Twin studies in schizophrenia with special emphasis on concordance figures

A R T I C L E

Twin Studies in Schizophrenia With SpecialEmphasis on Concordance FiguresEINAR KRINGLEN

Twin studies in schizophrenia have been reviewed with special emphasis on concordance rates in popula-tion-based investigations. Sources of error have been discussed with particular focus on sampling. Thepair-wise concordance rates in schizophrenia are 30–40% in MZ and 5–10% in DZ, with somewhat higherrates for proband concordance. The findings from twin studies support the diathesis stress model inschizophrenia, and it is argued that the polygenic model gives the best explanation for the empiricalfindings. Am. J. Med. Genet. (Semin. Med. Genet.) 97:4–11, 2000.Q 2000 Wiley-Liss, Inc.

KEY WORDS: concordance rate; methods; schizophrenia; twins

INTRODUCTION

Research in schizophrenia illustrateshow psychiatric thought is determinedby the “Zeit-geist.” Until the SecondWorld War psychiatry held the viewthat schizophrenia was a genetic disease.This hypothesis received strong supportfrom the influential large-scale twinstudy by Franz Kallmann [1946], reit-erated at the First International Con-gress of Psychiatry in Paris in 1950, thatreported concordance figures of 86% inmonozygotic (MZ) and 14% in dizy-gotic (DZ) twins. In the mid-1960s thisview changed, partly because of newtwin studies from the Nordic countriesthat de-emphasized the genetic compo-nent. But more important perhaps, wasthe influence of the postwar social psy-chiatry with provocative family hy-potheses of schizophrenia, and the em-phasis on social processes in general,that grew out of a liberal reform ideol-ogy created by the economic growth inthe United States and Western Europe.During the last two decades, stimulatedby the biological revolution, the pen-dulum has swung back to a stronger be-lief in the genetics of schizophrenia. It isinteresting to note that this extremechange in psychiatric thoughts has not

been based upon radically new empiri-cal facts.

Admittedly, our knowledge con-cerning schizophrenic disorders has in-creased during the last decades. Never-theless, one has to admit that a numberof fundamental questions are still un-solved. First of all, we are confrontedwith uncertainty as regards the clinicalboundaries of schizophrenia. We donot know if the syndromes that are de-fined according to ICD-10 or DSM-IVrepresent valid etiological distinctions.We know that schizophrenia is partlygenetically determined, but does thegenetic disposition apply to all forms ofthe disorder? Obviously, schizotypalpersonality disorder is related to schizo-phrenia, but is this relationship a geneti-cal one? In more than half of the casesof schizophrenia, patients have showndeviant behavior in childhood or earlyadolescence such as shyness or introver-sion, anxiety or sensitivity. The base-rate for such deviances in the generalpopulation, however, is quite high andthere is no evidence that these patientsdiffer radically from patients withoutsuch a premorbid history, neither withregard to genetics, clinical picture noroutcome. MZ-twins are paired withnormal co-twins in 30–40%. This find-ing is remarkable both from a geneticaland environmental point of view. Ifgenes play a significant role, one wouldexpect to observe if not identical clini-cal syndromes, at least some deviance in

schizophrenic direction in the co-twin.Also, if environmental factors such asfamily conflicts predispose to schizo-phrenia, one would expect that bothtwins in a MZ-pair would be affected inone way or another, because MZ-twinsare of the same gender and age. It is alsopuzzling that the offspring of twoschizophrenic parents are non-schizo-phrenic in 60–70%, despite their ex-treme genetical and environmental risks[Rosenthal, 1966; Kringlen, 1978].

RATIONALE OF THETWIN METHOD

Applied twin research is based on theassumption that there are two types oftwins, MZ who are identical in heredi-tary equipment, and DZ with half oftheir genes in common on an average.All differences between MZ twins haveto be attributed to environmental fac-tors. DZ on the other hand, are from agenetical point of view sibs who wereaccidentally born at the same time. Dif-ferences between DZ twins may there-fore be due to both hereditary and en-vironmental factors.

The so-called classical twin methodwas developed in the 1920s. One com-pares statistically MZ and DZ pairs inrespect of their concordance for thetrait or the illness in question. Concor-dance is usually expressed as the rate ofsimilar occurrence in both twins, or inthe case of measurable traits as the av-

*Correspondance to: ProfessorEinar Kringlen, M.D., Ph.D., Departmentof Psychiatry, University of Oslo, PO Box85, Vinderen, 0319 Oslo, Norway. E-mail:[email protected]

AMERICAN JOURNAL OF MEDICAL GENETICS (SEMIN. MED. GENET.) 97:4–11 (2000)

© 2000 Wiley-Liss, Inc.

erage intra-pair difference. A pair iscalled concordant if both have the sameillness, discordant if one is sick and theother is not sick. Significantly higherconcordance figures in the group ofidentical twins have been regarded asevidence in support of a hereditarybackground of the traits concerned. Ifthe concordance is the same in bothMZ and DZ, the illness under discus-sion has been thought to be mainly de-termined by environmental factors. Forinstance, in the case of infectious dis-eases, the genetic differences are of neg-ligible importance. One would here ex-pect clustering in families, but nomarked difference between the concor-dance rates for MZ and DZ.

