Childhood Cognitive Functioning in Schizophrenia Patients and Their Unaffected Siblings: A...

Childhood Cognitive Functioning inSchizophrenia Patients and Their Unaffected

Siblings: A Prospective Cohort Studyby Tyrone D. Cannon, Carrie E. Bearden, J. Megginson HoUister,

Isabelle M. Rosso, Laura E. Sanchez, and Trevor Hadley

Abstract

While it is known that children of schizophrenia par-ents perform more poorly on tests of cognitive func-tioning than children of normal parents, less certain isthe degree to which such deficits predict schizophreniaoutcome, whether cognitive functioning deterioratesduring childhood in preschizophrenia individuals, andwhether nongenetic etiologic factors (such as obstetriccomplications) contribute to these deficits. In the pres-ent study, 72 patients with schizophrenia or schizoaf-fective disorder, 63 of their siblings not diagnosed withschizophrenia, and 7,941 controls with no diagnosiswere ascertained from a birth cohort whose membershad been evaluated with standardized tests of cogni-tive functioning at 4 and 7 years of age. Adult psychi-atric morbidity was ascertained via a longitudinaltreatment data base indexing regional public healthservice utilization, and diagnoses were made by reviewof all pertinent medical records according to DSAf-TVcriteria. Both the patients with schizophrenia and theirunaffected siblings performed significantly worse thanthe nonpsychiatric controls (but did not differ fromeach other) on verbal and nonverbal cognitive tests at4 and 7 years of age. Preschizophrenia cases and theirsiblings were increasingly overrepresented acrossdecreasing quartiles of the performance distributions.There was not significant intra-individual decline, andthere were no significant relationships between obstet-ric complications and test performance among thepreschizophrenia subjects. These results suggest thatduring the period from age 4 to age 7 years, premor-bid cognitive dysfunction in schizophrenia represents arelatively stable indicator of vulnerability derivingfrom primarily genetic (and/or shared environmental)etiologic influences.

Keywords: Schizophrenia, cognition, premorbidfunctioning, genetics.

Schizophrenia Bulletin, 26(2):379-393,2000.

Findings of cytoarchitectural and morphologic brainchanges consistent with a prenatal and perinatal originhave led to the notion that neurodevelopmental distur-bances contribute to the etiology of at least some cases ofadult schizophrenia (Kovelman and Scheibel 1984; Jakoband Beckmann 1986; Weintraub 1987; Arnold et al. 1991;Akbarian et al. 1993; Cannon et al. 1993). However, theetiologic correlates of these disturbances and the propor-tion of the population with schizophrenia that they charac-terize remain to be determined (Lewis and Murray 1987;Cannon and Mednick 1991).

Prospective examination of cognitive functioning inindividuals who develop schizophrenia as adults provides anindirect means to address these questions. Longitudinal"high-risk" studies have consistently reported that childrenof patients with schizophrenia perform more poorly on neu-ropsychological tests than children of other people(Mednick and Schulsinger 1968; Landau et al. 1972;Asarnow et al. 1978; Rutschmann et al. 1980; Harvey et al.1981; Winters et al. 1981; Worland et al. 1982; Driscoll1984; Lifshitz et al. 1985; Sohlberg 1985; Hallett et al.1986; Fish 1987; Goodman 1987; Sameroff et al. 1987;Weintraub 1987; Erlenmeyer-Kimling et al. 1989; Schreiberet al. 1992; Marcus et al. 1993; Bergman and Walker 1995).These findings suggest that early signs of brain compromisein preschizophrenia subjects may be mediated at least in partby genetic predisposition to the disorder. However, few ofthe samples in these studies have passed through the periodof risk for onset of schizophrenia, and those that have con-tain only a handful of target outcomes. Thus, presently lim-ited information is available on the degree to which earlycognitive deviance predicts schizophrenia in such samples.It also remains unclear whether high-risk findings can begeneralized to the total population of schizophrenia subjects,or whether they are relevant to only the 5-10 percent ofpatients with a parent diagnosed with schizophrenia.

Reprint requests should be sent to Prof. T.D. Cannon, University ofCalifornia, Los Angeles, Dept. of Psychology, 1285 Franz Hall, Box951563, Los Angeles, CA 90095-1563; e-mail: [email protected].

379

by guest on September 22, 2016

http://schizophreniabulletin.oxfordjournals.org/D

ownloaded from

http://schizophreniabulletin.oxfordjournals.org/

Schizophrenia Bulletin, Vol. 26, No. 2, 2000 T.D. Cannon et aL

Recent population-based cohort studies suggest thatpremorbid deficits in general cognitive ability may in factcharacterize a majority of schizophrenia patients. In twoprospective British cohort studies, preschizophrenia chil-dren scored significantly lower than comparison subjectson standardized intelligence tests from middle childhoodthrough adolescence (Jones et al. 1994; Jones and Done1997). In one of these cohorts, Jones et al. (1994) foundthat cases had significantly lower intelligence quotients(IQs) than controls at 8, 11, and 15 years of age, with sig-nificant linear increases in risk for adult schizophreniaacross decreasing tertiles of the distributions of generalintellectual functioning at the 11- and 15-year assess-ments. This pattern suggests that most preschizophreniasubjects are below their expected level of cognitive func-tioning at these ages. That is, if only a subgroup of suchcases were cognitively impaired, risk should be elevatedamong individuals in the lowest tertile of the distributionbut should not differ between individuals in the upper ter-tiles. Power to reject a linear association in favor of a sub-group effect was limited in this study, however, since only30 schizophrenia subjects were ascertained and risk wasmodeled at only three levels of ability. In a Swedish Armycohort study with a larger number of schizophrenia out-comes, David et al. (1997) found a linear relationshipbetween IQ at 18 years and later schizophrenia, with riskprogressively increasing as IQ declined across the entirerange of intellectual ability. Unfortunately, the Armycohort findings can be generalized only to male cases ofschizophrenia.

Only a few studies have examined the developmentaltime course of the emergence of early signs of cognitivedeviance in preschizophrenia subjects. In an early meta-analytic review of the literature, Aylward et al. (1984)concluded that children not yet diagnosed with schizo-phrenia show significant IQ deficits when compared withage-matched normal controls from kindergarten throughhigh school but that there was little evidence for cognitivedecline before onset of psychosis. In contrast, the findingsof two recent cohort studies point to increasing cognitivedeviance as preschizophrenia individuals approach illnessonset. In the Jones et al. (1994) study, the linear trend ofincreased risk for schizophrenia across decreasing levelsof performance became stronger with age, such that thedistribution of preschizophrenia children was increasinglyshifted toward lower categories of intellectual function-ing. However, the issue of whether cases showed signifi-cant cognitive decline intra-individually was notaddressed. In a followup study of the Providence cohortof the National Collaborative Perinatal Project (NCPP), alarge decline in IQ scores between 4 and 7 years predictedthe presence of psychotic symptoms in young adulthood(Kremen et al. 1998). Unfortunately, the relevance of

these findings to schizophrenia is uncertain, as only 1 ofthe 54 subjects in the group with a large IQ decline wasdiagnosed with schizophrenia.

At present, information is limited on whether non-hereditary etiologic influences contribute to premorbidcognitive dysfunction in schizophrenia. Of the nongeneticinfluences proposed to exist in schizophrenia, obstetriccomplications (OCs)—particularly those associated withfetal oxygen deprivation, fetal underdevelopment, andmaternal infection during pregnancy—have the mostrobust association (see Cannon 1997 for a review). It hasbeen suggested that a history of such OCs is related to adistinct "neurodevelopmental" form of schizophrenia—involving early cognitive, behavioral, and motor abnor-malities; an early age at onset; and poor prognosis (Lewisand Murray 1987; Murray and Lewis 1988; Foerster et al.1991; O'Callaghan et al. 1992). Among the few high-riskstudies that examined the contribution of obstetric factorsto premorbid cognitive dysfunction in schizophrenia, find-ings are conflicting (Mednick and Schulsinger 1968;Rieder et al. 1977; Fish et al. 1992; Marcus et al. 1993).One investigation found no significant associationbetween a history of OCs and cognitive deficits duringinfancy and childhood among individuals at high geneticrisk for schizophrenia (Marcus et al. 1993). In contrast,Rieder and colleagues (1977) found that vaginal bleedingand maternal edema predicted lower 7-year verbal IQscores, particularly in the verbal domain, among the off-spring of schizophrenia subjects but not normal controls.In two other studies, low birth weight was associated withcognitive developmental deviance among high-risk butnot low-risk subjects (Mednick and Schulsinger 1968;Fish et al. 1992). However, it is unclear that the relation-ship observed between birth weight and IQ in these stud-ies reflected a nongenetic etiologic influence, because ofthe possibility of genotype-environment covariation. Suchcovariation occurs when a putative environmental riskfactor (e.g., low birth weight) is more common in high-risk compared with low-risk subjects overall, thus con-founding the genetic and presumptive nongenetic influ-ences.

In the present study we used a prospective cohortdesign to determine (1) the degree to which cognitive dys-function during childhood predicts schizophrenia, (2)whether similar deficits appear in the unaffected siblingsof patients with schizophrenia, (3) the developmental timecourse of these deficits, and (4) whether a history of fetalhypoxia and other OCs contribute to these deficits. Theoriginal cohort consisted of 9,236 individuals bom inPhiladelphia from 1959 to 1966 whose gestations andbirths were monitored prospectively with standardresearch protocols, with followup examinations usingstandardized tests of cognitive functioning at 4 and 7

380



ownloaded from


Childhood Cognitive Functioning and Schizophrenia Schizophrenia Bulletin, Vol. 26, No. 2, 2000

years of age. Psychiatric outcome was ascertained byscreening the cohort for contacts with local mental healthfacilities in adulthood, followed by chart reviews to deter-mine diagnoses according to DSM-IV (AmericanPsychiatric Association 1994) criteria.

Methods

Subjects and Psychiatric DiagnosesFormation of cohort From 1959 to 1966, the NCPP

enrolled for study 9,236 offspring of 6,753 mothers whodelivered at two inner-city hospital obstetric wards inPhiladelphia—the Pennsylvania Hospital and theChildren's Hospital of Philadelphia (Niswader andGordon 1972). The offspring in over 90 percent of all thedeliveries at these two sites during the sampling periodwere enrolled. 53.6 percent (n = 4,956) of the cohortmembers were the only children from their familiesenlisted in the study, and the remaining 46.3 percent (n =4,280) were from families with two or more children par-ticipating. The recruitment sites for the Philadelphiacohort were chosen to result in a predominantly African-American cohort (88%), thus permitting ethnic balanceacross the NCPP study sites taken together.

Diagnostic screening and evaluation. In January1996, we conducted a search of the Perm LongitudinalDatabase (Rothbard et al. 1990), a city wide data base forregistration of contacts with public mental health facilitiesin Philadelphia from 1985 to 1995, ascertaining 1,197individuals whose names and dates of birth matched thoseof subjects in the birth cohort In total, 339 (3.7%) of thecohort members had a psychotic disorder diagnosis (194with schizophrenia or schizoaffective disorder, and 145with affective or drug-induced psychosis), and 858 (9.3%)had nonpsychotic disorder diagnoses (i.e., affective, anxi-ety, adjustment, developmental, and substance abuse dis-orders). Because these register diagnoses were assignedby hundreds of different clinicians in scores of differenttreatment settings without an explicit attempt to standard-ize diagnostic procedures and criteria, they are unlikely tohave a sufficient degree of precision in differentiatingschizophrenia from other psychotic disorder categories.We therefore undertook a diagnostic validation studybased on a review of the psychiatric medical records ofthe psychotic-disordered probands. We reviewed medicalrecords for 144 such probands who had been treated at 15mental health facilities that agreed to cooperate with thestudy team in providing access to medical records. Theremaining psychotic-disordered probands whose chartswere not reviewed (n = 195) had been treated at facilitiesthat no longer existed, refused to cooperate, or were notcontacted because only one proband had been treated

there. The participating facilities included inpatient ser-vices, day treatment clinics, and case management ser-vices and were in all apparent ways equivalent to the non-participating facilities.

Six diagnosticians (two psychiatrists, two clinicalpsychologists, and two advanced graduate students inclinical psychology) performed the chart reviews. An ini-tial set of 10 charts was used in training to calibrate theless experienced with the more experienced diagnosti-cians. A standard coding form was used to record infor-mation pertinent to DSM-IV diagnostic criteria, differen-tial diagnosis of schizophrenia and affective disorders,age at first treatment contact, and course. In the vastmajority of cases the information present in the recordsspanned the patient's illness, and a clear picture of theduration and primacy of psychotic symptoms emerged. Ofthe 144 cases whose charts were reviewed, 72 received aDSM-IV diagnosis of schizophrenia or schizoaffectivedisorder, 41 were diagnosed as having a psychotic formof major depressive disorder or bipolar disorder; and theremaining 31 were given a primary diagnosis of substanceabuse, anxiety disorder, atypical psychosis, psychotic dis-order caused by a general medical condition, personalitydisorder, or adjustment disorder. For a randomly selectedsample of 56 of the cases (excluding those used in train-ing), charts were evaluated independently by two or moredifferent examiners, with good agreement on the diagno-sis of schizophrenia and schizoaffective disorder (K =0.85). There was only moderate agreement between thechart-based diagnoses of schizophrenia and schizoaffec-tive disorder with the original register diagnoses (K =0.63).

Demographics of comparison groups. The 72 caseswith chart-review-based DSM-FV diagnoses of schizo-phrenia or schizoaffective disorder had 63 siblings notdiagnosed with schizophrenia who were also NCPP studyparticipants. The 63 siblings included 7 with a history ofpsychiatric treatment (1 with psychosis not otherwisespecified, 1 with mental retardation, and 5 with affectiveand anxiety disorders) and 56 without such a history. Forcontrols we used the cohort members without a siblingdiagnosed with schizophrenia who, according to our psy-chiatric screen, had not been treated in a public mentalhealth facility in greater Philadelphia as an adult (n =7,941). The remaining 1,160 members of the originalcohort were excluded from the primary analyses. Therewere four categories of excluded subjects: (1) 72 caseswith a history of treatment for a psychotic disorder whowere determined by chart review not to have schizophre-nia or schizoaffective disorder, (2) 194 cases with a his-tory of treatment for a psychotic disorder whose chartswere not reviewed (i.e., 195 minus 1 included sibling ofan included schizophrenia patient), (3) 852 cases with a

381



ownloaded from


Schizophrenia Bulletin, Vol. 26, No. 2, 2000 T.D. Cannon et al.

history of treatment for a nonpsychotic psychiatric disor-der (i.e., 858 minus 6 included siblings of included sub-jects with schizophrenia), and (4) 42 cases of fetal orneonatal death.

Sixty-nine percent of the original cohort members (n =6,399) were followed up at 4 years of age and 79 percent (n= 7,326) at 7 years of age with standardized tests of cogni-tive functioning. Tables 1 and 2 provide demographic infor-mation on the subjects examined at the age 4 and age 7assessments, respectively, according to diagnostic group.

Cognitive and Behavioral Variables. Subjects wereevaluated with the full Stanford-Binet Intelligence Scale(Form L-M) (Terman and Merrill 1960; Thomdike 1973)at age 4 and with 7 of the 10 subtests of the WechslerIntelligence Scale for Children (WISC; Wechsler 1949) atage 7 (i.e., Information, Vocabulary, Comprehension,Digit Span, Picture Completion, Block Design, andCoding). Age-corrected standard scores (prorated forWISC measures) were computed using published normsfor each instrument. While the Stanford-Binet scale yieldsonly an overall measure of cognitive functioning, theWISC yields separate measures of verbal and nonverbalfunctioning (i.e., Verbal and Performance IQ scores,respectively), which were examined in addition to theWISC Full-Scale IQ score.

Lateral dominance was evaluated at each assessmentby recording which hand subjects used in writing andmanipulating objects. Cases of ambiguous or mixed hand-edness were collapsed with those of left-handedness forthe analyses. In addition, at each assessment, the psychol-ogist performing the cognitive evaluations also rated eachsubject for several aspects of behavioral adjustment (i.e.,emotional reactivity, cooperation, dependency, goal orien-tation, degree of activity, and communication), with anoverall summary rating indicating whether behavioraldeviance was absent, suspected, or definitely present. Wecounted ratings of suspected or definite abnormality asindicating behavioral deviance.

