SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Contrasting maternal and paternal exposure associations with offspring outcomes
  4. Timing of exposure
  5. Discordant sibling exposure
  6. Maternal genotype as an indicator of intrauterine environment: Mendelian randomization studies
  7. Non-genetic instrumental variables
  8. Conclusions
  9. Acknowledgements
  10. References

Abstract:  The influence of factors acting during the intrauterine period on health outcomes of offspring is of considerable research and public health interest. There are, however, methodological challenges in establishing robust causal links, because exposures often act many decades before outcomes of interest, may act before it is evident that women are pregnant and would enter pregnancy birth cohorts, and may also be strongly related to other factors, generating considerable degrees of potential confounding. The degree of confounding can sometimes be estimated by comparing the association between exposures experienced by the mother during pregnancy and outcomes among the offspring with the association of exposures experienced by the father during the pregnancy period and offspring outcomes. If the effects are due to an intrauterine exposure, then maternal exposure during pregnancy should have a clearly greater influence than paternal exposure. A different approach is that of Mendelian randomization, which utilizes genetic variants of known functional effect that can proxy for modifiable exposures. If carried by the mother, these variants would influence the intrauterine environment experienced by her offspring. These genetic variants are stable over time and can be assessed after pregnancy is complete or even after outcomes in the offspring have been observed. The variants would also not generally be related to potential confounding factors. Other epidemiological strategies are briefly reviewed. It is concluded that the naïve acceptance of findings utilizing conventional epidemiological methods in this setting is misplaced.

There is considerable toxicological and epidemiological interest in the proposition that exposures acting during pregnancy have long-term consequences for health in adulthood of offspring. Exposures that have been investigated include maternal diet, smoking, alcohol use and a variety of environmental toxins. Nearly all domains of later health experience – including obesity, cardiovascular disease, various cancers, respiratory disease, psychiatric disorders and cognitive function – have been associated with intrauterine exposures of one kind or another. In public health policy terms, it is important to separate out causal associations, which offer the possibility of intervention and disease prevention, from non-causal associations. If the non-causal associations are mistaken for causal associations, this could lead to misguided strategies that at best waste resources and divert attention from effective approaches, and at worst could have health-damaging consequences.

There are particular issues that render the epidemiological study of the influence of exposures acting during the intrauterine period on later health outcomes problematic. These relate particularly to the long time-gap between exposure and outcome. First, the assessment of exposure may be difficult, because retrospective approaches may be required to obtain information regarding exposures acting many decades before the health outcome is observed. There are few truly prospective studies with detailed and continuous data – including biological measures – from pregnancy (or before) through to late adulthood. Inaccuracy of exposure assessment can lead to random errors, which will attenuate effect estimates and make it more difficult to establish robust associations [1]. This will lead to studies being underpowered, but will not in general lead to spurious associations being observed. Perhaps more seriously, retrospective assessment may be biased by knowledge of later health outcomes. For example, if people with and without respiratory disease are asked about the smoking habits of their mothers then those with disease may have a tendency to report greater levels of parental smoking, because they will be aware of possible links between such exposures and their own disease state. Biased exposure reporting can, in this situation, generate spurious associations where none actually exist.

A further consequence of the long time-gap between exposure and outcome is that even when associations are observed from well-conducted prospective studies, and therefore likely to be robust, their relevance to contemporary pregnant women is unclear. Thus, studies of components of maternal diet (such as fish intake) many decades ago will have uncertain application to contemporary cohorts, because the types of fish being eaten, their nutrient composition and toxicological load, is likely to have changed substantially.

Confounding is an issue in all observational studies but could be a particular problem in studies of the intrauterine origins of disease, because it may be that confounding factors were inadequately measured at the time of exposure (particularly in historical cohort studies based on existing data). An intrauterine exposure of interest may be related to other factors that have health consequences, and the effects of these extraneous factors could be misinterpreted as causal effects of the exposure that is being investigated. Confounding and bias certainly generate links in observational epidemiological studies of adulthood risk factors and disease outcomes. Consider cardiovascular disease, where observational studies suggesting that β-carotene [1], vitamin E supplements [2,3], vitamin C supplements [4] and hormone replacement therapy [5] were protective were followed by large randomized controlled trials (RCT) showing no such protection [6–11]. In each case, special pleading was advanced to explain the discrepancy – were the doses of vitamins given in the trials too high, too low or of too short duration to be comparable to the observational studies [12,13]? Did hormone replacement therapy use start too late in the trials [14]? Were differences explained by duration of follow up or other design aspects [15]? Were interactions with other factors such as smoking or alcohol consumption key? Rather than such particular explanations being true (with the happy consequence that both the observational studies and the trials had got the right answers, but to different questions), it is likely that a general problem of confounding – by lifestyle and socio-economic factors, or by baseline health status and prescription policies – is responsible [16–19]. Indeed, in the vitamin E supplements example, the observational studies and the trials tested precisely the same thing. Figure 1A and B show the findings from observational studies of taking vitamin E supplements [2,3] and a meta-analysis of trials of supplements [20]. The point here is that the observational studies specifically investigated the effect of taking supplements for a short period (2–4 years) and found an apparent robust and large protective effect, even after adjustment for confounders. The trials tested randomization to essentially the same supplements for the same period, and found no protective effect.

image

Figure 1. (A) Observed effect of duration of vitamin E use compared to no use on coronary heart disease (CHD) events in the Health Professional Follow-up Study [2]. (B) Vitamin E supplement use and risk of CHD in two observational studies [2,3] and in a meta-analysis of RCTs [20] (relative risks and 95% confidence intervals for taking or not taking vitamin E supplements).

Download figure to PowerPoint

Similar evidence of confounding and bias operating in studies of intrauterine exposures and later health outcomes clearly exist. Consider, for example, the influence of maternal diet on offspring health and development. In the Southampton Women's Study, there are very strong associations between diet, socio-economic position and smoking [21]. A quality of diet score (labelled the prudent diet score) was generated, with low scores being characterized by, among other factors, low intakes of fruit and vegetables, wholemeal bread and breakfast cereals, but high intake of chips and roast potatoes, sugar, white bread, processed meat and full-fat dairy products. High scores (representing healthy/prudent diet) were characterized by the opposite dietary pattern. Figure 2 presents the percentage of women with a prudent diet score in the lowest quarter (i.e. the quarter of the study sample with the least healthy diet) according to educational attainment and smoking status. As can be seen there are extremely strong associations between diet and education, with women of child-bearing age with a university degree having around a tenth of the prevalence of poor diet compared to the women with no qualifications, and women who smoke having considerably worse diet than women who do not smoke. This magnitude of association is substantial and could generate strong, non-causal, associations between a ‘prudent’ maternal diet during pregnancy and a wide range of health outcomes among offspring.

image

Figure 2. Percentage of women of childbearing age with prudent diet scores in the lowest quarter according to educational attainment and smoking status [21]. Reprinted by permission.

