Birth Order Effects on Nonverbal IQ Scores in Autism Multiplex Families

Transcription

1 Journal of Autism and Developmental Disorders, Vol. 31, No. 5, 2001 Birth Order Effects on Nonverbal IQ Scores in Autism Multiplex Families Donna Spiker, 1,4 Linda J. Lotspeich, 1 Sue Dimiceli, 1 Peter Szatmari, 2 Richard M. Myers, 3 and Neil Risch 3 Lord (1992) published a brief report showing a trend for decreasing nonverbal IQ scores with increasing birth order in a sample of 16 autism multiplex families, and urged replication in a larger sample. In this report, analyses of nonverbal IQ scores for a sample of 144 autism multiplex families indicated that nonverbal IQ scores were significantly lower in secondborn compared with firstborn siblings with autism. This birth order effect was independent of gender as well as the age differences within sib pairs. No such birth order effects were found for social or communicative deficits as measured by the Autism Diagnostic Interview-Revised (ADI-R), but there was a modest tendency for increased scores for ritualistic behaviors for the firstborn sibs. Further, there were no gender differences on nonverbal IQ scores in this sample. Results are discussed in terms of implications for genetic studies of autism. KEY WORDS multiplex families; birth order; IQ scores. INTRODUCTION In recent years, researchers have become interested in studying families with more than one child with autism (multiplex families), focusing on this group because of the greater possibility of identifying genetic causes (Smalley, Asarnow, & Spence, 1988; Spiker, 1999; Szatmari et al., 1998). Lord (1992) published data for a sample of 16 autism multiplex families indicating that there was a trend for decreasing nonverbal IQ scores with increasing birth order. She suggested that if this trend is seen in other larger samples, it may 1 Division of Child Psychiatry and Child Development, Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, 401 Quarry Rd, MC 5719, Stanford, CA Department of Psychiatry, McMaster University Research Building, Chedoke-McMaster Hospitals, Box 2000, Station A, Hamilton, Ontario, Canada L8N 3Z5. 3 Department of Genetics, Stanford University School of Medicine, 300 Pasteur Drive, Stanford, CA Requests for reprints should be addressed to Donna Spiker, Ph.D., at her current affiliation: SRI International, Center for Education and Human Services, 333 Ravenswood, Menlo Park, CA suggest that certain environmental factors, such as maternal immunoreactions, may play a role in influencing cognitive ability in autism multiplex families. Lord s sample included some families in which siblings within the same family met criteria for a diagnosis of autism, whereas other families had one child with a diagnosis of autism and another with a diagnosis of pervasive developmental disorder, not otherwise specified (PDD-NOS), based on the Autism Diagnostic Interview (ADI) (LeCouteur et al., 1989; Lord, Rutter & LeCouteur, 1994; Lord et al., 1997). Severity of autism, as indexed by ADI scores, did not appear to be associated with birth order in any simple way. Because of the small sample size and the possibility of referral bias owing to the sample being drawn from two clinic samples, additional studies in other, independent samples of multiplex families would be useful. Lord noted that this birth order effect may interact with the gender of the affected siblings, but her sample was too small to draw definitive conclusions about a possible birth order gender relationship. In this paper, we report nonverbal IQ data for a significantly larger sample of siblings with autism in /01/ $19.50/ Plenum Publishing Corporation

2 450 Spiker et al. autism multiplex families. Although there are a great deal of data on singleton cases of autism, and there have been reports about IQ scores of siblings and parents in singleton autism families (see Szatmari & Jones, 1991; Freeman et al., 1989), the published data for autism multiplex families are limited. As Szatmari & Jones (1991) noted in a review of studies about IQ and autism with mainly singleton families, the majority of children with autism are mentally retarded, but their IQs range from severely retarded to well above average, with females more likely to be more delayed than males. We have recently completed a genome-wide linkage study of autism that included 139 autism multiplex families (those with more than one child with autism; Risch et al., 1999). All children in this study were evaluated with the Autism Diagnostic Interview-Revised (ADI-R) (LeCouteur et al., 1989; Lord, Rutter, & LeCouteur, 1994; Lord et al., 1997), and only those families with two or more siblings meeting ADI criteria for a diagnosis of autism were included. The children were also observed using the Autism Diagnostic Observation Schedule (ADOS) (Lord et al., 1989). As part of that study, records containing results for standardized IQ testing were obtained for these siblings with autism. In this paper, we addressed the following questions by using data from this large sample of autism multiplex families: (1) Is there evidence, in this larger multiplex sample, for a trend of decreasing nonverbal IQ with birth order, as suggested by Lord (1992)? (2) are there sex differences in IQ scores for autistic children in these multiplex families and, if so, are they also related to birth order? (3) is there evidence for greater severity of autistic symptoms associated with increasing birth order? (4) are nonverbal IQ or symptom severity differences in birth order associated with age differences between firstborns and secondborns? and (5) is an observed birth order effect due to autism birth order or actual birth order in the sibship? METHODS Subjects As part of an ongoing genetic study of autism (Hallmayer et al., 1996a, 1996b; Hallmayer et al., 1994; Risch et al., 1999; Rogers et al., 1999; Salmon et al., 1999; Spiker, 1999; Spiker et al., 1994) families were recruited based on previous clinical evaluations that indicated a diagnosis of autism or another pervasive developmental disorder (without any associated primary disorder such as fragile X). Referrals came directly from families who saw notices in parent newsletters or who referred other families to the study, and from clinicians, other professionals, and colleagues who work with families with children with autism and PDD-NOS. The sample for this report includes a