Obstetric Variables. We modeled three classes of obstet-ric influences that could be related to premorbid cognitivedeficits in schizophrenia: hypoxia-associated OCs, other(prenatal) OCs, and birth weight. A detailed description ofthe formation of the hypoxia-associated OCs scale andprenatal OCs scale is provided elsewhere (see Cannon etal., this issue). Briefly, the hypoxia-associated OC scalewas formed by selecting complications considered to beeither direct (blue at birth, required resuscitation, neonatalcyanosis, neonatal apnea) or indirect (abnormalities offetal heart rate or rhythm, umbilical cord knotted orwrapped tightly around neck, third trimester bleeding,

Table 1. Demographic and performance characteristics of cases examined at age 4 by adultpsychiatric outcome

Characteristic

Age at examination, mean (± SD)Years of parental education, mean (± SD)Socioeconomic status, mean (± SD)Stanford-Binet IQ, mean (± SD)Gender, n (%)

MalesFemales

Race, n (%)BlacksWhites

Handedness, n (%)Left/mixedRight

Behavioral deviance, n (%)YesNo

IQ quartile, n (%)1st2nd3rd4th

Nodiagnosis(n = 5,127)

4.0(0.1)10.5(2.1)3.4(1.9)

93.4(13.9)

2,532 (49)2,595(51)

4,614 (90)513(10)

1,051 (21)4,076 (79)

590(12)4,537 (88)

1,327(26)1,148(22)1,490 (29)1,162(23)

Schizophreniapatients(n = 46)

4.0(0.1)10.3(1.9)2.9(1.9)

84.2(15.6)

32 (70)14(30)

46(100)0(0)

21(46)25(54)

13(28)33(72)

20(43)13(28)9(20)4(9)

Unaffectedsiblings(n = 33)

4.0(0.1)10.1 (1.8)2.5 (2.3)

85.5(15.5)

16(49)17(51)

33(100)0(0)

6(18)27 (82)

4(12)29(88)

18(55)5(15)4(12)6(18)

Note.—IQ = intelligence quotient; SD = standard deviation.

382



ownloaded from



Table 2. Demographic and performance characteristics of cases examined at age 7 by adultpsychiatric outcome

Characteristic

Age at examination, mean (± SD)Years of parental education, mean (± SD)Socioeconomic status, mean (± SD)WISC Full-Scale IQ, mean (± SD)Gender, n (%)

MalesFemales

Race, n (%)BlacksWhites

Handedness, n(%)Left/mixedRight

Behavioral deviance, n (%)YesNo

IQ quartile, n (%)1st2nd3rd4th

Nodiagnosis(n = 5,829)

• 7.4(1.3)10.5(2.0)3.4(1.8)

90.6(11.7)

2,865 (49)2,964(51)

5,316(91)513(9)

687(12)5,142(88)

509(9)5,320(91)

1,313(23)1,583(27)1,423(24)1,510(26)

Schizophreniapatients(n = 57)

7.6(1.5)10.3(2.1)2.8(1.8)

84.4(13.1)

41 (72)16(28)

56 (98)1(2)

12(21)45(79)

10(18)47 (82)

21 (37)18(32)10(17)8(14)

Unaffectedsiblings(n = 51)

7.5(1.3)9.9 (2.2)2.6(2.1)

84.4 (9.8)

24(47)27(53)

51 (100)0(0)

5(10)46(90)

10(20)41 (80)

21 (41)14(28)13(25)3(6)

Note.—WISC = Weschler Intelligence Scale for Children; IQ = intelligence quotient; SD » standard deviation.

placental hemorrhaging or infarcts, polyhydramnios,meconium in amniotic fluid, underdeveloped lungs) indi-cators of fetal oxygen insufficiency. The indirect indica-tors were chosen based on empirical validation againstdirect hypoxia indicators in prior studies (Low et al. 1992,1995; Arabin et al. 1993; Maier et al. 1994; Adamson etal. 1995; Salafia et al. 1995). Each of the indirect indica-tors was also significantly predictive of one or more of thedirect hypoxia indicators in this cohort. We did not weightthe indirect indicators according to strength of this associ-ation, however, because the direct indicators reflectedhypoxia at or shortly after birth, whereas many of theindirect indicators could reflect hypoxia before birth (sen-sitivity to which would be reduced if the items wereweighted according to their relationship with birthhypoxia). Each complication contributed 1 point to thetotal hypoxia-associated OC score.

Five prenatal OCs that were not significantly associ-ated with the direct hypoxia measures in this sample (andwhose primary impact on the fetus was unlikely to beoxygen deprivation) were included in a prenatal OCsscale: maternal infection during pregnancy (includingviral, bacterial, and fungal), maternal cardiovascular ill-ness during pregnancy, maternal pulmonary illness duringpregnancy, maternal hematologic illness during preg-

nancy, and maternal endocrine illness during pregnancy.Again, each complication contributed 1 point to the scaletotal.

Statistical Analysis. The primary goal of the data analy-ses was to evaluate the relationships between the mea-sures of cognitive functioning during childhood and adultdiagnostic outcome while controlling for a number ofsociodemographic influences that could potentially con-found such relationships. Because it was possible forthere to be more than one subject from the same familywithin an outcome group, and because the cognitive testscores of such cases are not independent, it was first nec-essary to reduce the degrees of freedom available foranalysis to preserve the assumption of independence ofobservations. One way that this data reduction could beachieved would be by randomly selecting one subject perfamily within each outcome group and excluding theremaining subjects, but this approach does not guaranteethat the included subjects are representative of the remain-ing subjects when the group size is small (as in the case ofthe unaffected siblings in this study). To avoid this pitfallwe computed averages of the test scores (and covariates)within each family (separately for each diagnostic group).The procedure just described produced a total of 46

383



ownloaded from



unique observations in the proband group, 24 uniqueobservations in the sibling group, and 4,184 unique obser-vations in the control group for the age 4 analyses. For theage 7 analyses the corresponding numbers were 57, 32,and 4,557, respectively. To determine whether the familyanalyses obscured any relationships among variables thatwere present at the level of individuals, the analyses werealso performed using the individualized data. There wereno differences in the relationships that were or were notsignificant between the two analytic approaches.Therefore, we report only the results using the data inwhich each observation in each diagnostic group is uniquewith respect to family of origin.

We first evaluated whether the cognitive test scoresvaried significantly by diagnostic outcome using analysisof variance and covariance. Each test measure was ana-lyzed as the dependent variable in a separate analysis—with gender, race, and adult diagnostic group (probands,siblings, controls) as classification variables—and withage at examination, parental education (mean number ofyears of schooling), and socioeconomic status (reflectingthe occupation and income of primary wage earner on ascale of 1 [unemployed, on public assistance] to 9 [profes-sional, upper middle class]) as continuously scaled covari-ates. In the presence of a significant overall main effect ofdiagnostic group, independent sample t tests were per-formed comparing the schizophrenia cases with the con-trols and the siblings with the controls. The matched-pair tstatistic was used to compare the schizophrenia cases withthe unaffected siblings from their own families.

We also performed categorical data analyses to exam-ine whether the preschizophrenia cases and their siblingswere distributed unevenly among quartiles of the IQ distrib-utions at age 4 and age 7. These analyses used a generalizedlogits logistic regression model (Stokes et al. 1995). Theoutcome measure classified the cohort members into threegroups: probands, siblings, and controls. Gender and racewere included as categorical predictors; and age at examina-tion, parental education, socioeconomic status, and IQ cate-gory (1st, 2nd, 3rd, 4th quartile) were included as interval-scale predictors. Two analyses were conducted: one usingStanford-Binet IQ category at age 4 as the cognitive predic-tor and the other using WISC Full-Scale IQ category at age7 as the cognitive predictor. In the presence of a significantoverall effect of IQ category, odds ratios were computedcontrasting each outcome (i.e., schizophrenia cases, unaf-fected siblings of schizophrenia subjects) with the cohortmembers with no psychiatric diagnoses.

To determine whether subjects in any of the diagnos-tic groups changed their relative positions in the IQ distri-butions across the age 4 and age 7 assessments, arepeated-measures analysis of variance was conducted onthe percentile ranks for the IQ scores at age 4 and age 7,

with adult diagnostic outcome as a between-subject vari-able and assessment wave as a within-subject, repeated-measures variable. Percentile ranks were computed withinthe 4,562 unique observations (44 probands, 24 siblings,and 4,494 controls) with test scores available from both ofthe childhood assessments.

The contribution of OCs to premorbid cognitive func-tioning in schizophrenia was examined using analysis ofvariance. Separate analyses were carried out with eachmeasure of childhood IQ as the dependent variable. Ineach case, the hypoxia-associated OCs scale, the prenatalOCs scale, birth weight, adult diagnosis, and the interac-tions of the three obstetric variables with diagnosis wereincluded as independent variables. We also evaluatedwhether differences in childhood cognitive functioningamong the diagnostic groups were moderated by gender,handedness, or childhood behavioral disturbances andwhether childhood cognitive dysfunction in the schizo-phrenia group was specific to those with an early age atfirst treatment contact.