Download figure to PowerPoint

In observational studies, vitamin C intake during pregnancy has been associated with higher birth weight of offspring [22]; however, data such as those from Robinson et al. [21] would suggest that mothers with higher vitamin C intake during pregnancy would have much lower rates of smoking and be of more privileged socio-economic background, generating substantial confounding. Figure 3 contrasts the results of the association of vitamin C with birth weight from the observational study with those from the largest randomized controlled trial to date in which pregnant women were randomized to a supplement containing vitamin C and E [23]. Findings from the observational study and the randomized controlled trial are clearly not compatible.

image

Figure 3. Comparison of observational epidemiological evidence and randomized controlled trial evidence of the effect of maternal vitamin C intake during pregnancy on birth weight.

Download figure to PowerPoint

Given the inherent difficulties in relating intrauterine exposures to offspring health outcomes, in this MiniReview I will discuss several methods that can be applied in epidemiological studies to increase the ability to make causal inferences. These approaches include the use of maternal/paternal comparisons and the use of genetic and non-genetic instrumental variable approaches. The approaches illustrated here constitute a far from exhaustive list, but illustrate methods that have some general utility. Clearly, RCTs are the gold standard for assessing causality, but in many cases they cannot practically be applied and even when they can be the best observational evidence is required to ensure that the most promising candidates are chosen for evaluation in long-term and expensive RCT experiments.

Contrasting maternal and paternal exposure associations with offspring outcomes

  1. Top of page
  2. Abstract
  3. Contrasting maternal and paternal exposure associations with offspring outcomes
  4. Timing of exposure
  5. Discordant sibling exposure
  6. Maternal genotype as an indicator of intrauterine environment: Mendelian randomization studies
  7. Non-genetic instrumental variables
  8. Conclusions
  9. Acknowledgements
  10. References

Our concern is with the possibility that maternal exposures during pregnancy have a direct biological effect on offspring outcomes, through influencing the intrauterine environment in which foetal development of the offspring occurs. Thus maternal smoking may influence offspring obesity, or maternal alcohol use may lead to impairments in various domains of offspring functioning. However, there are many confounding factors that could generate non-causal links between the smoking, drinking and dietary behaviours of mothers and the health of their children. One approach to this issue is to compare the strength of associations between an exposure among mothers and offspring outcomes and the same exposure among fathers and the offspring outcomes. If there were a direct biological effect of intrauterine exposure on offspring health status, then the link with offspring health should be much stronger for exposure among mothers than for exposure among fathers. This can be illustrated with respect to an outcome where there is strong evidence of a causal influence of maternal smoking during pregnancy: birth weight. Figure 4A demonstrates that in the Avon Longitudinal Study of Parents and their Children (ALSPAC) maternal smoking during pregnancy is associated with lower offspring birth weight, while smoking by the father during pregnancy is only weakly associated (presumably because of some confounding with maternal smoking), and when both maternal and paternal smoking during pregnancy are taken into account (fig. 4B) the former shows a robust association that is little attenuated, while the latter association is essentially abolished.

image

Figure 4. (A) Effect of maternal and paternal smoking on offspring birth weight (difference in birth weight between offspring of smokers and non-smokers in grams). (B) Effect of maternal and paternal smoking during pregnancy on offspring birth weight (in grams) with mutual adjustment.

Download figure to PowerPoint

There has been considerable interest in the possibility that maternal smoking influences foetal development in a way that leads to higher body mass index (BMI) and risk of obesity in later life [24–26]. In the case of offspring BMI at age 7, initial examination of the ALSPAC data shows that the average BMI of children of smoking mothers is raised (fig. 5A) [27]. However, a similar sized association is seen with smoking by fathers, and including both maternal and paternal smoking behaviour in the same model leaves residual effects of similar magnitude (fig. 5B). These findings suggest that in the case of offspring BMI, maternal smoking during pregnancy does not have a direct intrauterine effect, rather confounding factors associated with parental smoking and offspring BMI generate an association of both maternal and paternal smoking with offspring BMI. Similar findings are seen with respect to offspring blood pressure [28] (fig. 6A and B). In contrast, for offspring leg length at age 7, maternal smoking shows stronger effects than paternal smoking [29] suggesting a biological effect of maternal smoking on femur development, as supported by other studies of this issue [30].

image

Figure 5. (A) Effect of maternal and paternal smoking during pregnancy on offspring body mass index with mutual adjustment. (B) Effect of maternal and paternal smoking during pregnancy on mutual adjustment.

Download figure to PowerPoint

image

Figure 6. (A) Effect of maternal and paternal smoking during pregnancy on blood pressure of offspring at age 7 [28]. (B) Effect of maternal and paternal smoking during pregnancy on systolic blood pressure of offspring at age 7 with mutual adjustment [28].

Download figure to PowerPoint

There is particular controversy regarding the potential influence of maternal smoking on cognitive development of offspring [31–36]. Are the associations found in most studies of this issue causal – as some suggest [31]– or do they reflect the considerable confounding that will exist in many databases, given the associations between maternal smoking and a wide variety of parental socio-economic, cognitive and behavioural factors? Julvez et al. [31] demonstrate a stronger association of maternal than paternal smoking during pregnancy on offspring cognitive abilities, although their study was small and the comparison between the two effects are under-powered. Larger studies of this issue may be informative.

It has been hypothesized that maternal obesity and metabolic profiles related to this may, during pregnancy, programme the offspring for greater risk of obesity in later life [37–39]. Such thinking lead to the postulation of the intergenerational acceleration hypothesis for obesity – that female offspring of women who are overweight will themselves be overweight mothers, and programme their offspring for yet greater levels of adiposity [39,40]. If true this is of considerable public health importance, because it would lead to the expectation of ever increasing levels of obesity in the population, independent of any further adverse shifts in energy balance. Under this hypothesis, the expectation would be that the association between the BMI of mothers and that of their offspring would be greater than the equivalent association between BMI of fathers and that of their offspring. Studies that have examined this have generally been of small sample size, or have reported the associations of offspring BMI with BMI of the mothers and fathers in such away as not to be directly comparable, or have had different sources of BMI data for mothers and fathers and the different degrees of validity of the measures among mothers and fathers would influence the strength associations with offspring BMI [41–48]. In the ALSPAC study, we found that maternal and paternal BMI related very similarly to the BMI of offspring at age 7 (fig. 7A and B) [49]. This finding, which is in line with similar data from the 1958 birth cohort [50], argues against a major specific effect on offspring obesity mediated through the intrauterine environment. The intergenerational acceleration hypothesis receives no support from these analyses.

image

Figure 7. (A and B) Offspring body mass index (BMI) according to deciles of maternal and paternal BMI (based on all parent–offspring pairs available). Values are mean and 95% CIs [49].

Download figure to PowerPoint

The influence of the diabetic intrauterine environment on offspring predisposition to type II diabetes has been investigated through examining the metabolic profile of offspring of type I diabetic mothers and type I diabetic fathers [51]. It was demonstrated that the offspring with type 1 diabetic mothers had less favourable metabolic profiles than offspring with type I diabetic fathers, and this was interpreted as indicating a specific effect of the diabetic intrauterine environment on the offspring.

When maternal and paternal behaviours – such as cigarette smoking, alcohol use or dietary patterns – are being utilized in this approach then it is probably not of concern whether the father is the biological father or not, because confounding by sociocultural background will apply to the associations between characteristics of fathers and offspring in the same way for biological and non-biological father figures. When the parental characteristics are biological ones, such as in the example of BMI discussed above, then non-paternity is clearly an issue, as genetic variants shared between parents and offspring will generate associations for biological fathers but not non-biological fathers. In this case, it is important to investigate the potential contribution of non-paternity to generating greater maternal–offspring then paternal–offspring associations, and statistical approaches to assess this are available [49].

In my view, the strong implication of finding that maternal and paternal lifestyle-related factors during pregnancy are associated in similar ways with offspring outcomes suggests that such associations are generated by underlying socially patterned environmental influences acting at the family level, and do not reflect direct biological influences of the exposures per se. However, it is possible to interpret the findings differently. For example, Pembrey et al. [52] suggest that the association between paternal smoking and offspring BMI reflects epigenetic influences, and that such male-line transgenerational responses have important health implications. While it is possible that these male-line epigenetic factors exactly match the biological influence of maternal smoking on the intrauterine environment, to generate very similar associations, this is unlikely. Informal or formal approaches to comparing explanatory models, which adopt the parsimony principle of Occam's razor [53], would suggest that the likelihood of such perfectly matched effects, being produced by mechanistically distinct processes, is rather low.

Timing of exposure

  1. Top of page
  2. Abstract
  3. Contrasting maternal and paternal exposure associations with offspring outcomes
  4. Timing of exposure
  5. Discordant sibling exposure
  6. Maternal genotype as an indicator of intrauterine environment: Mendelian randomization studies
  7. Non-genetic instrumental variables
  8. Conclusions
  9. Acknowledgements
  10. References

Another approach to elucidating whether an exposure has a specific intrauterine influence or not is to observe whether an influence of the exposure is related to when the exposure occurs. For example, maternal smoking before or after pregnancy would not be expected to have the same influence on offspring outcomes as smoking during pregnancy if the mechanism of influence is through a biological effect of the intrauterine environment. It could be argued that confounding of maternal smoking and other factors would be similar, independent of when maternal smoking occurs. Thus, in a Norwegian study, it was suggested that maternal smoking during pregnancy, but not after pregnancy, was related to adverse respiratory health among the offspring [54]. This was interpreted as indicating a direct, biological, influence of maternal smoking on foetal development that led to increased risk of adverse respiratory health among the offspring. There are two problems with this approach, however. First, there is generally a strong correlation between exposures before, during and after pregnancy. In the Norwegian study discussed above, the correlation between maternal smoking during pregnancy and maternal smoking after birth was 0.93. This makes it very difficult to disentangle the effects of smoking at different stages. Second, it is possible – indeed perhaps likely – that women who stopped smoking during pregnancy are different from women who continue smoking, even if the former recommence smoking after pregnancy. Therefore, the central aim of utilizing these methods to explore the degree to which associations are generated by confounding may be vitiated.

The timing of exposure approach may provide stronger evidence when influences on exposure patterns are externally generated. For example, in the well-known Dutch Hunger Winter studies short- and long-term outcomes among offspring have been related to whether women experienced nutritional deficiency, generated by deliberate withholding of food from parts of the Netherlands towards the end of the World War II, before pregnancy, during early pregnancy, during later pregnancy or after pregnancy was completed [55–57]. In this case, an association between timing of exposure and confounding factors is unlikely. This study demonstrated a specific effect of nutritional deprivation during early pregnancy on neural tube defects among the offspring and several other links between exposures to nutritional deprivation at specific periods and long-term health outcomes among offspring have been reported [55–57]. With externally generated influences on timing of exposure, the method can be similar to the instrumental variables method discussed later.

Discordant sibling exposure

  1. Top of page
  2. Abstract
  3. Contrasting maternal and paternal exposure associations with offspring outcomes
  4. Timing of exposure
  5. Discordant sibling exposure
  6. Maternal genotype as an indicator of intrauterine environment: Mendelian randomization studies
  7. Non-genetic instrumental variables
  8. Conclusions
  9. Acknowledgements
  10. References

A powerful approach to determining potential causal effects of intrauterine exposures is to study siblings who are discordant with respect to the exposure. This approach has been applied in investigations of the diabetic intrauterine environment and diabetes risk among offspring [58–60]. Among sibling pairs where the mother had been diagnosed with diabetes before the pregnancy offspring have considerably increased risk of diabetes. The member of a sibling pair who was born before the mother's diagnosis of diabetes. To investigate whether this was simply due to the mothers with diagnosed diabetes being older, a similar comparison was made with respect to offspring born to fathers before diagnosis of diabetes. As fig. 8 shows the discordance in risk was seen with respect to timing of diagnosis of diabetes in mothers but not with respect to timing of diagnosis of diabetes in fathers. This suggests that the intrauterine environment does indeed have a specific influence on diabetes risk.

image

Figure 8. Diabetes among offspring according to whether the mother (inline image) or father (inline image) was diagnosed with diabetes before or after the pregnancy [60].

Download figure to PowerPoint

Maternal genotype as an indicator of intrauterine environment: Mendelian randomization studies

  1. Top of page
  2. Abstract
  3. Contrasting maternal and paternal exposure associations with offspring outcomes
  4. Timing of exposure
  5. Discordant sibling exposure
  6. Maternal genotype as an indicator of intrauterine environment: Mendelian randomization studies
  7. Non-genetic instrumental variables
  8. Conclusions
  9. Acknowledgements
  10. References

Mendelian randomization is a term that has recently been used to describe methods for obtaining unbiased estimates of causal associations in observational studies in aetiological epidemiology [62–67]. These developments reflect a response to well-publicized cases of conventional observational epidemiology having produced misleading information about supposed causal or protective factors, as discussed in the introduction.

Mendelian randomization – the random assortment of genes from parents to offspring that occurs during gamete formation and conception – provides a method of assessing whether certain environmental exposures are causally related to a disease. The association between risk of a disease and a genetic variant that influences the exposure or mimics the biological link between a proposed exposure and disease is not generally susceptible to the ‘reverse causation’ or confounding that may distort interpretations of conventional observational studies [62]. A large number of examples of the application of this approach within epidemiological studies now exist. For example, the null variant of an aldehyde dehydrogenase genotype, which is related to large differences in alcohol drinking behaviour, relates in the expected way with oesophageal cancer risk [67], indicating that alcohol intake does indeed increase the risk of oesophageal cancer. Genetic variants related to cholesterol level are associated with the risk of coronary heart disease in the expected way, in agreement with considerable evidence that circulating cholesterol is causally related to coronary heart disease risk [66]. Many other examples are provided in recent reviews [62–66]. In all these cases, the inference relates to the environmentally modifiable exposure (e.g. alcohol intake or circulating cholesterol levels), rather than to the particular genetic variant. The approach is utilized to strengthen evidence regarding the causal role of the environmentally modifiable factor and therefore to highlight possibilities for disease prevention.

Mendelian randomization studies can provide unique insights into the causal nature of intrauterine environment influences on later disease outcomes. In such studies, maternal genotype is taken to be a proxy for environmentally modifiable exposures mediated through the mother that influencing the intrauterine environment. For example, it is now widely accepted that neural tube defects can in part be prevented by periconceptual maternal folate supplementation [68]. RCTs of folate supplementation have provided the key evidence in this regard [69,70]. But could we have reached the same conclusion before the RCTs were carried out, if we had access to evidence from genetic association studies? Studies have been carried out that have looked at the MTHFR 677C[RIGHTWARDS ARROW]T polymorphism (a genetic variant that is associated with methyltetrahydrofolate reductase activity and circulating homocysteine levels; TT genotype being associated with lower homocysteine levels) in newborns with neural tube defects compared to controls, and have found an increased risk in TT versus CC newborns, with a relative risk of 1.75 [95% confidence interval (CI): 1.41–2.18] in a meta-analysis of all such studies [71]. Studies have also looked at the association between this MTHFR variant in parents and the risk of neural tube defect in their offspring. Mothers who have the TT genotype have an increased risk of 2.04 (95% CI: 1.49–2.81) of having an offspring with a neural tube defect compared to mothers who have the CC genotype [72]. For TT fathers, the equivalent relative risk is 1.18 (95% CI: 0.65–2.12) [68]. This pattern of associations suggests that it is the intra-uterine environment – influenced by maternal TT genotype – rather than the genotype of offspring that is related to disease risk (fig. 9). This is consistent with the hypothesis that maternal folate intake is the exposure of importance.

image

Figure 9. Inheritance of MTHFR polymorphism, and neural tube defects.

Download figure to PowerPoint

In this case, the findings from observational studies, genetic association studies and an RCT are closely similar. Had the technology been available, the genetic association studies, with the particular influence of maternal versus paternal genotype on neural tube defect risk, would have provided strong evidence of the beneficial effect of folate supplementation, before the results of any RCT had been completed; although trials would still have been necessary to confirm the effect was causal for folate supplementation. Certainly, the genetic association studies would have provided better evidence than that given by conventional epidemiological studies that had to cope with the problems of accurately assessing diet and also with the considerable confounding of maternal folate intake with a wide variety of lifestyle and socio-economic factors that may also influence neural tube defect risk. The association of genotype with neural tube defect risk does not suggest that genetic screening is indicated, rather it demonstrates that an environmental intervention may benefit the whole population, independent of the genotype of individuals receiving the intervention.

Studies utilizing maternal genotype as a proxy for environmentally modifiable influences on the intrauterine environment can be analysed in a variety of ways. First, the mothers of offspring with a particular outcome can be compared to a control group of mothers who have offspring without the outcome, in a conventional case–control design, but with the mother as the exposed individual (or control) rather than the offspring with the particular health outcome (or the control offspring). Fathers could serve as a control group when autosomal genetic variants are being studied. If the exposure is mediated by the mother, maternal genotype, rather than offspring genotype, will be the appropriate exposure indicator. Clearly, maternal and offspring genotype are associated, but conditional on each other, it should be the maternal genotype that shows the association with the health outcome among the offspring. Indeed, in theory it would be possible to simply compare genotype distributions of mothers and offspring, with a higher prevalence among mothers providing evidence that maternal genotype, through an intrauterine pathway, is of importance. However, the statistical power of such an approach is low, and an external control group, whether fathers or women who have offspring without the health outcome, is generally preferable.

The influence of high levels of alcohol intake by pregnant women on the health and development of their offspring is well recognized for very high levels of intake, in the form of foetal alcohol syndrome [73]. However, the influence outside of this extreme situation is less easy to assess, particularly as higher levels of alcohol intake will be related to a wide array of potential sociocultural, behavioural and environmental confounding factors. Furthermore, there may be systematic bias in how mothers report alcohol intake during pregnancy, which could distort associations with health outcomes. Therefore, outside of the case of very high alcohol intake by mothers, it is difficult to establish a causal link between maternal alcohol intake and offspring developmental characteristics. Some studies have approached this by investigating alcohol-metabolizing genotypes in mothers and offspring outcomes.

Although sample sizes have been low and the analysis strategies not optimum they provide some evidence to support the influence of maternal genotype [74–76]. For example, in one study mental development at age 7.5 was delayed among offspring of mothers possessing a genetic variant associated with more rapid alcohol metabolism. Among these mothers there would presumably be more rapid clearance of alcohol and thus a reduced influence of maternal alcohol on offspring during the intra-uterine period [75]. Offspring genotype was not independently related to these outcomes, indicating that the crucial exposure related to maternal alcohol levels. As in the MTHFR examples, these studies are of relevance because they provide evidence of the influence of maternal alcohol levels on offspring development, rather than because they highlight a particular maternal genotype that is of importance. In the absence of alcohol drinking, the maternal genotype would presumably have no influence on offspring outcomes. The association of maternal genotype and offspring outcome suggests that alcohol levels in mothers, and therefore their alcohol consumption, has an influence on offspring development.

As mentioned above, studies utilizing the Mendelian randomization approach need to examine the influence of maternal genotype conditioned on offspring genotype, to demonstrate that the influence is intrauterine rather than potentially postnatal and due to differential responses to environmental exposures by the offspring. The approach has considerable potential, however, as genetic variants of known effect are rapidly being uncovered. For example, identification of genetic variants related to glucose level [77] and obesity [78] allow the demonstration of potential influences of maternal hyperglycaemia or obesity on offspring outcomes. For example, investigating the influence of maternal genotype related to obesity on offspring obesity risk will provide another mechanism for testing the intergenerational acceleration hypothesis for obesity, discussed above.

There is currently rapid progress in identifying genetic variants that are related to the biological processing of a wide variety of environmental exposures. Therefore, the scope of Mendelian randomization studies in the environmental toxicology field should expand greatly in the coming few years. Such studies require large sample size, as the influence of genetic variants on the phenotype of interest is generally small, but this cost is generally worth paying as it allows better specification of causal effects of environmentally modifiable risk factors. There are of course limitations to the approach, which have been discussed in detail elsewhere (fig. 10) [62,66].

image

Figure 10. Limitations of Mendelian randomization.

Download figure to PowerPoint

Non-genetic instrumental variables

  1. Top of page
  2. Abstract
  3. Contrasting maternal and paternal exposure associations with offspring outcomes
  4. Timing of exposure
  5. Discordant sibling exposure
  6. Maternal genotype as an indicator of intrauterine environment: Mendelian randomization studies
  7. Non-genetic instrumental variables
  8. Conclusions
  9. Acknowledgements
  10. References

The Mendelian randomization method is one example of an instrumental variables approach [66,79,80], as illustrated in fig. 11. This principle – that it is possible to utilize an instrumental variable that relates to exposure probability but not to confounding factors and is not influenced by reporting bias or health outcomes (reverse causation) – can have broader applicability [81–84]. The trick, of course, is finding an instrument that relates to the exposure but does not relate to the health outcomes through any other pathways. In the field of intrauterine inferences on later health, for example, the instrumental variable approach has been utilized to study the effect of cigarette smoking by mothers on offspring birth weight. As discussed earlier, there is good evidence that maternal smoking reduces offspring birth weight; however, maternal smoking is related to many other characteristics that could influence offspring birth weight. Evans and Ringel [85] utilized cigarette taxation rates as an instrument for smoking behaviour to investigate this further. Across US states, there are different levels of cigarette taxes, and it can be demonstrated that taxation level influences smoking behaviour. Analysing a large data file, they could relate state-level taxation levels to smoking behaviour and through to birth weight of offspring. Relatively straightforward methods exist for utilizing the joint associations to predict the effect of individual-level cigarette smoking on offspring birth weight. Evans and Ringel demonstrate a substantial influence of maternal smoking on offspring birth weight, of the order of 400 g, which is somewhat larger than the naïve estimates generated by simply relating reports of maternal smoking to birth weight of offspring. It is possible that the instrumental variable approach yields higher estimates because this approach is not susceptible to individual-level measurement error (including potential systematic error) in reporting of cigarette smoking by mothers. Indeed, the estimate made by Evans and Ringel is close to that from a RCT encouraging smoking cessation [86–88]. The estimate produced by the instrumental variable study, however, reflects the influence of smoking on birth weight among women who are encouraged not to smoke by higher cigarette taxation. There is no reason to believe that the causal effect of not smoking among women who are encouraged not to smoke by higher cigarette taxation will be different from the causal effect that would be observed among women who are not so influenced, but this assumption cannot be tested within the instrumental variable framework.

image

Figure 11. Mendelian randomization as an instrumental variable approach. The instrumental variable is not related to confounding factors nor is its ascertainment biased. The instrumental variable has no association with the outcome except through the modifiable exposure, in the above example maternal alcohol levels. The join associations of the black and grey arrows allow the correct estimate of the causal association indicated by the dotted arrow.

Download figure to PowerPoint

In another example utilizing routine data the influence of a bus strike on ability of women to get to their prenatal care appointments was investigated [89]. The bus strike did indeed reduce antenatal clinic attendance and the study suggested that antenatal clinical attendance had beneficial effects on birth outcomes.

Instrumental variable approaches are conceptually similar to natural experiments. Indeed, John Snow's famous investigation of cholera risk according to the company supplying water to households in London in the mid-19th century essentially utilized water company as an instrument for receiving contaminated water that bore the cholera-causing agent. It was clear that households did not systematically differ from one another except in terms of from where they received their water supply, therefore, there would be no expectation of greater cholera risk in one or the other group, unless cholera risk was influenced by water supply [90]. The challenge, of course, is identifying an appropriate instrumental variable. The strength of association between the instrumental variable and the exposure of interest determines the sample size required for reliably estimating the effects of the modifiable exposure of interest. In the case of cigarette taxation and birth weight, for example, the association between taxation rate and smoking is small and a sample size of millions is required for estimation. Fortunately, routine data sources can provide such samples. In other areas (such as the link between water company and water contamination), the instrument will be more strongly related to the exposure of interest and sample sizes do not need to be so massive. The issue of statistical power and other problems with instrumental variable approaches within epidemiology are discussed elsewhere [81–83].

Conclusions

  1. Top of page
  2. Abstract
  3. Contrasting maternal and paternal exposure associations with offspring outcomes
  4. Timing of exposure
  5. Discordant sibling exposure
  6. Maternal genotype as an indicator of intrauterine environment: Mendelian randomization studies
  7. Non-genetic instrumental variables
  8. Conclusions
  9. Acknowledgements
  10. References

There is considerable interest in the role of intrauterine exposures in influencing offspring disease risk in human beings. Drawing causal inferences in epidemiological studies is problematic, however, for several reasons. These include the possibility that exposure patterns present in the past are not relevant to contemporary cohorts; the difficulties in obtaining accurate measures of exposure and confounders and the problems of disentangling causal associations from those that are due to confounding or bias. Various approaches to increasing the strength of causal inference in studies of intrauterine exposure and later health outcomes have been discussed. They have one characteristic in common, and that is that they generally require large sample sizes. This is because some of the methods require tests between associations, rather than simply a test of whether an association exists or not – for example, tests between maternal and paternal effects. In other cases, the studies relate to situations of known small effect size (such as the association between common genetic variants and disease outcomes in the case of Mendelian randomization). However, the price of large sample sizes is certainly one worth paying to get closer to reliable estimates of causal effects in epidemiological studies.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Contrasting maternal and paternal exposure associations with offspring outcomes
  4. Timing of exposure
  5. Discordant sibling exposure
  6. Maternal genotype as an indicator of intrauterine environment: Mendelian randomization studies
  7. Non-genetic instrumental variables
  8. Conclusions
  9. Acknowledgements
  10. References

My work in this area is supported by the MRC centre for causal analyses and translational epidemiology. I have no conflicts of interest with respect to this paper. I would like to thank Sam Leary for analyses of ALSPAC data and Jonathan Sterne for an earlier version of fig. 11.