Twin studies might in principle beable to shed light on several questionsrelated to etiology. First of all, are ge-netic factors present in the disorder un-der discussion? Second, is it possible todiscover any vulnerability factors in thenon-disturbed co-twin? If so, this couldbring us closer to an understanding ofthe factors that predispose to psychopa-thology, and also contribute to a morevalid classification. Third, is it possibleto identify organic or social factors thatdiscriminate between affected and non-affected twin partners (Table I)?

An extremely valuable method isthe study of MZ twins reared apart. Be-cause genes and environment are oftenconfounded in twin studies, the studyof MZ twins separated at an early ageand reared apart, is from a scientificpoint of view the sovereign method be-

cause it combines the twin methodwith the adoption design. Such a studymight also test the family hypothesis ofschizophrenia. If the concordance fig-ures in such studies are the same as inordinary twin studies, this would sug-gest that environmental factors relevantto schizophrenia, are not strongly famil-ial. The drawback of this method is ofcourse, that it is hard to find suchpairs, and most of such published casesmight be biased because they perhapshave been published because they areconcordant.

Studying the incidence of schizo-phrenia in the offspring of MZ twinpairs, discordant for schizophrenia, pro-vides us with yet another method toobtain an impression of the magnitudeof the genetic contribution. The off-spring of the non-schizophrenic MZco-twin are biologically first degreerelatives of the schizophrenic probands,although they are legally their niecesand nephews. Because they were rearedin a family with no psychotic parent,they give the study the structure of across-fostering or adoption design. Agenetic hypothesis would indicate thatthe offspring of the two MZ-partnersdevelop psycho-pathological disorderof the same form and degree, becauseboth partners possess the same geneticdisposition. An environmental hypoth-esis would predict that the children ofthe non-psychotic co-twin developpsychosis less frequently. It goes with-out saying that few such studies havebeen carried out, because it is hard tofind such families.

Critiques of the twin method havesuggested that concordance rates inschizophrenia are flawed, because onecan not exclude the possibility that thedifference in concordance rates be-tween MZ and DZ may result from thegreater similarity of the social environ-ment of MZ. Kendler et al. [1993], inhis review of the literature, shows thatthis environmental bias is highly un-likely, because most studies show thatthe similarity of the social environmentof MZ is the result and not the cause ofsimilar behavior. There are studies,however, that weaken the equal envi-ronment hypothesis [Dalgard and Krin-glen, 1976]. In addition, co-twins of

DZ- schizophrenics normally have ahigher risk for morbidity than ordinarysibs.

COMPUTATION OFCONCORDANCE RATESAND SOURCES OF ERROR

There are several methods of computa-tion of concordance rates. The mostusual ones are the pairwise and probandmethod. The pairwise method issimple: One just calculates the percent-age of concordant pairs in the twinsample. This method has been mostlyused in somatic studies. The probandconcordance rate is the proportion ofaffected twins who have an affectedpartner. By this method each affectedpair doubly ascertained counts twice.There is no general agreement onwhich method should be used, al-though in psychiatry the probandmethod has been the most commonlyused in recent studies.

Earlier twin studies showed muchhigher concordance figures for MZthan for DZ twins with respect ofschizophrenia. This difference was mostconspicuous in Kallmann’s [1946] ex-tensive sample, but also Luxenburger[1930], Rosanoff et al. [1934], andSlater and Shields [1953], observed amarked difference.

Rosenthal [1962] drew special at-tention to the importance of samplingin twin studies, and argued convinc-ingly that the classical twin studies byand large showed misleadingly highconcordance figures with respect toschizophrenia.

Let us consider Kallmann’s [1946]study in more detail because his re-search illustrates the sources of errorthat might be inflicted upon twin stud-ies. Kallmann started out with a chronichospital population and later includedconsecutive admissions. His data arebased on reports from the staff of mentalhospitals in the state of New York, noton birth registers. Because his samplingis un-systematic he is bound to obtain apreponderance of concordance cases. Itis obvious that if one tries to obtain psy-chotic twins from a hospital population,by asking the director of the hospital toreport such cases, the staff of the hos-

TABLE I. Strategies inTwin Research

The classical twin methodStudies of concordance

(heredity)Studies of co-twins (nosology/

vulnerability)Studies of discordance

(environmental factors)Other methods

Studies of MZ twins rearedapart

Studies of offspring of MZ twinsdiscordant for disease

ARTICLE AMERICAN JOURNAL OF MEDICAL GENETICS (SEMIN. MED. GENET.) 5

pital will tend to remember concordantcases, whereas in cases where only oneof the twins in a pair has been hospital-ized, the staff might not know that thiscase is a twin. If one assumes that theprobability of being hospitalized and re-ported is 50% and that members of theconcordant pairs are admitted and re-ported independently, the probabilityof being in such a sample is 0.5 for adiscordant pair and 0.75 for a concor-dant pair (the probability that either oftwo outcomes will occur is the sum oftheir probabilities minus the probabilitythat both will occur together). Thus fora concordant pair P 4 0.5 + 0.5 − (0.5× 0.5) 4 0.75. If the probability is aslow as 10% P 4 0.1 and 0.2 respec-tively. This means that if the sample isso small that all concordant pair are rep-resented solely by one affected partner,then the direct pairwise concordancerate will give the double concordancerate for the population, simply becauseconcordant pairs have a double chanceof being represented in the sample[Kringlen, 1976].

