Abstract

Previous studies are inconsistent regarding whether there are independent effects of maternal and paternal age on the risk of autism. Different biologic mechanisms are suggested by maternal and paternal age effects. The study population included all California singletons born in 1989–2002 (n = 7,550,026). Children with autism (n = 23,311) were identified through the California Department of Developmental Services and compared with the remainder of the study population, with parental ages and covariates obtained from birth certificates. Adjusted odds ratios and 95% confidence intervals were used to evaluate the risk of autism associated with increasing maternal and paternal age. In adjusted models that included age of the other parent and demographic covariates, a 10-year increase in maternal age was associated with a 38% increase in the odds ratio for autism (odds ratio = 1.38, 95% confidence interval: 1.32, 1.44), and a 10-year increase in paternal age was associated with a 22% increase (odds ratio = 1.22, 95% confidence interval: 1.18, 1.26). Maternal and paternal age effects were seen in subgroups defined by race/ethnicity and other covariates and were of greater magnitude among first-born compared with later-born children. Further studies are needed to help clarify the biologic mechanisms involved in the independent association of autism risk with increasing maternal and paternal age.

Previous investigations have sought to determine whether the risk of autism in offspring increases with advancing maternal and/or paternal age and to estimate the magnitude of association (1–18). Although no studies have reported decreased risk with increasing parental age, early studies yielded inconsistent findings (1–11), as have more recent studies reporting maternal effects only (12), paternal effects only (13–16), or both maternal and paternal effects (17, 18). Differences across studies may, in part, be due to considerations of sample size and study design, but they may also reflect true differences across populations. Because age-related reproductive mechanisms differ between men and women, disentangling the interrelations between risk of autism and age of mothers and fathers is likely to be helpful in the search for etiologic factors.

We sought to further clarify the presence and magnitude of maternal and paternal age effects associated with autism risk in a very large and diverse North American population.

MATERIALS AND METHODS

Subjects

The study population included 7,550,026 singletons born in California between January 1, 1989, and December 31, 2002, to mothers residing in the state (fetal and infant deaths excluded). Children with autism were identified through electronic files of the California Department of Developmental Services (DDS) that operates a statewide system of regional centers and developmental centers that coordinate services for persons with autism, mental retardation, and other developmental disabilities. All remaining births in the study population constituted the comparison group for analyses.

DDS eligibility is determined and services are provided without regard to citizenship or financial status. Although the system is widely utilized across different socioeconomic levels and racial/ethnic groups, disparities in utilization may exist. A Client Development Evaluation Report is completed when an individual has met the eligibility requirements for active status in the DDS system and is updated periodically thereafter. Individuals with more than one qualifying condition may have one or multiple conditions coded on the Client Development Evaluation Report, depending on local practices. Autism/autistic disorder (Diagnostic and Statistical Manual of Mental Disorders (DSM), Third Edition-Revised, prior to 1994; DSM, Fourth Edition, starting in 1994) is a DDS-qualifying condition; individuals with Asperger's or pervasive developmental disorder, not otherwise specified, may not meet eligibility requirements for “substantial handicap.” Following DDS eligibility practice, we considered a child to have autism if the child was reported as being eligible for services based on autism (DDS codes 1 or 2) on at least one Client Development Evaluation Report prior to April 2006, irrespective of other qualifying conditions or diagnoses.

To identify DDS clients who were California resident births, we linked DDS electronic data to livebirth certificates using a series of matching algorithms based on personal identifiers. By use of predetermined probabilistic criteria, the algorithms classified the resulting matched records as “sure matches” or “possible matches.” Children whose records were not matched to a birth certificate were assumed to be out-of-state births and excluded from further consideration. Identifying variables for “possible matches” were manually reviewed for reclassification of the record as a “sure match” or “nonmatch” on the basis of obvious discrepancies of name spelling, hyphenation, and so on. Only children whose records were considered “sure matches” in either step were included in analyses. On the basis of a field validation study of DDS clients (n = 39,969) born in 1987–2000, we estimate that less than 0.2% of all records finally classified as “sure matches” may be inaccurately matched using these procedures.

Maternal and paternal age

Maternal and paternal ages at the time of the child's birth were obtained from birth certificates. Reported maternal age ranged from 7 to 96 years and paternal age from 0 to 97 years. To exclude improbable ages, we limited analyses to maternal age 15–44 years and paternal age 15–64 years. This restriction resulted in exclusion of maternal records for those less than 15 years for 12 cases (0.05%) and 17,571 noncases (0.23%), maternal records for those over 44 years for 51 cases (0.22%) and 8,220 noncases (0.11%), paternal records for those less than 15 years for 3 cases (0.01%) and 1,398 noncases (0.02%), and paternal records for those over 64 years for 11 cases (0.05%) and 2,340 noncases (0.03%). In addition, maternal age was missing for 5 cases (0.02%) and 1,678 noncases (0.02%) and paternal age for 1,225 cases (5.3%) and 522,982 noncases (7.0%).

Covariates

The covariates examined in adjusted models were those available from birth certificates and classified as follows: gender (male, female); year of birth (indicator variables for each year); maternal and paternal education (less than high school, high school graduate, some college, postgraduate); maternal race/ethnicity (white non-Hispanic, white Hispanic-US born, white Hispanic-not US born, black, Asian-US born, Asian-not US born, other); paternal race/ethnicity (white non-Hispanic, white Hispanic, black, Asian, other); birth weight (<1,500, 1,500–2,499, ≥2,500 g); gestational age (<32, 32–36, ≥37 weeks); parity (0, 1, ≥2); and, as a rough proxy for socioeconomic status, source of payment for delivery (private insurance (higher status) vs. self-insured/government/other).

Records with missing data or “improbable” maternal or paternal age were excluded from both unadjusted and adjusted statistical models (11.2% of cases and 13.6% of noncases).

Statistical analysis

We used logistic regression to investigate the association between autism and increasing maternal age and paternal age in cases compared with the remainder of the study population. First, we calculated the unadjusted odds ratios and 95% confidence intervals separately for maternal and paternal age, using 5-year age categories (25–29 years as the reference group). We then recalculated the odds ratios and confidence intervals in models adjusted only for the age of the other parent and in separate models adjusted for all covariates (in this large population, all covariates were significantly associated with autism, and all but child's sex were associated with both maternal and paternal age).

