Advancing paternal age and simplex autism

Abstract

De novo events appear more common in female and simplex autism spectrum disorder (ASD) cases and may underlie greater ASD risk in older fathers’ offspring. This study examined whether advancing paternal age predicts an increase in simplex (n = 90) versus multiplex ASD cases (n = 587) in 677 participants (340 families). Whether or not controlling for maternal age, results support a significant interaction of linear paternal age and sex of the child on simplex family type. Female ASD cases were significantly more likely to be simplex as paternal age increased, but the increase for males was not significant. Findings suggest that ASD arising from non-familial, de novo events may be far less prominent in males than in females, even if more prevalent in males, due to the substantially larger number of male cases attributable to other, more strongly male-biased risk factors.

Keywords

autism spectrum disorder de novo multiplex paternal age sex differences simplex

Several epidemiological studies suggest an increased risk of autism spectrum disorders (ASDs) in the progeny of older men (Croen et al., 2007; Durkin et al., 2008; Kolevzon, Gross, and Reichenberg, 2007; Lundstrom et al., 2010; Lauritsen, Pedersen and Mortensen, 2005; Reichenberg et al., 2006). In a historical, population-based cohort in Israel, Reichenberg and colleagues (2006) demonstrated a rise in ASD risk with increasing paternal age, such that fathers aged forty or older were over five times more likely to have a child with ASD than fathers under 30. This paternal age effect has not been easily explained by the presence of greater ASD-related traits in older fathers and their spouses (Puleo et al., 2008). Further, several studies suggest that associations between paternal age and ASD risk are distinct from maternal age (Kolevzon et al., 2007; Shelton, Tancredi, and Hertz-Picciotto, 2010), though some studies have not included this covariate (Reichenberg et al., 2006). More recent findings suggest that paternal age may contribute to ASD risk primarily in cases where maternal age is less than 30, augmenting rather than rivaling ASD risk associated with older maternal age (Shelton et al., 2010).

The etiological implications of these paternal age findings remain unclear. Causal gene mutations in male sperm cells (Crow, 1997, 2006), environmental exposure with mutagenic effects (Crow, 2006; Penrose, 1955), increasing use of infertility or assisted reproductive technologies (Wright et al., 2007) or the combination of these mechanisms may all be potential contributors. Such mechanisms may particularly play a role in ASD that appears less likely to be inherited, occurring in families with no prior history of the disorder, with spontaneous mutation hypotheses thus far gaining the most research attention and support (Miles et al., 2005; Sebat et al., 2007; Zhao et al., 2007).

Though a portion of ASD risk likely arises from the confluence of familial, genetic variants with increased penetrance in males and across family members (Szatmari et al., 2007), recent studies suggest that ASD risk may also be conferred via de novo genetic events (i.e. either base pair changes or copy number variations), which occur in a parental germ cell, though they are absent in parental somatic cells. Such events appear more evenly distributed in males and females with ASD (Miles et al., 2005; Sebat et al., 2007; Anello et al., 2009) and less common in youths with a family history of ASD as opposed to a sporadic presentation (Sebat et al., 2007). Though the preponderance of ASD occurs in males (Fombonne, 2003), reduced sex ratios are evident in phenotypically distinct groups of individuals with ASD, such as those with severe mental retardation (Volkmar, Szatmari and Sparrow, 1993) and abnormalities of early morphogenesis (Miles et al., 2005), where causal de novo events are more likely. De novo events may also underlie the novel occurrence of ASD in families without other affected siblings (i.e. simplex families – we will call affected cases from simplex families “simplex cases”; Zhao et al., 2007). In contrast to families where there are multiple incidences of ASD and thus a greater likelihood of genetic transmission from parent to offspring (i.e. multiplex families – we will call each affected case from multiplex families a “multiplex case”), de novo events allow offspring to inherit a disorder from their parents without phenotypic evidence of that disorder in the parents themselves and are often associated with sporadic diseases (Lupski, 2007).

This observed simplex/multiplex distinction may serve as a proxy for the underlying distinction between sporadic and familial cases. Using comparative genomic hybridization to compare simplex, multiplex and unaffected subjects, Sebat and colleagues (2007) detected de novo copy number variations (CNVs) in 10 per cent of individuals with simplex ASD compared to only 3 per cent of individuals with multiplex ASD and 1 per cent of typically developing participants. They also noted a higher proportion of affected females in the sample of cases with identified CNVs. This study and others (Zhao et al., 2007) suggest that simplex classification may indicate increased likelihood of a causative de novo event as opposed to a more traditional familial transmission of ASD risk.