The establishment of zygocity inKallmann’s work was based partly onhis personal observations (this was be-fore blood and serum testing could bedone). He used Sieman’s similaritymethod, rendering the diagnosis lessdefinite than it would have been if ablood work-up or a systematic ques-tionnaire had been included. Thus,most likely some MZ-pairs were incor-rectly labeled dizygotic. For obviousreasons a discordant MZ-pair mightlook dissimilar because the schizo-phrenic behavior also alters the person’sbodily behavior. This hypothesis is alsosupported by the fact that Kallmann re-ported on only 174 MZ pairs, butought to have found approximately250, according to Weinberg’s differen-tial method.

When Kallmann did his twin stud-ies, the concept of schizophrenia in theUS was rather broad due to the influ-ence of Adolf Meyer and psychoanaly-sis. This meant that his concordancerate would tend to be higher than if hehad applied a more strict concept ofschizophrenia, that was common in Eu-rope at that time. One would assumefor instance, that by applying a wide

concept of schizophrenia, a pair oftwins in which one presents a typicalschizophrenic picture, whereas theother who has massive neurotic symp-toms or borderline traits, might be clas-sified as concordant.

Finally, one has to make correctionfor age if the pair has not passed theperiod of risk as to schizophrenia. If onetwin is schizophrenic and for instance25 year of age, and his co-twin is non-schizophrenic, the co-twin might ofcourse later on will develop schizophre-nia. Weinberg’s abridged method is themethod most frequently used to correctfor age. This method was worked outfor estimating the incidence of an illnessin a sample of the general population.This method is only defensible if thereis no correlation between ages of onsetin probands and relatives, that is not thecase with regard to twins. Empiricalstudies show that in schizophrenia con-cordance usually develop within a shortperiod of time, in general within 2–3years. Thus by applying Weinberg’smethod on a twin sample of schizo-phrenia, one might over-correct forage. It should be noted that Kallmann’soriginal concordance figure for schizo-phrenia in MZ twins was 69%. The of-ten-cited figure of 86% was arrived atby using Weinberg’s short method. Thediscordantly affected pairs were on theaverage 33 years of age, and becausethey had been discordant for at least 8years, the risk for developing schizo-phrenia ought to be minimal.

In conclusion, it is hard to form anaccurate idea of the sources of error inKallmann’s work, but it seems likelythat all of the weaknesses mentioned,tend to inflate his concordance figures.This is fairly certain with regard to sam-pling procedure, clinical psychiatric di-agnoses and employment of the Wein-berg statistical method. Whether thetendency is the same when it comes tothe establishing of zygosity diagnosis isharder to decide, but it seems probable(it is my impression that the surround-ings tend to consider a pair as dizygoticif they are psychiatrically discordantlyaffected, and monozygotic if a pair isconcordant). In many clinical studiesthe various sources of error go in dif-ferent direction and thus might level

each other out. In Kallmann’s study,however, all the eventual sources of er-ror tend to point in the same direction,thus increasing the concordance figures,particularly in MZ twins.

EMPIRICAL TWIN STUDIES

The first twin studies of schizophreniawere carried out in Germany by Lux-enburger [1928], who was well awareof the problem of sampling. Rosanoff etal. [1934] in the US and Slater andShields [1953] in England, as well asInouye [1961] of Japan, all studied largesamples. These studies showed consid-erably higher concordance figures forMZ than for DZ, with respect toschizophrenia, although the concor-dance figures of MZ never reached thehigh Kallmann figures. Slater andShields [1953] reported concordancefigures of 65% in MZ and 14% in DZ.Inouye [1961] who dealt especiallywith diagnostic and nosological prob-lems, observed concordances uncor-rected for age to be 36% in MZ and 6%in DZ with a rather strict concept ofschizophrenia and 60% in MZ and 12%in DZ with a broader concept. None ofthese before-mentioned studies had ac-cess to twin registers. Samples werebased upon hospital series of schizo-phrenic twins, that obviously led to anover-identification of concordant andsevere cases.

During the 1960s, three popula-tion-based studies from the Nordiccountries came to de-emphasize theimportance of genetic factors inschizophrenia.

Tienari’s [1963] study caused con-siderable discussion because of his re-markable finding of zero concordancein 16 MZ twins. The author dealt witha population of twins, all males born inFinland between 1920 and 1929. In1957, the Finnish Foundation of Alco-hol Studies collected from birth regis-tration offices, the names of all maletwins born in the country at that time.In this sample there were 17 MZ twinpairs, in which the index case was clas-sified as psychotic. Fifteen cases wereclassified as schizophrenia, one as bor-derline schizophrenia and one as reac-tive psychosis. None of the 16 pairs was

6 AMERICAN JOURNAL OF MEDICAL GENETICS (SEMIN. MED. GENET.) ARTICLE

strictly speaking concordant for schizo-phrenia, but three co-twins had border-line features. If a wide concept ofschizophrenia is employed, the concor-dance figures is 19%. Because the twinshad passed most of the risk period ofschizophrenia and because the follow-up time was considerable in most cases,the risk of development of schizophre-nia in the co-twins was small. In a fol-low-up study with access to some fewmore cases, Tienari [1975] reported15% concordance in MZ and 7 % con-cordance in DZ by the pair-wisemethod, and 33% and 14% with theproband method.

