Genetics of autism: Characteristics of affected and unaffected children from 37 multiplex families

American Journal of Medical Genetics (Neuropsychiatric Genetics) 54:27-35 (1994)

Genetics of Autism: Characteristics of Affected and Unaffected Children From 37 Multiplex Families

Donna Spiker, Linda Lotspeich, Helena C. Kraemer, Joachim Hallmayer, William McMahon, P. Brent Petersen, Peter Nicholas, Carmen Pingree, Susan Wiese-Slater, Carla Chiotti, Dona Lee Wong, Susan Dimicelli, Edward Ritvo, Luigi L. Cavalli-Sforza, and Roland D. Ciaranello Autism Genetics Program (D.S., L.L., H.C.K., J R . , S.W.-S., C.C., D.L.W., S.D., L.L.C.S., R.D.C.), Nancy Pritzker Laboratory of Developmental and Molecular Neurobiology, Department of Psychiatry and Behavioral Sciences (D.L.W., R .D.C.), and Department of Genetics (J.H., L.L.C.-S.), Stanford University School of Medicine, Stanford Department of Psychiatry (E.R.), UCLA School of Medicine, Los Angeles, California; and Department of Psychiatry (W.M., P.B.P., P N . , C.P.), University of Utah School of Medicine, Salt Lake City, Utah

Evidence from twin and family studies strongly suggests that genetic factors play a prominent role in the etiology of some cases of infantile autism. Genetic factors would be expected to be especially strong in families with multiple autistic members (multiplex families). This report describes the identification and evaluation of 44 families with two or more autistic children collected as part of a genetic linkage study in autism. Families were referred with a presumptive classification of multiplex autism. Children referred as autistic, as well as their presumptively normal siblings, were assessed using the Autism Diagnostic Interview (ADI) and the Autism Diagnostic Observation Scale (ADOS). Thirty-seven of the 44 families (87%) had at least two children who met diagnostic criteria for autism on the ADI. Of the total group of 117 children evaluated in those families, 83 (71%) met all AD1 criteria and could be unambiguously classified as autistic (affected), 26 (22%) met none of the AD1 criteria and were classified as not autistic (unaffected), and 8 (7%) were classified as uncertain because they met one or more but not all of the AD1 cutpoints. Autistic siblings were not significantly concordant for most autism characteristics, for IQ, or for verbal ability. Significant concordances were found, however, for behaviors related to rituals and repetitive play, and for social impairments in the expression and understanding of facial

Received for publication July 12,1993; revision received October 5, 1993.

Address reprint requests to Donna Spiker, Ph.D., Autism Ge- netics Program, Department of Psychiatry and Behavioral Sci- ences, Stanford University School of Medicine, Stanford, CA 94305.

0 1994 Wiley-Liss, Inc.

expressions of emotion. The results of this study suggests two major points: first, there does not appear to be a highly variable autistic phenotype expressed in multiplex families; in the vast majority of cases, children are either clearly affected or clearly unaffected, so that linkage analysis should not be compli- cated by a large number of ambiguous or uncertain cases. Second, multiplex families do not appear to associate into subgroups defined by clustering of specific autism behaviors. 0 1994 Wiley-Liss, Inc.

KEY WORDS: multiplex autism, genetic heterogeneity, etiology

INTRODUCTION Accumulating evidence now strongly implicates ge-

netic factors in the etiology of autism. Recently published reviews have summarized the empirical data leading to this conclusion. This includes data from twin and family studies, as well as data showing associations between autism and other known genetic disorders [Lot- speich and Ciaranello, 1993; Folstein and Piven, 1991; Smalley, 1991; Bolton and Rutter, 1990; Rutter, 1990; Rutter et al., 1990; Smalley et al., 19881. The concordance rate reported for autism in monozygotic twins ranges from 36 to nearly loo%, while the concordance rate reported for dizygotic twins ranges from 0 to 24%; pooled estimates for these figures are 64% and 9%, respectively [Smalley et al., 19881. The sibling recurrence risk rate pooled across several studies in 2.7% [Smalley et al., 19883. Based on a prevalence rate for autism of4 in 10,000 births, the recurrence risk in siblings is greater than 50 times the population risk, and may approach 100, making a linkage strategy aimed at finding a gene or genes in this disorder feasible [Risch, 19901.

While there is a growing consensus about the genetic etiology of some cases of autism, there is also widespread

28 Spiker et al.

agreement that there are many importance obstacles to locating an autism gene or genes. Currently, linkage analysis using polymorphic DNA markers is the best available and most commonly used strategy for finding disease genes when the biology of a disorder is unknown [Cavalli-Sforza and King, 1986; Ott, 19911. There are, however, several problems which can hinder or even invalidate a linkage analysis. These include genetic heterogeneity (several genes may cause a disorder, but only one or a few is responsible in any individual case), polygenic inheritance (the action of several genes is re- quired to cause the disorder in each individual), and the presence of phenocopies, or nongenetic cases, which may appear identical to the genetic cases. Moreover, since linkage analysis requires dichotomization of cases into affected and unaffected categories, diagnostic accuracy and a correct definition of the affected phenotype are essential.