Given that men more often accrue and transmit de novo genetic events as they age (Crow, 1997), de novo events have been hypothesized to underlie the substantial rise in ASD risk found in the offspring of older fathers (Anello et al., 2009; Reichenberg et al., 2006; Croen et al., 2007; Durkin et al., 2008) – a hypothesis supported by the observation of reduced sex ratios in these probands. A diminishing male–female ratio with increasing paternal age has been noted in two large, historical cohorts conducted in Israel and the United States (Reichenberg et al., 2006; Croen et al., 2007) and found significant when systematically tested in a sample of affected children from multiplex families (Anello et al., 2009). Evidence of increasing simplex ASD cases with advancing paternal age would further implicate de novo events; however, epidemiological studies of the paternal age effect have not distinguished between multiplex and simplex cases due to the scarcity of multiplex cases in these samples.

The present study utilized a sample enriched with multiplex families to examine whether increased paternal age is associated with a higher proportion of offspring with simplex as opposed to multiplex ASD in males and females. If indicated, such an effect would lend credence to the hypothesis that de novo events underlie the increased risk of ASD evident in offspring of older men. Though de novo events are less implicated by increased maternal age, older maternity has been repeatedly, independently linked to greater ASD risk (Durkin et al., 2008; Shelton et al., 2010), diminishing the effect of paternal age in some cases (Shelton et al., 2010). Thus maternal age was also considered as an independent predictor of simplex family type. We hypothesized that paternal age would be associated with a higher proportion of simplex ASD in both female and male offspring, suggesting a de novo mode of inheritance less differentiated by sex.

Method

Participants

Families with at least one child meeting criteria for an ASD (specified below) were enrolled in family/genetic or medication studies conducted from 1994 to the present at the Seaver Autism Center for Research and Treatment in The Mount Sinai School of Medicine. Families were referred to us by the Autism Genetic Resource Exchange or via advertising, word of mouth, and physicians in the United States. Caregivers (predominantly mothers) were routinely asked whether their affected child was ever diagnosed with an autism-related medical condition, such as fragile X, tuberous sclerosis, Angelman Syndrome, Prader Willi syndrome, PKU, or Rett syndrome. Children with these conditions were excluded from the study. The majority of families had more than one child with an ASD (described below) because they were recruited for a genetic/family study specifically seeking such participants. However, a smaller sample of families with only one affected child (i.e. simplex families) was also included. Families were excluded from the sample unless there was at least one additional full-sibling of an ASD proband in the family. After a complete description of the study to the subjects, written informed consent was obtained.

Diagnosis of autism spectrum disorders

The Autism Diagnostic Interview–Revised (ADI-R; Lord, Rutter and Le Couteur, 1994) and, after its release, the Autism Diagnostic Observational Schedule–Generic (ADOS-G; Lord et al., 2000) were administered for all children with suspected ASD. Trained interviewers (all demonstrating reliability of 90% or better with either Dr. Catherine Lord’s group or trainers certified by them) administered the ADI-R to the primary caregivers of each affected child. ADI-R information determined ICD-10/DSM-IV autism diagnoses according to ADI-R algorithm. A research diagnosis of Asperger’s disorder was given to those who met criteria for Asperger’s disorder according to DSM-IV, but not autism. Individuals not classified as having autism or Asperger’s disorder, who also failed to meet the ADI-R algorithm criteria for autism by no more than one point in the social domain and no more than one point in either the communication or repetitive behavior domain (but not both) were classified as borderline autism (sometimes called “not quite autism”; International Molecular Genetic Study of Autism Consortium, 1998). Individuals who met ADI onset criteria and marginally sub-threshold criteria in the social domain and displayed marked deficits in both of the other two domains, or marginally sub-threshold deficits in one other domain, qualified for the autism spectrum category. Individuals with notable autism-related deficits in at least one domain as well as the ADI onset criteria qualified for the PDD-NOS category.

