Invited Commentary: Association between Restricted Fetal Growth and Adult Chronic Disease: Is It Causal? Is It Important?

The article by Leon et al. (1) adds to the large and expanding body of literature reporting associations between restricted fetal growth and subsequent adult chronic disease, particularly hypertension, type 2 diabetes, and coronary heart disease. In view of the long delay between exposure and outcome in these studies, the reported associations are remarkably robust. Beginning with ecologic studies by Forsdahl (2) in Norway and later by Barker and Osmond (3) in the United Kingdom, subsequent historical and cohort studies from Barker's group based on individual-level analyses have replicated the original findings (4–6). Moreover, the associations have also been confirmed by many different groups of investigators in several countries, including the United Kingdom (7, 8), the United States (9–12), Sweden (13, 14), Finland (15), India (16), Zimbabwe (17), South Africa (18), and Brazil (19). Initial concerns about the potential for selection bias (20) have been largely dispelled by studies with small losses to follow-up, such as the one reported Leon et al. (1) in this issue of the Journal. Besides its low risk of selection bias, the study by Leon et al. improves on previous studies in two additional respects. First, its large sample and use of birth weight for gestational age z scores enable the authors to isolate a small, but independent, effect of gestational age on systolic blood pressure, with advancing gestational age reducing systolic pressure by approximately 0.25 mmHg per week. Previous studies have either restricted their analyses to term births or have had inadequate sample sizes to detect such small effects. Second, the authors have examined the independent effect of birth length for gestational age above and beyond the effect of birth weight. Once weight for gestational age is accounted for in their multivariate regressions, length appears to play no additional role, which the authors interpret as showing that the “fetal programming” of later blood pressure is a function of “accretion of fetal soft-tissue mass rather than of linear bone growth” (1, p. 602). Now that the association between restricted fetal growth and subsequent adult chronic disease has proven robust and has been further refined, can it be accepted as causal? Animal studies have provided some support for a causal relation. The strongest evidence comes from an experimental model of maternal protein restriction in rats, which appears to inhibit placental 11 β-hydroxysteroid dehydrogenase type 2 and thereby increases fetal exposure to maternal glucocorticoids (21, 22). Unfortunately, most epidemiologic investigators in this field seem more content with replicating the associations predicted by the fetal programming hypothesis rather than devising critical attempts to falsify the hypothesis, and important doubts remain about the causal nature of the association. These doubts concern issues of confounding, causal pathways, and conflicting evidence from “natural experiments.” Confounding by socioeconomic status (SES) remains a concern in studies of the programming hypothesis. Many studies have controlled for social class, occupation, or education at adulthood, i.e., at the time the chronic disease outcome is ascertained. A few have also controlled for SES at the time of birth (7, 8, 13), but most have not. Although SES is strongly associated with both fetal growth and adult chronic disease, it is a proxy for more directly acting risk factors with which it is highly associated. For the adult chronic diseases under discussion, these include such potent risk factors as diet, physical activity, smoking, and other behaviors. Diet and physical activity, in particular, tend to “track” over time and accumulate over the life cycle, strongly influencing the risk of hypertension, type 2 diabetes, and coronary heart disease (23). None of the epidemiologic studies in this area, including the one by Leon et al. (1), have had information that would permit adequate control for these potent risk factors. Thus, part of the reported association may be due to low SES-associated family environments that predispose both to suboptimal fetal growth and to postnatal dietary and physical activity patterns (and cigarette smoking) that increase the risk of adult chronic disease. Another source of confounding that might be even more important than those associated with SES relates to the mechanism underlying restricted fetal growth. For example, maternal prepregnancy hypertension and pregnancy-induced hypertension both restrict fetal growth. If the mother's tendency to be hypertensive is passed on to her offspring (either genetically or by means of vascular programming due to the hormonal or metabolic milieu of pregnancy), her offspring will be more likely to develop subsequent hypertension, but the underlying cause of the hypertension would be the mother's hypertension, not the associated reduced fetal growth. Few studies have controlled for this important potential source of confounding (24, 25). Maternal and fetal insulin resistance could underlie the association with type 2 diabetes by reducing fetal growth (which depends heavily on insulin and insulin-like growth factors) and independently lead to glucose intolerance in adulthood (26). Similarly, maternal thrombophilia caused by genetic mutations in methylenetetrahydrofolate reductase, factor V, or prothrombin could (via inheritance or vascular programming) increase the risk of coronary heart disease. Studies by Davey Smith et al. (27, 28) indicating that mothers who give birth to growth-restricted fetuses are themselves at increased risk for coronary heart disease provide some empirical support for such a causal mechanism. Many studies in this area have “controlled” for body mass index, weight, and/or height during adulthood. Associations between fetal growth and adult chronic disease outcomes are usually strengthened (and occasionally only become statistically significant) when such “control” is introduced (7, 15–19, 24, 25, 29). Because infants who are smaller and shorter at birth have a tendency to remain thinner and shorter throughout life, this is tantamount to adjustment for a factor that lies on the causal path between fetal growth and adult chronic disease, although in this case, the causal intermediate (high adult body mass index) is inversely associated with exposure (restricted fetal growth). In other words, statistical “adjustment” for the smaller average adult size of growth-restricted newborns artificially “inflates” their height and body mass index during adulthood and thus inflates the association between restricted fetal growth and adult disease outcome (30). This is the mirror image of overcontrol for intermediates that