Social Determinants of Health and Child Development: An Interview with Dr. Henning Tiemeier

interview by Allie Oh

Associate Editor Allie Oh interviewed Henning Tiemeier, MA, MD, PhD, a Professor of Social and Behavioral Science and the Sumner and Esther Feldberg Chair of Maternal and Child Health at the Harvard T.H. Chan School of Public Health. Dr. Tiemeier received both his medical and sociological degree from the University of Bonn, Germany, and his PhD from the Erasmus University in Rotterdam, Netherlands. Since 2018, he has led the Maternal and Child Center of Excellence at Harvard Chan. As one of just 13 HRSA-funded Centers of Excellence in Maternal and Child Health in the United States, the center trains future leaders in the field. Dr. Tiemeier has worked broadly in pediatric epidemiology for more than 20 years with an emphasis on child developmental research. At Harvard his research focuses on high-risk children, such as preterm children and homeless families.

To ensure clarity, the interview below has been minimally edited.

Allie Oh (AO): My first question is what motivated you to specialize in child developmental research and how did you get interested in studying prenatal exposures as a factor in child development? How did you get started with your research?

Dr. Henning Tiemeier (HT): Thank you. I did my PhD in psychiatric epidemiology, working on depression in the elderly and vascular problems. So on a completely different topic – I was then thinking of going into clinical practice as a doctor and continuing my training as a psychiatrist. At the end of my PhD studies, I was offered the chance to set up a study, a large epidemiological data collection in children and their families. Indeed that came with the suggestion that I would change my research topic completely: do child psychiatric research and collect data. After some hesitation, I accepted and deeply engaged with this birth cohort. My work on this birth cohort started with participant inclusion and the design of data collection. Nearly 10,000 women were included while pregnant. Their children were followed from early fetal life onwards. And I had time to grow into child developmental research in those years.

AO: That’s great, thank you. Going off that, I can see that you’ve conducted a lot of different studies and written a lot of different articles in this field. However, your studies are all based on many different topics. How do you choose your areas of focus and how do you pivot between different topics? And out of your work so far, what personally to you has been the most meaningful or enlightening paper that you’ve written or study that you’ve conducted?

HT: Let me start by admitting that I chose many of my topics in a way that was opportunistic. Because I was managing a large part of the data collection, including childhood exposures, motor development of children, cognitive development, and behavioral development, I had many opportunities, and this allowed me to take a broader approach. I didn’t want to just test one hypothesis or work in one area, but I wanted to build a dataset which could answer many questions for many people so that we could link these the most different topics in child development together; for example, to study the genetics of mother-infant attachment.

For the data collection I took the approach to ask myself: what is innovative? What are the future topics in ten years? Where can we one day get subsidies?”

And so let me give you an example. I had children that were just a bit older than the children in my cohort. I always studied the problems that my own children or their peers, their friends experienced. One of those topics was picky eating. My youngest daughter was a picky eater. So, I told myself, I will introduce questions about picky eating at a very early age. And to this day, 15 years later this data is used by my colleagues and by myself. I studied many questions not because I had an area of research focus but because I had supervised and led this data collection. Sometimes the scale of the study, sometimes the combination of different data, sometimes the early life-intrauterine assessment, and sometimes the original data collection ideas (e.g. to assess lying, we played games with study participants in which young children had to cheat to win) made the research innovative.

Now it comes to what are my favorite studies. I think I would give you a different answer each day. I learnt a lot from my favorite studies. One of them was about thyroid hormones of the mother during pregnancy and the observation that even small variations inside the normal range of non-clinically altered thyroid metabolism would have an effect on the child. The thyroid hormone of the mother is important for the child's brain development, it regulates much of the brain development such as neuronal migration in the embryo. And we were able to relate the levels of thyroid hormones during pregnancy to brain development and have follow up assessments of the child 10 years later. In one of my favorite studies, we could show that this differs by the moment in pregnancy when we had assessed the thyroid hormone. We had assessed maternal thyroid hormones in some women at eight weeks, in others at ten weeks, or at 12 weeks pregnancy and we could relate that timing to the brain development. Many years later, as we assessed child neurodevelopment with brain imaging, we could identify when hormonal alterations mattered most. That corresponded to the time before the child's own thyroid hormones kicks in, which is around 20 weeks in pregnancy when hormonal production starts in the fetus. That was interesting epidemiologically and physiologically, and a massive data collection because we linked prenatal thyroid hormones with brain imaging at age ten of the children. That's one study.