Parental age, birth order and family size determination

Each family member’s birth date was collected independently from parents. With this information we determined family size, sibling birth order and maternal and paternal ages when each child with an ASD was born. Paternal and maternal ages were assessed long after enrollment in the study and played no role in ascertainment of families.

Concordance in monozygotic (MZ) and dizygotic (DZ) twins do not have the same implications for the presence of transmitted versus de novo genetic factors. For this reason, each genetically unique (GU) child, rather than each child (which would include both members of an MZ twin pair), was considered the unit of analysis. All GU children affected with an ASD were analyzed; however, families with only one GU child (i.e. either only one offspring or one pair of MZ twins) were excluded from the analyses, given the absence of any family information that would help determine their family type (i.e. either simplex or multiplex ASD). Families were classified simplex if only one GU child in the family was affected and family size, defined as the number of GU children in a family, was two or more. Affected cases were categorized as multiplex if there was at least one additional GU-child who also met criteria for an ASD.

Birth order was determined by the number of successful pregnancies in a given family, not the number of children delivered over the course of those pregnancies. Thus in a family where a mother gave birth to a singleton in her first pregnancy and MZ twins in her second pregnancy, we counted only two GU children and two births, such that the singleton male was the first born and the first-listed twin was selected as second born. As all MZ twins in our database presented with equivalent ASD diagnoses, selection of the first twin was deemed satisfactory. Notably, for the purposes of this study, a family with three children, two of whom were concordant MZ twins with ASD diagnoses (i.e. not GU, so only the first-listed was included), was classified as simplex unless their singleton sibling also met criteria for an ASD. Further, in a family where a mother gave birth to DZ twins and then a singleton child, family size was three for the three GU children, and birth order was considered 1 ½ (average of first and second) for each twin and 3^rd for the singleton.

Analytic plan

An overall logistic regression analysis, with steps defined a priori, was performed for all children. Sex of the child, birth order, family size, and all their interactions were entered as control variables in this analysis. We then tested the linear, pure quadratic, and pure cubic components of the effect of paternal age. Finally, we tested the interactions of these effects of paternal age with sex to see whether boys and girls differed in their paternal age effects. To clarify the interpretation of the interaction between sex and paternal age, we performed separate logistic regression analyses for males and females, with steps defined a priori, entering the three control variables and then testing the linear, pure quadratic, and pure cubic components of paternal age. For these analyses, a mixed model approach was not employed to distinguish between families and then among family members according to paternal age. As simplex families have only one affected case, distinguishing among families at the group level would preclude the inclusion of paternal age, an individual level variable, in the model for that same case.

Several strategies were employed to determine the relationship between maternal age and family type and to assess the utility of including maternal age as a control variable in the present analyses. To examine the relationship between paternal and maternal age, the correlation between these variables as well as the mean, median, and range of maternal age for each quartile of paternal age in males and females were calculated. To determine whether a relationship between maternal age and family type was present, both continuous and dichotomous (i.e. women 35 or older at maternity versus those under 35) maternal age variables were examined as predictors of family type via two separate logistic regressions, each controlling for birth order, family size, and their interaction. Logistic regressions assessing a possible three-way interaction of maternal age, paternal age and sex were also conducted. Finally, all principal analyses were run with and without maternal age to ensure that inclusion of this variable did not alter results.

Results

Descriptive statistics

The sample consisted of 677 individuals with an ASD disorder from 340 families. The diagnostic and demographic characteristics of ASD males (n = 553) and females (n = 124) in the total ASD sample are shown in Table 1. The percentage of multiplex cases was similar for male (86.3 %) and female children (88.7%). Eighty-nine individuals in the sample were twins (13.1%; 62 from DZ twinships and 27 from MZ twinships). In a subsample (n = 261) of cases for whom paternal education data was available, simplex families had higher paternal education than multiplex families (F = 6.10; df = 1, 137; p = .02), though there was no significant main effect of sex or significant interaction of sex and paternal age. Paternal age and paternal education were not significantly correlated (r = −.06, NS)