Genetic heterogeneity in autism has not been studied, and in the absence of any identified gene or linkage marker, is a problem of unknown magnitude. Genetic heterogeneity can arise by mutation at different points within a single gene locus (allelic heterogeneity), or by mutation in individual genes whose proteins subserve related functions. An example of the latter form of non- allelic heterogeneity is phenylketonuria, which can be caused by mutation in the phenylalanine hydroxylase gene, or in the gene for the enzyme tetrahydropteridine reductase, which is involved in formation of the cofactor essential to the phenylalanine hydroxylase reaction [Sugita et al., 1990; Takashima et al., 19911. As the biochemical bases for more genetic diseases are uncov- ered, both types of genetic heterogeneity are increas- ingly likely to be observed.

The mode of inheritance of autism is unknown. The results to date indicate that autism does not follow clas- sical Mendelian inheritance patterns, but several studies of the inheritance of autism have been contradictory. Smalley et al. U9881 proposed a polygenic model of inheritance, while Jorde et al. r19911, analyzing 185 nu- clear families with autism, proposed a mixed model. If autism is indeed the product of the action of several genes, then the probability of finding any disease genes by linkage strategies rests on the number of genes involved and the fractional contribution each makes to the disorder.

The proportion of phenocopies to genetic cases of autism is unknown. However, several nongenetic causes, including prenatal viral infection, are associated with autism [reviewed in Lotspeich and Ciaranello, 19931, so phenocopies are likely to be an important potential con- found in genetic analyses. In designing a linkage study attempting to identify disease genes in autism, we paid particular attention to diagnostic accuracy and the definition of the autistic phenotype. In the absence of genetic markers, elimination of phenocopies is extremely difficult. However, we expected that they would more likely occur in families with a single autistic offspring than in multiplex families. For this reason, we have collected only multiplex families for our linkage study.

This report describes the diagnostic procedures and the initial data accumulated for a group of 44 presump-

tively multiplex families. They are part of an ongoing genetic linkage study in which we are collecting behavioral data on all children using a standardized autism assessment instrument, the Autism Diagnostic Inter- view [ADI, LeCouteur et al., 19891 and carrying out DNA genotyping for linked markers [Hallmayer et al., 19931. This report presents clinical and descriptive data on the autistic subjects and their siblings using the ADI. It addresses two questions important in autism genetic research. First, is there evidence for a partial or sub- clinical autism phenotype in the nonautistic siblings of autistic children in multiplex families? This question is relevant in determining the boundary of the autistic phenotype for linkage analysis. Second, are there specific signs and symptoms for which autistic siblings within a given family show significant concordance? If so, could they be used to identify diagnostically heterogeneous forms of autism which have implications for underlying genetic heterogeneity?

MATERIALS AND METHODS Subject Recruitment

Forty-four families volunteered for this study. Fifteen are families who participated in the UCLA-University of Utah Epidemiologic Survey [Ritvo et al., 1989a,bl. The remainder are families seen for diagnostic evaluation in the Stanford Pervasive Developmental Disorders Clinic, families referred by professionals who conduct evaluations and/or research on autism, or families re- sponding to recruitment notices placed in autism parent group newsletters. To be eligible for the study, a psychia- trist or psychologist had to have performed a clinical evaluation and determined that a t least two children in the family were autistic, and those records had to be available for our review. The families we evaluated res- ide in California, Utah, Florida, Michigan, Illinois, Indi- ana, Maryland, and New Hampshire.

Data Collection Procedures After referral to the study, a phone call was made to

the family to explain the nature of the study, to obtain background information on the family members, to determine the presence of records to confirm a clinical diagnosis of autism in at least two children, and to determine availability for the diagnostic assessments. Pre- vious diagnostic and medical records and written con- sents to participate were obtained prior to scheduling diagnostic appointments. Families were excluded if any presumed autistic member had a history of a neurologic disorder known to be associated with autism, such as Fragile X syndrome, Norrie syndrome, neurofibromato- sis, phenylketonuria, or tuberous sclerosis. All referred families who gave informed consent and who met the eligibility criteria were seen for evaluations. All children in these families were assessed with the Autism Diagnostic Interview [ADI, LeCouteur et al., 19891. Children under 17 years received the Autism Diagnostic Observation Schedule [ADOS, Lord et al., 19891.

One of us (D.K.S.) completed the AD1 training course given by Dr. Catherine Lord, and subsequently trained four additional evaluators (L.L., W.Mc.M., P.B.P., and P.N.). At the completion of the training period, all inter-

Autism in Multiplex Families 29

stricted interests. For each area there are, respectively, 5,5, and 6 criteria, and multiple AD1 items (questions) are used to evaluate each criterion. Therefore scores may be obtained at the level of individual items, criteria or areas. For these analyses, the full range of item scores (0-3) were used to provide maximal information. We performed the analysis on the scores for the 49 AD1 items, the 16 ICD-10 autism criteria, and on the total scores for the three major symptom areas covered by the ICD-10 diagnostic scheme. With a components of variance model, the proportion of total variance that was intrafamilial was estimated. The F-test comparing families was used to test the null hypothesis of random variation within families.

A high concordance could reflect nongenetic familial traits, in particular, maternal reporting bias. Mothers could be describing their autistic and non-autistic children in a similar way on a particular behavior. To test for this, we compared scores for the autistic and nonautistic siblings in the same family and using a matched pair t-test procedure. The standardized mean difference (d) between autistic and non-autistic siblings was used to describe the difference; the greater the d, the greater the difference between autistic and nonautistic siblings.

