David Barker (1938–2013) – a giant in reproductive epidemiology

Barker was a pioneer in advocating the role of fetal life for understanding diseases that may manifest themselves much later in life. Barker stressed that the important factors that “program adult diseases” during fetal life are often of a non-genetic nature; in his view mainly related to nutrition. In a way, he predicted epigenetic modifications of gene expression before anybody knew about it. At least epigenetic modulations provided a mechanism for long-term “disease programming” that does not have to wait for Darwin's forces of selection. Epigenetic modulations may be induced by many other different factors than nutrition, such as stress, chemical exposures or lifestyle factors, and this broadens the field of “fetal programming” beyond Barker's early work. Research should not only focus upon nutrition, fetal growth and the metabolic syndrome 1. Almost all diseases and pre- or perinatal exposure could be subjects for investigation and we may hope for important new breakthroughs in the years to come, especially since we are now beginning to get the data we need from pregnancy cohorts. Barker had no time to wait for new cohorts to become informative but looked for existing data, always a clever thing to do. The Hertfordshire cohort and the Helsinki cohort became important data sources in his early landmark studies on fetal programming 2-5. The Hertfordshire cohort was not seen as a cohort in the beginning, but as a set of records collected by Ethel Burnside, chief health visitor and lady inspector of midwives, who was concerned about the health conditions among people, including pregnant women, in this poor region of the UK in the beginning of the 20th century. One of 10 children died in the first year of life in this area. Her records became available for research, for follow- up, only because of Barker's personal engagement and a family connection to some of the people in the registry. It is notable that the important research these records permitted would not have been done under the current suggested EU regulations on data protection. The people of Hertfordshire did not give their informed consent for this study, and this was not even considered at the time of data collection. With these data, Barker could show that Forsdahl's observation 6 that being born in a poor region in Norway led to a high mortality risk in middle age life, also held in the UK. The link was not only due to social tracking, but was also mediated by low birthweight – not low birthweight in the demographic sense but low with reference to “optimal” birthweight wherever that was located on the birthweight scale. Barker strongly believed this was related to poor fetal nutrition and the adaptive fetal response was to induce insulin resistance to reduce fetal growth and to prepare the fetus for a postpartum life with an expected shortage of food. Such a shift in fetal growth would preserve brain and heart growth at the expense of other less valuable parts of the body. An asymmetrical growth pattern where the placenta plays an important role would be a marker of fetal growth impairments, even if the birthweight was in the normal range. This brought birthweight back into mainstream perinatal epidemiology where it had lived a shadowy existence after Allen Wilcox′s landmark articles on the importance and unimportance of birthweight 7, 8. Both researchers may well be right, i.e. in that birthweight reflects important underlying causes, but Barker would claim that fetal growth is a critical mediator of these exposures, e.g. leading to a lower number of nephrons in the kidney at birth, which again would be of importance for hypertension later in life 9. Wilcox would probably claim that birthweight is often an epiphenomenon in another chain of causation and that chain is what we should study. For Wilcox, birthweight was a measure of convenience, often used because the data exist, but not always of interest. For Barker, birthweight was the measure of interest, or rather fetal growth. Upcoming studies of fetal growth, based on longitudinal ultrasound data, will probably show whether he was right to suggest that impaired fetal growth is an intermediate and proximal cause of impairment that has long-term consequences. However, both Wilcox and Barker would agree that the association between birthweight and child health operates over the full scale of weights and is not restricted to a low birthweight of less than 2500 g 10, 11. A birthweight that is short of the genetic predicted birthweight is a risk indicator. For Wilcox this is illustrated in a population by birthweights falling outside the Gaussian birthweight distribution 11. Since smoking slows down fetal growth, even early in pregnancy, it should also play a role, but Barker did not think smoking would have long-lasting health effects on the offspring. Smoking induces asymmetric growth impairment and a loss of fat, but not an impairment of more vital organs. We still do not yet know the full consequences of the smoking epidemic for pregnant