Table 1

Diagnostic and demographic characteristics of sample

Characteristic	Males		Females		Total
n	553 (82%)^a		124 (18%)^b		677
Age (at assessment) mean years (SD)	8.06	(5.37)	8.71	(6.26)	8.18	(5.55)
ASD disorder
Autism	439 (79%)		93 (75%)		532 (79%)
Asperger’s	20 (4%)		6 (5%)		26 (4%)
Probable autism	37 (7%)		4 (4%)		41 (6%)
Autism spectrum	27 (5%)		8 (6%)		35 (5%)
PDD-NOS	30 (5%)		13 (10%)		43 (6%)
Ethnicity
European ancestry (non-Hispanic)	447 (80%)		98 (79%)		545 (82%)
Hispanic	39 (7%)		12 (10%)		51 (7%)
African ancestry	9 (2%)		1 (1%)		10 (.6%)
Asian/Pacific Islander	3 (1%)		0 (0%)		3 (.4%)
Other/unknown	55 (10%)		13 (11%)		68 (10%)
Family type
Familial	477 (86%)^a		110 (89%)^b		587 (87%)
Simplex	76 (14%)		14 (11%)		90 (73%)
Family size
Familial mean (SD)	2.88	(1.05)	3.15	(1.64)	2.93	(1.19)
Simplex mean (SD)	2.38	(.69)	2.64	(.93)	2.42	(.73)
Birth order
Familial mean (SD)	1.99	(.99)	2.21	(1.42)	2.03	(1.09)
Simplex mean (SD)	1.75	(.64)	1.92	(.94)	1.78	(.73)

Note: ASD = autism spectrum disorder; PDD-NOS = pervasive developmental disorder not otherwise specified; SD = standard deviation.

Selected from 259 families.

Selected from 19 families.

Maternal age

There was a correlation between paternal age and maternal age (r = .61, CI .56 – .66, p < .01), consistent with the correlation noted in several other studies (r = .74, Croen et al., 2007; r = .74, Shelton et al., 2010). Calculations of the mean, median, and range of maternal age by paternal age quartile for males and females revealed considerable variation in maternal age across paternal age quartiles (see Table 2). Notably, over 50 per cent of mothers in the uppermost quartile (Q4 in Table 2) of paternal age for both males and females were under age 35 at the time of their deliveries. With or without the inclusion of paternal age, preliminary analyses revealed no significant relationship between either continuous or dichotomous (i.e. maternity above or below age 35) maternal age variables and family type in either the overall or gender-stratified samples after controlling for birth order, family size and their interaction. Additional preliminary analyses also including paternal age in the model revealed no evidence of significant interactions between maternal age and paternal age, maternal age and sex, or a significant three-way interaction between maternal age, paternal age and child sex. Additionally, inclusion of dichotomous maternal age (chosen over continuous maternal age to maximize power) as a control variable in the principal analyses did not alter the findings; thus, maternal age was not included in the final model.

Table 2

Maternal age by paternal age quartiles for males and females

	Maternal age
Paternal age	Mean	Median	Minimum	Maximum
Males
Q1 (20–29 y/o)	27.7	27.7	17.2	39.6
Q2 (30–33 y/o)	30.7	30.8	21.1	42.6
Q3 (34–36 y/o)	32.3	32.4	21.9	42.9
Q4 (37–62 y/o)	34.8	34.8	24.1	45.6
Females
Q1 (22–30 y/o)	27.9	28.1	20.5	35.5
Q2 (31–33 y/o)	30.1	29.5	22.1	37.9
Q3 (34–37 y/o)	33.1	32.8	27.4	40.6
Q4 (38–52 y/o)	34.8	34.2	27.5	42.1

Note: Q = quartile; y/o = years old.

Paternal age

After controlling for sex, birth order, family size, and all their interactions, there was a significant linear association of family type with paternal age (chi-square = 4.09, df = 1, p = .04) in the overall linear regression. Quadratic and cubic associations between family type and paternal age as well as sex and paternal age were not statistically significant. There was a significant interaction of sex with paternal age as a linear term (χ²= 3.98, df = 1, p = .046; Nagelkerke R²= .08; see Table 3). The significant interaction indicates that the association of paternal age with simplex/multiplex family type differed for males versus females.

Table 3

Logistic regression equation prediction family classification (simplex or multiplex) by paternal age and child sex

Variable	B	SE	χ² Model	Nagelkerke R²
Regression model			3.98 (df = 1)*	.08
Step 1
Birth order	−0.63	0.37
Family size	0.08	0.45
Birth order x Family size	−0.03	0.14
Step 2
Paternal age	0.02	0.02
Child sex	−4.22	2.15
Step 3
Paternal age x Child sex	0.11	0.06

p = .046.