AO: That’s very interesting. So when you think of these, how do you get started in these topics? Is it more a hypothesis you have and then you collaborate with other disciplines on this, or do you work on existing questions as posed by the other departments?

HT: I believe that most hypotheses do not come to you when you sit alone in your room and think “what hypothesis could I test?” The hypotheses mostly live in some form in the sphere of all your colleagues and in the literature. And there are themes that we're interested in that may be shaped by previous studies, and you follow up your -or other people’s- studies. Hypothesis development and knowledge generation are both incremental. Questions – at least in my case – and the best hypotheses are slowly developed and carefully framed. It's in conversation with colleagues, often students that we come up with the questions. Hypotheses are ideally posited very clearly before we conduct any analysis. But sometimes we are in the business of retrospective hypothesis framing, which is very problematic but admittedly happens.

AO: Yeah, that's very interesting. And it sounds like you work with a lot of departments and your work is very interdisciplinary. Could you talk about how you collaborate with colleagues from different disciplines in terms of your different studies and the benefits of doing that?

HT: Yes. Big data sets are by nature multidisciplinary endeavors because you need many colleagues contributing expertise and funding for the data collection. In the US there are big NIH grants which fund data collections of large multidisciplinary consortia which teamed up together to get funding. In Europe, cohort studies are often funded by many small grants pieced together by colleagues who only then come to collaborate. And there is another model. Those are international consortia that work together because they have similar data, similar questions. Somebody took the lead, reached out, advertised at conferences and thus they find data and often expertise that is complementary to address certain questions together. This is very common in genetic and imaging research - different groups with similar data pool data. And sometimes collaboration occurs very differently; it's that you read a publication and you think, that's a brilliant idea, brilliant. I remember several instances that I reached out and offered collaboration or simply asked for advice. I might have data or ideas related to this work and suggest a project. Finally, surprisingly often, you just have friends from times when you were both PhD students or who you met at conferences. You simply collaborate because you're good friends and manage to find some common ground. That can work extremely well and is perhaps the most fun.

AO: Yeah, that's great. Thank you so much. For my last two questions I will pivot and ask about the Generation R study, which I saw that you helped lead. It seems very interesting. So you helped lead the Generation R study in Rotterdam and it enrolled a very large population, more than 10,000 mothers and children. Could you describe the motivations behind starting this study and your experience with working with such a vast population?

HT: The study itself was started because in the 1990s, increasingly there was evidence that the development across the life course and the occurrence of disorders is related to prenatal influences, growth and nutrition. A team of senior epidemiologists decided to study this early origins of disease hypothesis. The Erasmus University in Rotterdam was very good in large scale data collections and the researchers were skilled in managing large data collections. There was also a tradition of employing quite young researchers to lead these studies, as early career scientists would stick many years to execute the data collection, essentially younger associate or assistant professors, quite unusual. Even postdocs would run parts of these studies. That's how it started. And the funding was initially from the university and later directly from the government. Since, I've started a few new cohorts myself, another study in adolescent medicine or an aging study in India. I've always liked data collection. At Harvard, however, I work mostly with established cohort data. Yet, understanding data collection well, helps my research to this day.

AO: Could you describe some of the most interesting findings from that study or any lessons that you took away?

HT: One finding of my work is that there are clearly intrauterine influences on health, but often enough these effects are actually quite small. Sometimes - I think I'll say that a bit provocatively - that there's a cottage industry of people who relate prenatal factors to childhood and adulthood outcomes. Causality is hardly ever established. We need more critical voices to better interpret these findings.