In a hospital sample of 20 maleschizophrenic twin pairs Kringlen[1964] observed a concordance of 25%(2/8) in MZ and 16% (2/12) in DZ.This surprising finding led the author toembark on a large-scale nation-widetwin study, that was based upon amatching of a national twin register forthe period 1901–1930, comprisingmore than 25,000 pairs, with the Cen-tral register of psychoses. These twinswere in the age group of 35–64 years,and one or both of each pair had atsome time of their lives been hospital-ized for schizophrenia, manic depres-sive illness or reactive psychosis. Thestudy includes the largest number ofpsychotic twins personally investigatedand included both same-gender andopposite-gender DZ pairs, as well as or-dinary siblings. Zygosity diagnosis wasbased on information of similarity andblood- and serum-tests. The majority ofthe families of the MZ pairs includingthe parents and the siblings, were per-sonally investigated by the author.Forty-two pairs of DZ twins of thesame gender were also personally inves-tigated, but not their families. Kring-len’s [1967] two volume work com-prises clinical analysis, pedigree, ratingscale data and case histories of the MZpairs.

Of the 55 schizophrenic MZ pairs,45 were typical schizophrenic, whereas10 were classified as schizophreniformpsychosis, according to Langfeldt[1953]. Pairwise concordance was 25%in MZ and 4% in DZ, based upon hos-pitalization, and 38% and 10% whenbased upon personal investigation of

both twins in a pair and by applying abroader concept of schizophrenia (bor-derline schizophrenia).

After considering admission ratesto Norwegian hospitals, Kringlen wasunable to demonstrate any differencebetween twins and the general popula-tion in frequency of functional psycho-ses combined or for subgroups ofschizophrenia, manic depressive illnessand reactive psychosis. In the main itshould be possible to conclude safelythat the evidence available from thisstudy and the literature gives very littlesupport for the hypothesis that func-tional psychosis is more frequent intwins than in non-twins [Luxenburger,1928; Essen-Møller, 1941; Fischer,1973; Kendler and Robinette, 1983;Cannon et al., 1998]. Recently Klaning[1999] reported, however, that the rateof first admission to hospital for schizo-phrenia in DZ was 40% greater thanthat in the general population. The ratein MZ was, however, not increased,which one would expect if obstetriccomplication was a risk factor [Rydh-strøm, 1990].

Fischer et al. [1969] reported on anunselected Danish series of 21 MZ and41 DZ same-gender pairs of twins. Thesample was based upon a national twinregister, containing same-gender pairs,born 1870–1920, and a national psychi-atric in-patient register. Only probandswho met strict criteria for chronicschizophrenia were included. Personalinterviews was conducted in case of un-certain information from the hospitalrecords. The sample was rather old—35% of the twins were dead at the timethe study was reported—so age correc-tion was unnecessary. With a strict con-

cept of schizophrenia the MZ pairwiseconcordance was 24% (4/21), with 10%(4/41) in same-gender DZ. Probandconcordance was 36% for MZ and 18%for DZ. If concordance includedschizophreniform, paranoid and atypi-cal psychosis, pairwise concordance was40% in MZ and 19% in DZ [Fischer,1973] (Table II).

Gottesman and Shields [1966]studied 24 MZ and 33 DZ same-genderpairs of twins obtained through con-secutive admissions to the Maudsleyand the Bethlem hospitals in London.The ascertainment seemed to be rathercomplete because every patient admit-ted to those hospitals since 1948, hadbeen routinely asked whether he wasborn a twin. The medium age of thetwins was 37 years, with the range of19–64 years. Zygocity was diagnosedby a combination of blood-group andfinger-print analysis and the resem-blance in appearance. The twins wereinvestigated clinically by means of hos-pital records, tape-recorded interviewsand psychological tests [see also Gottes-man and Shields, 1972].

Clear-cut pairwise concordancewas 42% in MZ and 9% in DZ, withrespect to schizophrenia. The authorscollected 24 MZ pairs, of whom 10were concordant and 14 discordant. Inonly 4 of 10 concordant pairs, bothpartners were patients at the Maudsleyand Bethlem hospitals. In the remaining6 pairs, only one of the partners was apatient at the Maudsley, and the co-twin had been diagnosed as schizo-phrenic by follow-ups. Concordantpairs will have a greater chance of ap-pearing in a hospital sample. Particu-larly this is true if the hospital sample is

TABLE II. Etiological Importance of Genetic Factors in SchizophreniaBased Upon Population-Based Twin Studies from the Nordic Countries

Study G SE H’C SE

Kringlen [1967] 0.61 0.20 0.68 0.09Fischer [1973] 0.41 0.29 0.54 0.14Tienari [1975] 0.53 0.33 0.57 0.21

Adapted from Kendler [1983]. G is the coefficient of genetic determination. SE,standard error. H’C is a modification of the Hollinger statistic (heritability index) thatcan vary from 0–1.


small compared with the total twinpopulation. Hence, if one wants to cal-culate the real pair-wise concordancerate in the population on the basis ofthis study, the 4 of the 10 concordantpairs can be counted fully, whereas 6 of10 must be counted half. Thus the realconcordance for MZ with respect toschizophrenia might be 33% in thisstudy, a percentage that deviates con-siderably from the classical earlier stud-ies and that is in accordance with morerecent works from the Nordic coun-tries. Concordance measured with theproband method is of course higher(Table III).

The study by Kendler and Robi-nette [1983] was based upon the Na-tional Academy of Sciences-NationalResearch Council Registry, and psy-chiatric diagnoses were collected from avariety of clinical settings. Because thissample is based upon twins who hadserved in the military, there was a pau-city of cases with early onset. Also, arather wide diagnostic standard forschizophrenia was employed. Thus datafrom such a registry will likely under-estimate the true population concor-dance. In this large sample the probandconcordances are 30.9% in MZ and6.5% in DZ.

