table2.We began by fitting a fully saturated Cholesky model,which fit well but required the estimation of 63 sepa-rate parameters. We then attempted to fit two verysimple models. The common pathway model allowedfor a single factor that influenced all six traits by actingthrough a common latent variable. This produced arelatively poor fit. The second simple model assumedone genetic, one family environmental, and one uniqueenvironmental common factor that acted on the sixvariables independently of one another. This model(independent pathway model 1 in table2) fit muchbetter than the common pathway model and nearly aswell as the saturated model.We then added to this very simple independent path-way model a second genetic factor (independent path-way model 2), a second family environmental factor(independent pathway model 3), or a second uniqueenvironmental factor (independent pathway model 4).All three of these models produced an improvement inAkaike’s information criterion, but the best fit wasfound with independent pathway model 2, which wassubstantially better than that found with the saturatedmodel.We then explored whether we could improve the fityet further with additions to independent pathwaymodel 2. We added a second family environmental fac-tor (independent pathway model 5) and a secondunique environmental factor (independent pathwaymodel 6). By Akaike’s information criterion, indepen-dent pathway model 5 was clearly inferior, but inde-pendent pathway 6 produced a slightly superior fit.We then tested two further variations on indepen-dent pathway model 6. We tried dropping the one fam-ily environmental factor (independent pathway model7) and adding a second such factor (independent path-way model 8). Neither improved upon the fit of inde-pendent pathway model 6, which we concluded wasthe model that provided the best overall explanation of the data. CorrelationTwin 2PhysicalHealthNon-conflictualRelationshipsLowAnxiety/ DepressionLowSubstanceUseSocial SupportSelf-Esteem0.28 b b b b 0.09– b 0.080.10–0.010.13– b 0.290.300.–0.050.06–0.070.25 b Within-trait, cross-twin correlation.  510 Am J Psychiatry 157:4, April 2000  MENTAL HEALTH IN WOMEN Parameter Estimates  The path coefficients that were estimated from thebest-fitting model (independent pathway 6) are seen infigure1. Table3 presents the proportion of variance inthese six aspects of mental health accounted for by thevarious aspects of the model.Seven aspects of these results are noteworthy:1. The first genetic factor has substantial loadings onall six variables, with the highest loadings on anxiety/ depression and self-esteem and the lowest loading onsubstance use.2. The second genetic factor—which, by definition,did not load on physical health—had its highest load-ings on anxiety/depression and substance use and neg-ative loadings on social support and self-esteem.3. Genetic risk factors for three of the variables (in-terpersonal relationships, anxiety/depression, and self-esteem) were entirely accounted for by the two geneticcommon factors. By contrast, substantial variable-spe-cific genetic loadings were found for physical health,substance use, and social support.4. The single common family environmental factorhad substantial loadings on only two variables: inter-personal relationships and social support.5. Variable-specific family environmental effectswere found for two variables—interpersonal relation-ships and substance use—one of which (substance use)was particularly large.6. The first unique environmental common factorhad substantial loadings on physical health, interper-sonal relationships, and anxiety/depression; a modestloading on substance use; and very small loadings onsocial support and self-esteem. By contrast, the secondunique environmental common factor had its highestloadings on social support and self-esteem.7. Substantial variable-specific unique environmen-tal factors were seen for all of the putative dimensionsof mental health. DISCUSSION We sought, in this report, to explore the pattern of genetic and environmental factors that influence im-portant dimensions of mental health in women. We at-tempted to address three specific questions and will ex-amine our results in turn. Genetic Factors That Influence Mental Health  Our first goal was to clarify the structure of the ge-netic factors that influence healthy mental functioning.One plausible hypothesis is that the genetic componentof mental health has a very simple structure; there is asingle set of mental health genes that are entirely re-sponsible for the genetic contribution to the dimen-sions of mental health. This hypothesis might seemanalogous to a genetic version of the unitary hypothe-sis of mental illness (24). An alternative hypothesis is FIGURE 1. Path Estimates of the Best-Fitting Model for Ge-netic, Family Environmental, and Individual-Specific Environ-mental Factors in Mental Health Indices for 794 Pairs of Fe-male Twins a a Path estimates are standardized partial regression coefficientsand must be squared to determine the proportion of variance ac-counted for in the dependent variable. Observed variables are de-picted in squares and latent variables in circles. The following ab-breviations are used: A, additive genetic effects; C, common orfamily environment; E, individual-specific or unique environment.The subscript C indicates a common factor, and the subscript Sindicates a factor that is unique to a single variable. . A C1 A S .46 A S .40 A S .35 –.07 –.33.41.23 A C2 NonconflictualRelationshipsPhysicalHealth Low Anxiety/ DepressionLow Substance UseSocialSupportSelf-Esteem Genetic NonconflictualRelationshipsPhysicalHealth Low Anxiety/ DepressionLow Substance UseSocialSupportSelf-EsteemNonconflictualRelationshipsPhysicalHealth Low Anxiety/ DepressionLow Substance UseSocialSupportSelf-Esteem. –.19 C C C S .57 C S Family Environmental . –. E C1 E C2 Individual-Specific Environmental .78 E S .65 E S .57 E S .50 E S .61 E S .72 E S
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks