- DOHaD
developmental origins of health and disease
The field of research currently known as the developmental origins of health and disease (DOHaD) grew from initial observations that infant mortality rates at the turn of the last century were highly correlated with standard mortality rates approximately 60 years later(Reference Forsdahl1, Reference Barker and Osmond2). One explanation given for these findings is that a ‘nutritional deficit’ operating in early life results in ‘a lifelong vulnerability to aspects of an affluent adult lifestyle’(Reference Forsdahl1). After 30 years the field of DOHaD has grown into a vibrant multi-disciplinary research area, although controversy remains about the nature of the early-life nutritional deficits and the mechanisms through which they might affect later health.
CVD is one of the main diseases of affluence faced by the world today. Evidence that CVD risk may be related to early-life, specifically intrauterine, exposures was initially derived from the inverse association between birth weight and CVD mortality(Reference Barker3–Reference Leon, Lithell and Vagero5) and/or risk factors such as blood pressure(Reference Huxley, Shiell and Law6, Reference Adair and Dahly7) insulin resistance(Reference Barker8) and the metabolic syndrome(Reference Barker, Hales and Fall9). One of the most-widely-cited hypotheses, termed the ‘thrifty phenotype’(Reference Hales and Barker10), suggests that when exposed to deprived conditions during development fetal adaptations (metabolic, endocrine and/or anatomical) occur that allow for immediate survival but may be maladapted to cope with the demands of a more-affluent lifestyle in later life (Fig. 1). Maternal nutrition is one of the potentially-modifiable determinants of fetal nutrition and is thus a focus of research seeking to understand the association between early life and later disease.
Schematic diagram of the ‘thrifty phenotype’ hypothesis showing potential pathways to type 2 diabetes and hypertension. (Adapted from Hales & Barker(Reference Hales and Barker10).)
A wealth of animal studies, extensively reviewed elsewhere(Reference Fernandez-Twinn and Ozanne11–Reference McArdle, Andersen and Jones13), have confirmed that perturbations in the maternal diet during pregnancy can have a profound impact on offspring disease risk. Pregnant rats fed a low-protein diet give birth to offspring that develop hypertension, which persists into adult life(Reference Langley-Evans, Welham and Jackson14). Indeed, some or all the components of the metabolic syndrome can be produced by a variety of maternal exposures, including global nutrient restriction(Reference Vickers, Breier and Cutfield15), placental insufficiency(Reference Simmons, Templeton and Gertz16) and a low-Fe diet(Reference Gambling, Dunford and Wallace17) as well as a low-protein diet. The animal data have allowed for increasingly sophisticated exploration of potential mechanisms and have greatly enhanced the DOHaD field. Nevertheless, the translation of these findings into human studies remains challenging.
The majority of data from human subjects still centres on the inverse association between birth weight and later disease. These observations are primarily from retrospective cohort studies and the evidence base has been criticised for inappropriate statistical methods(Reference Lucas, Fewtrell and Cole18), publication bias and inadequate adjustment for socio-economic status(Reference Huxley, Neil and Collins19). Perhaps the most important criticism is that birth weight is a poor measure of exposure, primarily as it is multifactorial in origin, prompting calls to move towards more direct assessments of exposure. Studies relating maternal diet in pregnancy and offspring CVD risk are rare, however, and have shown little or no associations to date(Reference Leary, Ness and Emmett20, Reference McGarvey, Zinner and Willett21). Furthermore, cohort studies investigating diet during pregnancy retain many of the same study design issues as those that have utilised birth weight. The Dutch Hunger Winter Study (1944–5) has been used as a pseudo-experiment with which to study the impact of nutritional deprivation during pregnancy on offspring disease risk. Results focusing on CVD risk have been inconsistent(Reference Roseboom, de Rooij and Painter22–Reference Ravelli, Stein and Susser24), however, and by their nature the cohorts have widely-recognised design issues; including large loss to follow-up and a lack of data on individual dietary and/or nutrient intakes.