women that occurred in the 1960s to the late part of the 20th century. The consequences will become apparent when the children born of these smoking mothers reach the end of their life cycles events, and perhaps only when the children born in cohorts with rich data on prenatal confounding factors reach that age. Wilcox and Barker are two giants in reproductive epidemiology with substantial influence on Nordic epidemiology coming from different backgrounds: Barker from infectious disease epidemiology and chronic disease epidemiology, and Wilcox from reproductive epidemiology with a much stronger basis in theoretical epidemiology. These different levels of analytical sophistication and reading of data continue to be important and a great asset in reproductive health science. Barker′s way of formulating his hypothesis predicted that countries with the highest risk for developing an epidemic of the metabolic syndrome (and related health consequences) would be the countries where fetal nutrition was suboptimal 30–50 years ago; these now 30–50-year-old people were programmed to live in a setting predicted to be characterized by a shortage of food. Had this situation changed in the opposite direction, with food in excess, then obesity, diabetes, acute myocardial infarction and premature death may have been expected to follow. India could well be such a high risk area 12-14 and Barker was active in this research. The future will be more complicated. Studying long-term consequences of prenatal exposures carries numerous challenges that we are only now becoming aware of. Causal links may be direct or mediated and these mediators should not be taken for confounders and should therefore not be adjusted for. An adjustment may not only block a causal path but also induce collider bias. Further, all the “fetal programming” studies assume that the pregnancy will end in a live-born child, often neglecting that 30% of pregnancies may end in a miscarriage and that the exposures of interest may also influence the risk of fetal death and then cause selection bias if the competing event (fetal death) is not properly adjusted for. The problem is that even moderate to large associations may reflect causal paths that differ from those we think we are studying. Barker inspired research all over the world and not least in the Nordic Countries. We have the data, or we will get the data with time, to follow up on the fetal programming theories. We used Barker as a consultant when we established the Danish National Birth Cohort (DNBC), and it is because of his advice that in Denmark we now on a routine basis record placental weight and abdominal circumference. However, the birth cohorts established at the end of the 20th century and beginning of the 21th century were established for many reasons. Before the ideas of fetal programming were established we knew that exposures during fetal life (infections, alcohol, environmental exposures like mercury and many more) could have long-lasting effects. When epigenetics was discovered, a whole new set of avenues opened for research, especially since epigenetic mortification of gene expression may be long-lasting and may perhaps even be carried over to the next generation. The slogan for DNBC (www.DNBC.DK) “Nine months that last a lifetime” may even present too narrow a time slot. Some pregnancy or birth cohorts were established before the Baker hypothesis, but these cohorts grew rapidly as a result of his ideas and substantial funding was given to the fetal programming research from large funding agencies such as NIH or the EU Biomedical Research program. Barker had a brilliant career starting as a student with a publication in Nature on testosterone and bone cancer. He was more interested in biology than in statistics and he believed in observational research. He believed that the associations he saw were causal and had a biological mechanism behind them. He considered it strange to keep testing null-hypotheses when a potential causal pattern had emerged. He was upset to see how much of the research funding in the early days went to “anybody who could spell DNA”. He was a “old school shoe leather epidemiologist” who in the beginning of his career went to Uganda to study Mycobacterium ulcerans infections (Buruli ulcer), only leaving when the political tension in the country made research impossible. As many paradigm turners, he did not spend much time in attempts to falsify his theories. He was very aware of the importance of communicating his ideas not only to his scientific colleagues but also to a wider audience. He published a series of books, with BMJ as publisher, starting in later life (from 1994) with Mothers, Babies and Diseases. Time may well show that many of his ideas are wrong but the research they inspired is important and it is to a large extent his achievement that we have advanced the field as much as we have. Fetal programming is an area of research that provides a good fit for what we can do in the Nordic Countries in the years to come. We should put up a statue or name a street after him.