A Clarification of the Association Between Parental Age and the Risk of Autism

Background:   Previous reports on the risk of autism among children born to older parents have yielded conflicting results as to how increased paternal/maternal age affects risk.

Objectives:   To quantify the association between increasing parental ages and the risk of autism and to evaluate the percent attributable risk due to older average maternal age in California between 1990 and 1999.
Methods:   By analyzing restricted subgroups of parental age in a California birth cohort of over 5.6 million children, we were able to clarify the independent effects of parental age in a population large enough to evaluate discordant parental age pairings.  Multivariate logistic regression models were adjusted for the other parent’s age, maternal and paternal education, maternal and paternal race/ethnicity, year of birth, parity, and insurance payment type.  When age effects were modeled categorically, the maternal/ paternal reference age group was 25-29.  A variance inflation factor was used to adjust confidence intervals to account for the possibility of repeated births from within the same family. 
Results: We observed consistent stepwise increased risk for autism with advancing maternal age across all strata of paternal age.  The risk of autism for fathers 30 –34 increased from an OR_a of 0.96 (95% CI 0.76-1.2) among mothers < 25 to 1.92 among mothers 40+ (95% CI 1.31-2.79), a trend consistent across subgroups of paternal age.  The increased risk of autism from advancing paternal age was larger for mothers < 30 than among mothers > 30, contributing almost no pattern of increased risk to mothers over 40.  Among mothers < 25, fathers <25 had an OR_a of 0.84 (95%CI 0.69-1.01) and fathers 40+ produced the highest OR_a of 1.91 (95% CI 1.32-2.74), increasing monotonically per five-year subgroups of maternal age.  Among mothers over 40, paternal age showed a steady trend 30-34 (OR_a =1.39), 35-39 (OR_a =1.34), and 40+ ( OR_a =1.31) of slightly elevated risk.  Additionally, we calculated the number of expected cases in 1999 if the age distribution of mothers had been the same as 1990, and estimate 3.8% of the observed increase in autism cases in California between 1990 and 1999 can be attributed to the trend towards older maternal age observed over that time period.

Conclusions:   Because maternal and paternal ages are highly correlated, previous analysis on smaller populations have produced mixed results as to which parent contributes the increased risk of autism.  Through stratified analysis, we were able to detect heterogeneity of the paternal age effect, decreasing in magnitude as mothers age increased.  Effect measure modification of paternal age by maternal age indicates that the maternal age effect overwhelms the paternal age effect among mothers over 30, while increased paternal age contributes a stepwise risk among mothers under 30.  We conclude that advanced maternal age increases the risk of autism independent of paternal age, and paternal age increases the risk of autism dependent on the mother’s age.  Approximately 3.8% of the cases observed in 1999 may be attributable to risk factors associated with older average maternal age between 1990 and 1999.