To determine if maternal or paternal age were linearly associated with the log(odds ratio) of autism, and, if so, over what age ranges, we performed assessments of age as a continuous versus a categorized variable, using likelihood ratio tests. We also assessed whether the association between autism and 10-year increments in maternal or paternal age persisted among subgroups defined by the covariates; we ran logistic models for the association between the risk of autism and parental age separately for each stratum of each covariate (adjusted for all other covariates). We then evaluated whether the magnitude of a maternal or paternal age effect was similar across the strata of each covariate by testing the equality of the log(odds ratio) with a chi-square test of homogeneity (19).

This study was conducted with approval from the California Committee for the Protection of Human Subjects.

RESULTS

From this study population of over 7.5 million singletons, we identified 23,311 children with DDS-reported autism, yielding a prevalence of autism of 3.1/1,000 for the total time period. Children with autism were more likely to be male than female (odds ratio (OR) = 4.6, 95% confidence interval (CI): 4.4, 4.8). After exclusion of observations with missing data and improbable parental ages, the analytical file included 20,701 singletons with DDS-reported autism and 6,506,555 singletons without autism. The median age at delivery for mothers at the beginning of the study period in 1989 was 27 years and increased to 28 years by 2002. For fathers, the median age at delivery was 29 years in 1989 and 31 years in 2002.

Maternal and paternal age as categorical variables

When parental ages were treated as 5-year categorical variables in unadjusted models, the odds ratio for autism increased with each increasing maternal age category over the age span 15–44 years and with each increasing paternal age category over the age span 15–64 years, except for the 55–59 age group (Table 1). When considered in multivariate models adjusted for only the other parent's age, the magnitude of the associations with increasing maternal age, and, independently, increasing paternal age, was considerably attenuated but remained statistically significant (Table 1). Adjustment for additional covariates resulted in minimal further attenuation of the estimates of association; for both maternal and paternal age, the odds ratios continued to be significantly different from the reference group (Table 1), with reduced risk for young parents and increased risk for older parents.

Table 1.

Risk of Department of Developmental Services-reported Autism and Maternal and Paternal Age (Categorical), California Resident Births, 1989–2002

 Singletons Without Autisma (n =  6,506,555)
 
Singletons With Autisma (n = 20,701)
 
Crude Odds Ratioa 95% Confidence Interval Adjusted Odds Ratioa,b 95% Confidence Interval Adjusted Odds Ratioa,c 95% Confidence Interval 
 No. No. 
Maternal age, years           
    15–19 651,054 10.0 960 4.6 0.49 0.46, 0.52 0.62 0.57, 0.67 0.65 0.59, 0.70 
    20–24 1,556,113 23.9 3,531 17.1 0.75 0.72, 0.79 0.84 0.81, 0.88 0.86 0.82, 0.90 
    25–29 1,852,900 28.5 5,581 27.0 1.00 Referent 1.00 Referent 1.00 Referent 
    30–34 1,558,709 24.0 6,144 29.7 1.31 1.26, 1.36 1.18 1.14, 1.23 1.14 1.10, 1.19 
    35–39 738,308 11.3 3,659 17.7 1.64 1.58, 1.72 1.37 1.31, 1.44 1.33 1.27, 1.40 
    40–44 149,471 2.3 826 4.0 1.84 1.70, 1.97 1.44 1.33, 1.56 1.43 1.32, 1.55 
Paternal age, years           
    15–19 280,837 4.3 385 1.9 0.52 0.47, 0.58 0.74 0.66, 0.83 0.76 0.67, 0.85 
    20–24 1,208,883 18.6 2,416 11.7 0.76 0.72, 0.80 0.89 0.84, 0.94 0.89 0.84, 0.94 
    25–29 1,717,336 26.4 4,534 21.9 1.00 Referent 1.00 Referent 1.00 Referent 
    30–34 1,690,238 26.0 5,944 28.7 1.33 1.28, 1.38 1.17 1.12, 1.22 1.12 1.07, 1.17 
    35–39 1,028,251 15.8 4,432 21.4 1.63 1.57, 1.70 1.31 1.25, 1.37 1.23 1.17, 1.30 
    40–44 408,114 6.3 2,063 10.0 1.92 1.82, 2.02 1.46 1.38, 1.55 1.39 1.30, 1.47 
    45–49 123,373 1.9 649 3.1 1.99 1.84, 2.16 1.50 1.38, 1.64 1.41 1.29, 1.54 
    50–54 35,083 0.5 201 1.0 2.17 1.88, 2.50 1.64 1.42, 1.90 1.53 1.32, 1.77 
    55–59 10,799 0.2 52 0.3 1.82 1.39, 2.40 1.39 1.06, 1.83 1.35 1.02, 1.77 
    60–64 3,641 0.1 25 0.1 2.61 1.76, 3.86 2.00 1.35, 2.97 2.05 1.38, 3.05 
 Singletons Without Autisma (n =  6,506,555)
 
Singletons With Autisma (n = 20,701)
 