total of 144 autism multiplex families; it includes 132 of the 139 families from the genome screen reported in Risch et al., (1999). Seven families were excluded from the 139; these included 6 sets of DZ twins and one set of DZ triplets. Another 12 families were added to the 132 who were comparable to those families, but were assessed after the genome screen was underway. Five families included in this study had affected MZ twins and another affected sibling; one of the MZ twins was randomly selected for analyses. Data Collection Procedures If both the initial telephone intake interview of parents and collected medical records were consistent with a presumptive diagnosis of autism, then a follow-up visit to the family s home by a diagnostician was arranged with the family. During these visits, the children were assessed using the ADI-R (LeCouteur et al., 1989; Lord et al., 1994, 1997), and by the ADOS (Lord et al., 1989), to determine a research diagnosis of autism. The ADI-R and ADOS diagnostic assessments were videotaped and a subset were independently scored to assess reliability (see Spiker et al., 1994). To determine a research diagnosis of autism, an individual was required to meet the prespecified cutoff scores in all three symptom areas of the ADI-R (social impairment, language and communication impairment, and unusual and restricted interests), as well as have an age at onset of less than 3 years. In addition, two or more diagnosticians reviewed the ADOS tape to exclude children who do not show significant impairments in social and communicative reciprocity, irrespective of the results of the ADI interview. This procedure resulted in the exclusion of some individuals who would have a clinical diagnosis of PDD-NOS. Thus, families were excluded if, on the basis of all available information, there was no consensus that at least two affected children had deficits consistent with a diagnosis of autism. Diagnostic records were reviewed to obtain nonverbal IQ and mental-age (MA) information. When available, nonverbal IQ scores were obtained from performance subtests of either the Wechsler Intelligence Scales for Children (Revised or III or the preschool version) or the Stanford-Binet Intelligence Scale (4th Ed.), from the Leiter International Performance Scale, or from the Merrill-Palmer Scale (these were available for 50% of the sample). When such scores were unavailable, we used scores from such tests as the Stanford-

3 Birth Order Effects in Autism 451 Binet Intelligence Scale (3rd Ed.), Slossen Intelligence Scale, or McCarthy Scales (for 13% of the sample). For children for whom results of these tests were not available, MAs and ratio IQ scores were derived from nonverbal scales of various developmental instruments (for example, Daily Living Scale of the Vineland Adaptive Behavior Scale, the Bayley Scales of Infant Development Mental Scale, and Developmental Profile II for 37% of the sample). For inclusion, at least one affected child had to have both an MA of 18 months and a nonverbal IQ 30 (i.e., families in which all affected children had severe mental retardation were excluded). Data Analysis A matched pairs t-test was used to examine the effects of birth order on the following outcomes: nonverbal IQ, severity of autistic symptoms measured by the ADI-R (social impairment scores, language scores, rituals scores, verbal-nonverbal status at the time of the ADI-R administration). An ANOVA was used to test for gender effects on these same variables. Because the secondborn child would be younger than the firstborn child, we examined correlations between nonverbal IQ and age within each birth order group to determine if adjustments would be needed when examining the birth order effects. Each ADI item is rated for two different time periods in the subject s life: at the current time and at age 4 to 5 years or ever. In the current study, only the scores from the 4 to 5 years/ever ratings were used, as required by the standard scoring algorithm. The scoring of individual ADI-R items ranges from 0 to 3; scores of 3 were converted to 2 (as specified in the standard scoring algorithm). The full range of ADI-R items was used: 22 social impairment items (scores ranging up to 44); 8 language impairment items for nonverbal subjects (scores ranging up to 16); 15 language impairment items (i.e., 8 nonverbal items, plus 7 verbal items) for verbal subjects (scores ranging up to 30); and 12 rituals and routines items (scores ranging up to 24). Because the verbal and nonverbal subjects have different language impairment scores and thus are not comparable, we used only the 7 nonverbal items that are comparable between the verbal and nonverbal subjects. Higher ADI-R scores indicate greater severity of autistic symptoms. RESULTS The ADI-R and the IQ tests were administered at different ages (Table I). As expected, for the ADI-R, the secondborn siblings with autism were significantly younger at the time of administration (t 4.06; p.0001). The mean ages of subjects at the time of ADI- R administration was not significantly different for males (N 222) compared with females (N 66) (t 0.59). For IQ testing, the age at which the test was administered was also significantly younger for secondborn compared with firstborn siblings, with a mean age difference of 1.5 years (SD 3.8 years) and the 25th to 75th percentiles being from 0 to 2.8 years (t 2.611; p.01). The mean age at IQ testing was also significantly younger for females compared with males (t 1.20, p.05). The actual mean age difference between the firstborn and secondborn autistic siblings was 3.1 years (SD 2.3 years), with the 25th to 75th percentiles ranging from 1.6 to 3.8 years (t 15.7, p.0001). A Spearman rank order correlation between the nonverbal IQ scores and the age at which the IQ test was administered was not significant for the firstborns (r.12), but there was a significant negative Spearman rank order correlation for the secondborn siblings (r.30 p.0002). There was no significant Pear- Table I. Ages of Administration of IQ Tests and ADI-R of Autism Sib Pairs by Birth Order and Gender 1st 1st 2nd 2nd 1st 2nd Male Female Male Female Male Female Total Total Total Total TOTAL Variable (N 115) (N 29) (N 107) (N 37) (N 144) (N 144) (N 222) (N 66) (N 288) Age at IQ Test Mean SD percentile Age at ADI-R Mean SD percentile ADI-R, Autism Diagnostic Interview-Revised; SD, standard deviations.