Results

Comparisons of Mean Differences by DiagnosticGroup. Table 3 gives the results of the analyses of vari-ance and covariance comparing the three diagnosticgroups (probands, siblings, controls) on Stanford-Binet IQat age 4 and WISC Full-Scale IQ at age 7. Each of thedemographic factors accounted uniquely for a significantproportion of the variance in test scores at age 4, and allbut gender accounted uniquely for a significant proportionof the variance in test scores at age 7. In general, childrenwith higher scores were older, came from more affluentfamilies, and had better educated parents. Caucasians hadhigher scores than African-Americans. In the case ofStanford-Binet IQ at age 4, females had higher scoresthan males. Importantly, the differences in test scores byadult diagnostic outcome that were apparent in unadjusteddata (tables 1 and 2) remained significant even after con-trolling for these sociodemographic influences.

Figure 1 gives the least-squares mean ± standard errorof the mean (SEM) scores (adjusting for the effects of age,socioeconomic status, parental education, gender, and race)of the patient, sibling, and control groups on Stanford-BinetIQ at age 4 and WISC Full-Scale IQ at age 7. At bothassessments, the nonpsychiatric controls had significantlyhigher scores than schizophrenia cases (age 4: t = 4.1, d$ =•4229, p < 0.0001; age 7: t = 3.53, df= 4613, p < 0.0001) andtheir siblings (age 4: t = 2.1, df= 4207, p < 0.02; age 7: t =2.3, df= 4588, p < 0.01). Nearly identical results wereobtained in relation to WISC Verbal IQ (controls vs.patients: t = 2.8, df= 4613, p < 0.002; controls vs. siblings: t

384



ownloaded from



Table 3. Contributions of demographic factors and adult diagnostic group to Stanford-Blnet IQ atage 4 and WISC Full-Scale IQ at age 7

Source of variation

Age at examination

Socioeconomic statusParental education

Sex

RaceDiagnosis groupError

at111

11

24267

Age 4MS

4116.8

2697.917987.33538.7

6518.81817.9

175.5

F

23.5*15.4*

102.5*20.2*37.1*

10.4*

Of

1111 .

1

24641

Age 7MS

9183.55977.7

16167.739.4

9676.41026.4

117.9

F

77.9*

50.7*

137.2*

0.382.1*8.7*

Note.—WISC - Weschler Intelligence Scale for Children; IQ - intelligence quotient.*p< 0.0001

Figure 1. Mean (± standard error) Stanford-BinetIQ scores at age 4 and WISC Full-Scale IQ scoresat age 7 among the proband, sibling, and controlgroups1

Controls

Note.—IQ « Intelligence quotient; SEM = standard error of themean; WISC = Weschler Intelligence Scale for Children.1 Effects of age, gender, race, parental education, and social class

are controlled. Probands and siblings performed significantlymore poorly than controls at both assessments.

= 1.8, df= 4588, p < 0.04) and WISC Performance IQ (con-trols vs. patients: t = 3.2, df = 4613, p < 0.001; controls vs.siblings: t = 2.2, <% = 4588, p < 0.02) at the age 7 assess-ment When the patients and siblings were compared witheach other (using the matched-pair t statistic), there were notsignificant differences between pairs of discordant siblingsin terms of Stanford-Binet IQ score at age 4 (mean ± SEMpatient minus sibling difference = -0.40 + 4.91, t = -0.08, df= 20,p = 0.93) or WISC Full-Scale (-2.58 ± 1.98, t = -1.3,df= 29, p = 0.20), Verbal (-1.37 ± 1.83, t = -0.75, df= 29, p= 0.46), and Performance IQ scores (-3.32 ± 2.69, t = -1.23,# = 29, p = 0.23) at age 7.

On all of the subsequent analyses, the results forWISC Verbal and Performance IQ paralleled thoseobtained for Full-Scale IQ. Thus, only the results for Full-Scale IQ are reported.

Comparisons Across Quartiles of IQ Distributions.Tables 1 and 2 give the distributions of subjects by quar-tiles of the IQ distributions at age 4 and age 7 accordingto diagnostic outcome, hi the logistic regression analy-sis—after controlling for age at examination, gender, race,parental education, and socioeconomic status—Stanford-Binet IQ category at age 4 was a significant predictor ofoutcome (x2 = 10.8, df= 2,p = 0.005), achieving a signif-icant negative association with schizophrenia (x2 = i.l,df= 1, p - 0.004, OR [odds ratio] = 0.66, 95% CI [confi-dence interval] = 0.50-0.88) and marginal negative asso-ciation with sibling status (x2 = 2.7, df=\,p = 0.09, OR= 0.72, 95% CI = 0.49-1.06). The odds of schizophreniaincreased by 1.52 times (1/0.66) per unit decrease in abil-ity level, such that individuals scoring in the lowest quar-tile of functioning were 5.34 times more likely to be diag-nosed with schizophrenia than those in the highestquartile. As shown in figure 2, this effect was approxi-mately linear across the four ability levels for schizophre-nia but was not linear for the unaffected siblings. Whileunaffected siblings were overrepresented in the lowestquartile of functioning, they were approximately equallyrepresented among the three higher levels of functioning.

WISC Full-Scale IQ category at age 7 was also a sig-nificant predictor of outcome (x2 = 11.5, df = 2, p =0.003), achieving a significant negative association withschizophrenia (x2 = 6.2, df= \,p = 0.01, OR = 0.73, 95%Q = 0.57-0.93) and with sibling status (x2 = 5.4, df=\,p= 0.02, OR = 0.66, 95% CI = 0.47-0.94). The odds ofschizophrenia increased by 1.37 times (1/0.73) per unitdecrease in ability level, such that individuals in the low-est quartile of functioning were 3.5 times more likely tobe diagnosed with schizophrenia than those in the highest

385



ownloaded from



quartile. The odds of being an unaffected sibling of aschizophrenia subject increased by 1.52 times per unitdecrease in ability level, such that individuals scoring inthe lowest quartile of functioning were 5.34 times morelikely to be an unaffected sibling of a schizophrenia sub-ject than those in the highest quartile. As shown in figure3, the relationship between IQ at age 7 and adult psychi-atric outcome was approximately linear across the fourperformance levels for both schizophrenia and unaffectedsibling status.

Figure 2. Rates (± standard errors) of adultschizophrenia (A) and of being an unaffectedsibling of a schizophrenia subject (B) byStanford-Blnet IQ category at age 41

Change in Percentile Rankings Across Assessments. Asshown in table 4, there was not a significant effect of assess-ment wave (i.e., no change between age 4 and age 7) onmean IQ percentile rankings, nor an interaction of assess-ment wave with diagnostic outcome. In fact, none of thediagnostic groups showed significant change in percentilerankings across assessments. There was also not a significantdifference in the proportions of each group who improved,stayed the same, or declined in terms of percentile rankingsacross the two assessments (x2 = 2.4, df = 4,p = 0.66).

Figure 3. Rates (± standard errors) of adultschizophrenia (A) and of being an unaffectedsibling of a schizophrenia subject (B) by WISCFull-Scale IQ category at age 71

2.0-

1.0-

O 8 -

0.01-26 26-49 50-74 76-99

Laval of Performance (PercentJie Range)

3.0

IS

2.0

' 1.8-

1.0"

O8-

0.01-26 26-40 60-74 78-09

Level of Performance (Percentile Range)

1.0-

oa-

0.0

I1 Hi1-28 26-49 60-74 75-99


Note.—IQ => intelligence quotient.1 In each graph, the line indicated by circles gives the observedprevalence of the outcome at each level of performance, and theline indicated by triangles gives the prevalence expected underthe null hypothesis of no association between level of perfor-mance and the outcome.

u

1.0-

aO'

3f

1-28 23-49 50-74 78-99


Note.—IQ - intelligence quotient; WISC - Weschler IntelligenceScale for Children.1 See note to figure 2.