RESULTS Reliability Measures

Table I compares the diagnoses obtained by our group with those obtained by the North Carolina evaluators. The kappa coefficient for autistic vs. non-autistic was 0.90 (95% confidence limit > 0.80). The sensitivity was 0.90 and the specificity was 1.00. The kappa coefficient for not-autistic vs. all others was 0.90 (95% confidence limit > 0.76). The sensitivity was 0.91 and the specificity was 1.00. These findings suggest that good interrater reliability can be obtained using blinded, trained evaluators.

Comparison of Referral and AD1 Diagnoses We recruited 44 families presumptive for multiplex

autism with a total of 142 children into the study. We classified all children as autistic, not-autistic, or uncertain, and compared the results with the referring classifications (Table 11). There was agreement in 123 of 142 cases between the referring diagnosis and our AD1 as-

viewers scored 90% or better agreement on AD1 practice tapes, which were scored by Dr. Lord or her staff a t the University of North Carolina, Greensboro. Almost all the data collection took place in hospital or clinic set- tings, or schools; a few assessments were conducted in subjects’ homes. Parents, usually the mothers, were in- terviewed with the AD1 about each child in the family, and all interviews were videotaped. To achieve the greatest possible blinding of evaluators, different interviewers were used for each child in a family, when this was feasible. This was achieved 73% of the time in families with two children, while in larger families independent interviews for all children were achieved 21% of the time. The ADOS was administered after the AD1 was completed, either by the AD1 interviewer or by another trained person. The AD1 interviewers did not have ac- cess to other clinical records or other family members’ AD1 data when administering or scoring a given ADI. Independent reliability checks (described below) served as a check for possible diagnostic bias associated with non-blindness.

Intelligence test scores were obtained from existing records for the autistic children in 30 of the 37 families. Nonverbal or performance I& scores were used when available, otherwise full scale I& scores were used to classify children as mentally retard (IQ<70) or non- retarded (IQ>70).

Testing for Fragile X All children were tested for the fragile X mutation

using polymerase chain reaction amplification of the (CGG)n trinucleotide repeat region of the fragile X locus. No occult cases of fragile X were detected in this sample [Hallmayer et al., 19931.

Scoring the AD1 and Diagnostic Classifications The AD1 uses the ICD-10 criteria for autism [WHO,

19921. To meet these, the child must have a score above a prespecified cutpoint in the three symptom areas of the ICD-10 diagnostic system, and also have an age of onset prior to 3 years of age. In this study, we classified a child as autistic or affected if s/he met all four cutpoints. The child was classified as not autistic or unaffected if s h e did not meet any of the cutpoints. A child was classified as uncertain if s h e met one or more but not all the AD1 cutpoints .

Reliability Estimates To assess the reliability of our AD1 data, we selected

37 AD1 videotapes at random, with oversampling for uncertain cases. These were sent to Dr. Catherine Lord at the University of North Carolina, Greensboro, for independent diagnosis. Dr. Lord was blinded to any information about the children.

Data Analysis: Sibling Concordance for Autism Signs and Symptoms

The AD1 data from autistic siblings were analyzed using a one-way analysis of variance (ANOVA) model to estimate the within and between family variance. The ICD-10 criteria for autism are divided into three areas: social impairments, language impairments and re-

TABLE I. Reliability of Diagnoses* Diagnosis by North Carolina

Not N Autistic autistic Uncertain

Diagnosis by Stanford Autistic 20 18 0 2 Not autistic 11 0 10 1 Uncertain 6 0 0 6

Total 37 18 10 9 ~~ ~~ ~

* Thirty-seven videotapes of children from multiplex autistic families were randomly selected for reliability determination by Dr. Catherine Lord and her group at the University of North Carolina, Greensboro, as described in Materials and Methods. An oversampling of children classed as uncertain was done.

30 Spiker et al.

TABLE 11. Comparison of Presumptive and AD1 Diagnoses* Presumptive diagnosis

AD1 Diagnoses N Autistic Not autistic Uncertain A. Relationship between presumptive (referral) diagnosis and AD1 diagnosis for children in presumed autism multiplex families (N = 44 families)

Autistic 90 90 0 0 Not autistic 32 0 32 0 Uncertain 20 7 12 1 Total 142 97 44 1

B. Relationship between referral diagnosis and AD1 diagnosis for children in strictly defined multiplex families (N = 37 families)

Autistic 83 83 0 0 Not autistic 26 0 26 0 Uncertain 8 0 7 1 Total 117 83 33 1

* A shows the relation between the referral diagnosis and AD1 diagnosis for all 44 families. B provides the same information after 8 families who had one but not two autistic children were excluded (“strict” definition of multiplex).

sessments. The referral diagnosis of autism matched the AD1 diagnoses in 90 of 97 cases (sensitivity = 0.93), and in 32 of 44 non-cases (specificity = 0.73). On the basis of the research diagnoses, seven presumptively multiplex families were excluded. In these seven families, one child of the presumed affected pair met all AD1 cutpoints, but the other did not. This latter sib was classified as uncertain. In most cases s h e would have been given a clinical diagnosis of pervasive developmental disorder, (PDD-NOS) [American Psychiatric Associa- tion, 19871, a category many autism researchers believe is related to, if not part of, the autistic syndrome.

Thus 37 families could be strictly defined as multiplex for autism because they had two or more children who met AD1 criteria for the disorder. In this group of 37 families, 83 of 117 children (70.9%) were autistic, 26 (22.2%) were not-autistic, and only 8 (6.8%) had an uncertain classification. Results for the 37 eligible families are shown in Figure 1.