4 452 Spiker et al. son correlation between IQ and age for the firstborn siblings (r.15, p.06), nor for the secondborns (r.04). After a square root transformation, the Pearson correlation for the group of firstborns was.06, and it was.22 (p.01) for the group of secondborns. Taken together, the correlations indicate that, for the secondborn group only, the younger the child s age at the time of testing, the higher the nonverbal IQ score (and this effect is taken into account in later analyses). The nonverbal IQ scores for the autism sib pairs by birth order are shown in Table II. The average difference between the IQ scores within sib pairs was 12.9 points (SD 32.6), and was significantly different using a matched pairs t-test (t 4.78, p.001). The firstborns were more likely to have higher nonverbal IQ scores. The nonverbal IQ scores within sib pairs were modestly, but significantly, positively correlated (r.20, p.05). Fig. 1 and Fig. 2 show the distribution of the nonverbal IQ scores for the firstborns and secondborns, respectively. Fig. 3 depicts the within sib pair nonverbal IQ score differences for this sample of 144 families, with positive differences indicating that the firstborn sib had a higher score, and negative differences indicating that the score of the secondborn was higher within a sib pair. Figure 3 is skewed toward the right, showing that there are more families for which the nonverbal IQ score of the firstborn sib with autism was higher than that of the secondborn sib with autism (N 90; 62%). Table II also shows nonverbal IQ scores for autism sib pairs by gender. To test for a gender effect for nonverbal IQ scores, we performed an analysis of covariance, with age as a covariate and using the IQ and age at IQ testing with square root transformation (to control for the correlations between IQ and age at testing described previously). There was no significant gender difference (F 0.09), nor was there a significant interaction between birth order and gender (F 0.28). As expected, this analysis also showed the significant birth order effect (F 17.11, p.0001), as well as a significant effect for age (p.02). Post-hoc comparisons to investigate the significant birth order main effect for this analysis revealed that the four gender comparisons were significant. Table III shows the distribution of the nonverbal IQ scores for firstborns, secondborns, males and females, to provide an indication of the degree to which the nonverbal IQ scores are in the mental retardation range. For birth order, there was a significant chi-square (15.76, p.001), with a distribution indicating more firstborns with higher scores and more secondborns with lower scores. A greater proportion of the secondborns had nonverbal IQs in the range of mental retardation (70.8%) compared with firstborns (47.9%). Similar proportions of males and females had nonverbal IQ scores in the range of mental retardation (59.0% versus 60.6%, respectively) (! , ns). Furthermore, as shown in Table IV, mean nonverbal IQ scores were similar for males and females, regardless of whether they were in same sex or mixed sex sib pair families. To determine if the sib pairs with the most discrepant IQ scores (largest differences in IQ scores) tended to have the greatest age differences between the firstborn and secondborn sibs, we calculated the correlation between the sib pair IQ score differences and the sib pair age differences at the time of IQ testing. This correlation was nonsignificant (.07) indicating that the age at testing does not explain the observed birth order effect. We did a similar analysis correlating the sib pair IQ difference with actual sib pair age difference between firstborn and secondborn autistic sibs. Here also the correlation (.05) was not significant. Furthermore, Fig. 4, showing mean IQ scores by the actual birth order compared with the autistic birth order, indicates that the decreasing IQ scores is due to autistic birth order rather than actual birth order. Table II. Nonverbal IQ Scores for Autism Sib Pairs (N 144) by Birth Order and Gender 1st 1st 2nd 2nd 1st 2nd Male Female Male Female Male Female Total Total Total Total TOTAL Variable (N 115) (N 29) (N 107) (N 37) (N 144) (N 144) (N 222) (N 66) (N 288) Nonverbal IQ Mean SD percentile % SD, standard deviations.