386



ownloaded from



Obstetric Influences. In the analyses of the obstetricvariables in relation to IQ scores at the two childhoodassessments, there were not significant effects of thehypoxia OC scale (age 4: F = 0.3, df= 1, 4553, p = 0.61;age 7: F = 0.1, df= 1, 4951, p = 0.73) or the prenatal OCscale (age 4: F = 0.7, df = 1, 4553, p = 0.39; age 7: F =0.3, df= 1, 4951, p = 0.59), nor were there significantinteractions of diagnosis with hypoxia-associated OCs(age 4: F = 1.4, df= 2,4553, p = 0.25; age 7: F = 0.2, df=2,4951, p = 0.81) or with prenatal OCs (age 4: F = 0.6, df= 2, 4553, p = 0.54; age 7: F = 0.1, df = 2, 4951, p =0.95). Birth weight had a marginal association with IQ atage 4 (F = 2.5, df= 1, 4553, p = 0.12) and a significantassociation with IQ at age 7 (F = 6.7, df = 1, 4951, p =0.004), and in both cases birth weight interacted signifi-cantly with diagnosis (age 4: F = 6.2, df = 2, 4553, p =0.002; age 7: F = 5.1, df = 2,4951, p = 0.006). In post hocanalyses, birth weight was found to correlate significantlypositively with IQ at both age 4 (r = 0.14, df = 4488, p =0.0001) and age 7 (r = 0.14, df = 4866, p = 0.0001)among the controls but was not significantly related to IQamong the schizophrenia subjects (age 4: r = -0.13, df =50, p = 0.36; age 7: r = 0.07, df= 62, p = 0.59) or unaf-fected siblings (age 4: r = 0.14, df= 25, p = 0.52; age 7: r= 0.19, df = 28, p = 0.33).

Potential Moderating Variables. In the analyses ofhandedness, behavior disturbance, and gender as potentialmoderators of the effects of diagnosis in relation to child-

hood IQ, there were not significant interactions of diagno-sis with handedness (age 4: F = 1.2, df = 2, 4556, p =0.31; age 7: F = 0.1, df = 2, 4949, p = 0.90), behavior dis-turbance (age 4: F = 2.3, df= 2,4556, p = 0.10; age 7: F =1.7, df = 2,4949, p = 0.19), or gender (age 4: F = 1.2, df=2, 4556, p = 0.29; age 7: F = 0.4, df=2, 4949, p = 0.68).There were also not significant differences in IQ betweenschizophrenia cases with early and later ages at first treat-ment contact (age 4: t = 0.45, df =44, p = 0.65; age 7: t =-0.57, df= 55, p = 0.57).

Finally, to determine whether the childhood intellectualdeficits in the sibling group could be explained by the pres-ence of non-schizophrenia-related psychiatric morbidity inadulthood, the sibling-control contrasts were repeated afterexcluding siblings with any psychiatric diagnoses. Whenthis was done, there continued to be significant differencesbetween the nonpsychiatric controls and the unaffected sib-lings of schizophrenia patients in terms of Stanford-Binet IQat age 4 (t = 2.07, df= 4207, p = 0.02) and WISC Full-ScaleIQ at age 7 (r = 2.08, df = 4588, p = 0.03).

Discussion

Childhood Cognitive Dysfunction in SchizophreniaPatients and Their Siblings. Our study demonstrates thatcognitive dysfunction is evident in preschizophrenia sub-jects as early as age 4 and thus precedes onset of formaldiagnostic symptoms and signs by 10-15 years.

Table 4. Results of repeated measures analysis of percentlle ranks for Stanford-Blnet IQ at age 4and WISC Full-Scale IQ at age 7

Time

Diagnostic outcome

Time x diagnostic outcome

Error

Diagnostic outcome

Schizophrenia

Unaffected siblings

No psychiatric diagnosis

Source of variation

n44

24

4,068

df

Percentlle Ranks,Mean (± SD)

Age 4

34.9 (26.8)

37.6 (33.4)

50.3 (28.6)

ANOVAMS

Age 7

35.0 (27.9)

33.2(25.1)

50.4 (28.7)

ResultsF

f

-0.03

0.86

-0.34

P

P0.98

0.39

0.73

1

2

2

4133

421.315544.1

384.1

1239.1

0.34

12.5

0.31

0.56

0.00010.73

Note.—ANOVA - analysis of variance; IQ - intelligence quotient; SD - standard deviation; WISC - Weschler Intelligence Scale forr^hiHrnnChildren

387



ownloaded from



Moreover, the distribution of functioning is unimodallyshifted downward in these cases, such that preschizophre-nia subjects are underrepresented at the highest levels ofability and are progressively more overrepresented acrossdecreasing ability levels at both 4 and 7 years of age. Thispattern suggests that most preschizophrenia subjects failto reach their expected level of cognitive attainment dur-ing childhood. If such deficits characterized only a sub-group of the schizophrenia population, one would expecta large increase in risk for the disorder among individualsin the lowest level of functioning but a flat distribution ofrisk across the higher ability levels. Thus, to the extentthat cognitive dysfunction during childhood reflects dis-turbances of brain development, our results replicate andextend those of Jones et al. (1994) and David et al. (1997)in suggesting that such neurodevelopmental compromiseis more broadly characteristic of this syndrome than haspreviously been supposed (Lewis and Murray 1987).

Consistent with the results of numerous prior studiesof children at elevated genetic risk for schizophrenia(Mednick and Schulsinger 1968; Landau et al. 1972;Asamow et al. 1978; Rutschmann et al. 1980; Harvey etal. 1981; Winters et al. 1981; Worland et al. 1982;Driscoll 1984; Lifshitz et al. 1985; Sohlberg 1985; Hallettet al. 1986; Fish 1987; Goodman 1987; Sameroff et al.1987; Weintraub 1987; Erlenmeyer-Kimling et al. 1989;Schreiber et al. 1992; Marcus et al. 1993; Bergman andWalker 1995), in this study cognitive deficits were alsoobserved among the unaffected siblings of preschizophre-nia patients at 4 and 7 years of age, regardless of whethersiblings with adult psychiatric disorders were included.Given that a substantial number of the first degree rela-tives of schizophrenia patients carry a predisposing geno-type without manifesting the disorder phenotypically(Fisher 1973; Gottesman and Bertelsen 1989), this patternis consistent with an association between presence of agenetic diathesis to schizophrenia and expression of cog-nitive dysfunction during childhood. Because theyoungest siblings in this study were 30 years old at thetime of the most recent psychiatric register informationavailable, and because onset of schizophrenia occurs afterage 30 in approximately 10-20 percent of cases, somemembers of the sibling group may in fact be diagnosedwith schizophrenia in their lifetimes. However, "futurecaseness" is not likely to account completely for thechildhood intellectual deficits observed in the siblinggroup. Only about 5 percent of siblings of a proband withschizophrenia are expected to develop the condition them-selves, and the majority of these will have their onsetsbefore age 30. Even if we assume that 2 of the 33 subjectsin the sibling group (i.e., approximately 5%) tested at theage 7 assessment will be diagnosed with schizophreniaafter age 30 and that both came from the lowest quartile

of functioning, siblings of schizophrenia patients wouldstill be overrepresented in the lowest quartile of function-ing (i.e., 37% versus an expected value of 25%) andunderrepresented in the highest quartile of functioning(i.e., 13% versus an expected value of 25%).

There are numerous contributors to variability in. intelligence test scores besides predisposition to schizo-phrenia (Sattler 1982). In this study we evaluated the rela-tionships between childhood cognitive functioning andadult psychiatric outcome controlling for the knownmajor sources of individual differences on such tasks,including age, gender, race, parental education, and socialclass. That the associations were present even after con-trolling these factors provides one indication of the inter-nal validity of these findings. At the same time, it isimportant to emphasize that schizophrenia patients andtheir unaffected siblings do not account for all cases ofcognitive dysfunction during childhood, even when con-sidered collectively. In view of this pattern, general intel-lectual measures cannot be used as a screening device forfuture schizophrenia or genetic predisposition to schizo-phrenia in general population samples. Possibly, morerefined measures of deficits, in particular neural systems,could be used in this manner, but further study will beneeded to make this determination.

Course of Cognitive Development in PreschizophreniaPatients. In this study preschizophrenia subjects did notdiffer from their unaffected siblings at the age 4 and age 7assessment points in terms of IQ scores. By early adulthood,nearly all schizophrenia patients, even those without expo-sure to neuroleptics, perform more poorly on neuropsycho-logical tests than their own unaffected siblings and cotwins(Goldberg et al. 1990; Cannon et al. 1994). It is thus possi-ble that preschizophrenia cases increasingly diverge fromthe level of functioning expected in their own families withage. However, it is not clear if this increasing divergencereflects a factor intrinsic to the causes of the illness, theeffects of acute psychosis, or an artifact of differences in thepsychometric properties of cognitive tests in adults com-pared with children.