Clinical Characteristics of Autism Multiplex Families

Table I11 shows the ages, sex, and family sizes of the 37 autism multiplex families. Thirty-three (89%) of the families had two autistic children, two had four and two had five. It should be noted that 15 families (40.5%) had no non-autistic children. Thirty-four of the families were Caucasian and 3 were African-American. Table IV shows the AD1 area scores by diagnostic status. Note that the ranges and medians for the autistic versus other two groups are quite different. The implications of this for a linkage study are important. Children in the uncertain group met some but not all the AD1 cutoffs. The observation that their AD1 scores were much lower than the autistic group and not different from the unaffected group indicates that including the uncertain children in either the affected (autistic) or unaffected group in a linkage analysis would not be justified.

Characteristics of Unaffected Siblings in Autism Multiplex Families

We next examined further the clinical characteristics of the 26 individual unaffected siblings. The AD1 area scores for this group are shown in Table V. Although this group was judged, both clinically and by the AD1 area scores, as being non-autistic, there were a few slightly elevated but still subthreshold scores.

Characteristics of Uncertain Siblings in Autism Multiplex Families

Table VI shows the AD1 area scores for the 8 children in 6 families who were classified as uncertain because they met one or more but not all of the AD1 cutpoints. One of these (258-3) was referred to the study with an uncertain diagnosis; the remainder were considered unaffected by their referring source. We examined in more detail why children in the uncertain group received this classification. Five met the AD1 cutpoint for age of onset only, meaning parents or professionals identified some developmental problem before 3 years of age. These were usually language delays or behavioral irritability. Two met only the AD1 cutpoint for ritualistic behaviors. One (258-3) met the social, rituals, and age of onset cutpoints, and also had language deficits, but with a subthreshold score.

Concordance of Autistic Behaviors in Multiplex Families

As previously described, one of the important problems facing a linkage study is genetic heterogeneity. We attempted to address this question in a preliminary fashion by examining the concordance of autistic behaviors within multiplex families. Our hypothesis was that diagnostic heterogeneity might be revealed by a clustering of certain behaviors within families. We tested this by comparing the intrafamilial vs. interfamilial variation for each AD1 item, criterion and area scores. The results are shown in Table VIIA-C. For all AD1 items,

TABLE 111. Ages, Sex, and Family Size in Autism Multiplex Sibling Families (N = 37)*

Autistic Not autistic Uncertain N = 83 N = 26 N = 8

Age (yr)

Sex

Mean (S.D.) Median

Male N Female N M/F

Autistic children 2 (33) 4 (2“) 5 (2)

13.6 (9.2) 16.2 (9.9) 13.8 (6.9) 11 14 13

59 14 3 24 12 5 2.45 1.16 0.6

Number of non-autistic children

0 1 2 3 4 5 6 1 4 1 0 5 3 0 0 1 1 1 0 0 0 0 0 0 1 l b O O O O

* The ages, sex ratios, and family sizes for the 37 strictly defined autism multiplex families are shown. “Includes one family with four autistic children, two of whom are monozygotic twins.

One child in this family died a t 18 months of age.


An Male A u t i s t i c f) S o c i a l

0 Fema le 0 U n a f f e c t e d C o m m u n i c a t i o n

@ U n t e s t e d @ R i t u a l s and R o u t i n e s

8 Age o f Onse t

Family ID

C

351 a-6 1 1 2 3

276 h-4 1 2 3 4 5

346 h w h 1 2 3 4 5

356 6-d 1 2 3 4 5

396 h- 1 2 3 4 5

197 & I I 4 I I3 1 2 3 4 5 6

344 4 I I I 6 1 2 3 4 5 6

6 6 I 6 I3 6 1 3 4 5 6 7 8

B Family Family ID ID

111 h i 258 h 7 i b 1 2 1 2

133 h-h 1 2

217 h-h

143 d T h 1 2 3

199 1 2 1 2 3

234 221 &ji--&-a 1 2 1 2 3

325 227 h T b 1 2 1 2 3

339 h-h 236 1 2 1 2 3

398 280 d T b

195 b-4 352 h----a 1 2 1 2 3

1 2 3

163 h-6 225 & T b 1 2 1 2 3

167 1-3 1 2

290 I-& 1 2

152 6-h 1 2

139 H 1 2

Fig. 1. Sibship representations for 37 autism multiplex families. Completely shaded circles or squares are for autistic children; completely unshaded forms indicate not autistic, unaffected children; and partially shaded forms indicate children with an uncertain classification. The legend indicates which AD1 cutpoints were met, with each quadrant representing the four areas evaluated.

TABLE IV. AD1 Area Scores for Children in Autism Multiplex Sibling Families (N = 37 Families, 117 Children)*

Autistic (N = 83) Not Autistic (N = 26) (In 37 families) (In 22 families) (In 6 families)

Uncertain (N = 8)

Social total Range Median

Language total (VY No. of children Range Medyan

No. of children Language total (NWb

Range Median

Ritual total Range Median

17-44 36

58 11-29 21

25

15 9-16

3-12 8

Onset present?" N (9% Yes) 83 (100)

0-9 1

26 0-8 0.5

0 - -

0-2 0

0-14 1.5

8 0-7 0

0 - -

0-4 1.5

0 (0) 6 (75) _ _ \ _ . ---,

* Social cutpoint = 12, range = 0-42; language cutpoint = 10 for verbal subjects, range = 0-30, and cutpoint for nonverbal subjects is 8, range = 0-16; rituals cutpoint = 3, range = 0-12. "V = verbal: having phrase speech by age 5 years. bNV = non-verbal: not having phrase speech by age 5 years. 'Meets onset criteria; developmental delays or deviance 4 3 years.