Crude Odds Ratioa 95% Confidence Interval Adjusted Odds Ratioa,b 95% Confidence Interval Adjusted Odds Ratioa,c 95% Confidence Interval 
 No. No. 
Maternal age, years           
    15–19 651,054 10.0 960 4.6 0.49 0.46, 0.52 0.62 0.57, 0.67 0.65 0.59, 0.70 
    20–24 1,556,113 23.9 3,531 17.1 0.75 0.72, 0.79 0.84 0.81, 0.88 0.86 0.82, 0.90 
    25–29 1,852,900 28.5 5,581 27.0 1.00 Referent 1.00 Referent 1.00 Referent 
    30–34 1,558,709 24.0 6,144 29.7 1.31 1.26, 1.36 1.18 1.14, 1.23 1.14 1.10, 1.19 
    35–39 738,308 11.3 3,659 17.7 1.64 1.58, 1.72 1.37 1.31, 1.44 1.33 1.27, 1.40 
    40–44 149,471 2.3 826 4.0 1.84 1.70, 1.97 1.44 1.33, 1.56 1.43 1.32, 1.55 
Paternal age, years           
    15–19 280,837 4.3 385 1.9 0.52 0.47, 0.58 0.74 0.66, 0.83 0.76 0.67, 0.85 
    20–24 1,208,883 18.6 2,416 11.7 0.76 0.72, 0.80 0.89 0.84, 0.94 0.89 0.84, 0.94 
    25–29 1,717,336 26.4 4,534 21.9 1.00 Referent 1.00 Referent 1.00 Referent 
    30–34 1,690,238 26.0 5,944 28.7 1.33 1.28, 1.38 1.17 1.12, 1.22 1.12 1.07, 1.17 
    35–39 1,028,251 15.8 4,432 21.4 1.63 1.57, 1.70 1.31 1.25, 1.37 1.23 1.17, 1.30 
    40–44 408,114 6.3 2,063 10.0 1.92 1.82, 2.02 1.46 1.38, 1.55 1.39 1.30, 1.47 
    45–49 123,373 1.9 649 3.1 1.99 1.84, 2.16 1.50 1.38, 1.64 1.41 1.29, 1.54 
    50–54 35,083 0.5 201 1.0 2.17 1.88, 2.50 1.64 1.42, 1.90 1.53 1.32, 1.77 
    55–59 10,799 0.2 52 0.3 1.82 1.39, 2.40 1.39 1.06, 1.83 1.35 1.02, 1.77 
    60–64 3,641 0.1 25 0.1 2.61 1.76, 3.86 2.00 1.35, 2.97 2.05 1.38, 3.05 
a

Excluded are observations with missing values for child's sex and birth weight, maternal age, maternal race/ethnicity, paternal race/ethnicity, maternal education, paternal education, parity, gestational age, delivery method of payment, and birth year.

b

Adjusted for age of other parent only.

c

Adjusted for child's sex and birth weight, maternal age, paternal age, maternal race/ethnicity, paternal race/ethnicity, maternal education, paternal education, parity, gestational age, delivery method of payment, and birth year.

Maternal and paternal age as continuous variables

Figures 1 and 2 present the log(odds ratio) for autism plotted against maternal and paternal ages based on fully adjusted multivariate models. In the figures, the line represents the log(odds ratio) when parental age was modeled as a continuous variable, and the dots represent the log(odds ratio) at each year of age when parental age was represented as indicator variables. For maternal age, a linear trend was observed within maternal ages 20–39 years (86.4% of children) but not ages 15–19 years (11.4% of children) or ages 40–44 years (2.3% of children) (Figure 1). Paternal age demonstrated a similar pattern, exhibiting a linear trend for ages 20–59 years (95.4% of children) but not for ages 15–19 years (4.5% of children) or for ages 60–64 years (0.1% of children) (Figure 2).

Figure 1.

Maternal age and risk of autism, California resident births, 1989–2002. Children with autism are clients of the California Department of Developmental Services; the denominator population is all remaining resident livebirths alive at 1 year of age. The line (—) represents the log(odds ratio) when parental age was modeled as a continuous variable; the dots (•) represent the log(odds ratio) at each year of age when parental age was represented as indicator variables.

Figure 1.

Maternal age and risk of autism, California resident births, 1989–2002. Children with autism are clients of the California Department of Developmental Services; the denominator population is all remaining resident livebirths alive at 1 year of age. The line (—) represents the log(odds ratio) when parental age was modeled as a continuous variable; the dots (•) represent the log(odds ratio) at each year of age when parental age was represented as indicator variables.

Figure 2.

Paternal age and risk of autism, California resident births, 1989–2002. Children with autism are clients of the California Department of Developmental Services; the denominator population is all remaining resident livebirths alive at 1 year of age. The line (—) represents the log(odds ratio) when parental age was modeled as a continuous variable; the dots (•) represent the log(odds ratio) at each year of age when parental age was represented as indicator variables.

Figure 2.

Paternal age and risk of autism, California resident births, 1989–2002. Children with autism are clients of the California Department of Developmental Services; the denominator population is all remaining resident livebirths alive at 1 year of age. The line (—) represents the log(odds ratio) when parental age was modeled as a continuous variable; the dots (•) represent the log(odds ratio) at each year of age when parental age was represented as indicator variables.

In adjusted models with maternal age restricted to 20–39 years (and paternal age restricted to 20–59 years), the odds ratio for autism associated with each 1-year increase in maternal age was 1.03; for each 10-year increment in maternal age, the risk increased by 38% (ORadjusted = 1.38, 95% CI: 1.32, 1.44). For paternal age 20–59 years (with maternal age 20–39 years), the odds ratio for autism associated with each 1-year increase in age was 1.02, and for each 10-year increment in paternal age, the odds ratio for autism increased by 22% (ORadjusted = 1.22, 95% CI: 1.18, 1.26). The results were similar when the age ranges were expanded to include maternal ages 15–44 years (ORadjusted = 1.39, 95% CI: 1.34, 1.44) and paternal ages 15–64 years (ORadjusted = 1.22, 95% CI: 1.18, 1.25), despite deviations from linearity.

Subgroup analyses

In subgroup analyses in which maternal age was restricted to 20–39 years and paternal age to 20–59 years, the odds ratios for autism associated with a 10-year increase in both maternal age and paternal age remained statistically significant within all subgroups defined by the covariates (Table 2). The magnitude of autism risk associated with 10-year increments in parental age was statistically similar (P > 0.05) for boys compared with girls and for subgroups of gestational age, birth weight, maternal race, maternal education, and paternal race. The risk of autism associated with paternal age showed some variation across subgroups of paternal education (P  = 0.009), without a clear pattern. The risk of autism associated with maternal age was of somewhat greater magnitude for children whose delivery was not paid for through private insurance compared with those with private insurance (P = 0.016) and across birth year categories (P = 0.032).

Table 2.