5 Birth Order Effects in Autism 453 Fig. 1. Distribution of nonverbal IQ scores of firstborn sibling with autism in autism multiplex families. Table V shows the means and standard deviations of the ADI-R variables by birth order and gender. The results of matched pair t-tests for the four ADI-R variables revealed no significant birth order effect for the social total scores (t 0.72) or the nonverbal language scores (t 0.49). The rituals total scores were significantly higher for the firstborn sibs (t 2.07, p.05). There was also a significant birth order effect for verbal-nonverbal status at the time of ADI-R administration, with higher proportions of secondborns classified as nonverbal and thus more impaired (t 4.50, p.0001). Being verbal at the time of ADI-R administration was significantly positively correlated with nonverbal IQ scores (r.54 for total sample; r.55 for firstborn sibs; r.43 for secondborn sibs; all with p.0001). Because of the possible confounding of birth order and nonverbal IQ with verbal status on the ADI-R, we examined the correlation between the nonverbal IQ score differences within sib pairs and the verbal status differences within sib pairs. This correlation was.48 (p.001), indicating that verbal status is contributing significantly to the IQ birth order relationship that was found. In Table VI, the IQ scores and within sib pair IQ differences show that although verbal status accounts for much of the relationship between birth order and nonverbal IQ scores, an IQ difference persists after controlling for verbal status. In sib pairs who were discordant for verbal status, nonverbal IQ scores were higher for the siblings who were verbal compared with their nonverbal sibling, regardless of birth order, although the differences were somewhat greater when the first child was verbal compared with the second child being verbal. When both sibs were nonverbal, the mean IQs were similarly low for both firstborns and secondborns. However, when both sibs were verbal (the largest subgroup), the nonverbal IQs of the firstborns were much higher than those of their secondborn siblings (by 15 points). We also noted that verbal status is only modestly familial, because the observed number of concordant pairs was only slightly increased versus expectation under independence (56 are both verbal versus 52 expected pairs, 27 are both nonverbal versus 23 expected pairs). To determine the degree to which the birth order effect on nonverbal IQ scores is related to the particu-

6 454 Spiker et al. Fig. 2. Distribution of nonverbal IQ scores of secondborn sibling with autism in autism multiplex families. lar test administered, we examined the mean scores by test for the firstborn and secondborn children with autism. We grouped the IQ tests into eight categories (Table VII). There is a significant main effect for type of test (F 26.74, p.0001). Higher scores were obtained by children administered the Wechsler Scales, Stanford-Binet (4th Ed.), and Leiter Scale, whereas lower scores were obtained for children administered the Vineland Daily Living Scale or other developmental inventories. This difference among tests is a reflection of the fact that the lower functioning children were not capable of performing on the former, more standardized tests, and were therefore usually assessed using the latter instruments which are commonly used with younger or more developmentally delayed children. Examination of the data in Table VII reveals that even stratifying on test type, the firstborn children performed better than did the secondborn children. Within test type, the difference in mean between firstborn and secondborn sibs was positive for each test type, with a median of 7 points across tests. The difference tended to be greater for the more standardized full IQ tests (e.g., Wechsler Scales) administered to higher functioning children. In the overall sample of firstborn and secondborn sibs, the difference in mean nonverbal IQ was 12 points. From this table, we can derive that 7 of those 12 points come from the difference by birth order within test type, with the remaining 5 points owing to the fact that firstborn sibs tend to receive the more standardized full IQ test (e.g., Wechsler Scales) than the secondborn sibs. This latter difference is not a reflection of a random effect of test administered, but rather is due to the fact that secondborn sibs often were more impaired than the firstborn sibs and therefore could not be administered the higher level tests. DISCUSSION The data from this large sample of 144 autism multiplex families show that there is a significant correlation between birth order and nonverbal IQ scores in autism multiplex sibships. Firstborns tended to have higher nonverbal IQ scores than secondborns or laterborns. There was no effect of gender, nor was there a birth order by gender interaction. Although the secondborn sibs were younger when their IQs were measured, there was a significant but negative correlation