32 Spiker et al.

TABLE V. Distributions of AD1 Area Scores for Non-Autistic Children in Autism Multiplex Sibling

Families (N = 22 Families, 26 Children)

Total AD1 scoresa Social Language Rituals

Area scorea N (%I N (%I N (%I 0 6 (23.1) 13 (50.0) 21 (80.8) 1 9 (34.6) 5 (19.2) 4 (15.4) 2 4 (15.4) 3 (11.5) 1 (3.8) 3 2 (7.7) 0 (0) 4 1 (3.8) 2 (7.7) 5 2 (7.7) 2 (7.7) 6 1 (3.8) 0 (0) 7 0 (0) 0 (0) 8 0 (0) 1 (3.8) 9 1 (3.8)

~

"Social cutpoint = 12, range = 0-42; language cutpoint = 10 for verbal subjects, range = 0-30, and cutpoint for nonverbal subjects = 8, range = 0-16; rituals cutpoint = 3, range = 0-12.

criteria and area scores tested, the effect sizes (d) were large (>0.8), indicating that mothers were describing their autistic children differently than their unaffected children, so reporting bias could be eliminated as a fac- tor in these results. For the majority of AD1 items and criteria, and for the three area scores, concordance scores were non-significant, indicating that intrafamilial variation among autistic siblings was similar to interfamilial variation among multiplex families.

For a few items and criteria, we observed significant concordances between sibling scores (R2 = 0.16-0.33). In the area of social impairment (Table VIIA), there was significant concordance for three individual items tap- ping behaviors concerning the display and understanding of emotional expression: limited facial expressions, seeking to share one's own enjoyment and showing inappropriate facial gestures.

In the language impairment area (Table VIIB), one criterion, the use of gestures, showed a significant clustering within families, as did one item in this criterion, pointing to express interest. Two items under the repeti-

tive speech criterion showed evidence of concordance as well: use of stereotyped utterances and pronominal reversals.

The greatest number of significant concordances was in the area concerning restricted interests and rituals (Table VIIC). Two criteria and four items scores were significant. These included ritualistic behaviors and play, and repetitive sensory behaviors. To illustrate this, we plotted data for individual autistic children, ordered by mean within each family for the item about compulsions and rituals and for the criterion about idiosyncratic and repetitive play. While some families showed evidence of clustering, in general the results are not impressive (data not shown).

Concordance for I& Scores Among Autistic Siblings

I& scores were available for both autistic siblings in 30 of the 37 families. As an overall indicator of intellectual functioning, autistic children in multiplex families may show concordance for I&. To examine this hypothesis, we first examined the I& data by classifying children as retarded (IQ<70) or normal (IQ>70). In seven families, all autistic sibs had I& scores below 70. In eleven families, all autistic siblings had IQs in the normal range (>70). In the remaining twelve families, some autistic children had normal and some had retarded scores. Overall, using a binomial test, families were not more likely to be concordant than discordant (chi-square = 1.18, n.s.) Because the classification of normal vs. retarded might be too broad, we next examined the I& scores within a more restricted numerical range. In 15 families the autistic sibs had I& scores within 15 points (1 standard deviation) of each other. In the other 15 families, the I& scores of autistic sibs differed by more than 15 points. Again, families were no more likely to be concordant when the data were examined in this fashion (chi-square = 0.00, n.s.1.

Since language ability correlates with I& in autistic children [Bartak and Rutter, 1976; Wing, 19811, we examined concordance for verbal ability in these multiplex families. On the ADI, the child is classified as verbal if

TABLE VI. AD1 Scores for Uncertain Cases*

Onset Subject Age (yr), Verbal ID Sex typeb Social Language Rituals (YiN)'

Total AD1 Area Scores"

258-3 3,M 337-2 20,F 337-7 10,F 381-2 26,F 346-3 14,F 346-4 12,M 328-4 15,F 265-3 10,M

V 12 V 2 V 0 V 5 V 3 V 1 V 0 V 0

4 Yes 3 No 3 No 1 Yes 0 Yes 2 Yes 0 Yes 0 Yes

* Siblings of autistic subjects who were classed as uncertain are shown here; the subject ID matches the pedigree drawings in Figure 1. "Social cutpoint = 12, range 0-42; language cutpoint = 10 for verbal subjects, range = 0-30, and cutpoint for nonverbal subjects is 8, range 0-16; rituals cutpoint = 3, range 0-12. bV = verbal, NV = nonverbal status on ADI: presence vs. absence of phrase speech by age 5 years. meets onset criteria: developmental delays or deviance <3 yr.