Risk of Department of Developmental Services-reported Autism Associated With 10-Year Increase in Maternal and Paternal Ages for Demographically Defined Subgroups, California Resident Births, 1989–2002a

Subgroups No. of Cases No. of Noncases Maternal Age, 20–39 Years
 
P Valueb Paternal Age, 20–59 Years
 
P Valueb 
Odds Ratio 95% Confidence Interval Odds Ratio 95% Confidence Interval 
Child's sex     0.298   0.421 
    Male 15,573 2,882,025 1.37 1.31, 1.43  1.21 1.17, 1.25  
    Female 3,210 2,761,075 1.45 1.31, 1.59  1.25 1.17, 1.35  
Parity     0.003   0.017 
    0 7,879 1,977,055 1.49 1.40, 1.59  1.27 1.22, 1.33  
    1 7,123 1,921,604 1.32 1.24, 1.41  1.21 1.15, 1.27  
    ≥2 3,781 1,744,441 1.26 1.15, 1.37  1.13 1.06, 1.20  
Gestational age, weeks     0.931   0.246 
    <32 289 58,800 1.46 1.08, 1.97  1.01 0.80, 1.28  
    32–36 1,494 408,070 1.37 1.19, 1.57  1.20 1.08, 1.33  
    ≥37 17,000 5,176,230 1.38 1.32, 1.44  1.23 1.19, 1.27  
Birth weight, g     0.561   0.112 
    <1,500 192 31,690 1.69 1.17, 2.46  0.92 0.69, 1.24  
    1,500–2,499 768 201,223 1.36 1.13, 1.65  1.30 1.12, 1.49  
    ≥2,500 17,823 5,410,187 1.38 1.32, 1.44  1.22 1.18, 1.26  
Maternal race     0.562    
    White non-Hispanic 7,987 2,219,660 1.40 1.31, 1.49     
    White Hispanic, US born 2,211 678,211 1.43 1.26, 1.62     
    White Hispanic, not US born 4,000 1,700,688 1.38 1.27, 1.50     
    Black 1,441 342,563 1.45 1.26, 1.67     
    Asian, US born 329 72,785 1.68 1.19, 2.37     
    Asian, not US born 2,418 494,898 1.31 1.17, 1.47     
    Other 397 134,295 1.31 0.98, 1.74     
Paternal race        0.079 
    White non-Hispanic 8,446 2,218,193    1.19 1.13, 1.24  
    White Hispanic 5,983 2,367,264    1.19 1.13, 1.26  
    Black 1,527 412,178    1.35 1.24, 1.48  
    Asian 2,397 505,120    1.28 1.17, 1.39  
    Other 430 140,345    1.12 0.91, 1.38  
Maternal education     0.762    
    Less than high school 2,647 1,482,071 1.34 1.22, 1.48     
    High school graduate 5,076 1,688,114 1.39 1.29, 1.50     
    College 8,350 1,947,044 1.40 1.32, 1.49     
    Postgraduate 2,710 525,871 1.46 1.29, 1.66     
Paternal education        0.009 
    Less than high school 2,644 1,459,374    1.18 1.10, 1.28  
    High school graduate 5,060 1,743,094    1.25 1.18, 1.33  
    College 7,729 1,785,742    1.15 1.10, 1.21  
    Postgraduate 3,330 654,890    1.32 1.23, 1.42  
Delivery payment     0.016   0.428 
    Insurance 11,402 2,871,823 1.32 1.25, 1.39  1.23 1.18, 1.28  
    Self, government, other 7,381 2,771,277 1.46 1.38, 1.56  1.20 1.15, 1.26  
Birth year     0.032   0.071 
    1989 649 419,447 1.55 1.24, 1.94  1.25 1.07, 1.47  
    1990 772 451,666 1.62 1.32, 1.98  1.14 0.98, 1.33  
    1991 988 447,614 1.58 1.32, 1.89  1.14 0.99, 1.30  
    1992 1,183 439,926 1.63 1.38, 1.92  1.08 0.96, 1.23  
    1993 1,190 425,181 1.48 1.26, 1.74  1.25 1.11, 1.41  
    1994 1,352 413,745 1.28 1.10, 1.50  1.15 1.03, 1.30  
    1995 1,361 398,649 1.21 1.04, 1.41  1.34 1.20, 1.49  
    1996 1,452 387,804 1.30 1.12, 1.50  1.24 1.11, 1.38  
    1997 1,517 374,334 1.33 1.16, 1.53  1.33 1.20, 1.47  
    1998 1,646 372,815 1.37 1.20, 1.57  1.27 1.15, 1.41  
    1999 1,720 370,804 1.29 1.13, 1.48  1.22 1.11, 1.35  
    2000 1,775 382,640 1.41 1.24, 1.60  1.21 1.09, 1.33  
    2001 1,703 379,904 1.31 1.14, 1.50  1.18 1.07, 1.31  
    2002 1,475 378,571 1.41 1.22, 1.63  1.19 1.07, 1.33  
Subgroups No. of Cases No. of Noncases Maternal Age, 20–39 Years
 