References

  1. Top of page
  2. Abstract
  3. Contrasting maternal and paternal exposure associations with offspring outcomes
  4. Timing of exposure
  5. Discordant sibling exposure
  6. Maternal genotype as an indicator of intrauterine environment: Mendelian randomization studies
  7. Non-genetic instrumental variables
  8. Conclusions
  9. Acknowledgements
  10. References
  • 1
    Manson J, Stampfer MJ, Willett WC et al. A prospective study of antioxidant vitamins and incidence of coronary heart disease in women. Circulation 1991;84 (suppl II):II546.
  • 2
    Rimm EB, Stampfer MJ, Ascherio A, Giovannucci E, Colditz GA, Willett WC. Vitamin E consumption and the risk of coronary heart disease in men. N Engl J Med 1993;328:14506.
  • 3
    Stampfer MJ, Hennekens CH, Manson JE, Colditz GA, Rosner B, Willett WC. Vitamin E consumption and the risk of coronary disease in women. N Engl J Med 1993;328:14449.
  • 4
    Osganian SK, Stampfer MJ, Rimm E et al. Vitamin C and risk of coronary heart disease in women. J Am Coll Cardiol 2003;42:24652.
  • 5
    Stampfer MJ, Colditz GA. Estrogen replacement therapy and coronary heart disease: a quantitative assessment of the epidemiologic evidence. Prev Med 1991;20:4763 (reprinted Int J Epidemiol 2004;33:445–53).
  • 6
    Omenn GS, Goodman GE, Thornquist MD et al. Effects of a combination of beta carotene and vitamin A on lung cancer and cardiovascular disease. N Engl J Med 1996;334:11505.
  • 7
    Alpha-Tocopherol, Beta Carotene Cancer Prevention Study Group. The effect of vitamin E and beta carotene on the incidence of lung cancer and other cancers in male smokers. N Engl J Med 1994;330:102935.
  • 8
    Dietary supplementation with n-3 polyunsaturated fatty acids and vitamin E after myocardial infarction: results of the GISSI-Prevenzione trial. Gruppo Italiano per lo Studio della Sopravvivenza nell’Infarto miocardico. Lancet 1999;354:44755.
  • 9
    Heart Protection Study Collaborative Group. MRC/BHF Heart Protection Study of antioxidant vitamin supplementation in 20536 high-risk individuals: a randomised placebo-controlled trial. Lancet 2002;360:2333.
  • 10
    Beral V, Banks E, Reeves G. Evidence from randomized trials of the long-term effects of hormone replacement therapy. Lancet 2002;360; 94244.
  • 11
    Manson JE, Hsia J, Johnson KC et al. Estrogen plus progestin and the risk of coronary heart disease. N Engl J Med 2003;349:52334.
  • 12
    Steinberg D, Witztum JL. Is the oxidative modification hypothesis relevant to human atherosclerosis? Do the antioxidant trials conducted to date refute the hypothesis? Circulation 2002;105:210711.
  • 13
    Willett WC, Stampfer MJ. What vitamins should I be taking, doctor? N Engl J Med 2001;345:181924.
  • 14
    Stampfer MJ. Commentary: Hormones, heart disease, and the definition of hormone ‘initiation’. Int J Epidemiol 2006;35:7389.
  • 15
    Stampfer MJ, Rimm E, Willett W. Folate supplementation and cardiovascular disease. Lancet 2006;367:12378.
  • 16
    Davey Smith G, Ebrahim S. Epidemiology: is it time to call it a day? Int J Epidemiol 2001;30:114.
  • 17
    Lawlor DA, Davey Smith G, Bruckdorfer KR et al. Those confounded vitamins: what can we learn from the differences between observational versus randomised trial evidence? Lancet 2004;363:17247.
  • 18
    Vandenbroucke JP. Commentary: the HRT story: vindication of old epidemiological theory. Int J Epidemiol 2004;33:4567.
  • 19
    Davey Smith G, Ebrahim S. Folate supplementation and cardiovascular disease. Lancet 2006;366:167981.
  • 20
    Eidelman RS, Hollar D, Hebert PR, Lamas GA, Hennekens CH. Randomized trials of vitamin E in the treatment and prevention of cardiovascular disease. Arch Intern Med 2004;164:15526.
  • 21
    Robinson SM, Crozier SR, Borland SE, Hammond J, Barker DJP, Inskip HM. Impact of educational attainment on the quality of young women's diets. Eur J Clin Nutr 2004;158:117480.
  • 22
    Matthews F, Yudkin P, Neil A. Influence of maternal nutrition on outcome of pregnancy: prospective cohort study. BMJ 1999;319:33943.
  • 23
    Poston L, Briley AL, Seed PT, Kelly FJ, Sheenan AH, for the Vitamins in Pre-eclampsia (VIP) Trial Consortium. Vitamin C and vitamin E in pregnant women at risk for pre-eclampsia (VIP trial): randomised placebo-controlled trial. Lancet 2006;367:1145.
  • 24
    Vik T, Jacobsen G, Vatten L, Bakketeig LS. Pre- and postnatal growth in children of women who smoked in pregnancy. Early Hum Dev 1996;45:24555.
  • 25
    Power C, Jefferis BJ. Fetal environment and subsequent obesity: a study of maternal smoking. Int J Epidemiol 2002;31:4201.
  • 26
    Mamun AA, Lawlor DA, Alarti R, O’Callaghan MJ, Williams GM, Najman JM. Does maternal smoking during pregnancy have a direct effect on future offspring obesity? Evidence from a prospective birth cohort study. Am J Epidemiol 2006;164:31725.
  • 27
    Leary S, Davey Smith G, Ness A, ALSPAC Study team. Smoking during pregnancy and components of stature in the offspring. Am J Hum Biol 2006;18:50212.
  • 28
    Brion MJA, Leary SD, Davey Smith G, Ness AR. Similar associations of parental prenatal smoking suggest child blood pressure is not influenced by intrauterine effects. Hypertension 2007;49:14228.
  • 29
    Leary SD, Davey Smith G, Rogers IS, Reilly JJ, Wells JCK, Ness AR. Smoking during pregnancy and offspring fat and lean mass in childhood. Obesity 2006;14:228493.
  • 30
    Jaddoe VWV, Berburg BE, De Ridder MAJ et al. Maternal smoking and fetal growth characteristics in different periods of pregnancy. Am J Epidemiol 2007;165:120715.
  • 31
    Julvez J, Ribas-Fitó N, Torrent M, Forns M, Garcia-Esteban R, Sunyer J. Maternal smoking habits and cognitive development of children at age 4 years in a population-based birth cohort. Int J Epidemiol 2007; doi: 10.1093/ije/dym107
  • 32
    Breslau N, Paneth N, Lucia VC, Paneth-Pollack R. Maternal smoking during pregnancy and offspring IQ. Int J Epidemiol 2005;34:104753.
  • 33
    Batty GD, Der G, Deary IJ. Effect of maternal smoking during pregnancy on offspring's cognitive ability: empirical evidence for complete confounding in the US national longitudinal survey of youth. Pediatrics 2006;118:94350.
  • 34
    Butler NR, Goldstein H. Smoking in pregnancy and subsequent child development. BMJ 1973;4:5735.
  • 35
    Baghurst PA, Tong SL, Woodward A, McMichael AJ. Effects of maternal smoking upon neuropsychological development in early childhood: importance of taking account of social and environmental factors. Paediatr Perinat Epidemiol 1992;6:40315.
  • 36
    Lawlor DA, Najman JM, Batty GD, C’Callaghan MJ, Williams GM, Bor W. Early life predictors of childhood intelligence: findings from the Mater-University study of pregnancy and its outcomes. Paediatr Perinat Epidemiol 2006a;20:14862.
  • 37
    Whitaker RC, Dietz WH. Role of the parental environment in the development of obesity. J Pediatr 1998;132:76876.
  • 38
    Oken E, Gillman MW. Fetal origins of obesity. Obes Res 2003;11:496506.
  • 39
    Levin BE. The obesity epidemic: metabolic imprinting on genetically susceptible neural circuits. Obes Res 2000;8:3427.
  • 40
    Ebbeling CB, Pawlak DB, Ludwig DS. Childhood obesity: public-health crisis, common sense cure. Lancet 2002;360:47382.
  • 41
    Garn SM, Clark DC. Trends in fatness and the origins of obesity. Pediatrics 1976;57:44356.
  • 42
    Esposito-Del Puente A, Scalti L, De Filippo E et al. Familial and environmental influences on body composition and body fat distribution in childhood in southern Italy. Int J Obes 1994;18:596601.
  • 43
    Keiller SM, Colley JRT, Carpenter RG. Obesity in school children and their parents. Ann Hum Biol 1979;6:44355.
  • 44