People always ask me what are big factors? What are big risk factors? Are there big risk factors in child development outside the joint effect of genetic variations and severe maltreatment and institutionalization? We have not identified any new big risk factors. Yet, what always strikes me is how strongly ongoing family conflict impacts child development. So much more than family separation or divorce. The other factor that I found had an overwhelming influence on the child was bullying. Both probably being bullied is not only a cause of child anxiety and aggression – but also an indicator – a non-causal marker of poor child development that is already ongoing. However, a child that is bullied repeatedly and chronically will very likely not do well. And a third finding reflects insights from my more recent work on brain development. We always conceived that the characteristics of the brain, function and structure, underlie behavior. I worked in aging research, and it was clear that the brain is a determinant of behavioral problems and cognitive decline. In children, we've got repeated measures of brain imaging: we now see, not very strong and not in all periods consistently, but some evidence that the child’s behavior can shape their brain. The connections between parts of the brain, the white matter tracts, develop less quickly, if children are more anxious.

This finding is not only fascinating, it is highly relevant for clinical medicine. And it makes life more complicated for researchers: it goes both ways: the brain is the biological basis for our behavior, but our behavior can change the brain structure. That means research becomes much more susceptible to confounding and reverse causality. That is a big challenge.

Cohort research is humbling. It is often impossible to translate the many associations we observe, even if we think they are causal, into policy and interventions or clinically relevant knowledge. Only a fraction of what is published has a reasonably clear link with relevant knowledge needed for intervention. And that has many reasons. Associations are not causal. Or exposures cannot be translated into possible interventions, e.g. it is very hard to address family conflict on a population level. Effect sizes are small. And often there are simply no interventions for many of the risk factors. Genetic risk factors and brain development, at least currently, are not amenable to interventions. It is surprising how difficult translation has been and continues to be, as an epidemiologist you have to be humble- or fool yourself.

And there are totally different takeaways from this study: we have had very good response rates, certainly prenatally. 70% of the women in the general population that we approached participated. That largely is unheard of in the United States and is also very hard to replicate these days, currently it is often around 10-15%. It always surprised me that if you approach people very systematically, if you offer them a good service – we offer them free ultrasounds of the embryo and fetus – then you can get very good response rates and very loyal participants. That was another thing. Admittedly, this is hard to translate to other settings.

AO: Great, thank you so much. Yeah, I think you already touched on my final question, which was asking about the practical and translational effects and programs aimed at maternal and child health outcomes. So you're mentioning that although you're producing so many findings, you're finding so many correlations in theory, even though you're finding these identifiers, it's hard to actually translate them in policy. Have you had experience trying to or has it mostly just been theoretical?

HT: It's an interesting theoretical question to try to understand how the brain and childhood problems develop. That's by itself an interesting question for society. But of course, we do want to improve public health and we have worked together with clinical groups developing interventions. There was some work in the thyroid field which I mentioned earlier that was translated into guidelines and there are now many clinicians that think you should treat hypothyroidism during pregnancy more aggressively. We've worked very closely together with groups that target bullying in schools, developing school programs. We developed a bullying assessment program that is now used in intervention studies. It has been further refined to assess bullying both in the Netherlands and even abroad. So there are some findings which I am very proud of. But mostly, knowledge translation is a highly cumulative and slow process. We have published many studies indicating how relevant depression in mothers is for child problems. But to link any public health measure on this front to a single study is impossible. There are hundreds of studies that contribute to that insight.

AO: Yeah, that makes sense. It sounds like it's an ongoing effort and it does sound like it's hard to pinpoint a single study and correlate it to a policy change. But again, it's very interesting to learn about the theory behind it.

HT: Negative studies can have the greatest impact. We conducted several studies on how antidepressant use during pregnancy has likely no causal impact on child brain development. There was the idea that this might be quite harmful for the child. And to show that that's likely not the case is of course very important for clinical practice. These are small contributions and even they do not stand alone, many others have confirmed that now. So that knowledge is part of clinical practice. If you prescribe antidepressants during pregnancy, there is little reason to think that is harmful to a child.

AO: That's all very interesting. Thank you so much. That concludes our interview for now. I just want to thank you again so much for sharing all that with me. It was very enlightening and I'm sure our readership will really appreciate learning about it.