Figure 1 presents the estimated odds of simplex ASD occurrence for males and females as a function of mean quartile paternal age, based on the logistic regression model for the complete sample including the paternal age by sex interaction term. Comparing female offspring of fathers in the 4^th to 1^st quartiles of paternal age, the ratio in odds of simplex to multiplex ASD occurrence was 6.69. In other words, an ASD female whose father was in the highest quartile was over six times more likely to be a simplex than a multiplex case than one whose father was in the lowest quartile. In contrast, an odds ratio of only 1.44 was observed in male offspring. It should be noted that in Figure 1 the effects of family size, birth order and their interaction were fitted separately for males and females, corresponding to the inclusion of sex interacting with family characteristics in the overall logistic regression analysis.

Figure 1

The odds of an individual’s autism spectrum disorder occurring in isolation (i.e. being a simplex case) estimated at the mean quartile ages (Q1 = 27.5, Q2 = 31.9, Q3 = 35.1, Q4 = 41.1) of paternity for each sex

Figure 2 shows the cumulative distributions of paternal age for the four groups: multiplex males, simplex males, multiplex females, and simplex females. This unadjusted analysis shows that the interaction (i.e. among females, simplex offspring are more likely than multiplex offspring to have older fathers at the time of their birth), is not due to a few outliers in a small sample. It is noteworthy, and reflective of the overall finding, that all four groups have similar distributions; however, the distribution of simplex females is shifted to the right. Though the distribution of simplex females occupies a different location, its similar shape negates the role of influential outliers in these findings.

Figure 2

Cumulative distributions of paternal age for the four groups: multiplex males, simplex males, multiplex females, and simplex females

To better interpret the interaction of sex with paternal age interaction, we performed separate logistic regression analyses of the association of family type with paternal age for males and females. Females had a significant linear association of paternal age with family type (χ²= 7.06, df = 1, p = .01), with greater odds of being from a simplex family occurring as paternal age increased after controlling for birth order, family size and their interaction. In contrast, for male cases, as paternal age increased, the odds of being from a simplex family were nominally higher but not statistically significant (χ²= .92, df = 1, p = .34), suggesting that the overall association of paternal age with family type for both males and females was attributable primarily to the association among females. Since the sample size was 4.46 times larger for males than females, this difference in significance was not attributable to lower power to detect an effect in males. The discrepancy between the respective effect sizes (calculated from chi-square tests; Cohen, 1988; w = 0.24 for females and w = 0.04 for males) was statistically significant (z = 1.99, p < .05).

The comparison of male and female odds in Figure 1 does not reflect the discrepancy in quartile sample sizes, with 4.46 as many males as females. Notably, comparison of the estimated number of simplex ASD cases for each sex separately as a function of mean quartile paternal age suggests that males and females had similar increases in the number of simplex cases for the successive paternal age quartiles, despite the larger number of affected males.

Discussion

The interrelationships of advanced paternal age, simplex ASD, and ASD risk in males and females require clarification. Findings suggest a linear rise in the odds of female, simplex ASD with increasing paternal age in a sample enriched with multiplex families, the infrequency of which have prevented such investigations in epidemiological studies. Though the findings should be considered preliminary given the small number of female simplex cases in the sample, this rise is consistent with an increase in de novo event transmission as men age – a phenomenon Reichenberg et al. (2006) suggested may explain why paternal age is a risk factor for ASD. The proposed relevance of this mechanism to both sexes may also suggest why a reduced sex ratio in the offspring of older fathers has been described in two large cohorts, one population based (Reichenberg et al., 2006) and one hospital-based (Croen et al., 2007).

