“When should I have children?” “Am I too old
to have a child?” “How old is too old?” “Is my daughter destined to have
problems because I had her when I was older?” These are just some of the questions
that many people, including colleagues, friends, and anonymous individuals (via
email), posed to both of us after the news media covered the results of two large-scale
studies on the association between paternal age at childbearing and mental-health
problems of offspring. One study, which was conducted by John McGrath and
colleagues, was based on data from Denmark. 1 The second study, based on
data from Sweden, was conducted by the two of us and our colleagues. 2 How did we answer these
questions from the public? What does the science say—and not say—about the
At the outset, let us first acknowledge that these
two studies address very sensitive issues. Decisions about whether to have
children and when to have children are incredibly private. In addition, the
prospect that one’s children will have both mental-health and physical problems
is a major concern for parents, and caring for offspring with such problems can
cause considerable pain and suffering.
Overview of the Studies
The two recent studies had many similarities.
First, both analyzed comparable data sets to explore the associations with
parental age at childbearing via epidemiologic methods that use large,
representative samples to examine the distribution and determinants of health.
The data sets in both studies were based on extensive population records kept
by the governments of Scandinavian countries, including Denmark and Sweden. For
example, these countries have extensive records of all inpatient and outpatient
visits to hospitals. Scandinavian governments also are able to use personal
identification numbers to link several databases. As a result, separate
databases can be combined such that they contain information across different generations
and domains (e.g., psychiatric diagnoses and age at childbearing). To
facilitate the analysis of these population data sets while protecting personal
information, government agencies in both countries alter the personal
identifiers after merging the data sets, a practice that has led to extensive
psychiatric epidemiology research.
Unfortunately for research
purposes, no such data sets are available in the United States.
Second, both studies took advantage of the
size and scope of the national data sets. Access to these large data sets
enabled both research groups to explore a fuller range of parental age at
childbearing because the data sets included high numbers of individuals at the
extreme ends of childbearing (e.g., older than 45 years). The large data sets had
sufficient numbers of exposed cases (e.g., offspring diagnosed with autism
whose parents were older than 45 years at childbirth), enabling both groups to
predict relatively rare outcomes in the offspring (e.g., schizophrenia and
autism). Previous research teams, which relied on smaller data sets, could not
explore advancing paternal age at childbearing with as much precision.
Third, both studies presented population and
adjusted estimates of the magnitude of the association between paternal age at
childbearing and disorders of offspring. Notably, the two estimates address different
research questions. The population estimates respond to this question: How
common is it that an offspring born to an older father has psychiatric problems?
Both research teams sought to answer this question by comparing the rates of several
disorders among offspring born to fathers of different ages in the entire
population. The adjusted estimates, on the other hand, respond to this
question: When it comes to psychiatric problems in offspring, how much risk is specificallydue to advancing paternal age?
The studies addressed the latter question by
using statistical techniques and other approaches to estimate the magnitude of
the association between paternal age at childbearing and offspring disorders
while trying to hold constant the highest possible number of factors that
differ among fathers who have children at different ages (e.g., parents’ socioeconomic
status). Researchers must use various techniques to study advancing paternal
age at childbearing because men who have children when they are younger differ,
sometimes greatly, from men who delay childbearing. Plus, it is impossible to
conduct a randomized experiment in humans—you can’t randomly assign men to have
children at different ages.
It is important to stress that both research
questions (based on population and adjusted estimates) are valid and might be
of interest for public-health policy. Both questions can also have important
implications for subsequent basic research because the analysis of large
population data sets can provide critical information about mechanisms at
multiple levels of analysis that researchers should explore in the future.