TABLE VII. Concordances for AD1 Criteria, Items, Area Scores

Criteriditems R2 d Criteriditems R2 d A: Concordances for AD1 criteria and items: social behavior impairments

B1 Nonverbal communication Direct gaze Social smiling Range of facial expressions Arms up to be lifted B2 Peer friendships Imaginative play with peers Interest in children Response to other children Group play with peers (4-10 yrs) Friends (10-15 yrs) B3 Comfort and affection Offers comfort Coming for comfort Affection Secure base Separation anxiety B4 Shared enjoyment Showing and directing attention Offering to share Seeking to share own enjoyment Sharing others’ pleasure B5 Integration of social behaviors Greeting Quality of social overtures Inappropriate facial expressions Appropriateness of social response Use of others’ body Total: All B criteria

-0.03 0.07 0.05 0.17* 0.05

-0.02 0.08

-0.05 0.08 0.05 0.05 0.02 0.15

- 0.03 0.14 0.02 0.14 0.08 0.10 0.02 0.24** 0.02 0.07 0.04

- 0.02 0.30** 0.03

0.07

B. Concordances for AD1 criteria and items: language and communication impairments

All children C1 Gestures 0.16* Pointing to express interest 0.19 Conventional gestures 0.04 Nod head to mean ‘yes’ 0.04 Shake head to mean ‘no’ 0.12

3.23 3.24 2.24 2.35 1.51 4.59 4.80 2.95 2.58 3.59 3.59 2.96 2.47 2.10 1.87 1.83 0.93 3.93 2.82 2.45 2.81 2.82 4.39 2.60 2.63 2.22 3.28

5.42

3.51 2.32 3.38 2.40 2.20

C5 Make-believe or imitative play Spontaneous imitation Imaginative play Imitative social play

C2V Conversational interchanges Chat Reciprocal conversation C3V Stereotyped or repetitive

Stereotyped utterances Inappropriate questions Pronominal reversal Neologisms/idiosyncratic language C4V Prosody Intonatiodvolumelrhythdrate Vocal expression Total: All C criteria

Verbal children

speech

0.02 0.09 0.00 0.10

0.11 0.01 0.13 0.16

0.33** 0.12 0.28* 0.04 0.04 0.12 0.03 0.12

C: Concordances for AD1 criteria and items: restricted interests and unusual repetitive behaviors

R1 Restricted interests -0.01 Circumscribed interested -0.04 Unusual preoccupations 0.05 R2 Idiosyncratic play 0.21* Repetitive use of objects 0.07 Unusual sensory interests 0.19* Abnormal idiosyncratic neg response 0.18* R3 Attachment to unusual objects 0.09 Unusual attachments 0.08 R4 Distress over trivial changes 0.12 Minor changes in routines 0.09 Minor changes in environment 0.18* R5 Stereotyped hand or body

Hand and finger mannerisms 0.09 Complex mannerisms 0.11

0.10 movements

R6 Compulsive routines and rituals

Compulsions/rituals 0.25** Total: All R criteria 0.03

0.03 Verbal rituals -0.05

5.26 2.66 4.07 3.32

3.36 3.26 3.07 2.16

2.00 1.62 1.26 0.68 1.95 1.74 1.70 1.87

1.61 0.79 1.57 2.54 2.53 1.39 0.87 1.09 1.20 2.01 2.50 1.21 1.87

1.38 1.65 1.13 0.84 1.18 2.64

* p < .05. ** p < .001.

s h e has phrase speech by 5 years of age, and as nonverbal if s h e does not. In 19 families the autistic siblings were all verbal, in three they were nonverbal and in 15 families the autistic siblings had discrepant verbal status. Once again, no significant concordance was observed using a one-sample chi-square test (chi-square = 1.31, n.s.1. Taken together, the results from these three analyses do not support the hypothesis that there are strong concordance of IQs and verbal ability within autism multiplex families.

DISCUSSION Problems in a diagnostic accuracy and phenotypic def-

inition have had an adverse impact on genetic studies of psychiatric disorders. Genetic linkage analyses have proven to be very sensitive to diagnostic errors resulting in subject misclassification, and to the presence of phenocopies or false positives [Ott, 19911. For these reasons,

in designing our genetic study we employed several strategies aimed at minimizing diagnostic error. These included (1) the use of a standardized diagnostic instrument, the Autism Diagnostic Interview, (2) evaluation by multiple trained diagnosticians with blinding of evaluators, and (3) independent verification of our diagnoses by a separate, blinded diagnostic group. To minimize the problem of subject misclassification, we limited the definition of affected individuals to those meeting all of the AD1 criteria for autism.

This study is the first to determine the presence of autistic behaviors in all children, both the presumed affected and unaffected, in a large sample of multiplex autism families, and the results of our analyses are of interest in several ways. First is the question of presumptive or referral diagnosis vs. research diagnoses. Second is the definition of affected status, and third is the question of genetic heterogeneity.

34 Spiker et al.

Referral vs. Research Diagnoses In genetic studies, it may be difficult or costly for the

research team to examine blindly every subject. To what extent can the evaluations done by professionals outside the study be used in linkage analyses? Our data suggest that there is good agreement between our referral sources and our own diagnoses in classifying children as autistic rather than not autistic and vice versa, but we classified some cases in both groups as uncertain (Table 11). Of 20 children whom we classified as uncertain, outside clini- cians classified 7 as autistic, 12 were presumed to be unaffected, and only one was referred with an uncertain diagnosis. One important consequence of the differences in classification seen in Table I1 is that we had to exclude seven presumptively multiplex families from the study because there were seven false positive presumptive diagnoses using the research diagnoses. For purposes of genetics studies, the research team must treat the referral diagnosis as presumptive and arrive at its own diagnostic conclusions. The practical implications of this conclusion are that classification by chart review or clinical records of individuals who are unavailable for assessment is probably unwise. Additionally, clinical information is unlikely to be obtained by individuals who are blinded to status of other family members, and for the presumably unaffected siblings, clinical records are especially likely to be sparse or unavailable.