In the present study, the rise in the odds of simplex ASD with increasing paternal age did not reach significance in males. This does not necessarily suggest that de novo mechanisms are less apparent in males, however. Rather, the present study provides further evidence that the balance of mechanisms underlying ASD in females is substantially different from that in males. ASD is approximately four times more common in boys than girls (Fombonne, 2003), a discrepancy that may, in part, be explained by the greater susceptibility of males to certain ASD risk factors, such as fetal testosterone exposure (Knickmeyer and Baron-Cohen, 2006) or genetic factors involving the transmission of ASD liability from one generation to the next. In their efforts to subtype children with autism, Miles and colleagues (2005) observed that ASD children without abnormalities of early morphogenesis, or those without clear ASD-related insults to early embryiologic development, were twice as likely to be male (5.6:1 versus 3.2:1), to have a family history of autism (20% vs. 9%) and to have siblings with autistic traits (12% vs. 6%), suggesting that autism arising from heritable causes is more prominent in males than females. Consistently, in several molecular genetic studies, evidence for ASD-related genes has depended on excluding families with cases of female ASD (Schellenberg et al., 2006; Ma et al., 2007). ASD arising from non-familial, de novo events may be more similarly prevalent in male and female cases (Miles et al., 2005) and yet be far less prominent in males than females due to the substantially larger number of male cases attributable to these other, male-biased risk factors. In the current sample of females, where the base rates of male-biased mechanisms are presumably lower, the potential contributions of de novo events were striking when compared to a male sample. There, a similar increase in the number of simplex ASD cases with increasing paternal age was relatively small and insignificant (see Figure 1). It follows from this interpretation of the findings, that the effect of paternal age may be similar for males and females, as reported in a prior epidemiological study (Shelton et al., 2010), and simultaneously more influential in females, where ASDs attributable to other risk factors are less common.

Associations between maternal age and simplex family type were not evident in the present study, and maternal and paternal age were not as highly correlated as in some prior research (Croen et al., 2007; Shelton et al., 2010). Given that paternal age may confer greater ASD risk when maternal age is below age 30 (Shelton et al., 2010), this lower correlation may reflect the particular contributions of paternal age to ASD risk in this sample. Though both advanced paternal and maternal age have been independently associated with increased ASD risk (Kolevzon et al., 2007; Shelton et al., 2010), the risk mechanisms conferred by age may differ for mothers and fathers. While de novo events arise, replicate, and thus accumulate in the paternal germline overtime (Crow, 1997), eggs are fully formed at the time of a mother’s birth. Thus the risks posed by advanced maternal age might well arise from alternative mechanisms that may or may not be related to simplex family type, such as increased obstetric complications (Kolevzon et al., 2007), increased autoantibody production (Larbi, Fulop and Pawelec, 2008), accumulating environmental exposures (Hertz-Picciotto et al., 2006), and epigenetic changes (Shelton et al., 2010). Though these mechanisms have been hypothesized to underlie maternal age risk, it remains unclear whether such complications confer ASD risk independent of genetic vulnerability, compound, confound or co-vary with it (i.e. both obstetric complications and ASD risk arise from the same genetic information).

This study had limitations. Female simplex cases, though sufficient to demonstrate a significant association with increasing paternal age, were few (n = 14). Several precautions were taken to offset potential risks associated with small group size and possible misclassfication, including the incorporation of potential confounds, family size and birth order, as well as careful examination of the raw data for potential outliers. Our multiplex/simplex distinction is meant to serve as a proxy for familial versus sporadic ASD (i.e. ASD resulting from a de novo event), but it is plausible that some truly familial ASD cases were classified as simplex in families where there were too few offspring to express the familiality of the disorder. As a precaution against such misclassification, cases were categorized as simplex only if they had at least one unaffected sibling. The effect of misclassification of truly multiplex cases as simplex would be to dilute the distinctions among truly simplex cases by inclusion of misclassified multiplex cases – reducing the strength of the observed results. Further, we have no reason to suspect that such misclassification is systematic (i.e. more frequent in one sex than the other). Thus, these findings may be conservative. The alternative misclassification of a truly sporadic family as multiplex seems less probable, as it requires two truly sporadic cases to occur in the same family.

Our sample was one of convenience, ascertained opportunistically and predominated by families with two or more affected children. As a result, the sample cannot be deemed representative of the ASD population as a whole and paternal age was not assessed as a risk factor for ASD overall. Nonetheless, the proportions of simplex and multiplex families do not essentially affect the comparison between them. In Figure 1, the objective is not to evaluate the odds of simplex to multiplex, but rather the relative odds for comparisons within the figure: males and females for each paternal age, and paternal ages within males or females. Even if all the odds are greatly affected by oversampling, the relative odds for subsets are not likely to be strongly affected. Further, as Figure 2 compares males and females within multiplex or within simplex, oversampling should not affect the cumulative distributions.