If there is increasing divergence from within-family(or population) norms among preschizophrenia patientswith age, such divergence does not necessarily implyintra-individual decline. Our ability to detect intra-indi-vidual decline is more limited in children than in adultsbecause cognitive abilities in the former undergo bothqualitative and quantitative modification with age.Normal cognitive development appears to proceed byadding new skills and knowledge on top of those previ-ously acquired; the same process is evident in individualswith currently classified neurodevelopmental conditionsbut at a lower rate—such that by the time most cognitive

388



ownloaded from



functions have crystallized in young adulthood, affectedindividuals are maximally deviant compared with theirage-matched peers (Sattler 1982). In this study, percentilerankings within the study population did not change sig-nificantly within subjects across the two assessments inany of the groups, nor did the groups differ in the rate ofintra-individual change. Thus, if the rate of cognitivedevelopment is lower than normal in preschizophreniasubjects, it will be necessary to employ assessment pointsand/or tests different from those used in this study todemonstrate such an effect.

Role of Obstetric Complications. Numerous epidemio-logical studies have demonstrated that OCs are associatedwith increased liability to schizophrenia and with an ear-lier age of its clinical onset (Cannon 1997). The extent towhich OCs may contribute to the neurobehavioral fea-tures and precursors of schizophrenia has received com-paratively little attention. In this study, we examined therelationship between a history of hypoxia-associated OCs,prenatal OCs, and birth weight to measures of cognitivefunctioning at 4 and 7 years of age. While preschizophre-nia children and their unaffected siblings had significantlylower IQ scores than normal control children at bothassessment points, these deficits were not significantlyassociated with any of the obstetric influences. Further,the cognitive test scores of children not yet diagnosedwith schizophrenia were not significantly more depressedthan those of their unaffected siblings, who also per-formed significantly worse than controls at both assess-ment points. This pattern suggests that the OCs examinedin this study do not contribute to lower IQ scores amongpreschizophrenia children in early and middle childhood,but rather that cognitive impairment at these ages reflectsthe expression of schizophrenia susceptibility genes(and/or shared environmental influences not accounted forby parental education or social class).

Limitations. The procedures used to ascertain and diag-nose cases in this study were not ideal. We address themajor threats to validity and ability to generalize posed bythese methods below.

Are there biases related to the assignment of diag-noses by chart review? We used a two-stage diagnosticprocedure whereby we first screened the cohort for anypsychotic disorder diagnosis via treatment sources andsubsequently performed detailed evaluations of the psy-chiatric medical records for nearly half of the psychotic-disordered probands for the determination of diagnosesaccording to DSM-IV criteria. Only those cases withchart-review-based DSM-IV diagnoses of schizophreniaor schizoaffective disorder were classified as "cases" inthe analyses; the remaining psychotic-disordered

probands were excluded. The chart-review diagnoses arereliable; independent evaluations of a random sample ofmedical records by different reviewers produced a highrate of diagnostic agreement (k - 0.85, 93% simple agree-ment). Diagnostic reliability is a necessary but not suffi-cient condition for diagnostic validity. Optimally, thechart-review diagnoses should be validated against thoseobtained using direct structured psychiatric interviews,but we have not yet interviewed a sufficient number ofcases from this cohort to make this determination. It isimportant to note that invalid assignment of schizophreniaor schizoaffective diagnoses by chart review would pro-duce a false positive result only if the cognitive dysfunc-tion-outcome association were restricted to cases whoseactual diagnoses were not schizophrenia or schizoaffec-tive disorder. The diagnoses most likely to compete withthose assigned are psychotic forms of bipolar or unipolaraffective illness. In secondary analyses conducted toaddress this question, the 41 psychotic-disordered caseswho were screened out of the proband sample becausethey received DSM-IV diagnoses of bipolar or unipolaraffective disorder by chart review did not differ signifi-cantly from nonpsychiatric controls in terms of Stanford-Binet IQ at age 4 (F = 2.8, df = 1, 4253, p = 0.10) orWISC Verbal IQ at age 7 (F = 1.5, df = 1, 4617, p = 0.22)but did differ significantly from controls on WISCPerformance IQ at age 7 (F = 9.7, df = 1, 4617, p =0.002). The unaffected siblings of these cases did not dif-fer from nonpsychiatric controls on any measure(Stanford-Binet IQ: F = 1.1, df= 1, 4253, p = 0.29; WISCVerbal IQ: F = 0.5, df = 1, 4617, p = 0.48; WISCPerformance IQ: F = 0.1, df=l, 4617, p = 0.75). Giventhis pattern, if there are misdiagnosed cases of affectivepsychosis in the DSM-IV schizophrenia/schizoaffectivegroup, the effect on the results should be an underestima-tion rather than overestimation of the associationsbetween childhood cognitive dysfunction and schizophre-nia (and genetic risk for schizophrenia).

Are there biases related to incomplete ascertain-ment of schizophrenia cases? Direct followup of all sur-viving cohort members in adulthood would provide themost complete ascertainment of outcome, but such anapproach is possible (albeit infeasible because of expense)only in a country that maintains a central register indexingcurrent residence of all members of its population. A rea-sonable alternative—and the approach taken in thisstudy—is to screen the birth cohort for a history of localpsychiatric service utilization. This approach missed caseswho were deceased at the time of followup, who did notcome to treatment, or who because of emigration orchanges in social class used psychiatric facilities otherthan those whose patient rolls were screened. We also lostcases who were identified as having a history of treatment

389



ownloaded from



for a psychotic disorder but who had been treated at facili-ties that were no longer operating, did not cooperate withthe chart reviews, or had treated only one proband.Incomplete ascertainment of schizophrenia in the cohortcould produce biasing effects on the results in two majorways. First, if the identified cases of schizophrenia do notresemble the unidentified cases in terms of demographicfactors and premorbid cognitive functioning, selectioneffects could limit the ability to generalize the findings.Second, a false positive association could result if a sub-stantial number of unidentified cases of schizophrenia areincluded in the sample classified as nonpsychiatric con-trols and such cases had substantially better cognitivefunctioning than the cases who were ascertained.

In secondary analyses conducted to address the firstissue, the 72 cases with chart-review-based DSM-IV diag-noses of schizophrenia or schizoaffective disorder werefound not to differ significantly from the remaining 121probands with register diagnoses of schizophrenia orschizoaffective disorder (whose charts were not reviewed)in terms of gender, year of birth, birth order, mother's ageat birth, years of parental education, socioeconomic sta-tus, number of unaffected siblings enrolled in the NCPPstudy, or history of OCs (see Cannon et al., this issue).The groups also did not differ significantly from eachother on Stanford-Binet IQ at age 4 (t = 1.3, df = 121, p =0.21) or WISC Full-Scale IQ at age 7 (t = 1.5, df= 145, p= 0.11). Finally, when the primary analysis was repeatedusing the 194 cases with register diagnoses of schizophre-nia or schizoaffective disorder and their siblings, the con-trols continued to have significantly higher tests scores atboth assessments than the probands (age 4: t = 3.94, df =5214, p = 0.0001; age 7: t = 3.89, df = 5927, p = 0.0001)and their siblings (age 4: t = 2.82, af=5H3,p = 0.003;age 7: t = 1.81, df= 5895, p = 0.03). There thus appears tobe no significant biasing effect related to which of theavailable probands were ascertained for chart review.