In this study we used a standardized diagnostic instrument, the ADI/ADOS, for diagnostic purposes. There are several advantages to this approach. First, without such standardized procedures, it is unclear whether the children being classified as autistic are comparable from study to study, from cohort to cohort, or even within a single study where different diagnosticians have participated. Second, although affected status might be appropriately defined by the referral diagnosis (e.g., autism), a decision to define affected status by individual signs and symptoms or clusters of these might later be warranted for linkage analysis. Such data is only available when standardized instruments are used and information about individual behaviors is collected systematically for each subject.

Definition of Affected Status A second important question addressed by these find-

ings is the matter of the autistic phenotype. There is general consensus that an autistic spectrum exists, with variation both in severity and clinical expression [Rut- ter and Schopler, 1992; Szatmari, 19921. If this is true, then who, in a linkage study, should be included as affected? There have been numerous studies attempting to define the boundaries of the autism phenotype. Clas- sically autistic subjects, those meeting criteria for pervasive developmental disorder, as well as those with developmental language disorders have all been included in the definition of affected [Wing, 1988; Bolton and Rutter, 19901. One of the goals of our study was to determine the proportion of children in multiplex families who could be unambiguously classified, so that we could arrive at a suitable definition of the affected phenotype for linkage analysis.

At the outset of this study we hypothesized that AD1

analysis of all children in multiplex families would yield a broad distribution of phenotypes, spanning more or less a continuum from unaffected to affected, without clear breakpoints. We expected to find a large group of children falling into an ambiguous, or uncertain, phenotype. Such a phenotypic distribution would present a daunting challenge to a linkage analysis.

The data indicated otherwise: children were either unambiguously autistic (-71%), or unaffected (-22%), with only a small uncertain group. These proportions were not greatly affected by excluding the seven families who were not strictly multiplex; taking all 44 families, the autistic group was 63%, the unaffected group 22% and the uncertain group 14% of all children (Table 11). Some important points emerge from this. First, a con- servative or strict definition of autism can be used without undue concern for the presence of an ambiguous group. Linkage can then be performed by defining affected individuals as solely those meeting strict criteria for autism. The uncertain group can be treated as “unknown,” and excluded from linkage analysis. Some information is lost this way, but the risk of false positive classification is avoided.

Second, the distinction between affected and unaffected is clear. A few children in the unaffected group had mildly elevated AD1 scores in some areas. In the absence of additional data from AD1 evaluations of children taken randomly from the general population, the significance of subthreshold scores is unclear, but there is no empirical reason to classify these children as other than unaffected. Third, because there is only a very small uncertain group, we may not need to collect a great number of families to accumulate sufficient num- :hers for linkage analysis. Power calculations to determine the needed sample size are dependent on estimates of genetic heterogeneity, which is unknown in autism. Therefore these calculations are tenuous at best, but we have estimated that 100-200 families should provide sufficient power to detect linkage, even in the presence of substantial genetic heterogeneity (Hallmayer, un- published observations). Fourth, because of the problems of phenotypic spectrum, Pauls [19931 has advo- cated the use of multidimensional or continuous approaches to diagnosis in psychiatric genetics. While our results indicate this might not be necessary in autism, the data from the ADI, which includes information on the degree of severity, as well as the presence or absence of a large number of autism symptoms is ideally suited for such an approach.

Genetic Heterogeneity A third point which we attempted to address was

genetic heterogeneity. Our hypothesis was that there might be diagnostically distinct subgroups of autistic subjects that could be defined on the basis of high intrafamilial concordance for certain autistic behaviors. From the analyses of individual AD1 items, criteria, and areas, we could find only limited evidence in support of this hypothesis. The familial concordances for specific autistic symptoms were generally not strong, and high intrafamilial variability was the rule. There were, however, significant intrafamilial concordances for autistic


tersen PB, Nicholas P, Pingree C, Kraemer HC, Wong DL, Ritvo E, Cavalli-Sforza LL, Ciaranello RD (In preparation): Exclusion of linkage between the Fragile X gene and familial autism.

Jorde LB, Hasstedt SJ,-Ritvo ER, Mason-Brothers A, Freeman BJ, Pingree C, McMahon WM, Peterson B, Jenson WR, Moll A (1991): Complex segregation analysis of autism. Am J Hum Genet 49: 932-938.

Lecouteur A, Rutter M, Lord C, Rios P, Robertson S, Holdgrafer M, McLennan J (1989): Autism Diagnostic Interview: A standardized investigator-based instrument. J Autism Dev Disord 19:363-387.

Lord C, Rutter M, Good S, Heemsbergen J , Jordan H, Mawhood L, Schopler E (1989): Autism Diagnostic Observation Scale: A standardized observation of communicative and social behavior. J Au- tism Dev Disord 19:185-212.

Lotspeich W, Ciaranello RD (1993): The neurobiology and genetics of infantile autism. Int Rev Neurobiol 35:87-129.

Ott J (1991): “Analysis of Human Genetic Linkage,” Revised Edition. Baltimore: The Johns Hopkins University Press.

Pauls DL (1993): Behavioural disorders: lessons in linkage. Nature 3:4-5.