One emerging resource in this field is the follow up of trials of maternal supplementation during pregnancy. Although primarily conducted to evaluate the effectiveness of interventions to improve pregnancy and birth outcomes, these trials represent a useful resource for the DOHaD field by providing randomised controlled trial data. To date, the only exposures that have been studied in this manner are protein–energy, multiple-micronutrient and Ca supplementation. The remainder of the present paper will summarise these follow-up studies insofar as they relate to offspring CVD risk.
Protein–energy
In the late 1960s the Institute of Nutrition of Central America and Panama conducted a community trial of protein supplementation in rural Guatemala. Two villages received the ‘Atole’ supplement, a protein–energy-dense drink, whilst two control villages received the ‘Fresco’ supplement, a drink that contained no protein and approximately one-third of the energy(Reference Martorell, Habicht and Rivera25). The study design is summarised in Table 1; supplement drinks were provided twice daily and their consumption by pregnant women and children up to the age of 7 years was recorded. A number of follow-up studies have been conducted on the trial participants and the findings have been extensively summarised(Reference Martorell26). Findings most relevant to CVD risk include a marginally-greater fat-free mass in adolescent girls born in intervention villages(Reference Rivera, Martorell and Ruel27), reduced fasting glucose in 25-year-old women born in intervention villages(Reference Conlisk, Barnhart and Martorell28) but no difference in blood pressure between individuals born in intervention and control villages(Reference Webb, Conlisk and Barnhart29) (Table 2). A major limitation of this study in relation to the current review is that it is not possible to distinguish between pre- and postnatal enhanced nutrition in the intervention analysis. Furthermore, because of the small number of villages the cluster design of the original trial has not been accounted for in the analysis. Although often viewed as a trial of protein supplementation, the two drinks also provided different levels of certain micronutrients (Ca, P, Zn, folic acid and vitamin B12), which may affect interpretation of the results(Reference Martorell26).
Characteristics of trials and follow-up studies of maternal protein–energy supplementation
IGF, insulin-like growth factor; IGFBP-3, IGF-binding protein.
Effect of maternal protein–energy supplementation on CVD risk factors in the offspring from trials conducted in Guatemala(Reference Rivera, Martorell and Ruel27–Reference Webb, Conlisk and Barnhart29), Wales, UK(Reference Ben-Shlomo, Holly and McCarthy30) and The Gambia(Reference Hawkesworth, Prentice and Fulford34, Reference Hawkesworth, Prentice and Fulford35)Footnote *
FFM, fat-free mass; IGF, insulin-like growth factor; IGFBP-3, IGF-binding protein; M, males; F, females; NR, not reported; IQR, interquartile range.
* S Hawkesworth, AM Prentice, AJC Fulford and SE Moore (unpublished results).
† Whole-body fat assessed by bioelectrical impedance analysis utilising population-specific prediction equations.
‡ Calculated from prediction equations using arm diameter for males and waist circumference for females.
§ Because of skewness of data these values are the median and IQR.
‖ Unadjusted regression analysis unless otherwise stated.
¶ Generalised estimating equations used to take village clustering into account.
** Regression adjusted for age and gender.
The Barry Caerphilly Growth Study is a follow up of children whose mothers took part in a trial to improve their cow's milk intake during pregnancy that was conducted in South Wales, UK in the early 1970s(Reference Ben-Shlomo, Holly and McCarthy30) (Table 1). Women were randomised to receive milk tokens throughout pregnancy and for the first 5 years of their child's life or to receive no tokens (control). Milk tokens were equivalent to 284 ml milk/d from the milkman. At 25 years of age the offspring were enrolled into a follow-up study and it was found that those whose mothers had received the intervention had lower serum levels of insulin-like growth factor 1 compared with those born to control mothers (Table 2)(Reference Ben-Shlomo, Holly and McCarthy30). Insulin resistance(Reference Sandhu, Heald and Gibson31) and IHD(Reference Juul, Scheike and Davidsen32) have been associated with low insulin-like growth factor 1 concentrations in observational studies. For the purposes of the present review milk tokens have been viewed as a protein–energy supplement, but it should be acknowledged that the increased consumption of Ca and/or growth factors may also be relevant.