Previous epidemiological research on
advancing paternal age had suggested that advancing paternal age is associated
with increased risk for psychiatric problems in the offspring, but the findings
researchers suggested that any association between advancing paternal age and
offspring psychiatric problems was not due to the specific consequences of
delaying childbearing; rather, differences associated with advancing paternal
age, such as personality and psychiatric problems in the fathers
birth-order effects in the offspring,
could better explain any
Given these concerns, the two recent studies based on the large,
population-based registries were intended to provide more understanding of the
associations between paternal age at childbearing and offspring psychiatric
The Danish Study
The study by McGrath and colleagues included
all individuals born in Denmark from 1955 through 2006, a sample of almost 3 million
people. The study indexed psychiatric problems based on inpatient
hospitalizations and outpatient visits to psychiatric departments. The study presented
population estimates for several psychiatric disorders. For instance, researchers
found that compared to offspring born to fathers 25 to 29 years old (the reference
group in the study), offspring born to fathers over the age of 45 were 1.4
times more likely to have schizophrenia and 1.7 times more likely to have autism.
However, the offspring of older fathers were less likely to have hyperkinetic
disorders, which is similar to the diagnosis of attention-deficit/hyperactivity
disorder (ADHD): offspring born to men 30 to 34 years old (21 percent less
likely), 35 to 39 years old (26 percent less likely), 40 to 44 years old (21
percent less likely), and older than 45 years old (5 percent less likely) had
lower rates of these disorders than offspring in the reference group.
In order to provide adjusted estimates that
assessed the magnitude of the association between paternal age at childbearing
and the offspring disorders that was independent of other factors, the
researchers ran a series of analyses that statistically controlled for multiple
factors that are correlated with advancing paternal age. For example, the
researchers statistically controlled for maternal age at childbearing because
men who have children when they are older are more likely to have children with
women who are older. In those models, advancing paternal age was still associated
with the disorders of offspring, and the magnitudes of the associations were as
large or larger than the population estimates. The researchers also accounted
for urbanization at place of birth and family history of mental illness. The
same pattern or results emerged. For example, offspring born to men over the
age of 45 were 1.4 times more likely to have schizophrenia and 1.7 times more
likely to have autism. Notably, offspring born to fathers older than 45 were 1.2
times more likely to have a hyperkinetic disorder when statistically controlling
for the other factors, despite the fact that in the population, offspring born
to men above age 45? were less likely to have the disorder.
The results from the Danish study indicate
that in the population offspring of older fathers, some psychiatric problems,
such as schizophrenia and autism, are more likely, and that hyperkinetic
disorders are less likely. In trying to examine the pattern of associations
when accounting for correlated factors, such as maternal age at childbearing
and family history of psychiatric problems (the adjusted estimates), the
researchers found that advancing paternal age at childbearing was still
associated with psychiatric problems. In other words, these measured factors do
not explain the associations between advancing paternal age and offspring
psychiatric problems. The differences between the population and the adjusted estimates
also indicated that researchers must take into account other differences among men
who have children at different ages.
The Swedish Study
Our study—conducted separately from the
Danish study—included all individuals born in Sweden from 1973 through 2001, a sample
of just over 2.6 million people. We similarly explored psychiatric problems,
indexed by inpatient hospitalizations and outpatient visits, but our study also
included information regarding criminal convictions, low academic achievement
based on school grades at age 15, and dropping out of school early. The first
analysis provided population estimates. We found that the population of offspring
born to older fathers (above 45) had higher rates of schizophrenia (1.6 times
more likely) and autism (1.4 times more likely) compared to offspring born to
men 20 to 24 years old (the reference group in the study). For many of the
other outcomes, however, offspring born to older fathers had fewer problems
than offspring in our reference group. For instance, offspring born to men over
the age of 45 were 43 percent less likely to have a diagnosis of ADHD.
Our team also used statistical techniques to obtain
adjusted estimates that were independent of factors that could correlate with
advancing paternal age, but we controlled for a more extensive list of measured
variables than the Danish study. We accounted for maternal age at childbearing,
as well as paternal and maternal nationality, highest level of education,
lifetime history of serious psychiatric conditions, and lifetime history of
criminality. Similar to the results in the Danish study, the associations
between advancing paternal age and the outcomes in the adjusted group were as
large or larger than the overall population estimates. The results indicated
that the associations could not be explained by these factors; rather, when
accounting for these measured variables, the associations were in some cases
larger than the population estimates.