Though neither the sex of the affected children nor the age of their fathers played any active role in study recruitment, the sample may have reflected volunteer bias, affecting its demographic characteristics and family type. As the male to female ratios in simplex and multiplex cases were not significantly different and compared to ratios reported in the ASD population (Fombonne, 2003), our findings are likely not attributable to the oversampling of simplex females. Missing paternal education data in over half the sample precluded its use as a control variable. However, where data was available, simplex families had higher paternal education than multiplex families. Though paternal age and education were not correlated in the present study, paternal education may yet be an important variable to include in future analyses. Men with lengthier academic histories may in turn have children at later ages, resulting in an increased risk of ASD in their offspring, a hypothesis that may warrant future consideration in epidemiological studies.

In view of the small size (n = 14) of the simplex female group, the findings of the present study should be considered preliminary. A larger sample is needed to replicate and directly test the hypotheses discussed and supported herein. Rates of causative de novo events in the affected offspring of older versus younger fathers should be measured and compared between sexes to assess whether such mechanisms are in fact more “gender egalitarian” as we have proposed. Our findings are consistent with but not specific to de novo event hypotheses and may represent a variety of non-familial genetic, epigenetic or environmental risk factors for ASD that accumulate as men age. Other possible explanations, such as low fecundity, infertility treatment, dosage compensation and X-imprinting theories (Skuse, 2000; Brooks, 2005; Olsen and Zhu, 2009) should also be investigated and tested.

Although the precise mechanisms underlying increased ASD risk in the children of older fathers remain unclear, the preliminary findings of this study complement and add to accumulating evidence of etiological divergence related to advanced paternal age in subsets of individuals with ASD. Though sporadic susceptibility to ASD may be most evident in simplex cases, previous genetic susceptibility to ASD does not preclude the occurrence of a de novo event. As such, noted increases in female ASD risk with advancing paternal age in multiplex families (Anello et al., 2009) may arise from the same sporadic process implied by simplex families in this study. A family with two affected males may have a third affected daughter whose ASD risk is distinct from her siblings’. What remains unclear then – and a question for future research – is the impact of these de novo processes on changes in ASD susceptibility and female penetrance in the general population. In an epidemiological sample, increases in simplex ASD due to paternal age may be measured, rather than estimated. Future research is needed to develop a comprehensive understanding of the interplay and overlap of both novel and familial genetic risk factors underlying ASD.

References

Anello

Reichenberg

Luo

Schmeidler

Hollander

. (2009) Brief Report: Parental Age and the Sex Ratio in Autism. Journal of Autism and Developmental Disorders 39: 1487–1492.

Brooks

(2005) Autoimmune Disorders Result from Loss of Epigenetic Control Following Chromosome Damage. Medical Hypotheses 64: 590–598.

Cohen

(1988) Statistical Power Analysis for the Behavioral Sciences, 2nd edn. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

Croen

Najjar

Fireman

Grether

(2007) Maternal and Paternal Age and Risk of Autism Spectrum Disorders. Archives of Pediatric and Adolescent Medicine 16: 334–340.

Crow

(1997) The High Spontaneous Mutation Rate: Is It a Health Risk? Proceedings of the National Academy of Sciences U.S.A. 94: 8380–8386.

Crow

(2006) Age and Sex Effects on New Mutation Rates: An Old Problem with New Complexities. Journal of Radiation Research (Tokyo) 47: B75–B82.

Durkin

Maenner

Newschaffer

Lee

Cunniff

. (2008) Advanced Parental Age and the Risk of Autism Spectrum Disorder. American Journal of Epidemiology 168: 1268–1276.

Fombonne

(2003) Epidemiological Surveys of Autism and Other Pervasive Developmental Disorders: An Update. Journal of Autism Developmental Disorders 33: 365–382.

Hertz-Picciotto

Croen

Hansen

Jones

van de Water

Pessah

(2006) The CHARGE Study: An Epidemiologic Investigation of Genetic and Environmental Factors Contributing to Autism. Environmental Health Perspectives 114: 1119–1125.

10.

International Molecular Genetic Study of Autism Consortium (1998) A Full Genome Screen for Autism with Evidence for Linkage to a Region on Chromosome 7q. Human Molecular Genetics 7: 571–578.

11.