A study from Norway by Onstadet al. [ 1991] used structured psychi-atric interviews. Pairwise concordancewas 33% in MZ and 1% in DZ, and48% and 4% with the proband method.The concordance in DZ is exception-

ally low, that might be due to smallsample size. (DZ might in principlehave from 0–100% of their genes incommon). There was an increase ofschizophrenia and schizotypal personal-ity disorders in the co-twin of DZschizophrenics and in other first-degreerelatives, but no excess in other types ofpsychopathology.

In a study of same-gender twinsborn from 1940–1957 in Finland, thelifetime prevalence of schizophreniawas 2.0% [Cannon et al., 1998]. Pro-bandwise concordance was 46% in MZand 9% in DZ. Model fitting indicatedthat 83% of the variance in liability wasdue to additive genetic factors, and theremaining 17% was due to unique en-vironmental factors. The clinical diag-noses were based upon register infor-mation without systematic personalinvestigation of the twins. One had alsoexcluded twin pairs where one of thetwins had died before the investigationtook place. How these factors mighthave influenced the results is hard toknow.

In a study of 22 MZ and 23 same-gender DZ twin pairs, born after 1930and hospitalized for mental illness in aregion of Germany, overall pairwiseconcordance rates were 46% in MZ and17% in DZ, with probandwise 61% inMZ and 24% in DZ [Franzek andBeckmann., 1998]. Both MZ and DZconcordance rates are relatively highcompared with other studies from theliterature. The authors were particularlyinterested in the Leonard [1979] classi-fication and observed that MZ concor-dance was significantly higher than DZconcordance in strict schizophrenia ac-cording to DSM-III-R and unsystem-atic schizophrenia according to theLeonard classification. There was a ten-dency for strict schizophrenia to bemore genetically determined than theheterogeneous group of schizophreni-form, schizoaffective, delusional andpsychotic disorders not otherwise speci-fied. According to the Leonard classifi-cation unsystematic schizophreniaseemed to be heavily inherited.

In a recent study based on twinsadmitted to the Maudsley and BethlemHospitals, London, between 1948–1993, probandwise concordance for

schizophrenia was 40.9% (20/49) inMZ and 5.3% (3/57) in DZ [Cardno etal., 1999]. Probandwise concordancefor schizoaffective disorder was 39.1%(9/23) in MZ and 4.5% (1/22) in DZ.Heritability estimates for schizophrenia,schizoaffective disorder, mania were allbetween 82% and 85%, when ResearchDiagnostic Criteria, DSM-III-R andICD-10 were employed.

Based on a summary of the Nordicpopulation based studies, pair-wiseconcordance was on the average around30–35% for MZ and 5–10% for DZwith regard to schizophrenia. Proband-wise concordances were around 40% inMZ and 15% in DZ. Because of slightsampling errors even in the Nordicstudies, however, these rates must beconsidered to be maximum figures withregard to MZ concordance. In additionone should note that there is also dif-ference in concordance between DZand ordinary siblings, the rates for theDZ being higher. Clear-cut differencesare observed in the studies of Luxen-burger [1928], Slater and Shields[1953], Kringlen [1967] and Fischer[1973] (Table IV).

CO-TWINS

Because the discordance of schizophre-nia in MZ twins is considerable, onecould argue that it is not schizophreniaas such that is inherited, but a kind ofpersonality structure, that predisposecertain individuals to the developmentof the disorder.

In all modern twin studies ofschizophrenia, considerable disparity inthe clinical pictures of MZ pairs hasbeen reported. The most remarkablefinding is, however, the high preva-lence of normal co-twins, that was alsonoted by the pioneering twin research-ers. Based upon the studies of Essen-Moller [1941], Slater and Shields[1953], Tienari [1963], Kringlen[1967], Gottesman and Shields [1966],Fischer [1973] and Onstad et al [1991],32% of the co-twins of MZ schizo-phrenics were schizophrenic. Seven-teen percent were possible schizo-phrenics or borderline, 21% wereinflicted by a neurotic-like disorders,and 30% were clinically normal. Based

TABLE III. Co-Twins ofDefinite Schizophrenic MZTwins in Six Systematic Studies

N %

Schizophrenia 45 31.7Borderline schizophrenia 24 16.9Neurotic-like disorder 30 21.1Clinically normal 43 30.3Total 142 100.0

Based upon studies of Essen-Møller[1941], Slater and Shields [1953],Tienari [1963], Kringlen [1967],Gottesman and Shields [1972], andFischer [1973].


upon the three earlier Nordic popula-tion-based studies, as many as 40% wereclassified as normal.

A priori, one might surmise thatthe subtypes of schizophrenia thatappeared with schizophrenic or schizo-phrenic-like co-twins have a stronggenetic component, whereas theschizophrenics with neurotic or normalco-twins are more environmentally de-termined. According to the availablestudies, the various sub-types of schizo-phrenia seem to be more or less ran-domly paired with various co-twins.Particularly the normal co-twin may bepaired with any type of schizophrenia.Furthermore, the normal co-twin maybe paired not only with the moderatelysevere cases, but with extremely dete-riorated partners. This is also in accor-dance with findings that show thatthere are no genetic factors determiningthe outcome of schizophrenia [Kring-len, 1986].