7 Birth Order Effects in Autism 455 Fig. 3. Distribution of nonverbal IQ discrepancies between firstborn and secondborn siblings in autism multiplex families. between nonverbal IQ scores and age at testing for the group of secondborn sibs with autism. This result indicates that the birth order effect on nonverbal IQ scores found here was not confounded by age and, if anything, the magnitude of the effect may be underestimated. Less than half of the firstborn sibs had nonverbal IQs in the mental retardation range, whereas 70% of the secondborn sibs had such scores. This birth order effect on nonverbal IQ scores did not extend to the autistic symptoms indicative of social and nonverbal communicative deficits, although scores for ritualistic behaviors measured by the ADI- R were slightly higher for the firstborn sibs with autism. There was, however, a very significant birth order effect on verbal status: Firstborn autistic sibs tended to be verbal more often than secondborns or laterborns. This birth order difference in verbal status may contribute to the birth order effect on nonverbal IQ because verbal status is significantly correlated with nonverbal IQs, but it does not fully account for it, because we still found significantly lower nonverbal IQs among secondborn sibs in sib pairs in which both sibs were ver- Table III. Distribution of Nonverbal IQ Scores of Autism Sib Pairs (N 144) by Birth Order Firstborns (144) Secondborns (144) Total (288) Males (222) Females (66) Nonverbal IQ Scores N (%/%)* N (%/%)* N (%) N (%/%)* N (%/%)* IQ (12.8/25.7) 57 (19.9/39.6) 94 (32.6) 72 (25.0/32.4) 22 (7.6/33.3) IQ (11.1/22.2) 45 (15.6/31.2) 77 (26.7) 59 (20.5/26.6) 18 (6.3/27.3) IQ (26.0/52.1) 42 (14.6/29.2) 117 (40.6) 91 (31.6/41.0) 26 (9.0/39.4) *percentage of total sample of families/percentage of the column For birth order:! (p.001) For gender:! (ns)

8 456 Spiker et al. Table IV. Means and Standard Deviations for Nonverbal IQ Scores for Same and Mixed Sex Autism Sib Pairs Males Females Same sex Mixed sex Same sex Mixed sex Nonverbal IQ (N 184) (N 38) (N 28) (N 38) Mean SD SD, standard deviations. bal. Thus, the results suggest that the birth order effect is a more generalized deficit than simply affecting verbal status. Finally, there was a small but significant correlation in nonverbal IQ scores for affected sibs from the same sibship, but little correlation in verbal status. The data reported here support and extend the data reported by Lord (1992) suggesting lower nonverbal IQ scores of laterborn siblings with autism in autism multiplex families. The meaning of these birth order effects is not clear. Lord suggested that maternal immunoreactions may be associated with this decline in nonverbal IQ with parity (Gualtieri & Hicks, 1985). Presumably, maternal antibodies are sensitized to fetal brain tissue and affect the neuronal substrate of cognitive abilities in the secondborn or laterborn sibs. Perhaps the maternal histocompatibility complex system of genes is involved in altering the phenotype of affected children. These data are consistent with evidence of increasing weight with birth order among typical children as well (Seidman et al., 1988; Wilcox, Chang, & Johnson, 1996). Although the immune system has been studied in autism for many years and subtle changes have been reported (e.g., Singh et al., 1993; Warren et al., 1996), genes in the histocompatibility complex system of genes have been ruled out as major susceptibility genes in children with autism (Rogers et al., 1999). Genetic studies indicate that autism is caused by multiple genes, many of which may have small effects (Barrett et al., 1999; International Consortium, 1998; Phillippe et al., 1999; Risch et al., 1999), and it is possible that some of these genes may involve the immune system in both the fetus and the mother in interaction. This would be an example of possible environmental factors operating in utero that might interact with the underlying genetic liability to cause increased cognitive impairment in laterborn affected siblings. This observed birth order effect also could be produced by interactions between autism susceptibility genes and postnatal environmental factors. One way to test these hypothesized relationships would be to compare autism multiplex families with and without clear evidence of immune dysfunction. The birth order effects reported here did not extend to some of the other behaviors measured by the ADI-R, specifically social impairments and nonverbal communicative behaviors. The lack of birth order effects on these ADI-R variables is also similar to the data reported earlier by Lord (1992). However, scores Fig. 4. Mean nonverbal IQ scores by actual and autistic birth order.

9 Birth Order Effects in Autism 457 Table V. Means and Standard Deviations for ADI-R Scores for Autism Sib Pairs by Birth Order and Gender 1st 1st 2nd 2nd 1st 2nd Male Female Male Female Male Female Total Total Total Total Variable (N 115) (N 29) (N 107) (N 37) (N 144) (N 144) (N 222) (N 66) Social Total Score Mean SD Nonverbal Language Score Mean SD Rituals Score Mean SD Verbal Status Mean SD SD, standard deviations. indicating greater severity of ritualistic/repetitive behaviors in the firstborn sibs were found, an effect that warrants further investigation. It may be the case that higher levels of ritualistic/repetitive behaviors, on average, are positively associated with greater awareness of contingencies in the environment. Because the nonverbal IQs of the firstborns were higher, on average, this may contribute to the birth order effect for this domain of behaviors. An item that did show a more significant birth order effect, verbal-nonverbal status, was highly correlated with nonverbal IQ. It is not surprising, therefore, that this variable would show results similar to those found for nonverbal IQ. Either autistic behaviors Table VI. Nonverbal IQ Scores of Autism Sib Pairs (N 144) by Birth Order and Verbal Status on ADI-R Sib Pair Verbal Status Standard Nonverbal IQ Mean Deviation Both Verbal (N 56) Firstborn IQ Secondborn IQ Both Nonverbal (N 27) Firstborn IQ Secondborn IQ st Verbal, 2nd Nonverbal (N 47) Firstborn IQ Secondborn IQ st Nonverbal, 2nd Verbal (N 14) Firstborn IQ Secondborn IQ in the social and ritualistic domains are not sensitive to birth order effects, or these behaviors are influenced by a different set of mechanisms. Alternatively, these variables were measured differently the ADI symptoms via interviews using parental report, the IQ scores via tests by psychologists or