In 2005 a follow-up study was conducted in The Gambia, West Africa involving offspring (aged 11–17 years) who had been born during a randomised trial of supplementation with protein–energy-dense biscuits during pregnancy or lactation. The original trial(Reference Ceesay, Prentice and Cole33) provides some of the strongest evidence yet that protein–energy supplements can improve birth weight in nutritionally-deprived populations. The trial used a cluster randomised design in twenty-eight villages; pregnant women from intervention villages received protein–energy-dense biscuits from 20 weeks of gestation until delivery whilst women in control villages received the same supplement for 20 weeks post partum. The supplement given during pregnancy was shown to increase birth weight by 136 g overall, with a greater difference of 201 g during the nutritionally-poor ‘hungry season’ (June–October)(Reference Ceesay, Prentice and Cole33). The follow-up study has found no evidence that these early-life differences correspond to later differences in risk profile in the offspring; no effect of maternal supplementation during pregnancy on offspring blood pressure(Reference Hawkesworth, Prentice and Fulford34), body fat(Reference Hawkesworth, Prentice and Fulford35) or fasting insulin and glucose was found (S Hawkesworth, AM Prentice, AJC Fulford and SE Moore, unpublished results; Table 2). The limitation of this study is that the women in the control arm were provided with the same protein–energy-dense biscuits during lactation. However, as there is little evidence to suggest that supplementation of lactating women affects breast-milk quality or quantity(Reference Prentice, Roberts and Prentice36, Reference Gonzalez-Cossio, Habicht and Rasmussen37) the comparison is considered to be appropriate. Although no overall effect of maternal protein–energy supplementation on offspring CVD risk was found, an interaction with body composition was found. For offspring who were relatively lean (in the lowest quartile of percentage body fat) at follow up the intervention was associated with raised systolic blood pressure (Fig. 2). It is possible that providing supplements to the mother whilst the fetus is developing provides better conditions than those that are experienced postnatally, although only for offspring who remain particularly lean. This interaction may reflect the ‘mismatch’ theory, which suggests that if conditions during in utero development are ‘mismatched’ to the later environment this mismatch promotes the development of disease(Reference Gluckman and Hanson38). The pattern more often seen in the literature is the interaction between low birth weight and overweight in adulthood(Reference Gluckman and Hanson38); in the less-developed setting of The Gambia it seems that the reverse of this pattern may have been observed.
Interaction between body composition (percentage body fat) and maternal protein–energy supplementation for systolic blood pressure (BP) in Gambian adolescents at follow up. Values are means with 2 se represented by vertical bars. (Reproduced from Hawkesworth et al. (Reference Hawkesworth, Prentice and Fulford34))
Multiple micronutrients
In recent years there has been considerable interest in the potential of maternal micronutrient supplementation during pregnancy to provide important benefits for both the mother and her infant. Recently, the results of the many trials have been summarised, with the focus on short-term outcomes such as birth weight and infant survival(Reference Bhutta, Ahmed and Black39). Only one trial has so far published findings of follow up beyond infancy (Table 3), reporting lower systolic blood pressure and slightly greater triceps skinfold thickness for 2–3-year-old children born to Nepalese mothers who received a combination of fifteen micronutrients compared with ‘control’ women receiving only Fe and folic acid(Reference Vaidya, Saville and Shrestha40) (Table 4). The authors suggest caution in the interpretation of this finding from a single trial and recognise the importance of replication in other studies(Reference Vaidya, Saville and Shrestha40). It should also be noted that there was no difference in diastolic blood pressure, and that blood pressure is notoriously difficult to measure in very young children. Similar follow-up studies in Bangladesh(Reference Tofail, Persson and El Arifeen41) and a different area of Nepal(Reference Christian, Khatry and Katz42) are currently nearing completion and will soon add to the current limited knowledge.
Characteristics of a Nepalese trial and follow-up study of maternal multiple-micronutrient supplementation
Effect of maternal micronutrient supplementation on CVD risk factors in the offspring in the follow up(Reference Vaidya, Saville and Shrestha40) to the original Nepalese trial(Reference Osrin, Vaidya and Shrestha62)
* Intervention arm: fifteen multiple micronutrients (800 μg vitamin A, 10 mg vitamin E, 5 μg vitamin D, 1·4 mg thiamin, 1·4 mg riboflavin, 18 mg niacin, 1·9 mg vitamin B6, 2·6 μg vitamin B12, 400 μg folic acid, 70 mg vitamin C, 30 mg Fe, 15 mg Zn, 2 mg Cu, 65 μg Se, 150 μg iodine).
† Control arm: 60 mg Fe and 400 μg folic acid.
‡ Unadjusted regression analysis.
Calcium
A number of maternal Ca supplementation trials have been conducted on pregnant women in recent years, primarily to investigate the potential for reducing the risk of pre-eclampsia(Reference Villar, Abdel-Aleem and Merialdi43, Reference Hofmeyr, Duley and Atallah44). To date three trials have published follow-up data on the offspring, focusing on blood pressure as an outcome(Reference Belizan, Villar and Bergel45–Reference Hiller, Crowther and Moore47) (Table 5). The follow-up of a trial conducted in the USA has provided evidence that maternal Ca supplementation is associated with lower systolic blood pressure in the offspring at 2 years, although only 10% of eligible participants were recruited(Reference Hatton, Harrison-Hohner and Coste46). Data from The Gambia (S Hawkesworth, Y Sawo, AJC Fulford, GR Goldberg, LMA Jarjou, A Prentice and SE Moore, unpublished results), Australia(Reference Hiller, Crowther and Moore47) and Argentina(Reference Belizan, Villar and Bergel45) have shown no association between maternal Ca supplementation and offspring blood pressure at 5–10, 4–7 and 5–9 years respectively (Table 6). Despite no overall association with mean blood pressure, in Argentina the intervention was found to be associated with a reduced risk of having high systolic blood pressure (defined by age- and height-specific cut-offs)(Reference Belizan, Villar and Bergel45).
Characteristics of trials and follow-up studies of maternal calcium supplementation
* S Hawkesworth, Y Sawo, AJC Fulford, GR Goldberg, LMA Jarjou, A Prentice and SE Moore (unpublished results).
Effect of maternal calcium supplementation on CVD risk factors in the offspring from trials conducted in USA(Reference Hatton, Harrison-Hohner and Coste46), Argentina(Reference Belizan, Villar and Bergel45), Australia(Reference Hiller, Crowther and Moore47) and The GambiaFootnote *
BP, blood pressure.
* S Hawkesworth, Y Sawo, AJC Fulford, GR Goldberg, LMA Jarjou, A Prentice and SE Moore (unpublished results).
† Defined by age, gender and height centile specific cut-offs(67).
‡ Unadjusted regression analysis unless otherwise stated.
In the Argentinian study an interaction was also found between childhood BMI and maternal Ca supplementation on offspring blood pressure(Reference Belizan, Villar and Bergel45). For children with a BMI above the mean at follow up the intervention was shown to be associated with lower blood pressure(Reference Belizan, Villar and Bergel45). In the data from The Gambia an interaction with body composition was observed, but in the opposite direction; for individuals in the highest quartile of percentage trunk fat at follow up maternal Ca supplementation was found to be associated with raised diastolic blood pressure (S Hawkesworth, Y Sawo, AJC Fulford, GR Goldberg, LMA Jarjou, A Prentice and SE Moore, unpublished results).
Interpretation and implications
The interpretation of the data presented in the present review is that, to date, with the possible exception of multiple micronutrients, there is little evidence to suggest that supplementation of pregnant women affects their offspring's risk factors for CVD. It is therefore important to explore the reasons for this outcome. The first point that should be highlighted is that there are very little data in this area and more studies are required, including studies with nutrients for which there is currently no evidence (e.g. Fe, vitamin A, long-chain PUFA) or for which the evidence base is weak (protein–energy).
The animal data in this field of DOHaD in relation to CVD risk are extremely strong and many mechanisms have been identified that explain the impact of nutrient restriction during pregnancy on disease risk in the offspring(Reference Fernandez-Twinn and Ozanne11, Reference Langley-Evans12). It may therefore seem surprising that, to date, these findings have not been replicated in human subjects. However, the majority of animal studies involve interventions that are at the extreme end of nutrient restriction; equivalent treatments will not be given in human intervention studies because they are both unethical and not relevant to real life. In addition, animal studies have focused on restricting nutrients whereas studies of human pregnancy are designed to increase intake relative to control or placebo, therefore making the results difficult to compare. An additional explanation may be that the incremental nutrient requirements for human reproduction are very low compared with other species (because human subjects have evolved to have very slow pre- and postnatal growth rates) and the developing fetus is protected by a number of evolved adaptations(Reference Prentice and Goldberg48).
One relevant issue for interpretation is that of sample size. For example, in order to detect a difference of 2 mmHg in systolic blood pressure (commonly recognised as the population effect size associated with 1 kg higher birth weight(Reference Adair and Dahly7)) with 80% power at P<0·05 a study would require a sample size of 948 (474 per arm) if for blood pressure the sd is 11 mmHg, a value reported recently for Swedish military conscripts aged 18 years(Reference Lawlor, Hubinette and Tynelius49). Only a small number of follow-up studies in the present review have larger sample sizes than 474 per arm, highlighting again the importance of further research in this area.
Another issue is that studies often suffer from large losses to follow up (for example, see Hatton et al. (Reference Hatton, Harrison-Hohner and Coste46)), particularly if conducted in developed countries where individuals are difficult to trace. It has recently been discussed that it is necessary to allow for much larger attrition rates in DOHaD studies than are usually deemed to be acceptable for randomised controlled trials involving drugs(Reference Fewtrell, Kennedy and Singhal50). This requirement is mainly the result of the practicalities of tracing individuals several years after their mothers were enrolled into a trial. It has been argued that provided information on attrition rates and the association of characteristics with loss to follow up are reported, then a judgement can be made on the quality of the data presented, even with large losses to follow up(Reference Fewtrell, Kennedy and Singhal50). The majority of studies presented in the current review were conducted in developing country settings where close family ties can mean that tracing subjects is less problematic. This factor is reflected in the retention rates, most of which are >60%.
A number of the original trials presented here have their own limitations, which question the usefulness of associated follow-up data. Two of the protein–energy randomised controlled trials were randomised at the community level(Reference Martorell, Habicht and Rivera25, Reference Ceesay, Prentice and Cole33) and one of these trials (Institute of Nutrition of Central America and Panama, Guatemala(Reference Martorell26)) was unable to take clustering into account for statistical analysis. Both the trial in Guatemala(Reference Martorell, Habicht and Rivera25) and that in Wales(Reference Ben-Shlomo, Holly and McCarthy30) provided supplementation for infants and children as well as pregnant women, making it impossible to distinguish between pre- and postnatal intervention. In the multiple-micronutrient trial in Nepal(Reference Vaidya, Saville and Shrestha40) ‘control’ women received Fe and folic acid supplements that contained double the amount of Fe received by the multiple-micronutrient group. The authors raise this factor as an issue; any effect of supplementation could be the result of a lower Fe intake rather than an increase in other micronutrients(Reference Vaidya, Saville and Shrestha40). Only the Ca trials are true double-blind placebo-controlled trials and even then the background Ca intakes vary greatly between the populations studied.
Another factor that may explain the lack of association seen between maternal supplementation and offspring CVD risk is the timing of the follow-up studies. In the majority of studies subjects are children or adolescents. It may be that the effect of maternal supplementation only becomes apparent when individuals reach adulthood. In addition, the context of the study is important. The lack of association seen in the studies in rural areas of The Gambia (a transition country) may reflect the fact that the prevailing nutritional exposures experienced postnatally have yet to reach the levels seen in more affluent countries; such levels may be required to reveal a disease response. Finally, pregnancy represents just one time point within the life course and may not be the optimum time to intervene to influence offspring health. Fetal nutrition is influenced by the mother's nutrient stores and by her own exposure in utero (Reference Hypponen, Power and Smith51); the window of opportunity during pregnancy may actually be too late to affect offspring risk.
The final explanation is that it may actually be the early postnatal environment that explains the inverse association between birth weight and disease, a possibility that features strongly in the DOHaD literature. It has been proposed that this association, which is often only apparent after adjustment for current size, can be explained by the rate of growth of the individual after birth(Reference Lucas, Fewtrell and Cole18, Reference Singhal, Cole and Fewtrell52). The evidence from randomised controlled trials of infant feeding indicates that faster postnatal growth is associated with a range of CVD risk factors including blood pressure(Reference Singhal, Cole and Lucas53) and insulin resistance(Reference Singhal, Fewtrell and Cole54).
The data presented here may question the importance of the fetal environment for the programming of human CVD risk in later life. However, only some aspects of the fetal environment have been reviewed, that of the maternal diet insofar as it is affected by supplementation with protein–energy, multiple micronutrients and Ca. Other exposures such as maternal smoking(Reference Brion, Leary and Lawlor55), maternal age(Reference Brion, Leary and Lawlor55), stress(Reference Mastorci, Vicentini and Viltart56) or placental function(Reference Salafia, Charles and Maas57) during pregnancy have not been considered here and may influence fetal systems as they develop. However, supplementation has the potential to be more easily adopted into public health strategy than altering other exposures that influence the fetal environment.
It has been suggested that the DOHaD field is of particular relevance to less-developed countries that carry the greatest burden of growth-retarded infants whilst concurrently experiencing an increasingly rapid nutrition transition(Reference Prentice and Moore58). Concerns have been raised that intervening in pregnancy in such settings may act to increase the risk of obesity and related disorders(Reference Yajnik59). Perhaps one of the positive conclusions that can be drawn from the present review is that maternal interventions during pregnancy do not seem to promote adverse CVD risk in the offspring, at least within the age-groups presented here. Furthermore, supplementation during pregnancy confers short-term benefits, and may confer long-term benefits, to the offspring that are outside of the scope of the review. For example, in Bangladesh there is some evidence that food and multiple-micronutrient supplementation of undernourished women during pregnancy may improve offspring cognitive development(Reference Tofail, Persson and El Arifeen41).
Concluding remarks
Conducting follow-up studies of offspring born to mothers who have participated in supplementation trials during pregnancy will provide a useful evidence base for the DOHaD field. To date there is little evidence to suggest that maternal protein–energy or Ca supplementation affects offspring CVD risk factors (blood pressure, body composition, fasting glucose), but further data are required. One study of the impact of maternal multiple-micronutrient supplementation suggests an effect on offspring systolic blood pressure, but requires replication in other settings.
Acknowledgement
S. H. prepared the manuscript and has no conflict of interest to declare. S. H. is a research degree student at the London School of Hygiene and Tropical Medicine, supervised by Dr Sophie Moore and Professor Andrew Prentice. Dr Ann Prentice and Dr Gail Goldberg, MRC Human Nutrition Research, Cambridge, UK, are also closely involved in her research and have provided valuable input into this manuscript. Fieldwork in The Gambia conducted by S. H. was financially supported by the EU Sixth Framework Programme for Research and Technical Development of the EU Community ‘Early Nutrition Programming Project’ (FOOD-CT-2005–007036) and by the UK Medical Research Council. This fieldwork was made possible through the dedicated work of staff at MRC Keneba, The Gambia, particularly Landing Jarjou, Yankuba Sawo, Kabiru Ceesay, Morikebba Sanyang, Kalilu Sanneh, Saul Jarjou and Sheriff Kolley.