P Valueb Paternal Age, 20–59 Years
 
P Valueb 
Odds Ratio 95% Confidence Interval Odds Ratio 95% Confidence Interval 
Child's sex     0.298   0.421 
    Male 15,573 2,882,025 1.37 1.31, 1.43  1.21 1.17, 1.25  
    Female 3,210 2,761,075 1.45 1.31, 1.59  1.25 1.17, 1.35  
Parity     0.003   0.017 
    0 7,879 1,977,055 1.49 1.40, 1.59  1.27 1.22, 1.33  
    1 7,123 1,921,604 1.32 1.24, 1.41  1.21 1.15, 1.27  
    ≥2 3,781 1,744,441 1.26 1.15, 1.37  1.13 1.06, 1.20  
Gestational age, weeks     0.931   0.246 
    <32 289 58,800 1.46 1.08, 1.97  1.01 0.80, 1.28  
    32–36 1,494 408,070 1.37 1.19, 1.57  1.20 1.08, 1.33  
    ≥37 17,000 5,176,230 1.38 1.32, 1.44  1.23 1.19, 1.27  
Birth weight, g     0.561   0.112 
    <1,500 192 31,690 1.69 1.17, 2.46  0.92 0.69, 1.24  
    1,500–2,499 768 201,223 1.36 1.13, 1.65  1.30 1.12, 1.49  
    ≥2,500 17,823 5,410,187 1.38 1.32, 1.44  1.22 1.18, 1.26  
Maternal race     0.562    
    White non-Hispanic 7,987 2,219,660 1.40 1.31, 1.49     
    White Hispanic, US born 2,211 678,211 1.43 1.26, 1.62     
    White Hispanic, not US born 4,000 1,700,688 1.38 1.27, 1.50     
    Black 1,441 342,563 1.45 1.26, 1.67     
    Asian, US born 329 72,785 1.68 1.19, 2.37     
    Asian, not US born 2,418 494,898 1.31 1.17, 1.47     
    Other 397 134,295 1.31 0.98, 1.74     
Paternal race        0.079 
    White non-Hispanic 8,446 2,218,193    1.19 1.13, 1.24  
    White Hispanic 5,983 2,367,264    1.19 1.13, 1.26  
    Black 1,527 412,178    1.35 1.24, 1.48  
    Asian 2,397 505,120    1.28 1.17, 1.39  
    Other 430 140,345    1.12 0.91, 1.38  
Maternal education     0.762    
    Less than high school 2,647 1,482,071 1.34 1.22, 1.48     
    High school graduate 5,076 1,688,114 1.39 1.29, 1.50     
    College 8,350 1,947,044 1.40 1.32, 1.49     
    Postgraduate 2,710 525,871 1.46 1.29, 1.66     
Paternal education        0.009 
    Less than high school 2,644 1,459,374    1.18 1.10, 1.28  
    High school graduate 5,060 1,743,094    1.25 1.18, 1.33  
    College 7,729 1,785,742    1.15 1.10, 1.21  
    Postgraduate 3,330 654,890    1.32 1.23, 1.42  
Delivery payment     0.016   0.428 
    Insurance 11,402 2,871,823 1.32 1.25, 1.39  1.23 1.18, 1.28  
    Self, government, other 7,381 2,771,277 1.46 1.38, 1.56  1.20 1.15, 1.26  
Birth year     0.032   0.071 
    1989 649 419,447 1.55 1.24, 1.94  1.25 1.07, 1.47  
    1990 772 451,666 1.62 1.32, 1.98  1.14 0.98, 1.33  
    1991 988 447,614 1.58 1.32, 1.89  1.14 0.99, 1.30  
    1992 1,183 439,926 1.63 1.38, 1.92  1.08 0.96, 1.23  
    1993 1,190 425,181 1.48 1.26, 1.74  1.25 1.11, 1.41  
    1994 1,352 413,745 1.28 1.10, 1.50  1.15 1.03, 1.30  
    1995 1,361 398,649 1.21 1.04, 1.41  1.34 1.20, 1.49  
    1996 1,452 387,804 1.30 1.12, 1.50  1.24 1.11, 1.38  
    1997 1,517 374,334 1.33 1.16, 1.53  1.33 1.20, 1.47  
    1998 1,646 372,815 1.37 1.20, 1.57  1.27 1.15, 1.41  
    1999 1,720 370,804 1.29 1.13, 1.48  1.22 1.11, 1.35  
    2000 1,775 382,640 1.41 1.24, 1.60  1.21 1.09, 1.33  
    2001 1,703 379,904 1.31 1.14, 1.50  1.18 1.07, 1.31  
    2002 1,475 378,571 1.41 1.22, 1.63  1.19 1.07, 1.33  
a

Excluded are observations with missing values for any covariates. Subgroup-specific models were adjusted for all the other covariates listed in Table 2.

b

Chi-square test of homogeneity.

The increased risk of autism associated with older maternal and paternal age was of greater magnitude for first-born children compared with second or later-born children (maternal age, P = 0.003; paternal age, P = 0.017) (Table 2). Among first-born children, the increase in adjusted risk of autism associated with a 10-year increment in maternal age was approximately 49%, decreasing to 32% among second-born, and 26% among later-born children. The risks associated with paternal age followed a similar pattern, although the odds ratios were smaller.

All covariate subgroup results were similar when the maternal age range was extended to 15–44 years and the paternal age range was extended to 15–64 years (data not shown).

DISCUSSION

In this very large and diverse contemporary North American population, both increasing maternal age and increasing paternal age were independently associated with increased risk of autism. These associations were statistically linear with the log(odds) over a typical reproductive age range. For each 10-year increment in age, the odds ratio for autism associated with maternal age increased by 38% and that with paternal age by 22%. These maternal and paternal age effects persisted within racial/ethnic and gender subgroups, as well as within subgroups defined by other demographic and pregnancy-related covariates. Maternal age effects were of somewhat greater magnitude among children presumed to be at the lower end of the socioeconomic spectrum and among children born in the earlier study years. Both maternal and paternal age effects were greater for first-born compared with later-born children.

Results of early studies focusing only on the age of mothers, without consideration of the age of fathers, were inconsistent with regard to whether children born to older mothers were more likely to have autism (1–9, 11). More recent population-based studies have evaluated the risk of autism in models that considered both maternal and paternal age, adjusting for the age of the other parent and additional covariates. In Western Australia, children of mothers aged 30–34 years or 35 years or more were found to have an elevated risk for autism compared with children of mothers aged 25–29 years, with the magnitude of risk similar to our current findings; older paternal age was not a significant risk factor in this population (12). In contrast, 2 studies from Denmark, with overlapping populations born in 1973–1994 (13) and 1984–1998 (14), reported an association of autism with increasing paternal, but not maternal, age. When the models were further adjusted for parental psychiatric history, the increase in risk seen for older fathers remained but was statistically significant only in the later cohort. In these 2 Danish studies, the prevalence of autism was low, 0.52/1,000 for the earlier cohort and 0.86/1,000 for the later one. Reichenberg et al. (15) evaluated maternal and paternal age in an historical population-based cohort of Israeli draft board registrants with an autism prevalence of 0.84/1,000. When evaluated as a continuous variable, the odds ratio associated with each 10-year increase in paternal age was 2.14 (95% CI: 1.44, 3.16); no association was found with maternal age.

In a birth cohort study of autism in singleton children enrolled with Kaiser Permanente of Northern California (∼1.6% overlap with the current study population) (17), autism was significantly and independently associated with both maternal age and paternal age, with the magnitude of risk similar to that found in the present report. The prevalence of autism in this population was 4.46/1,000. Analyzing data from 10 US study sites participating in the Autism and Developmental Disabilities Monitoring Network, Durkin et al. found that maternal age and paternal age were independently associated with autism in adjusted models (18). Site-specific prevalence ranged from 3.3 to 10.6/1,000 (20); the parental age effects were similar in magnitude to our findings and were also greater for first-born than for later-born children.

The extent to which these reported inconsistencies in maternal and paternal age effects may reflect a true difference across populations is difficult to evaluate in light of unexplained but substantial differences in the observed prevalence of autism in the different study populations.