other professionals and such measurement differences may influence the data obtained. Furthermore, it is reasonable to expect that, because it is likely that several or many genes interact to increase susceptibility to autism, some of these genes may influence the variance of some phenotypic characteristics and not the variance of others. The results reported here are consistent with data reported by MacLean et al. (1999) that level of cognitive functioning shows familial correlation, but that ADI-R autistic behaviors do not. Thus, the birth order effect we observed is primarily for overall severity of cognitive and language functioning. Several interpretations of these results should be considered. First, this effect of increasing impairment with increasing birth order within sibships argues against a purely genetic mechanism and for an environmental-genetic interaction or a purely environmental mechanism for susceptibility to autism. Second, genetic stoppage may be operating to account for the observed birth order effect. That is, if the firstborn child is less severely affected phenotypically, then the parents may be more likely to have children subsequently, whereas decreased rates of childbearing may occur if the firstborn child is severely affected (Jones & Szatmari, 1988). The average age difference between the first and second autistic sibs of about 3 years in this sample (and the 25th and 75th percentiles of 1.6 to 3.75

10 458 Spiker et al. Table VII. Means and Standard Deviations for Nonverbal IQ Scores by Type of Test Firstborns Secondborns Total sample TEST* N Mean SD N Mean SD N Mean SD Wechsler Scales Stanford-Binet (4th Ed.) Stanford-Binet (3rd Ed.) Leiter Scale Merrill-Palmer Scale Other IQ tests VABS Daily Living Scale Developmental Tests SD, standard deviations. *The categories of test include the following tests: Wechsler Scales all versions of the Wechsler intelligence tests; other IQ tests nonverbal subscales of the Slossen Scale, Kaufman ABC, McCarthy Scale, or the Woodcock-Johnson; developmental tests a variety of nonverbal subscales of developmental tests and inventories such as the Bayley Scales, Psychoeducational Profile, Developmental Profile, Brigance Inventory, PEP-R, and Mullen Scales. years) suggests that genetic stoppage may not have been a major factor, because the existing literature indicates that the usual age of diagnosis of autism is after age 3 years (e.g., Sullivan, Kelso, & Stewart, 1990). Most of the children with autism in the current study were born before 1995 and would not have had the benefit of very early diagnosis, which has become more common in the past several years. Parents would generally not have had the diagnosis of the first sib with autism prior to the conception of the second sib with autism. Furthermore, we observed no greater IQ difference between firstborn and secondborn autistic sibs who were born far apart in time (e.g., greater than 3 years apart) than those born near in time (e.g., less than 2years apart). Third, there may be decreased parental attention and/or stimulation for laterborns compared with firstborns. This situation could occur as a byproduct of the protracted and time-consuming process that families endure to obtain a diagnosis, and then to identify and begin treatment with the first affected child. That is, parental stimulation with the secondborn child could be reduced when parents are expending considerable time and attention to confirm a diagnosis and obtain services for the firstborn child. The often-reported finding that females with autism tend to have a greater degree of mental retardation than males (see Lord & Schopler, 1987) was not replicated in our sample of autism multiplex families for nonverbal IQ. The overall average nonverbal IQs, as well as the range of scores and the percentage classified as mentally retarded, were similar for males and females in these multiplex sibships. This was also true whether the females were in same sex or mixed sex sib pairs. Lord & Schopler (1987) reported on males and females from autism singleton families; thus, the difference between the two studies could be accounted for by the phenotypic differences of males and females in multiplex versus singleton autism families. It is also possible that there were differences in ascertainment in these two samples. There are several limitations of the present study that may affect the generalizability of the findings. In the current study, although a relatively large sample of autism multiplex families was studied, it was not a population-based sample but rather a self-referred sample of families interested in participating in a genetic linkage study of autism. Thus, there is no way to know how well this sample represents a population-based sample of autism multiplex families on the critical variables examined. Second, our decision to use a rather strict definition of the affected phenotype for the genetic study led to the exclusion of other families that would be multiplex for a broader autism phenotype (e.g, autism PDD- NOS or Asperger s syndrome sib pairs). The present results may not apply to samples using a broader phenotype, a topic receiving a great deal of recent research attention, but an issue that complicates all genetic investigations of autism (Bailey et al., 1995, 1998; Folstein & Santangelo, 2000; Piven & Palmer, 1999; Piven et al., 1997; Szatmari et al., 1996). Third, the use of a variety of different tests to estimate nonverbal IQ is not an optimal design to test for birth order effects. However, the analyses presented indicate that the observed birth order effect was most likely a function of the lower functioning levels of the secondborn sibs than of the specific tests used to estimate their nonverbal IQs.

