Paternal Age Is Associated With Impaired Neurocognitive Outcome in Infancy : Results
Authors and Disclosures


There were 55,740 singleton pregnancies. Of these, 12,297 children were excluded because of (a) missing maternal and/or paternal age (1,542), (b) having indeterminate or unspecified sex (1,050), or (c) gestational age that was missing or less than 37 wk (9,705). After randomly selecting one live-born offspring per study mother, this left a total of 33,437 study offspring (17,148 males) available for the main analyses. Of these, 51% of the mothers were white, 39% black, and the remaining 10% were Asian and other racial groups. Finally, 6,355 children were missing information about age at testing at 8 mo, while 9,930 were missing age at testing at 4 y, and 9,109 were missing age at testing at 7 y. Those with missing paternal age were significantly more likely to have missing outcome variables at 8 mo, 4 y, and 7 y (each p < 0.001).

Table 1 shows descriptive statistics for paternal and maternal age and differences in parental age. On average, fathers were 3 to 4 y older than mothers, but the differences in parental age varied widely. Concerning the primary analyses, there was a statistically significant association between advanced paternal age and inferior performance on all neurocognitive tests (all p < 0.001) except for Bayley Motor score (Model 2, p = 0.104) (see Table 2). Concerning the influence of maternal age, there were statistically significant associations between advanced maternal age and superior performance on all measures. Figure 1 and Figure 2 show the mean adjusted score for paternal and maternal age for the outcome variables based on Models 1 and 2 respectively. Apart from the direction of the association between maternal and paternal age, the association between maternal age and the outcome variables at ages 4 and 7 y was curvilinear (generally steep at younger ages, then less steep at older ages), in contrast to the near-linear association with paternal age. Post-hoc analyses examining the goodness-of-fit of nonlinear versus linear models indicated that two of the variables were adequately capture by simple linear models (Bayley Mental score and Graham Ernhart Block Sort Test), but that nonlinear models were best suited for all other variables (unpublished data). Table 3 shows the estimated scores (and 95% CIs) for two paternal ages (20 and 50 y) based on the nonlinear modelling used in the primary analyses. For Model 2, the adjusted R-squared ranged from 2.4% (Bayley Motor) to 29.5% (WISC Full Scale IQ).


(Enlarge Image) Figure 1.
Primary Analyses: Model 1 -- Adjusted for Other Parent's Age, Mother's Race, Gestational Age, and Child Gender. Solid lines ranging from 15 to 45 y for maternal age, dotted lines ranging from 15 to 65 y for paternal age. Nonlinear model fit with 95% CIs.