In addition to using statistical techniques
to account for differences among fathers who had children at different ages, we
also used several advanced research designs to help account for factors that
could bias our estimates. In particular, we conducted a sibling-comparison
analysis, which estimated the association between paternal age at childbearing
and the outcomes while comparing offspring born to the same father. We explored
the rates of problems in offspring born when the father was younger compared to
their siblings born when the father was older. This design accounted for (or held
constant) all traits that made siblings similar, including unmeasured
environmental and genetic factors, thus arguably providing a more precise
estimate of the specific association with advancing paternal age.
When we compared siblings, the magnitude of
the associations with each outcome was as large as, and sometimes quite larger
than, the estimates in the other analyses. For instance, offspring born to
fathers over the age of 45 were 2.1 times more likely to have schizophrenia and
3.4 times more likely to have autism than their siblings who were born when
their father was younger. Furthermore, whereas the population estimates found
advancing paternal age to be correlated with lower incidence of ADHD, the
sibling comparisons indicated that advancing paternal age was more strongly
associated with the disorder (for example, offspring of fathers over the age of
45 were 13 times more likely to have the disorder than their siblings born when
the father was 20 to 24 years old). The sibling-comparison approach has many
advantages, but the design also has several limitations (for example, do the
results apply or generalize to other populations?).
To help address several concerns,
we conducted numerous additional analyses, including the comparison of cousins
and firstborn cousins, and found comparable results: advancing paternal age was
associated with more psychiatric problems, and the magnitude was stronger than
the population estimates.
The population estimates from our Swedish
study were generally consistent with the findings from the Danish study:
advancing paternal age was associated with greater risk for some disorders,
such as schizophrenia and autism, but less risk for others, including ADHD. The
adjusted estimates (when controlling for measured traits of both parents and
when comparing siblings and cousins born at different ages) suggested that
advancing paternal age was even more strongly associated with psychiatric
problems than previous estimates indicated.
Understanding the Underlying
What could explain the finding that advancing
paternal age at childbearing is associated with more offspring psychiatric
problems? The working hypothesis that guided both studies was that genetic
mutations during the production of sperm, referred to as de novomutations, increase as men get older and
have a causal influence on offspring psychiatric problems. Unlike in women who
are born with all of their eggs, in men sperm continue to replicate throughout their
lifetime. In fact, sperm cells undergo 20 to 30 divisions per
year—approximately 600 divisions by the age of 40.
Each cell division brings
the possibility of new mutations. Recent studies suggest that there are
approximately two new mutations each year, and there is an exponential increase
where mutations double every 16.5 years.
A growing number of
molecular genetic studies have found that these de novo mutations play a large
role in human diseases, including psychiatric problems.
For instance, several
studies have indicated that de novo mutations are associated with autism, s
uggesting a mediating biological pathway that could explain
the association between advancing paternal age and the disorder.
The two recent epidemiological studies we reviewed above,
which found independent associations between advancing paternal age and
offspring psychiatric problems, are consistent with the role of de novomutations.
But why would the adjusted estimates in the
studies be larger—sometimes quite larger—than the population estimates? To
understand the differences in magnitude of these two types of estimates, it may
be important to understand the genetic, psychological, educational, social, and
financial context associated with advancing paternal age because many of these
factors predict fewer psychiatric problems in the offspring. Twin, adoption,
and family-based studies have clearly shown that genetic factors influence age
at first childbearing. There are no genes “for” early or late childbearing; rather,
genetic factors influence personality traits, psychiatric problems, and other characteristics
that in turn influence age at childbearing.
For instance, a recent
family study found that women who have ADHD, their siblings who did not have
the disorder, and men who have children with women who have ADHD are all more
likely to have children as teenagers.
Offspring born to older parents, therefore, have lower
genetic risk for ADHD on average than offspring born to parents who were
teenagers at childbearing.
Furthermore (as we also showed in our study),
advancing paternal age at childbearing is also correlated with higher levels of
parental education and higher family income, both of which lead to increased social
and cultural capital.
In sum, delaying
childbearing is correlated with a host of factors that also predict fewer
psychiatric problems in offspring.