11 Birth Order Effects in Autism 459 These findings should be replicated in an independent sample to determine if this is an important potential clue about the underlying etiology of autism. The clinical implications in terms of genetic counseling also must await replication; at this point, families can be told only that the recurrence risk for siblings is 3% to 7% (Simonoff, 1998). Such future replication with another large sample of autism multiplex families should include a broader range of measures of functioning from multiple sources, if possible. The data reported here are most relevant to the suggestion that genetic investigations of autism need to consider possible etiological mechanisms that involve geneticenvironmental interactions. ACKNOWLEDGMENTS We are indebted to the parent support groups and our clinician colleagues who referred families to this project. Our gratitude goes to the families with individuals with autism who are our research partners in this endeavor. This research has been supported by National Institute of Mental Health (NIMH) grants MH52708, MH39437, MH00219, NIH grant HG00348, NHMRC grant , Research Scientist Development Award 1 K01 MH , and grants by the Scottish Rite, the Spunk Fund, Inc., the Rebecca and Solomon Baker Find, the APEX Foundation, the National Alliance for Research in Schizophrenia and Affective Disorders (NARSAD), the endowment fund to the Nancy Pritzker Laboratory (Stanford, California), and gifts from the Autism Society of America, the Janet M. Grace Pervasive Developmental Disorders Fund, and families and friends of individuals with autism. REFERENCES Bailey, A., LeCouteur, A., Bolton, P., Simonoff, E., Yuzda, E., Rutter, M. (1995). Autism as a strongly genetic disorder: Evidence from the British twin study Psychological Medicine, 25, Bailey, A., Palferman, S., Heavey, L., LeCouteur, A. (1998). Autism: The phenotype in relatives. Journal of Autism and Developmental Disorders, 28, Barrett, S., Beck, J. C., Bernier, R., Bisson, E., Braun, T. A., Casavant, T. L., Childress, D., Folstein, S. E., Garcia, M., Gardiner, M. B., Gilman, S., Haines, J. L., Hopkins, K., Landa, R., Meyer, N. H., Mullane, J. A., Nishimura, D. Y., Palmer, P., Piven, J., Purdy, J., Santangelo, S. L., Searby, C., Sheffield, V., Wang, K., & Winkolosky, B. (1999). An autosomal genomic screen for autism. American Journal of Medical Genetics (Neuropsychiatric Genetics), 88, Folstein, S. E., & Santangelo, S. L. (2000). Does Asperger s syndrome aggregate in families? In A. Klin, F. R. Volkmar, & S. S. Sparrow (Eds.), Asperger s syndrome. New York: Guilford Press (pp ). Freeman, B. J., Ritvo, E. R., Mason-Brothers, A., Pingree, C., Yokota, A., Jensen, W. R., McMahon, W. M., Petersen, P. B., Mo, A., & Schroth, P. (1989). Psychometric assessment of firstdegree relatives of 62 autistic probands in Utah. American Journal of Psychiatry, 146, Gualtieri, T., & Hicks, R. E. (1985). Immunoreactive theory of selective male affliction. Behavioral and Brain Sciences, 3, Hallmayer, J., Hebert, J. M., Spiker, D., Lotspeich, L., McMahon, W. M., Petersen, P. B., Nicholas, P. Pingree, C., Lin, A. A., Cavalli-Sforza, L. L., Risch, N., & Ciaranello, R. D. (1996a). Autism and the X chromosome: Multipoint sib pair analysis. Archives of General Psychiatry, 53, Hallmayer, J., Spiker, D., Lotspeich, L., McMahon, W. M., Petersen, P. B., Nicholas, P., Pingree, C., & Ciaranello, R. D. (1996b). Male-to-male transmission in extended pedigrees with multiple cases of autism. American Journal of Medical Genetics: Neuropsychiatric Genetics, 67, Hallmayer, J., Pintado, E., Lin, A., Hebert, J., Lotspeich, L., Spiker, D., McMahon, W. M., Petersen, P. B., Nicholas, P., Pingree, C., Kraemer, H. C., Wong, D., Ritvo, E. R., Ciaranello, R. D., & Cavalli-Sforza, L. L. (1994). Molecular analysis and test for linkage between the FMR-1 gene and infantile autism in multiplex families. American Journal of Human Genetics, 55, International Molecular Genetic Study of Autism Consortium (1998). A full genome screen for autism with evidence of linkage to a region of chromosome 7p. Human Molecular Genetics, 7, Jones, M. B., & Szatmari, P. (1988). Stoppage rules and genetic studies of autism. Journal of Autism and Developmental Disorders, 18, LeCouteur, A., Rutter, M., Lord, C., Rios, P., Robertson, S., Holdgrafer, M., & McLennan, J. (1989). Autism Diagnostic Interview: A standardized investigator-based instrument. Journal of Autism and Developmental Disorders, 19, Lord, C. (1992). Letter to the editor: Birth order effects on nonverbal IQ in families with multiple incidence of autism or pervasive developmental disorder. Journal of Autism and Developomental Disorders, 22, Lord, C., Pickles, A., McLennan, J., Rutter, M., Bregman, J., Folstein, S., Fombonne, E., Leboyer, M., & Minshew, N. (1997). Diagnosing autism: Analyses of data from the Autism Diagnostic Interview. Journal of Autism and Developmental Disorders 27, Lord, C., Rutter, M., & LeCouteur, A. (1994). Autism Diagnostic Interview-Revised: A revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders. Journal of Autism and Developmental Disorders, 