Risch N (1990): Genetic linkage and complex diseases, with special reference to psychiatric disorders. Genet Epidemiol 7:l-16.

Ritvo ER, Freeman BJ, Pingree C, Mason-Brothers A, Jorde L, Jenson WR, McMahon WM, Petersen PB, Mo A, Ritvo A (1989a): The UCLA-University of Utah epidemiologic survey of autism: Preva- lence. Am J Psychiatry 146:194-199.

Ritvo ER, Jorde LB, Mason-Brothers A, Freeman BJ, Pingree C, Jones MB, McMahon WM, Petersen PB, Jenson WR, Mo A (1989b): The UCLA-University of Utah epidemiologic survey of autism: Recur- rence risks estimates and genetic counseling. Am J Psychiatry 146:1032-1036.

Rutter M (1990): Autism as a genetic disorder. In McGufXn P, Murray R (eds): “The New Genetics of Mental Illness.” Stoneham, MA: Butter- worth-Heinemann, pp 225-244.

Rutter M, Schopler E (1992): Classification of pervasive developmental disorders: some concepts and practical considerations. J Autism Dev Disord 22:459-482.

Rutter M, MacDonald H, LeCouteur A, Harrington R, Bolton P, Bailey A (1990): Genetic factors in child psychiatric disorders - 11: empirical findings. J Child Psycho1 Psychiatry 31:39-83.

Smalley SL (1991): Genetic influences in autism. Psychiatr Clin North

Smalley SL, Asarnow RF, Spence MA (1988): Autism and genetics. Arch Gen Psychiatry 45953-961.

Sugita R, Takahashi S, Ishii K, Matsumoto K, Ishibashi T, Sakamoto K, Narisawa K (1990): Brain CT and MR bindings in hyperphenyl- alaninemia due to dihydropteridine reductase deficiency (variant of phenylketonuria). J Comp Assist Tomogr 14(5):699-703.

Szatmari P (1992): The validity ofautistic spectrum disorders: a litera- ture review. J Autism Dev Disord 22583-600.

Takashima S, Chan F, Becker LE (1991): Cortical dysgenesis in a variant of phenylketonuria (dihydropteridine reductase deficiency). Pediatr Path01 11(5):771-779.

Wing L (1981): Language, social and cognitive impairments in autism and severe mental retardation. J Autism Dev Disord 11:31-44.

Wing L (1988): The continuum of autistic characteristics. In Schopler E, Mesibov GB (eds.): “Diagnosis and Assessment in Autism.” New York Plenum Press, pp 91-100.

World Health Organization (1992): “ICD-10. Classification of Mental and Behavioural Disorders. Clinical Description and Diagnostic Guidelines.” Geneva: Author.

Am 14125-139.

behaviors concerning ritualistic and repetitive behaviors that may yield clues for identifying behaviorally defined subgroups of multiplex families. This possibility should be pursued with a larger sample of multiplex families. Additionally, the possibility of stronger concordances among distinct subgroups exhibiting patterns of behaviors needs to be explored. While we will continue to examine these possibilities as we collect additional families, we think it likely that further attempts at identifying genetic heterogeneity in autism must await the finding of a linkage marker or an actual gene locus.

Such a marker will answer many questions left unre- solved or raised by this study. Among the most important of these is the boundary of the autistic phenotype, which we have drawn tightly around the AD1 definition, but which others have argued quite persuasively should be extended. But where should the line be drawn? Ex- amining the “uncertain” and even the “unaffected” siblings of autistic subjects with a linked marker or gene probe will shed critical light on this question, as well as providing important independent biologic evidence for an autism classification.

ACKNOWLEDGMENTS This work was supported by a program project grant

from the National Institute of Mental Health (MH394371, and by grants from the Scottish Rite, the Spunk Fund, the Solomon and Rebecca Baker Fund, and the endow- ment fund of the Nancy Pritzker Laboratory. L.L. and J.H. are the recipients of Young Investigator Awards from the National Alliance for Research in Schizophre- nia and Affective Disorders (NARSAD). R.D.C. is the recipient of a Career Scientist Award from the NIMH (MH 00219). We are grateful to the families who are partici- pating in this research.

REFERENCES American Psychiatric Association (1987): “Diagnostic and Statistical

Manual of Mental Disorders 111-R’ (Third ed.). Washington, DC: American Psychiatric Association.

Bartak L, Rutter M (1976): Differences between mentally retarded and normally intelligent autistic children. J Autism Child Schizophr 4109-120.

Bolton P, Rutter M (1990): Genetic influences in autism. Int Rev Psy- chiatry 2:67-00.

Cavalli-Sfona LL, King MC (1986): Detecting linkage for genetically heterogeneous diseases and detecting heterogeneity with linkage data. Am J Hum Genet 38:599-616.

Folstein SE, Piven J (1991): Etiology of autism: Genetic influences. Pediatrics 87:767-773.

Hallmayer J, Underhill P, Spiker D, Lotspeich L, Kraemer HC, McMahon WM, Peterson PB, Nicholas P, Pingree C, Wong DL, Ciaranello RD, Cavalli-Sfona LL (1993): A linkage study of autism. Biol Psychiatry 33:104a.

Hallmayer J , Pintado E, Lotspeich L, Spiker D, McMahon WM, Pe-

Genetics of autism: Characteristics of affected and unaffected children from 37 multiplex families

Documents