There are several possible age-related biologic mechanisms through which increasing maternal and paternal age could affect fetal brain development leading to autism. For women, these include hormonal factors that alter the in utero environment (21), greater risk of infertility and exposure to assisted reproductive technologies (22–24), nucleotide repeat instability (25), and an increase in body burdens from cumulative toxic exposures (26, 27). These factors are not mutually exclusive and could work synergistically to increase the risk of atypical fetal brain development, but they have been little studied with regard to autism.

For men, the most likely age-related biologic explanation is increased de novo mutations in sperm, occurring more commonly in older fathers and perhaps affected by cumulative toxic exposures. Of possible relevance to autism, studies have shown an independent association of bipolar disorder (28) and schizophrenia (29–33) in offspring with advancing paternal, but not maternal, age. The association for schizophrenia is limited to sporadic cases that lack a family history of the disorder, consistent with the hypothesis of accumulating de novo mutations in the germ cells of older fathers (33). It has been hypothesized that de novo mutations are associated with nonfamilial or sporadic autism (34), and limited evidence indicates that de novo copy number variations may be associated with autism in some children (35).

Without further research, the extent to which these age-related biologic factors may contribute to increased risk of autism among children born to older mothers and fathers remains speculative, as is the possibility that variation in the frequency of these factors may explain differences in parental age effects from one population to another. However, our observation of substantially greater maternal and paternal age effects among first-born children would appear to be inconsistent with a strong role for age-related biologic factors unless there is a relation between these biologic factors and parity or birth order of the child. De novo mutations in sperm are unlikely to be related to parity, but maternal age effects could potentially be related to whether or not a woman has experienced prior pregnancies.

An alternative, but not mutually exclusive, explanation is that increased parental age may be a marker of preexisting genetic risk, as suggested by the earlier Danish study in which parental psychiatric history was associated with older age at parenthood (13). Men and women with a genetic predisposition for having a child with autism may simply be more likely to delay childbearing until they are older. Parents who delay childbearing may be more likely to have fewer children, but this could not explain the greater parental age effect among low-parity children unless delayed childbearing is related to risk of autism. Parents who have an autism spectrum disorder-affected child may decide to forgo further childbearing (i.e., “stoppage”), but stoppage could not explain the greater parental age effect among low-parity children unless stoppage is related to the age of parents. Finally, an independent age effect for both mothers and fathers may be consistent with a preexisting risk explanation, as genetic risk to the offspring could be transmitted through either or both parents.

Early studies that demonstrated preexisting risk in parents of children with autism, as defined by a history of psychiatric disorders (36–38) or characteristics of the broader autism phenotype (39–42), did not consider parental ages or parity as covariates. A recently reported study of preexisting risk and parental age did not observe an association between father's age at first paternity and characteristics of the broader autism phenotype in fathers or mothers of children with autism from multiplex families (34). Further studies are clearly needed.

We found a stronger maternal (but not paternal) age effect for children born during the early study years (1989–1993), compared with children born in more recent years. Although it is possible that this difference is related to differential age-related opportunities for obtaining an autism diagnosis, this seems unlikely because there is little difference in maternal age effects for children born in the years after 1993. The oldest children born during these years were well beyond the typical age at diagnosis in the DDS system. Alternatively, if maternal age is differentially associated with phenotypic subgroups within the autism spectrum, the difference between the early and later birth years may represent a change in the phenotypic profile for children receiving DDS services for autism after implementation of DSM, Fourth Edition, criteria in 1994 and other changes. Another possible explanation is that older mothers may have been more assertive in seeking services for their affected children during the earlier period when there was less public awareness and support. Our finding of a stronger maternal age effect among children of presumed lower socioeconomic status may indicate that, among families relatively disadvantaged by socioeconomic factors, older parents were more likely than younger ones to obtain services for their affected children.

Limitations in the data available through DDS records prevented us from further evaluating these and other possible explanations for the association between older parents and autism risk in children. Our study is also limited by reliance on a single source to ascertain autism cases, leading to underascertainment of affected children, particularly those who do not meet DDS eligibility criteria for “substantial handicap.” Despite this limitation, the DDS-based autism prevalence of 3.1/1,000 children is considerably higher than that reported in the Danish (13, 14) and Israeli (15) studies discussed above, although still lower than estimates obtained by several other recent studies (20, 43–49).

Strengths of our study include a very large cohort population, with comparable proportions of cases and noncases excluded for unlikely parental ages and missing data. By limiting analyses to singletons, we eliminated possible confounding by factors associated with multiple gestations. The very substantial size of this population permitted multivariate modeling and subgroup analyses with a high level of statistical precision. Diagnostic validation has been obtained for some DDS subjects through research evaluations using the Autism Diagnostic Observation Schedule (50) and the Autism Diagnostic Interview-Revised (51); less than 2% of over 300 children served by DDS for autism did not meet DSM, Fourth Edition, criteria for an autism spectrum disorder using these instruments (52).

Further understanding of the pathways that link childhood autism to older age of parents may require studies among subgroups of affected children defined by phenotypic and demographic characteristics, as well as studies focused on preexisting parental risk and age-related biologic factors.

Abbreviations

    Abbreviations
  • CI

    confidence interval

  • DDS

    Department of Developmental Services

  • DSM

    Diagnostic and Statistical Manual of Mental Disorders

  • OR

    odds ratio

Author affiliations: Environmental Health Investigations Branch, California Department of Public Health, Richmond, California (Judith K. Grether, Meredith C. Anderson, Daniel Smith, Gayle C. Windham); and Division of Research, Kaiser Permanente Northern California, Oakland, California (Lisa A. Croen).

This work was supported by the California Department of Public Health and by a Cooperative Agreement with the National Center for Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention (CDC U10/CCU920392).

The authors express appreciation to the California Department of Developmental Services for providing data sets for these analyses, for consultation on the complexities of the data, and for reviewing the manuscript for accuracy in description of departmental data.

Presented in part at the Fifth International Meeting for Autism Research in Montreal, Canada, in May 2006.

Conflict of interest: none declared.