24, 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. Journal of Autism and Developmental Disorders, 19, Lord, C., & Schopler, E. (1987). Neurobiological implications of sex difference in autism. In E. Schopler & G. B. Mesibov (Eds.), Neurobiological issues in autism. New York: Plenum Press. Lord, C., & Scholper, E. (1989). Stability of assessment results of autistic and non-autistic language-impaired children from preschool to early school age. Journal of Child Psychology and Psychiatry, 30, MacLean, J. E., Szatmari, P, Jones, M. B., Bryson, S. E., Mahoney, J. J., Bartolucci, G., & Tuff, L. (1999). Familial factors influence level of functioning in pervasive developmental disorders. Journal of the American Academy of Child and Adolescent Psychiatry, 38, Philippe, A., Martinez, M., Guilloud-Bataille, M., Gillberg, C., Rastam, M., Sponheim, E., Coleman, M., Zapella. M., Aschauer, H., van Malldergerme, L., Penet, C., Feingold, J., Brice, A., Leboyer, M, and the Paris Autism Research International Sibpair Study. (1999). Genome-wide scan for autism susceptibility genes. Human Molecular Genetics, 8,

12 460 Spiker et al. Piven, J., & Palmer, P. (1999). Psychiatric disorder and the broad autism phenotype: Evidence from a family study of multipleincidence autism families. American Journal of Psychiatry, 156, Piven, J., Palmer, P., Jacobi, D., Childress, D., & Arndt, A. (1997). Broader autism phenotype: Evidence from a family history study of multiple-incidence families. American Journal of Psychiatry, 154, Risch, R., Spiker, D., Lotspeich, L., Nouri, N., Hinds, D., Hallmayer, J., Kalaydjieva, L., McCague, P., Dimiceli, S., Pitts, T., Nguyen, L., Yang, J., Harper, C., Thorp, D., Vermeer, S., Young, H., Hebert, J., Lin, A., Ferguson, J., Chiotti, C., Wiese-Slater, S., Rogers, T., Salmon, B., Nicholas, P., Petersen, P. B., Pingree, C., McMahon, W., Wong, D. L., Cavalli-Sforza, L. L., Kraemer, H. C., & Myers, R. M. (1999). A genomic screen of autism: Evidence for a multi-locus etiology. American Journal of Human Genetics, 65, Rogers, T., Kalaydjieva, L., Hallmayer, J., Petersen, P. B., Nicholas, P., Pingree, C., McMahon, W., Spiker, D., Lotspeich, L. Kraemer, H. C., McCague, P., Dimiceli, S., Myers, R. M., Nouri, N., Peachy, T., Yang, J., Hinds, D., Risch, N. (1999). Exclusion of linkage to the HLA region in ninety multiplex sibships with autism. Journal of Autism and Developmental Disorders, 29, Salmon, B., Hallmayer, J., Rogers, T., Kalaydjieva, L., Petersen, P. B., Nicholas, P., Pingree, C., McMahon, W., Spiker, D., Lotspeich, L., Kraemer, H. C., McCague, P., Dimiceli, S., Nouri, N., Pitts, T., Yang, J., Hinds, D., Myers, R. M., & Risch, N. (1999). Absence of linkage and linkage disequilibrium to chromosome 15q11-q13 markers in 139 multiplex families with autism. American Journal of Human Genetics: Neuropsychiatric Genetics, 88, Seidman, D. S., Ever-Hadani, P., Stevenson, D. K., Slater, P. E., Harlap, S., & Gale, R. (1988). Birth order and birth weight reexamined. Obstetrics and Gynecology, 72, Simonoff, E. (1998). Genetic counseling in autism and pervasive developmental disorders. Journal of Autism and Developmental Disorders, 28, Singh, V. K., Warren, R. P., Odell, J. P., Warren, W. L., & Cole, P. (1993). Antibodies to myelin basic protein in children with autistic behavior. Brain, Behavior, and Immunology, 7, Smalley, S. L., Asarnow, R. F., & Spence, A. (1988). Autism and genetics: A decade of research. Archives of General Psychiatry, 45, Spiker, D. (1999). The role of genetics in autism. Infants and Young Children, 12, Spiker, D., Lotspeich, L., Kraemer, H. C., Hallmayer, J., McMahon, W. M., Petersen, P. B., Nicholas, P., Pingree, C., Wiese-Slater, S., Chiotti, C., Wong, D. L., Dimiceli, S., Ritvo, E. R., Cavalli- Sforza, L. L., & Ciaranello, R. D. (1994). Genetics of autism: Characteristics of affected and unaffected children from 37 multiplex families. American Journal of Medical Genetics: Neuropsychiatric Genetics, 54, Sullivan, A., Kelso, J., & Stewart, M. A. (1990). Mothers views of the ages of onset for four childhood disorders. Child Psychiatry and Human Development, 20, Szatmari, P., & Jones, M. B. (1991). IQ and genetics of autism. Journal of Child Psychology and Psychiatry, 32, Szatmari, P., Jones, M. B., Holden, J., Bryson, S., Mahoney, W., Tuff, L., MacLean, J., White, B., Bartolucci, G., Schutz, C., Robinson, P., & Hoult, L. (1996). High phenotypic correlations among siblings with autism and pervasive developmental disorders. American Journal of Medical Genetics: Neuropsychiatric Genetics, 67, Szatmari, P., Jones, M. B., Zaigenbaum, L., & MacLean, J. E. (1998). Genetics of autism: overview and new directions. Journal of Autism and Developmental Disorders, 28, Warren, R. P., Singh, V. K., Averett, R. E., Odell, J. D., Maciulis, A., Burger, R. A., Daniels, W. W., & Warren, W. L. (1996). Immunologic studies in autism and related disorders. Molecular Chemical Neuropathology, 28, Wilcox, M. A., Chang, A. M., & Johnson, I. R. (1996). The effects of parity on birth weight using successive pregnancies. Acta Obstetricia et Gynecologica Scandinavica, 75,