In regard to the second issue, it is virtually certain thatsome cases of schizophrenia were classified as nonpsychi-atric controls in the analyses (i.e., female patients whoselast names changed at marriage, patients who moved out ofPhiladelphia before 1985, patients who did not seek treat-ment, or patients whose treatment was exclusively via pri-vate practitioners). The extent of this misclassification can-not be quantified precisely. The prevalence of atreatment-ascertained psychotic disorder diagnosis in thiscohort is quite high (i.e., 3.7%), without even taking intoaccount the possibility of unascertained cases. Nevertheless,while acknowledging that some cases of schizophreniawere likely classified as controls, there is no reason to sus-pect that the premorbid functioning of the unascertainedschizophrenia subjects in the control group is different fromthat of the ascertained cases. Given the higher rate of schiz-

ophrenia among males in this cohort, an underascertainmentof female schizophrenia subjects represents the most plausi-ble candidate for such a biasing effect (but see Hambrechtet al. 1993). However, while males had lower IQ scoresthan females overall, gender did not interact with diagnosis(i.e., the cognitive deficits in schizophrenia were not differ-ential according to gender). It could be argued that morecognitively disabled members of the cohort were morelikely to stay in the vicinity where they were bom, in whichcase our ascertainment methods would have oversampledthe schizophrenia cases from the lower levels of function-ing. To address this possibility, we compared the schizo-phrenia cases to the sample of individuals who were identi-fied as having nonpsychotic forms of mental illness in thePhiladelphia area treatment data base. Any tendency for themore cognitively disabled subjects to remain in the loca-tions of their births (and/or for more cognitively intact sub-jects to have moved out of the area) should have affectedour ascertainment of these nonpsychotic cases as well. If abias toward ascertainment of cognitively disabled subjectsexists, the schizophrenia cases should not differ from thenonpsychotic cases ascertained in the register. However, theschizophrenia patients did in fact perform more poorly thanthe nonpsychotic register cases at both age 4 (t = 2.06, af=726, p = 0.02) and age 7 (t = 2.07, df= 848, p = 0.02).Moreover, when the primary analyses were repeated includ-ing the nonpsychotic register cases with the nonpsychiatriccohort members as controls, the controls continued to havehigher IQ scores than the patients with schizophrenia (age4: t = 3.87, df= 4611, p = 0.0001; age 7: t = 331, df= 5068,p = 0.0004) and their siblings (age 4: t = 2.01, df= 4655, p= 0.02; age 7: t = 2.13, ay= 5043, p = 0.02). Finally, it ispossible that cases with an earlier clinical onset of schizo-phrenia were more likely to be ascertained than cases with alater onset, in which case a false positive bias would be cre-ated if the early onset cases had lower IQs than later onsetcases. However, there were no differences in cognitivefunctioning at age 4 or 7 years as a function of age at firsttreatment contact among cases that were ascertained (seeResults), suggesting that incomplete ascertainment of later-onset cases is unlikely to have a significant biasing effect onthe results.

It is also important to emphasize that African-American subjects in this cohort were substantially over-represented (i.e., 88%). Unfortunately, the number ofCaucasian subjects with schizophrenia was too small topermit us to test for an interaction between race and diag-nosis in relation to childhood intellectual functioning. Arace-specific effect in schizophrenia patients and theirunaffected relatives appears unlikely, however, given thatprior studies of preschizophrenia subjects and at risk pop-ulations have reported similar findings in predominantlyCaucasian samples (David et al. 1997; Jones et al. 1994).

390



ownloaded from



References

Adamson, S.J.; Alessandri, L.M.; Badawi, N.; Burton,P.R.; Pemberton, PJ.; and Stanley, F. Predictors of neona-tal encephalopathy in full-term infants. British MedicalJournal, 311:598-602, 1995.

Akbarian, S.; Vinuela, A.; Kim, J.J.; Potkin, S.; Bunney,W.E.; and Jones, E.G. Distorted distribution of nicoti-namide-adenine dinucleotide phosphate-diaphorase neu-rons in temporal lobe of schizophrenics implies anom-alous cortical development. Archives of GeneralPsychiatry, 50:178-187,1993.

American Psychiatric Association. DSM—FV: Diagnosticand Statistical Manual of Mental Disorders. 4th ed.Washington, DC: APA, 1994.

Arabin, B.; Snyjders, R.; Mohnhaupt, A.; Ragosch, V.;and Nicolaides, K. Evaluation of the fetal assessmentscore in pregnancies at risk for intrauterine hypoxia.American Journal of Obstetrics and Gynecology,169:549-554,1993.

Arnold, S.E.; Hyman, B.T.; Van Hoesen, G.W.; andDamasio, A.R. Some cytoarchitectural abnormalities ofthe entorhinal cortex in schizophrenia. Archives ofGeneral Psychiatry, 48:625-632, 1991.

Asarnow, J.R.; Steffy, R.A.; MacCrimmon, D.J.; andCleghorn, J.M. An attentional assessment of foster chil-dren at risk for schizophrenia. In: Wynne, L.C.,Cromwell, R.L., and Matthysse, S., eds. The Nature ofSchizophrenia: New Approaches to Research andTreatment. New York, NY: John Wiley & Sons, 1978. pp.339-358.

Aylward, E.; Walker, E.; and Bettes, B. Intelligence inschizophrenia: Meta-analysis of the research.Schizophrenia Bulletin, 10(3):430-459,1984.

Bergman, A.J., and Walker, E. The relationship betweencognitive functions and behavioral deviance in children atrisk for psychopathology. Journal of Child Psychologyand Psychiatry, 36:265-278,1995.

Cannon, T.D. On the nature and mechanisms of obstetricinfluences in schizophrenia: A review and synthesis.International Review of Psychiatry, 9:387-397, 1997.

Cannon, T.D.; Rosso, I.M.; Hollister, J.M.; Bearden, C.E.;Sanchez, L.E.; and Hadley, T. A prospective cohort studyof genetic and perinatal influences in the etiology ofschizophrenia. Schizophrenia Bulletin, 26(2):351-366,2000.

Cannon, T.D.; Mednick, S.; Parnas, J.; Schulsinger, F.;Praestholm, J.; and Vestergaard, A. Developmental brainabnormalities in the offspring of schizophrenic mothers: I.Contributions of genetic and perinatal factors. Archives ofGeneral Psychiatry, 50:551-564,1993.

Cannon, T.D., and Mednick, S.A. Fetal neural develop-ment and adult schizophrenia: An elaboration of the para-digm. In: Mednick, S.A., Cannon, T.D., Barr, C.E., andLyon, M., eds. Fetal Neural Development and AdultSchizophrenia. New York, NY: Cambridge UniversityPress, 1991. pp. 227-237.

Cannon, T.D.; Zorrilla, L.E.; Shtasel, D.; Gur, R.E.; Gur,R.C.; Marco, E.J.; Moberg, P.; and Price, R.A.Neuropsychological functioning in siblings discordant forschizophrenia and healthy volunteers. Archives of GeneralPsychiatry, 51:651-661, 1994.

David, A.S.; Malmberg, A.; Brandt, L.; Allebeck, P.; andLewis, G. IQ and risk for schizophrenia: A population-basedcohort study. Psychological Medicine, 27:1311-1323, 1997.

Driscoll, R.M. Intentional and incidental learning in chil-dren vulnerable to psychopathology. In: Watt, N.F.,Anthony, E J., Wynne, L.C., and Rolf, J.E., eds. Children atRisk for Schizophrenia: A Longitudinal Perspective. NewYork, NY: Cambridge University Press, 1984. pp. 320-325.

Erlenmeyer-Kimling, L.; Golden, R.R.; and Cornblatt,B.A. A taxometric analysis of cognitive and neuromotorvariables in children at risk for schizophrenia. Journal ofAbnormal Psychology, 98:203-208,1989.

Fish, B. Infant predictors of the longitudinal course ofschizophrenic development. Schizophrenia Bulletin,13(3):395-4O9,1987.

Fish, B.; Marcus, J.; Hans, S.L.; Auerbach, J.G.; andPerdue, S. Infants at risk for schizophrenia: Sequelae of agenetic neurointegrative defect: A review and replicationanalysis of pandysmaturation in the Jerusalem InfantDevelopment Study. Archives of General Psychiatry,49:221-235, 1992.

Fisher, M. Genetic and environmental factors in schizo-phrenia: A study of schizophrenic twins and their families.Acta Psychiatrica Scandinavica, 238(Suppl):9-142, 1973.

Foerster, A.; Lewis, S.W.; Owen, M.J.; and Murray, R.M.Low birth weight and a family history of schizophreniapredict poor premorbid functioning in psychosis.Schizophrenia Research, 5:13-20, 1991.

Goldberg, T.E.; Ragland, J.D.; Torrey, E.F.; Gold, J.M.;Bigelow, L.B.; and Weinberger, D.R. Neuropsychologicalassessment of monozygotic twins discordant for schizophre-nia. Archives of General Psychiatry, 47:1066-1072,1990.

Goodman, S.H. Emory University project on children ofdisturbed parents. Schizophrenia Bulletin, 13(3):411-423,1987.

Gottesman, I.I., and Bertelsen, A. Confirming unex-pressed genotypes for schizophrenia: Risks in the off-spring of Fischer's Danish identical and fraternal discor-dant twins. Archives of General Psychiatry, 46:867-872,1989.