References

1.
Treffert
DA
Epidemiology of infantile autism
Arch Gen Psychiatry
 , 
1970
, vol. 
22
 
5
(pg. 
431
-
438
)
2.
O'Moore
M
A study of the aetiology of autism from a study of birth and family characteristics
J Ir Med Assoc.
 , 
1972
, vol. 
65
 
5
(pg. 
114
-
120
)
3.
Finegan
JA
Quarrington
B
Pre-, peri-, and neonatal factors and infantile autism
J Child Psychol Psychiatry
 , 
1979
, vol. 
20
 
2
(pg. 
119
-
128
)
4.
Gillberg
C
Maternal age and infantile autism
J Autism Dev Disord
 , 
1980
, vol. 
10
 
3
(pg. 
293
-
297
)
5.
Hoshino
Y
Kumashiro
H
Yashima
Y
, et al.  . 
The epidemiological study of autism in Fukushima-ken
Folia Psychiatr Neurol Jpn
 , 
1982
, vol. 
36
 
2
(pg. 
115
-
124
)
6.
Tsai
LY
Stewart
MA
Etiological implication of maternal age and birth order in infantile autism
J Autism Dev Disord
 , 
1983
, vol. 
13
 
1
(pg. 
57
-
65
)
7.
Steinhausen
HC
Göbel
D
Breinlinger
M
, et al.  . 
Maternal age and autistic children
J Dev Behav Pediatr
 , 
1984
, vol. 
5
 
6
(pg. 
343
-
345
)
8.
Eaton
WW
Mortensen
PB
Thomsen
PH
, et al.  . 
Obstetric complications and risk for severe psychopathology in childhood
J Autism Dev Disord
 , 
2001
, vol. 
31
 
3
(pg. 
279
-
285
)
9.
Hultman
CM
Sparén
P
Cnattingius
S
Perinatal risk factors for infantile autism
Epidemiology
 , 
2002
, vol. 
13
 
4
(pg. 
417
-
423
)
10.
Mouridsen
SE
Rich
B
Isager
T
Brief report: parental age in infantile autism, autistic-like conditions, and borderline childhood psychosis
J Autism Dev Disord
 , 
1993
, vol. 
23
 
2
(pg. 
387
-
396
)
11.
Croen
LA
Grether
JK
Selvin
S
Descriptive epidemiology of autism in a California population: who is at risk?
J Autism Dev Disord
 , 
2002
, vol. 
32
 
3
(pg. 
217
-
224
)
12.
Glasson
EJ
Bower
C
Petterson
B
, et al.  . 
Perinatal factors and the development of autism: a population study
Arch Gen Psychiatry
 , 
2004
, vol. 
61
 
6
(pg. 
618
-
627
)
13.
Larsson
HJ
Eaton
WW
Madsen
KM
, et al.  . 
Risk factors for autism: perinatal factors, parental psychiatric history, and socioeconomic status
Am J Epidemiol
 , 
2005
, vol. 
161
 
10
(pg. 
916
-
925
discussion 926–928
14.
Lauritsen
MB
Pedersen
CB
Mortensen
PB
Effects of familial risk factors and place of birth on the risk of autism: a nationwide register-based study
J Child Psychol Psychiatry
 , 
2005
, vol. 
46
 
9
(pg. 
963
-
971
)
15.
Reichenberg
A
Gross
R
Weiser
M
, et al.  . 
Advancing paternal age and autism
Arch Gen Psychiatry
 , 
2006
, vol. 
63
 
9
(pg. 
1026
-
1032
)
16.
Tsuchiya
KJ
Matsumoto
K
Miyachi
T
, et al.  . 
Paternal age at birth and high-functioning autistic-spectrum disorder in offspring
Br J Psychiatry
 , 
2008
, vol. 
193
 
4
(pg. 
316
-
321
)
17.
Croen
LA
Najjar
DV
Fireman
B
, et al.  . 
Maternal and paternal age and risk of autism spectrum disorders
Arch Pediatr Adolesc Med
 , 
2007
, vol. 
161
 
4
(pg. 
334
-
340
)
18.
Durkin
MS
Maenner
MJ
Newschaffer
CJ
, et al.  . 
Advanced parental age and the risk of autism spectrum disorder
Am J Epidemiol
 , 
2008
, vol. 
168
 
11
(pg. 
1268
-
1276
)
19.
Lui
KJ
Kelly
C
Tests for homogeneity of the risk ratio in a series of 2 × 2 tables
Stat Med
 , 
2000
, vol. 
19
 
21
(pg. 
2919
-
2932
)
20.
Prevalence of autism spectrum disorders—autism and developmental disabilities monitoring network, 14 sites, United States, 2002
MMWR Surveill Summ
 , 
2007
, vol. 
56
 
1
(pg. 
12
-
28
)
21.
Newschaffer
CJ
Croen
LA
Daniels
J
, et al.  . 
The epidemiology of autism spectrum disorders
Annu Rev Public Health
 , 
2007
, vol. 
28
 (pg. 
235
-
258
)
22.
Strömberg
B
Dahlquist
G
Ericson
A
, et al.  . 
Neurological sequelae in children born after in-vitro fertilization: a population-based study
Lancet
 , 
2002
, vol. 
359
 
9305
(pg. 
461
-
465
)
23.
Pinborg
A
Loft
A
Schmidt
L
, et al.  . 
Morbidity in a Danish national cohort of 472 IVF/ICSI twins, 1132 non-IVF/ICSI twins and 634 IVF/ICSI singletons: health-related and social implications for the children and their families
Hum Reprod
 , 
2003
, vol. 
18
 
6
(pg. 
1234
-
1243
)
24.
Lidegaard
O
Pinborg
A
Andersen
AN
Imprinting diseases and IVF: Danish national IVF cohort study
Hum Reprod
 , 
2005
, vol. 
20
 
4
(pg. 
950
-
954
)
25.
Persico
AM
D'Agruma
L
Maiorano
N
, et al.  . 
Reelin gene alleles and haplotypes as a factor predisposing to autistic disorder
Mol Psychiatry
 , 
2001
, vol. 
6
 
2
(pg. 
150
-
159
)
26.
Lawler
CP
Croen
LA
Grether
JK
, et al.  . 
Identifying environmental contributions to autism: provocative clues and false leads
Ment Retard Dev Disabil Res Rev.
 , 
2004
, vol. 
10
 
4
(pg. 
292
-
302
)
27.
Newschaffer
CJ
Fallin
D
Lee
NL
Heritable and nonheritable risk factors for autism spectrum disorders
Epidemiol Rev.
 , 
2002
, vol. 
24
 