    Ayatollahi SMT. Obesity in school children and their parents in southern Iran. Int J Obes 1992;16:84550.
  • 45
    Sorensen TIA, Holst C, Stunkard AJ, Skovgaard IT. Correlations of body mass index of adult adoptees and their biological and adoptive relatives. Int J Obes 1992;16:22736.
  • 46
    Greenland KJ, Liu K, Dyer AR, Kiefe CI, Burke GL, Yunis C. Body mass index in young adults: Associations with parental body size and education in the CARDIA study. Am J Public Health 1996;86:4805.
  • 47
    Duran-Tauleria E, Rona RJ, Chinn S. Factors associated with weight for height and skinfold thickness in British children. J Epidemiol Community Health 1995;49:46673.
  • 48
    Lawlor DA, Davey Smith G, O’Callaghan M et al. Epidemiologic evidence for the fetal overnutrition hypothesis: findings from the Mater-University Study of Pregnancy and its outcomes. Am J Epidemiol 2007;165:41824.
  • 49
    Davey Smith G, Steer C, Leary S, Ness A. Is there an intra-uterine influence on obesity? Evidence from parent-child associations in the Avon Longitudinal Study of Parents and Children (ALSPAC). Arch Dis Child 2007;92:87680.
  • 50
    Lake JK, Power C, Cole TJ. Child to adult body mass index in the 1958 British birth cohort: associations with parental obesity. Arch Dis Child 1997;77:37681.
  • 51
    Sobngwi E, Boudou P, Mauvais-Jarvis M et al. Effect of a diabetic environment in utero on predisposition to type 2 diabetes. Lancet 2003;361:18615.
  • 52
    Pembrey ME, Bygren LO, Kaati G et al. Sex-specific, male-line transgenerational responses in humans. Eur J Hum Genet 2006;14:15966.
  • 53
    Mackay DJ. Model comparison and Occam's razor. In: MackayDJ (ed.). Information Theory, Inference and Learning Algorithms. Cambridge University Press, Cambridge, UK, 2003.
  • 54
    Haberg SE, Stigum H, Nystad W, Nafstad P. Effects of pre- and postnatal exposure to parental smoking on early childhood respiratory health. Am J Epidemiol 2007;166:67986.
  • 55
    Susser E, Hoek HW, Brown A. Neurodevelopmental disorders after prenatal famine: the story of the Dutch Famine Study. Am J Epidemiol 1998;147:2136.
  • 56
    Lumey LH, Van Poppel FWA. The Dutch famine of 1944–1945: mortality and morbidity in past and present generations. Soc Hist Med 1994;7:22946.
  • 57
    Susser M, Stein Z. Timing in prenatal nutrition: a reprise of the Dutch famine Study. Nutr Rev 1994;52:8494.
  • 58
    Dabelea D, Pettitt DJ. Intrauterine diabetic environment confers risks for type 2 diabetes mellitus and obesity in the offspring, in addition to genetic susceptibility. J Pediatr Endocrinol Metab 2001;14:108591.
  • 59
    Dabelea D, Hanson RL, Lindsay RS et al. Intrauterine exposure to diabetes conveys risks for type 2 diabetes and obesity: a study of discordant sibships. Diabetes 2000;49:220811.
  • 60
    Dabelea D. The predisposition to obesity and diabetes in offspring of diabetic mothers. Diabetes Care 2007;30:16974.
  • 61
    Pettitt DJ, Knowler WC. Diabetes and obesity in the Pima Indians: a cross-generational vicious cycle. J Obes Weight Regul 1998;7:615.
  • 62
    Davey Smith G, Ebrahim S. ‘Mendelian randomization’: can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol 2003;32:122.
  • 63
    Davey Smith G, Ebrahim S. Mendelian randomization: prospects, potentials, and limitations. Int J Epidemiol 2004;33:3042.
  • 64
    Davey Smith G, Ebrahim S. What can Mendelian randomisation tell us about modifiable behavioural and environmental exposures. BMJ 2005;330:10769.
  • 65
    Davey Smith G. Cochrane Lecture: randomised by (your) god: robust inference from an observational study design. J Epidemiol Community Health 2006;60:3828.
  • 66
    Davey Smith G, Ebrahim S. Mendelian randomization: genetic variants as instruments for strengthening causal inference in observational studies. In: VaupelJW, WeinsteinM (eds). Bio-Social Surveys: Current Insight and Future Promise. The National Academies Press, National Research Council, Washington, DC, 2007.
  • 67
    Lewis S, Davey Smith G. Alcohol, ALDH2 and esophageal cancer: a meta-analysis which illustrates the potentials and limitations of a Mendelian randomization approach. Cancer Epidemiol Biomarkers Prev 2005;14:196771.
  • 68
    Scholl TO, Johnson WG. Folic acid: influence on the outcome of pregnancy. Am J Clin Nutr 2000;71 (suppl):1295S303.
  • 69
    MRC Vitamin Study Research Group. Prevention of neural tube defects: results of the Medical Research Council vitamin study. Lancet 1991;338:1317.
  • 70
    Czeizel AE, Dudás I. Prevention of the first occurrence of neural-tube defects by periconceptional vitamin supplementation. N Engl J Med 1992;327:18325.
  • 71
    Botto LD, Yang Q. 5,10-Methylenetetrahydrofolate reductase gene variants and congenital anomalies: a HuGE review. Am J Epidemiol 2000;151:86277.
  • 72
    Roseboom TJ, Van Der Meulen JH, Osmond C et al. Coronary heart disease after prenatal exposure to the Dutch famine, 1944–45. Heart 2000;84:5958.
  • 73
    Burd LJ. Interventions in FASD: we must do better. Child Care Health Dev 2006;33:398400.
  • 74
    Gemma S, Vichi S, Testai E. Metabolic and genetic factors contributing to alcohol induced effects and fetal alcohol syndrome. Neurosci Biobehav Rev 2007;31:2219.
  • 75
    Jacobson SW, Carr LG, Croxford J, Sokol RJ, Li TK, Jacobson JL. Protective effects of the alcohol dehydrogenase-ADH1B allele in children exposed to alcohol during pregnancy. J Pediatr 2006;148:307.
  • 76
    Warren KR, Li TK. Genetic polymorphisms: impact on the risk of fetal alcohol spectrum disorders. Birth Defects Res A Clin Mol Teratol 2005;73:195203.
  • 77
    Weedon MN, Clark VJ, Qian Y et al. A common haplotype of the glucokinase gene alters fasting glucose and birth weight: association in six studies and population-genetics analyses. Am J Hum Genet 2006;79:9911001.
  • 78
    Frayling TM, Timpson NJ, Weedon MN et al. A common variant in the FTO gene is associated with body mass index and predisposes to childhood and adult obesity. Science 2007;316:88994.
  • 79
    Thomas DC, Conti DV. Commentary: the concept of ‘Mendelian randomization’. Int J Epidemiol 2004;33:215.
  • 80
    Lawlor DA, Harbord RM, Sterne JAC, Davey Smith G. Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Stat Med 2007; doi: 10.1002/sim.3034
  • 81
    Newhouse JP, McClellan M. Econometrics in outcomes research: the use of instrumental variables. Annu Rev Public Health 1998;19:1734.
  • 82
    Angrist JD, Imbens GW, Rubin DB. Identification of causal effects using instrumental variables. J Am Stat Assoc 1996;91:44455.
  • 83
    Greenland S. An introduction to instrumental variables for epidemiologists. Int J Epidemiol 2000;29:7229.
  • 84
    Martens EP, Pestman WR, De Boer A, Belitser SV, Klungel OH. Instrumental variables. Application and limitations. Epidemiology 2006;17:2607.
  • 85
    Evans WN, Ringel JS. Can higher cigarette taxes improve birth outcomes? J Public Econ 1999;72:13554.
  • 86
    Sexton M, Hebel JR. A clinical trial of chance in maternal smoking and its effect on birthweight. J Am Med Assoc 1984;251:9115.
  • 87
    Permutt T, Hebel JR. Simultaneous-equation estimation in a clinical trial of the effect of smoking on birth weight. Biometrics 1989;45:61922.
  • 88
    Hamilton BH. Estimating treatment effects in randomized clinical trials with non-compliance: the impact of maternal smoking on birthweight. Health Econ 2001;10:399410.
  • 89
    Evans WN, Lien DS. The benefits of prenatal care: evidence from the PAT bus strike. J Econom 2005;125:20739.
  • 90
    Davey Smith G. Behind the Broad Street pump: aetiology, epidemiology and prevention of cholera in mid-19th century Britain. Int J Epidemiol 2002;31:92032.