In conclusion, several twin studiesclearly show that the clinical picture en-countered in the non-schizophrenicMZ co-twins is rather variable, rangingfrom a duplication of the schizophrenicpsychosis to schizotypal and paranoidpersonality disorder, neurosis and evenclinical normalcy. The findings lend noclear support to the idea that some sub-types of schizophrenia are more geneti-cally determined than others, exceptthat some studies show a slightly higherconcordance rate in non-paranoid thanparanoid schizophrenia [Kringlen,1967; Onstad et al., 1991]. Further-more, there is no evidence that the se-vere cases have a higher genetic load-ing. Finally, genetic factors do not

influence the course and outcome. Thislast point is in agreement with the studyby Ciompi [1980], where a subgroup ofschizophrenic patients with three ormore schizophrenics among first degreerelatives, had no worse outcome thanthe sub-group with no schizophrenicsat all in their families.

ENVIRONMENTALFACTORS

In theory, studies of discordance makeit possible to identify environmentalfactors that discriminate between dis-turbed and non-disturbed twins inMZ-pairs. Regrettably, up until nowfew research workers have capitalizedon such studies. Mainly, the resultsfrom the larger twin studies showthat birth order, birth weight, difficul-ties during birth, physical strength inearly childhood and psychomotoricdevelopment during the first years oflife are of practically no significancefor later development of schizophre-nia [Kringlen, 1987]. These findingsmight throw doubt on the validity ofthe neurodevelopmental hypothesis ofschizophrenia.

On the other hand, recent studieshave revealed minor but interesting dif-ferences in the brains of MZ discordantfor schizophrenia. In a CT-study of 12MZ pairs, clinically discordant forschizophrenia, Reveley et al. [1982]observed that the schizophrenic twinhad larger lateral ventricles comparedto the healthy co-twin. In a MRI-study of 15 MZ-pairs, also discordantfor schizophrenia, the schizophrenictwin, in contrast to the healthy co-

twin, had slight enlargement of thebrain ventricles, and slight reductionof hippocampus [Suddath et al.,1990]. Most likely these anatomicdeviations are congenital. Illness inpregnancy, birth complications or aninteraction between genetic and envi-ronmental organic factors might belikely explanations.

Slater and Shields [1953], Tienari[1963] and Kringlen [1967] all observedthat in discordant MZ-pairs, the moresubmissive, dependent and neurotictwin tends to be the one who developsschizophrenia. Although the differencesare small, the life history of schizo-phrenics and their co-twins seem to dis-play consistent early differences in per-sonality and parental treatment [Wahl,1976].

MZ-TWINS REAREDAPART

The study of MZ twins separated at anearly age and reared apart, is from a sci-entific point of view, the sovereignmethod. It goes without saying, how-ever, that it is hard to find such pairs.Only 6 MZ pairs have been recorded aspart of a consecutive series of schizo-phrenic subjects [Kringlen, 1990].There are more concordant than dis-cordant pairs, that might suggest thatenvironmental factors are usually notstrongly familial. A reservation is alsonecessary because a thorough examina-tion of concordant cases reveals that thechildhood experiences where usuallymiserable and stressful for both twins.Often the separated twins were broughtup in a closely related environment, forinstance by uncles and aunts.

OFFSPRING OFDISCORDANT PAIRS

There are few such studies, and theavailable samples are too small for a finalconclusion. In the thoroughly investi-gated sample by Kringlen and Cramer[1989] the number of schizophrenicsubjects are higher in the offspring ofMZ probands than in the offspring ofnon-schizophrenic co-twins. Thusthere were 3 schizophrenics (10.7%)among the 28 offspring of index cases in

TABLE IV. Morbidity Risk in DZ Twins and Ordinary Siblings*

Studies

Relation to index case

DZ co-twins Full sibs

Luxenburger [1928] 14.1 11.8Kallmann [1946] 10.3 10.2Slater and Shields [1953] 11.3 4.6Kringlen [1967] 8.1 3.0Fischer [1973) 26.6 10.0

*Uncorrected for age.


contrast to 1 case (2.2%) among the off-spring of non-schizophrenic co-twins.The difference is not statistically signifi-cant at the 5% level, which one wouldnot expect for such a small sample. Byincluding paranoid and schizotypal per-sonalities, the difference increases, inparticular when the data are based onMillon’s questionnaire (P < 0.01), thatmight point to environmental factorsconnected with rearing and family life.

The studies by Fischer [1973] andthe follow-up, including some addi-tional cases by Gottesman and Bertelsen[1989] had methodological weaknesses.For instance Fischer grouped togetheroffspring of both concordant pairs andoffspring of index cases in discordantpairs, and compared this group withoffspring of co-twins in discordantpairs. In addition, the information re-garding the offspring is scanty and prac-tically none of the offspring had beeninvestigated personally. See also critiqueby Torrey [1990, 1992].

DISCUSSION ANDCONCLUSIONIn this paper we have focussed attentionon concordance figures in twins withregard to schizophrenia. We havetouched upon various sources of errorand emphasized the importance of sam-pling with regard to concordance fig-ures. Based on a summary of recentpopulation based studies from the Nor-dic countries, pairwise concordance ison average around 30% for MZ and10% for DZ with regard to schizophre-nia, whereas probandwise concor-dances are around 40% in MZ and 15%in DZ. Because of slight sampling errorseven in the Nordic studies, these ratesmust be considered as maximum figureswith regard to MZ concordances. Inaddition one should note that there isalso differences in concordance be-tween DZ twins and ordinary siblings,the rates for the first ones in generalbeing higher, that indicate that thetwin-ship itself might enlarge the con-cordance figures. To quarrel about theexact figures is a waste of time, but it isobviously wrong when some Americantextbooks of psychiatry and scientificpapers state that concordance figures areabove 50% in MZ [Torrey, 1992].