Knickmeyer

Baron-Cohen

(2006) Fetal Testosterone and Sex Differences in Typical Social Development and in Autism. Journal of Child Neurology 21: 825–845.

12.

Kolevzon

Gross

Reichenberg

(2007) Prenatal and Perinatal Risk Factors for Autism: A Review and Integration of Findings. Archives of Pediatric and Adolescent Medicine 161: 326–333.

13.

Larbi

Fulop

Pawelec

(2008) Immune Receptor Signaling, Aging, and Autoimmunity. Advances in Experimental Medicine and Biology 640: 312–324.

14.

Lauritsen

Pedersen

Mortensen

(2005) Effects of Familial Risk Factors and Place of Birth on the Risk of Autism: A Nationwide Register-Based Study. Journal of Child Psychology and Psychiatry 46: 963–971.

15.

Lord

Risi

Lambrecht

Cook

Leventhal

. (2000) The Autism Diagnostic Observation Schedule–Generic: A Standard Measure of Social and Communication Deficits Associated with the Spectrum of Autism. Journal of Autism and Developmental Disorders 30: 205–223.

16.

Lord

Rutter

Le Couteur

(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: 659–685.

17.

Lundstrom

Haworth

Carlstrom

Gillberg

Mill

. (2010) Trajectories Leading to Autism Spectrum Disorders Are Affected by Paternal Age: Findings from Two Nationally Representative Twin Studies. Journal of Child Psychology and Psychaitry 51: 850–856.

18.

Lupski

(2007) Genomic Rearrangements and Sporadic Disease. Nature Genetics 39: S43–S47.

19.

Cuccaro

Jaworski

Haynes

Stephan

. (2007) Dissecting the Locus Heterogeneity of Autism: Significant Linkage to Chromosome 12q14. Molecular Psychiatry 12: 376–384.

20.

Miles

Takahashi

Bagby

Sahota

Vaslow

. (2005) Essential versus Complex Autism: Definition of Fundamental Prognostic Subtypes. American Journal of Medical Genetics A 135: 171–180.

21.

Olsen

Zhu

(2009) RE: Advanced Paternal Age and the Risk of Autism Spectrum Disorders. American Journal of Epidemiology 169: 1406–1406.

22.

Penrose

(1955) Parental Age and Mutation. Lancet 269: 312–313.

23.

Puleo

Reichenberg

Smith

Kryzak

Silverman

(2008) Do Autism-Related Personality Traits Explain Higher Paternal Age in Autism? Molecular Psychiatry 13: 243–244.

24.

Reichenberg

Gross

Weiser

Bresnahan

Silverman

. (2006) Advancing Paternal Age and Autism. Archive of General Psychiatry 63: 1026–1032.

25.

Schellenberg

Dawson

Sung

Estes

Munson

. (2006) Evidence for Multiple Loci from a Genome Scan of Autism Kindreds. Molecular Psychiatry 11: 1049–1060.

26.

Sebat

Lakshmi

Malhotra

Troge

Lese-Martin

. (2007) Strong Association of De Novo Copy Number Mutations with Autism. Science 316: 445–449.

27.

Shelton

Tancredi

Hertz-Picciotto

(2010) Independent and Dependent Contributions of Advanced Maternal and Paternal Ages to Autism Risk. Autism Research 3: 30–39.

28.

Skuse

(2000) Imprinting, the X-Chromosome, and the Male Brain: Explaining Sex Differences in the Liability to Autism. Pediatric Research 47: 9–16.

29.

Szatmari

Paterson

Zwaigenbaum

Roberts

Brian

. (2007) Mapping Autism Risk Loci Using Genetic Linkage and Chromosomal Rearrangements. Nature Genetics 39: 319–328.

30.

Volkmar

Szatmari

Sparrow

(1993) Sex Differences in Pervasive Developmental Disorders. Journal of Autism and Developmental Disorders 23: 579–591.

31.

Wright

Chang

Jeng

Chen

Macaluso

(2007) Assisted Reproductive Technology Surveillance – United States, 2004. MMWR Surveillance Summary 56: 1–22.

32.

Zhao

Leotta

Kustanovich

Lajonchere

Geschwind

. (2007) A Unified Genetic Theory for Sporadic and Inherited Autism. Proceedings of the National Academy of Sciences U.S.A. 104: 12831–12836.