Commentary: a hypothesis challengedBMJ 1997; 315 doi: https://doi.org/10.1136/bmj.315.7119.1348 (Published 22 November 1997) Cite this as: BMJ 1997;315:1348
- a Department of Ambulatory Care and Prevention, Harvard Medical School and Harvard Pilgrim Care, 126 Brookline Avenue, Suite 200, Boston, MA 02215, USA
- Correspondence to: Dr Rich-Edwards
Critics have charged David Barker, champion of the hypothesis that the environment in the uterus programmes the risk of adult disease,1 with overenthusiastic inductive reasoning as well as inattention to issues of selection bias and confounding.2 A growing number of independent epidemiologists, however, have confirmed associations between birth weight and hypertension,3 4 diabetes mellitus,3 5 6 7 cardiovascular disease,8 9 and breast cancer.10 11 12 These associations have been observed in various settings and have proved robust to adjustment for socioeconomic and lifestyle factors. Thus increasing attention is being paid to what underlies these observations; Barker himself has invoked “undernutrition” of the fetus to explain the cardiovascular outcomes.1 Now Stanner and colleagues put this explanation to the test by measuring cardiovascular risk factors among adults who were exposed to severe famine in the siege of Leningrad while in utero or in their first year of life. Their results are resoundingly null. Yet, before proclaiming this study a lethal blow to the hypothesis of in utero programming, consider whether this was a fair test of a clearly delineated hypothesis.
Stanner and colleagues' study is ecological as it lacked data on the exposure status of individuals. Thus the authors were forced to assume that every mother pregnant during the siege was malnourished and that every mother who delivered before the siege was not. The greater the deviation from this assumption, the more a bias to the null would result. At the appalling siege ration of 300 carbohydrate calories/day, however, little cause exists to second guess the extraordinary deprivation experienced by those enduring the siege. Indeed, the mean birth weight of those exposed to the siege in utero was 700 g less than those born before the siege. During the siege 27% of pregnancies delivered in Leningrad's hospitals were stillborn or perished in their first month.13 The selective survival implied by such a grim mortality might also raise our suspicions of bias. Yet, according to the Barker hypothesis, programming should be most evident among those fetuses who successfully downregulated their growth and survived. Among those who survived infancy during the siege, 64% could not be traced or declined to participate. However, the authors found no evidence for the rather implausible scenario necessary to produce biased results from loss to follow up: that they had inadvertently “over-enrolled” healthy adults who had been exposed to the siege in utero or had “under-enrolled” ill adults who were not exposed in utero.
With some assurance of a fair test, then, what was the hypothesis? As both exposure groups experienced famine during infancy, the study measures the additional impact of having been starved in utero. One interpretation of the null result is that starvation at any point between conception and an infant's first birthday is sufficient to exert programming effects. At the other extreme, perhaps only starvation during a certain period of gestation is relevant, and the exposed group was too broadly defined to detect an effect. Or specific dietary imbalances, rather than sheer starvation, may programme the fetus. Another alternative is that intrauterine programming results from factors other than maternal diet. Finally, these null results could be evidence that intrauterine programming is a chimera of confounding by social class, as rich and poor seem to have been equally starved in the Leningrad siege.
Such broad latitude in interpretation invites new, more specific hypotheses to explain the observed associations between measures of fetal growth and the risk of adult disease. Several recent theories implicate maternal and fetal hormones14 15 as well as trimester-specific effects of maternal diet.1 Historical data may prove too blunt to test these increasingly specific hypotheses. Other study designs, including animal experiments and prospective studies of mothers and their offspring, are likely to yield more definitive results. These studies are important to do, because ensuring optimal wellbeing for young women may prove a powerful way to promote the lifelong health of their children.