are positively associated with both the exposure and the outcome. In addition, Lucas et al. (31) have recently pointed out that control for size during adulthood transforms the exposure from restricted fetal growth to enhanced postnatal growth. Infants who are born small but become large by adulthood have experienced larger-than-average growth from birth to adulthood, irrespective of growth that occurred in utero. As with other epidemiologic associations, additional insight about causality can be gained by examining the results of so-called “natural experiments.” For the fetal programming hypothesis, these include studies of twins, famine victims, and migrants. The results of these studies are extremely mixed, but do not lend strong support to the programming hypothesis. Although Dwyer et al. (32) found stronger associations between fetal growth and systolic pressure at age 8 years in twins than in singletons in southern Tasmania, the Dunedin Twin Study (33) found that during adolescence, twins had a significantly lower blood pressures than did otherwise similar singletons. Coronary heart disease (CHD) mortality does not appear to be elevated among Swedish twins (34). Studies of survivors of the Dutch and Leningrad famines have reported opposite findings. Small, but significant, differences were observed in blood pressure among survivors of the Dutch hunger winter of 1944–1945 (35), but no effects were seen when infants born prior to versus during the Leningrad siege of 1941–1944 were compared (36). The British Regional Heart Study found that for both within-United Kingdom and foreign immigrants, the region of the country moved to has a greater effect on a CHD incidence than does the place moved from; nonimmigrants, within-United Kingdom immigrants, and foreign immigrants all exhibited the robust south-to-north gradient in CHD incidence that is well-known in the United Kingdom (37). The results of a more recent study by Strachan et al. (38), however, suggest that place of birth and place of later residence may both play a role. Even if the association between restricted fetal growth and adult chronic disease is causal, how important is the association in terms of public health impact? Here the evidence is consistent in suggesting that restricted fetal growth plays a minor role. Data from Finland clearly indicate a period effect on CHD mortality, with increases in the years after World War II until 1970, after which all birth cohorts experienced a substantial decrease (39). Data from rural Georgia in the early 1960s indicate a higher prevalence of CHD in White men than in Black men and, in Whites, a higher prevalence among those of high SES than among those of low SES (40, 41). These SES and racial gradients are opposite to those observed in more recent decades and would, under the fetal programming hypothesis, implicate an earlier period of lower birth weight in high-SES Whites than in low-SES Whites or in Blacks, but there is no evidence that this occurred. Similarly, the Atherosclerosis Risk in Communities Study has shown decreases in Whites in Washington County, Maryland; Minneapolis, Minnesota; and Forsyth County, North Carolina, and a similar pattern for Blacks in Forsyth County (42). For Blacks in rural Jackson, Mississippi, however, trends between the late 1960s and late 1970s were opposite, with a steady and significant increase in CHD mortality (42). If such an increase were driven primarily by changes in fetal growth, it should have been preceded many decades earlier by a trend toward reduced birth weight over time, a trend for which no evidence exists. Although ecologic studies provide weak evidence for causal relations, a disparity (“ecologic dissonance”) between observed geographic and temporal trends and those expected under the programming hypothesis undermines the public health importance of the effects suggested by the hypothesis. CHD mortality is far higher in northern Europe than in southern Europe despite higher birth weights in the North for many decades (39). CHD mortality has been increasing steadily in Eastern Europe, but no evidence exists of an earlier corresponding downward trend in birth weight as the underlying cause (39). Finally, the rising CHD prevalence in developing countries (43, 44) seems largely explainable by adult-level risk factors, with little “room” available for independent contributions from restricted fetal growth (45). This ecologic dissonance does not preclude a true (causal) effect of fetal growth on subsequent CHD, but it does suggest that such an effect is dwarfed by adult lifestyle factors, such as diet, physical activity, and smoking. This ecologic evidence is actually consistent with the very small, albeit statistically significant, effects reported by Leon et al. (1) and other epidemiologic studies. In the study by Leon et al., a 1-kg increase in gestational age-adjusted birth weight was associated with a fall of 1.6 mmHg in systolic pressure. Although a relatively small reduction in mean blood pressure can substantially reduce the prevalence of hypertension (46), a 1-kg difference in birth weight is enormous; it is equivalent to five or six times the difference in mean birth weight between infants of smoking and nonsmoking mothers. A more achievable public health target would be about 100 g. Such an increase, however, would prevent only about 2.0–2.5 percent of CHD deaths (47). The article by Leon et al. (1) is an excellent contribution to a large and growing body of evidence demonstrating an association between restricted fetal growth and later adult chronic disease. Concerns about confounding by lifestyle factors and for the mechanism underlying the fetal growth restriction, statistical adjustments for adult body size, and conflicting evidence from natural experiments raise questions about whether such an association is truly causal, however. Moreover, even if the association is causal, the evidence of period versus birth cohort effects and of geographic and temporal trends suggests that the public health importance of such an association is likely to be small, consistent with the small effect on blood pressure reported by Leon et al. Future epidemiologic studies should devise critical tests of the fetal programming hypothesis rather than merely replicate its predicted associations. More basic scientific investigation should also be encouraged in order to understand the mechanisms underlying the reported associations, since such an understanding may provide valuable clues to the prevention of hypertension, type 2 diabetes, and CHD irrespective of the effects of fetal programming. Dr. K. S. Joseph provided helpful comments on an earlier draft of this commentary. Dr. Kramer is a Distinguished Scientist of the Canadian Institutes of Health Research.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.