Thus, one conceivable
explanation for the discrepancies between the population and adjusted estimates
stems from the possibility that the population estimates could be an amalgam of
the deleterious effects associated with de novo mutations and the protective factors
(e.g., genetic inheritance, personality traits, and social/cultural capital)
associated with delaying childbearing. The models that adjusted for measured traits
and compared siblings and cousins, therefore, may have held constant many of
these protective factors. As a result, the analyses may have provided a clearer
estimate of the specific influence of delaying childbearing.
Implications of the
The Danish and Swedish studies are examples
of translational epidemiology, which can help guide subsequent basic research.
The provocative findings
regarding advancing paternal age at childbearing suggest that more research
needs to be conducted on de novo mutations, especially given recent
technological advances that enable researchers to better measure and
characterize genetic mutations.
It is important to note, however, that there could be other
mechanisms through which advancing paternal age at childbearing comes to be
associated with offspring psychiatric problems. Therefore, research into other
biological and social factors is needed to better clarify the processes that
account for the findings in these studies. Furthermore, the two studies also
highlight the need to understand the complex and multifaceted factors that are
associated with delaying
So, finally, what do these studies mean for people
who are making decisions about childbearing? How did we answer the questions we
received? As you might imagine, we were reluctant to provide concrete advice. We
were not evading the questions. Rather, we do not think science can provide a definitive
answer to these questions. (Plus, it is unethical to give medical advice to
strangers via email). But here is how we responded to the questions:
First, these are only two studies. Our study
in particular, which was one of the first studies to use sibling and cousin comparisons
to study advancing paternal age, needs to be replicated. We think it was a good
study (we know we are biased), but there is never one definitive study, especially
in this area of research, because each and every study has limitations. Again,
we cannot conduct randomized controlled studies of paternal age at
childbearing. As a result, researchers must obtain consistent findings using many
different designs and samples before they can make strong causal inferences
(e.g., advancing paternal age causes offspring psychiatric problems),
especially regarding implications for family policy.
Second, researchers need to conduct more
studies on the topic before any professional group can make explicit
recommendations. In particular, researchers must expressly examine if (and then
how) physicians and couples could incorporate information on paternal age at
childbearing into their decision-making process. This examination requires
studies using predictive models.
Third, both studies provide evidence that the
overwhelming majority of offspring born to older dads will not have a major
psychiatric or related problem. Both studies found that the risk that a child
will have psychiatric problems was correlated with advancing father’s age, but most
of the outcomes were quite rare. For instance, in our study less than 3 percent
of the offspring had psychiatric problems. Therefore, advancing paternal age
does not mean that any particular child will definitely develop problems, nor does
it mean that a psychiatric problem in an offspring born to an older father was
actually caused by the father’s age at childbearing.
Fourth, our study and others also have indicated
that there can be advantages to delaying childbearing. Some factors that are
correlated with delaying childbearing (e.g., attaining a higher level of education
or gaining financial security) predict better outcomes in children.
These two studies, therefore, add to a growing body of
research suggesting that families, doctors, and society as a whole must
consider both the potential advantages and the potential disadvantages of
when trying to weigh the possible risks and benefits of delaying childbearing,
there is no set age at which advancing paternal age suddenly becomes
problematic. Our study’s reports of increased risk for offspring born to men
over the age of 45, for example, were just one illustration. Both studies found
increasing risk for many of the disorders as paternal age increased, referred
to as a dose-response relationship (i.e., offspring born to men 40 to 44 years
old also had higher rates of the disorders compared to the reference groups). In
fact, each research paper provides graphical representations of magnitude of
the associations between paternal age at childbearing and the outcomes across
the entire range of parental age.
Sixth, individuals and couples concerned
about the consequences of delaying childbearing should consult with their
physician or a specialist in their area. For instance, concerned individuals
and couples can meet with genetic counselors, who can provide more detailed and
personal information regarding the risks associated with delaying childbearing.
Finally, and most important, there are many
personal circumstances and values that go into making the decision of when to
have a child or children. Yes, we think that research can help inform personal
decision-making. But no study, set of studies, or science in general should
unduly influence the decision of when someone should have children.