2
(pg. 
137
-
153
)
28.
Frans
EM
Sandin
S
Reichenberg
A
, et al.  . 
Advancing paternal age and bipolar disorder
Arch Gen Psychiatry
 , 
2008
, vol. 
65
 
9
(pg. 
1034
-
1040
)
29.
Hare
EH
Moran
PA
Raised parental age in psychiatric patients: evidence for the constitutional hypothesis
Br J Psychiatry
 , 
1979
, vol. 
134
 (pg. 
169
-
177
)
30.
Malaspina
D
Harlap
S
Fennig
S
, et al.  . 
Advancing paternal age and the risk of schizophrenia
Arch Gen Psychiatry
 , 
2001
, vol. 
58
 
4
(pg. 
361
-
367
)
31.
Brown
AS
Schaefer
CA
Wyatt
RJ
, et al.  . 
Paternal age and risk of schizophrenia in adult offspring
Am J Psychiatry
 , 
2002
, vol. 
159
 
9
(pg. 
1528
-
1533
)
32.
Zammit
S
Allebeck
P
Dalman
C
, et al.  . 
Paternal age and risk for schizophrenia
Br J Psychiatry
 , 
2003
, vol. 
183
 
5
(pg. 
405
-
408
)
33.
Sipos
A
Rasmussen
F
Harrison
G
, et al.  . 
Paternal age and schizophrenia: a population based cohort study [electronic article]
BMJ
 , 
2004
, vol. 
329
 
7474
pg. 
1070
 
34.
Puleo
CM
Reichenberg
A
Smith
CJ
, et al.  . 
Do autism-related personality traits explain higher paternal age in autism?
Mol Psychiatry
 , 
2008
, vol. 
13
 
3
(pg. 
243
-
244
)
35.
Sebat
J
Lakshmi
B
Malhotra
D
, et al.  . 
Strong association of de novo copy number mutations with autism
Science
 , 
2007
, vol. 
316
 
5823
(pg. 
445
-
449
)
36.
Mouridsen
SE
Rich
B
Isager
T
, et al.  . 
Psychiatric disorders in the parents of individuals with infantile autism: a case-control study
Psychopathology
 , 
2007
, vol. 
40
 
3
(pg. 
166
-
171
)
37.
Daniels
JL
Forssen
U
Hultman
CM
, et al.  . 
Parental psychiatric disorders associated with autism spectrum disorders in the offspring
Pediatrics
 , 
2008
, vol. 
121
 
5
(pg. 
e1357
-
e1362
)
38.
Piven
J
Chase
GA
Landa
R
, et al.  . 
Psychiatric disorders in the parents of autistic individuals
J Am Acad Child Adolesc Psychiatry
 , 
1991
, vol. 
30
 
3
(pg. 
471
-
478
)
39.
Piven
J
Wzorek
M
Landa
R
, et al.  . 
Personality characteristics of the parents of autistic individuals
Psychol Med
 , 
1994
, vol. 
24
 
3
(pg. 
783
-
795
)
40.
Piven
J
Palmer
P
Jacobi
D
, et al.  . 
Broader autism phenotype: evidence from a family history study of multiple-incidence autism families
Am J Psychiatry
 , 
1997
, vol. 
154
 
2
(pg. 
185
-
190
)
41.
Piven
J
Palmer
P
Psychiatric disorder and the broad autism phenotype: evidence from a family study of multiple-incidence autism families
Am J Psychiatry
 , 
1999
, vol. 
156
 
4
(pg. 
557
-
563
)
42.
Bailey
A
Palferman
S
Heavey
L
, et al.  . 
Autism: the phenotype in relatives
J Autism Dev Disord
 , 
1998
, vol. 
28
 
5
(pg. 
369
-
392
)
43.
Baird
G
Charman
T
Baron-Cohen
S
, et al.  . 
A screening instrument for autism at 18 months of age: a 6-year follow-up study
J Am Acad Child Adolesc Psychiatry
 , 
2000
, vol. 
39
 
6
(pg. 
694
-
702
)
44.
Chakrabarti
S
Fombonne
E
Pervasive developmental disorders in preschool children
JAMA
 , 
2001
, vol. 
285
 
24
(pg. 
3093
-
3099
)
45.
Bertrand
J
Mars
A
Boyle
C
, et al.  . 
Prevalence of autism in a United States population: the Brick Township, New Jersey, investigation
Pediatrics
 , 
2001
, vol. 
108
 
5
(pg. 
1155
-
1161
)
46.
Chakrabarti
S
Fombonne
E
Pervasive developmental disorders in preschool children: confirmation of high prevalence
Am J Psychiatry
 , 
2005
, vol. 
162
 
6
(pg. 
1133
-
1141
)
47.
Fombonne
E
Zakarian
R
Bennett
A
, et al.  . 
Pervasive developmental disorders in Montreal, Quebec, Canada: prevalence and links with immunizations
Pediatrics
 , 
2006
, vol. 
118
 
1
(pg. 
e139
-
e150
)
48.
Mental health in the United States: parental report of diagnosed autism in children aged 4–17 years—United States, 2003–2004
MMWR Morb Mortal Wkly Rep
 , 
2006
, vol. 
55
 
17
(pg. 
481
-
486
)
49.
Prevalence of autism spectrum disorders—autism and developmental disabilities monitoring network, six sites, United States, 2000
MMWR Surveill Summ
 , 
2007
, vol. 
56
 
1
(pg. 
1
-
11
)
50.
Lord
C
Rutter
M
DiLavore
PC
, et al.  . 
The Autism Diagnostic Observation Schedule (ADOS)
 , 
2000
Los Angeles, CA
Western Psychological Services
51.
Lord
C
Rutter
M
Le Couteur
A
Autism Diagnostic Interview-Revised: a revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders
J Autism Dev Disord
 , 
1994
, vol. 
24
 
5
(pg. 
659
-
685
)
52.
Hertz-Picciotto
I
Croen
LA
Hansen
R
, et al.  . 
The CHARGE Study: an epidemiologic investigation of genetic and environmental factors contributing to autism
Environ Health Perspect
 , 
2006
, vol. 
114
 
7
(pg. 
1119
-
1125
)