391



ownloaded from



Hallett, S.; Quinn, D.; and Hewitt, J. Defective interhemi-spheric integration and anomalous language lateralizationin children at risk for schizophrenia. Journal of Nervousand Mental Disease, 174:418-427, 1986.

Hambrecht, M.; Maurer, K.; and Hafner, H. Evidence fora gender bias in epidemiological studies of schizophrenia.Schizophrenia Research, 8:223-231, 1993.

Harvey, P.; Winters, K.; Weintraub, S.; and Neale, J.M.Distractibility in children vulnerable to psychopathology.Journal of Abnormal Psychology, 90:298-304, 1981.

Jakob, J., and Beckmann, H. Prenatal developmental dis-turbances in the limbic allocortex in schizophrenics.Journal of Neural Transmission, 65:303-326,1986.

Jones, P.; Rodgers, B.; Murray, R.; and Marmot, M. Childdevelopmental risk factors for adult schizophrenia in theBritish 1946 birth cohort. Lancet, 344:1398-1402, 1994.

Jones, P.B., and Done, D.J. From birth to onset: A devel-opmental perspective of schizophrenia in two nationalbirth cohorts. In: Keshavan, M.S., and Murray, R.M., eds.Neurodevelopment and Adult Psychopathology.Cambridge, U.K.: Cambridge University Press, 1997. pp.119-136

Kovelman, J.A., and Scheibel, A.B. A neurohistologicalcorrelate of schizophrenia. Biological Psychiatry,19:1601-1621, 1984.

Kremen, W.S.; Buka, S.L.; Seidman, L.J.; Goldstein,J.M.; Koren, D.; and Tsuang, M.T. IQ decline duringchildhood and adult psychotic symptoms in a communitysample: A 19-year longitudinal study. American Journalof Psychiatry, 155:672-677, 1998.

Landau, R.; Harth, P.; Othnay, N.; and Scharfhertz, C. Theinfluence of psychotic parents on their children's develop-ment. American Journal of Psychiatry, 129:70-75, 1972.

Lewis, S.W., and Murray, R.M. Obstetric complications,neurodevelopmental deviance, and risk of schizophrenia.Journal of Psychiatric Research, 21:413-421,1987.

Lifshitz, M.; Kugelmass, S.; and Karov, M. Perceptual-motor and memory performance of high-risk children.Schizophrenia Bulletin, ll(l):74-84, 1985.

Low, J.A.; Simpson, L.L.; and Ramsey, D.A. The clinicaldiagnosis of asphyxia responsible for brain damage in thehuman fetus. American Journal of Obstetrics andGynecology, 167:11-15, 1992.

Low, J.A.; Simpson, L.L.; Tonni, G.; and Chamberlain, S.Limitations in the clinical prediction of intrapartum fetalasphyxia. American Journal of Obstetrics andGynecology, 172:801-804, 1995.

Maier, R.F.; Gunther, A.; Vogel, M.; Dudenhausen, J.W.;and Obladen, M. Umbilical venous erythropoietin andumbilical arterial pH in relation to morphologic placental

abnormalities. Obstetrics and Gynecology, 84:81-87,1994.

Marcus, J.; Hans, S.L.; Auerbach, J.G.; and Auerbach,A.G. Children at risk for schizophrenia: The JerusalemInfant Development Study: n. Neurobehavioral deficits atschool age. Archives of General Psychiatry, 50:797-809,1993.

Mednick, S.A., and Schulsinger, F. Some premorbid char-acteristics relating to breakdown in children with schizo-phrenic mothers. In: Rosenthal, D., and Kety, S.S., eds.The Transmission of Schizophrenia. Oxford, U.K.:Pergamon Press, 1968. pp. 267-291.

Murray, R.M., and Lewis, S.W. Is schizophrenia a neu-rodevelopmental disorder? British Medical Journal,296:63, 1988.

Niswader, K.R., and Gordon, M. The CollaborativePerinatal Study of the National Institute of NeurologicalDiseases and Stroke: The Women and Their Pregnancies.Philadelphia, PA: W.B. Saunders, 1972.

O'Callaghan, E.; Gibson, T.; Colohan, H.A.; Buckley, P.;Walshe, D.G.; Larkin, C ; and Waddington, J.L. Risk ofschizophrenia in adults bom after obstetric complicationsand their association with early onset of illness: A controlledstudy. British Medical Journal, 305:1256-1259,1992.

Rieder, R.O.; Broman, S.H.; and Rosenthal, D. The off-spring of schizophrenics: II. Perinatal factors and IQ.Archives of General Psychiatry, 34:789-799,1977.

Rothbard, A.B.; Schinnar, A.P.; Hadley, T.R.; and Rovi,J.I. Integration of mental health data on hospital and com-munity services. Administration and Policy in MentalHealth, 18:91-99, 1990.

Rutschmann, J.; Cornblatt, B.A.; and Erlenmeyer-Kimling, L. Auditory recognition memory in adolescentsat risk for schizophrenia: Report on a verbal continuousrecognition task. Psychiatry Research, 3:151-161, 1980.

Salafia, C M . ; Minior, V.K.; Lopez-Zeno, J.A.;Whittington, S.S.; Pezzullo, J.C.; and Vintzileos, A.M.Relationship between placental histologic features andumbilical cord blood gases in preterm gestations.American Journal of Obstetrics and Gynecology,173:1058-1064, 1995.

Sameroff, A.; Seifer, R.; Zax, M.; and Barocas, R. Earlyindicators of developmental risk: Rochester longitudinalstudy. Schizophrenia Bulletin, 13(3):383-394, 1987.

Sattler, J.M. Assessment of Children's Intelligence andSpecial Abilities. Boston, MA: Allyn and Bacon, 1982.

Schreiber, H.; Stolz-Born, G.; Heinrich, H.; Kornhuber,H.H.; and Bom, J. Attention, cognition, and motor preser-vation in adolescents at genetic risk for schizophrenia andcontrol subjects. Psychiatry Research, 44:125-140, 1992.

392



ownloaded from



Sohlberg, S.C. Personality and neuropsychological perfor-mance of high-risk children. Schizophrenia Bulletin,ll(l):48-60, 1985.

Stokes, M.E.; Davis, C.S.; and Koch, G.G. CategoricalData Analysis Using the SAS System. Cary, NC: SASInstitute, 1995.

Terman, L.M., and Merrill, M.A. Stanford-BinetIntelligence Scale. Boston, MA: Houghton Mifflin, 1960.

Thomdike, R.L. Stanford-Binet Intelligence Scale, Form L-M,1972 Norms Tables. Boston, MA: Houghton Mifflin, 1973.

Wechsler, D. Manual for the Wechsler Intelligence Scale forChildren. New York, NY: Psychological Corporation, 1949.

Weintraub, S. Risk factors in schizophrenia: The StonyBrook high-risk project. Schizophrenia Bulletin,13(3):439-449, 1987.

Winters, K.C.; Stone, A.A.; Weintraub, S.; and Neale,J.M. Cognitive and attentional deficits in children vulner-able to psychopathology. Journal of Abnormal ChildPsychology, 9:435-453, 1981.

Worland, J.; Weeks, D.G.; Weiner, S.M.; and Schechtman, J.Longitudinal, prospective evaluations of intelligence in chil-dren at risk. Schizophrenia Bulletin, 8(1):135-141,1982.

Acknowledgement

This research was supported by grants from the March ofDimes Birth Defects Foundation, White Plains, NY, andthe Stanley Foundation, Bethesda, MD.

The Authors

Tyrone D. Cannon, Ph.D., is Staglin Family Professor ofPsychology, Psychiatry, and Human Genetics, Departmentof Psychology, University of California, Los Angeles,CA. Carrie E. Bearden, Ph.D., is Postdoctoral Fellow inChild Psychiatry; Laura E. Sanchez, M.D., is AssistantProfessor; and Trevor Hadley, Ph.D., is Professor ofPsychology in Psychiatry and Chief of the Center forMental Health Policy Research, Department ofPsychiatry, University of Pennsylvania, Philadelphia, PA.J. Megginson Hollister, Ph.D., is Research AssistantProfessor, Department of Psychiatry, University ofUtrecht, Utrecht, The Netherlands. Isabelle M. Rosso,M.A., is Ph.D. Candidate, Department of Psychology,University of Pennsylvania.

393



ownloaded from




ownloaded from


Childhood Cognitive Functioning in Schizophrenia Patients and Their Unaffected Siblings: A...

Documents

Transcript of Childhood Cognitive Functioning in Schizophrenia Patients and Their Unaffected Siblings: A...