That genes contribute considerablyto the etiology of schizophrenia is,however, beyond doubt. What onemight quarrel about is the magnitude ofthe genetic contribution. Obviously thegenes are not sufficient to produce theschizophrenic disorder. If schizophreniawas due entirely to the genes the con-cordance rates for MZ and DZ wouldapproach 100% and 50% respectively.The fact that the concordance figuresfor MZ fall far below 100% show thatenvironmental factors, organic or socialones, are involved.

Several twin studies clearly showthat the clinical picture encountered inthe non-schizophrenic MZ co-twins israther variable, ranging from a duplica-tion of the schizophrenic psychosis toschizotypal and paranoid personalitydisorder to ordinary neurotic symptomsand even clinical normalcy. There is noevidence that the severe cases of schizo-phrenia have a higher genetic loadingthan the more benign cases, and geneticfactors do not influence the course andoutcome.

What is then inherited in schizo-phrenia? We do not know. Most likelywe should conceive of the genetic dis-position as a certain potential for re-sponse pattern in the cognitive and per-ceptual field. The mode of inheritancealso remains uncertain. Schizophreniashows neither a broad or a narrow defi-nition of a Mendelian pattern of inher-itance [Gottesman and Shields, 1982].

According to most experts, themost plausible assumption is that themajority of cases of schizophrenia arepolygenetically determined. A poly-genic model would give a range of dys-function from mild to severe. The riskwhen it comes to relatives would in-crease as the number of affected familymembers increases, and the risk woulddrop as one goes from close relatives todistant ones. A polygenic inheritancecan also best explain the gliding transi-tion from normality to severe mentalillness found in twin partners and sib-lings and in parents of schizophrenics.This hypothesis also explains the factthat these illnesses occur rather fre-quently. Single gene diseases are ex-tremely rare.

In late year virus infection in preg-

nancy, time of birth (January–April atthe Northern hemisphere), and birthcomplications have been implicated inthe etiology of schizophrenia. Themagnitude of these factors is, however,unclear. The adoptive study by Tienari[1991] seems to indicate that social fac-tors associated with family life may beof etiological importance in the pres-ence of a genetic predisposition to thedisorder. It would seem that subjectswith the genetic predisposition are par-ticularly vulnerable to a noxious familymilieu.

The genetic liability to schizophre-nia is supported by several compellingfindings from methodologically soundfamily, twin and adoption studies. Twinstudies also suggest, however, that en-vironmental factors are of importance.

REFERENCES

Cannon TD, Kaprio J, Lønnquist J, Huttunen M,Koskenvuo M. 1998. The genetic epidemi-ology of schizophrenia in a Finnish twincohort. Arch Gen Psychiatry 55:67–74.

Cardno AG, Marshall EJ, Coid B, MacdonaldAM, Ribchester TR, Davies NJ, Venturi P,Jones LA, Lewis SW, Sham PC, GottesmanII, Farmer AE, McGuffin P, Reveley AM,Munroy RM. 1999. Heritability estimatesfor psychotic disorder. Arch Gen Psychiatry56:162–168.

Ciompi L. 1980. Catamnestic long-term studieson the course of life of schizophrenics.Schizophr Bull 6:604–618.

Dalgard OS, Kringlen E. 1976. A Norwegiantwin study of criminality. Br J Criminol 16:213–232.

Essen-Møller E. 1941. Psychiatrische Untersu-chungen an einer Serie von Zwillingen.Acta Psychiatr Neurol Scand (suppl 23)16:7–201.

Fischer M, Harvald B, Hauge M. 1969. A Danishtwin study of schizophrenia. Br J Psychiatry115:981–990.

Fischer M. 1973. Genetic and environmental fac-tors in schizophrenia. Acta Psychiatr Scand(suppl 238) 49:1–158.

Franzek E, Beckmann H. 1998. Different geneticbackground of schizophrenia spectrum psy-choses: a twin study. Am J Psychiatry 155:76–83.

Gottesman II, Shields J. 1966. Schizophrenia intwins: 16 years consecutive admissions to apsychiatric clinic. Br J Psychiatry 112:809–818.

Gottesman II, Bertelsen A. 1989. Confirming un-expressed genotypes for schizophrenia: risksin the offspring of Fischer’s Danish identicaland fraternal discordant twins. Arch GenPsychiatry 46:867–872.

Gottesman II, Shields J. 1972. Schizophrenia andgenetics: a twin study vantage Point. NewYork: Academic.

Gottesman II, Shields J. 1982. Schizophrenia: the


epigenetic puzzle. Cambridge MA: Cam-bridge University Press.

Inouye E. 1961. Similarity and dissimilarity ofschizophrenia in twins. In: Proceedings ofthe Third World Congress of Psychiatry,vol. 1. Montreal: University of TorontoPress. p 524–530.

Kallmann FJ. 1946. The genetic theory of schizo-phrenia: an analysis of 691 schizophrenictwin index families. Am J Psychiatry 103:309–322.

Kendler KS. 1983. Overview: a current prospec-tive on twin studies of schizophrenia. Am JPsychiatry 140:1413–1425.

Kendler KS, Robinette CD. 1983. Schizophreniain the National Academy of Sciences-National Research Council Twin Registry:a 16-year update. Am J Psychiatry 140:1551–1563.

Kendler KS. 1993. Twin studies of psychiatricillness. Current status and future directions.Arch Gen Psychiatry 50:905–915.

Kendler KS, Neale MC, Kessler RC, Heath AC,Eaves LJ. 1993. A test of the equal-environment assumption in twin studies ofpsychiatric illness. Behav Gen 23:21–27.

Klaning U. 1999. Greater occurrence of schizo-phrenia in dizygotic but not monozygotictwins. Register-based study. Br J Psychiatry175:407–409.

Kringlen E. 1964. Schizophrenia in male mono-zygotic twins. Acta Psychiatr Scand (suppl178) 40:1–76.

Kringlen E. 1967. Heredity and environment inthe functional psychoses: an epidemiologi-cal-clinical twin study, vol. I and II (casehistories). Oslo, University Press and Lon-don: William Heinemann 1968.

Kringlen E. 1976. Twins—still our best method.Schizophr Bull 2:429–433.

Kringlen E. 1978. Adult offspring of two psy-chotic parents, with special reference toschizophrenia. In: Wynne IC, CromwellRI, Mathysse S, editors. The nature of

schizophrenia: new approaches to researchand treatment. New York: Wiley. p 9–24.

Kringlen E. 1986. Status of twin research in func-tional psychoses. Psychopathology 19:85–92.

Kringlen E. 1987. Contributions of genetic stud-ies in schizophrenia. In: Hafner H, GattazWF, Janzarik W, editors. Search for thecauses of schizophrenia. Berlin: Springer. p123–142.

Kringlen E, Cramer G. 1989. Offspring of mono-zygotic twins discordant for schizophrenia.Arch Gen Psychiatry 46:873–877.

Kringlen E. 1990. Genetic aspects of schizophre-nia with special emphasis on twin research.In: Kringlen E, Lavik NJ, Torgersen S, edi-tors. Etiology of mental disorder. Oslo: De-partment of Psychiatry, Vinderen, Univer-sity of Oslo. p 63–88.

Langfeldt G. 1953. Some points regarding thesymptomatology and diagnosis of schizo-phrenia. Acta Psychiatr Scand (suppl)80:7–26.

Leonard K. 1995. Aufteilung der endogenen Psy-chosen und ihre differenzierte Atiologie.7th ed. New York: Thieme.

Luxenburger, H. 1928. Vorlaufiger Bericht uberpsychiatrische Serieuntersuchungen anZwillingen. Z ges Neurol Psychiat 116:297–326.

Luxenburge r H. 1930 . P sych i a t r i s ch-neurologische Zwillingpathologie. Z gesNeurol Psychiat 56:145–180.

Onstad S, Skre I, Torgersen S, Kringlen E. 1991.Twin concordance for DSM-III-R schizo-phrenia. Acta Psychiatr Scand 83:395–401.

Onstad S, Skre I, Torgersen S, Kringlen E. 1992.Birthweight and obstetric complications inschizophrenic twins. Acta Psychiatr Scand85:70–73.

Revely AM, Revely MA, Clifford CA, MurrayRM. 1982. Cerebral ventricular size intwins discordant for schizophrenia. Lancet1:540–541.

Rosanoff AJ, Handy LM, Plesset IR, Brush S.1934. The etiology of so-called schizo-phrenic psychoses. With special reference totheir occurrence in twins. Am J Psychiatry91:247–296.

Rosenthal D. 1962. Problems of sampling anddiagnosis in the major twin studies ofschizophrenic twins. J Psychiatr Res 2:116–134.

Rosenthal D. 1966. The offspring of schizo-phrenic couples. J Psychiatr Res 4:169–188.

Rydhstrøm H. 1990. The effects of maternal age,parity, and sex of the twins on twin perina-tal morbidity. A population based study.Acta Genet Med Gemellol 39:401–408.

Slater E, Shields J. 1953. Psychotic and neuroticillnesses in twins. Medical Research Coun-cil Special Report Series No. 278, London:Her Majesty’s Stationary Office.

Suddath RL, Chritison GW, Torrey EF, Casa-nova MF, Weinberger DR. 1990. Ana-tomical abnormalities in the brains ofmonozygotic twins, discordant for schizo-phrenia. N Engl J Med 322:789–794.

Tienari P. 1963. Psychiatric illness in identicaltwins. Acta Psychiatr Scand 171:1–195.

Tienari E. 1975. Schizophrenia in Finnish maletwins. In: Lader MH, editor. Studies ofschizophrenia. Br J Psychiatry, Spec PublNo. 10.

Tienari E. 1991. Interaction between genetic vul-nerability and family environment: theFinnish adoptive family study of schizo-phrenia. Acta Psychiatr Scand 84:460–465.

Torrey EF. 1990. Offspring of twins with schizo-phrenia. Arch Gen Psychiatry 47:976–977.

Torrey EF. 1992. Are we overestimating the ge-netic contribution to schizophrenia?Schizophr Bull 18:159–170.

Wahl OF. 1976. Monozygotic twins discordantfor schizophrenia: a review. Psychol Bull83:91–106.


Documents

Twin studies in schizophrenia with special emphasis on concordance figures