SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. How can we use data from twins to shed light on the causal pathways underlying the association between birthweight and later disease risk?
  5. Data from studies examining within-cohort and within-pair associations
  6. Generalisability of findings in cohorts of twins
  7. Acknowledgements
  8. References

There has been much interest in evidence that people with lower birthweight have higher risk of adult cardiovascular disease, but the causal pathways underlying such observations are uncertain. Study of twins offers an opportunity to shed light on the underlying causal pathways, in particular by investigating the role of ‘shared’ factors vs. factors affecting each individual fetus. This involves comparing results of within-cohort vs. within-pair analyses.

Twins share many factors during gestation but birthweight discordance (difference in birthweight within a twin pair) cannot be determined by these shared factors and must relate to factors affecting growth of each individual fetus. If associations seen in a cohort of twins remain in within-pair analyses, then factors specific to each individual must be involved in the underlying causal pathways. Conversely, if the relationships disappear or substantially diminish in within-pair analyses, then factors common to the pair must be involved. Comparison of findings in monozygotic vs. dizygotic twins may provide insights into the role of genetic factors, although issues related to chorionicity need to be taken into account.

We tabulate published data and conclude that differences in methodology and analyses preclude informative meta-analysis, and that analysis of pooled data would provide more useful information.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. How can we use data from twins to shed light on the causal pathways underlying the association between birthweight and later disease risk?
  5. Data from studies examining within-cohort and within-pair associations
  6. Generalisability of findings in cohorts of twins
  7. Acknowledgements
  8. References

There is a large body of literature describing a link between lower birthweight and an increase in cardiovascular risk factors or cardiovascular disease (CVD) in later life.1,2 Other outcomes such as cognition, bone health or mental health have also been studied in the context of birthweight, though the last two to a lesser extent.3–5 However, if there is a causal link between birthweight per se and later risk of CVD, and if cross-sectional associations represent the magnitude of the associations, then the public health implications would be limited. However, as we and others have indicated,6–8 we need to understand the causal pathways involved before we can accurately assess the risk associated with this pathway and the public health implications of these epidemiological findings.

This evidence regarding the ‘fetal origins of adult disease’ also needs to be set in the context of life course, since we know that postnatal exposures are also important, and there is evidence that exposures may interact and/or accumulate over time.9,10

We have suggested that there are a number of different types of factors operating during gestation (Fig. 1).6 Some factors (type A) influence birthweight but have no long-term effect on health, whereas others (type B) influence birthweight and thereby later health. Type C factors influence both birth size and outcome independently, with an apparent association between birthweight and outcome, and finally type D factors, for which there is some evidence, influence later risk of disease without any effect on birthweight.

image

Figure 1. Gestational factors and later disease risk.

Download figure to PowerPoint

How can we use data from twins to shed light on the causal pathways underlying the association between birthweight and later disease risk?

  1. Top of page
  2. Summary
  3. Introduction
  4. How can we use data from twins to shed light on the causal pathways underlying the association between birthweight and later disease risk?
  5. Data from studies examining within-cohort and within-pair associations
  6. Generalisability of findings in cohorts of twins
  7. Acknowledgements
  8. References

We suggest that data from twins can be used in three different ways.

Study of maternal factors and offspring outcome

Resources required of the mother to sustain a twin pregnancy are greater than those required to sustain a singleton pregnancy. We hypothesise that twin pregnancies may be a more efficient group in which to study the impact of maternal nutrition, for example, on offspring health, whether or not birthweight is affected. This requires prospective studies of twin pregnancies.

Comparison of twin vs. singleton cohorts

As a group, twins have shorter mean gestation length and lower birthweight for gestational age. The latter has been ascribed to ‘maternal constraint’, though the mechanism is uncertain.

Given that low birthweight is linked to an increase in CVD or its risk factors in most populations, we might expect to see an increase in cardiovascular mortality or morbidity in cohorts of twins vs. singletons. However, in studies to date this has not proved to be the case.11–15 What can we infer from this?

We can infer that there is no causal link between whatever underlies the particular constraint on growth of twins (as a group) and CVD risk in these twin populations

We suggest that whatever underlies this constrained growth is a type A factor in Fig. 1.

However, this does not imply that factors of types B to D are not operating in individual twins; indeed there is evidence that such factors are operating with respect to blood pressure. A number of studies have shown that within a population of twins there is a cross-sectional association between birthweight and blood pressure (Table 1, β within cohort), similar in magnitude to that seen in singletons.14,16–23 Very few studies to date have published data on the cross-sectional association between birthweight and other CVD risk factors or established CVD in twin cohorts.

Table 1.  Regression coefficients (β) for difference in systolic blood pressure (mm Hg) per 1 kg increase in birthweight in cohorts and within pairs, in twins. Adjusted value (generally for one or more measures of current size, ± age and sex in within-cohort analyses)
  β within cohortN individualsβ within pairsN pairs
  • a

    Two sets of triplets were included.

  • b

    Zygosity by questionnaire.

  • Birthweight by cmaternal or dself-report.

Dwyer16Childrena,b−7.0104−5.355
Poulter14Adult femalesd−3.0812−4.3406
IJzerman17Adolescents (at rest)c−1.9228−3.3114
Baird18Adultsb,c1.15083.3198
Zhang19Children; same-sex pairs: first twin−1.1361  
 second twin−1.3361−0.5210
Christensen20Adolescents−1.92622−0.61311
Loos21Adult males (at rest)−0.93791.3153
Adult females (at rest)−4.2389−2.7158
Nowson22Adolescent femalesa,b,c−2.6244−0.9118
Adult femalesb,d−2.3467−0.6178
Johansson-Kark23Young adult malesb−1.71772−0.2617
We can infer that there is no causal link between reduced mean gestation length and CVD risk in these populations

This is of interest as some studies of singletons have suggested an inverse relationship between gestation and later blood pressure. However, there is evidence that maternal factors associated with gestation length may differ between twins and singletons.24

We can infer that there is no causal link between the generally increased postnatal growth rate of young twins (catch-up growth) and CVD risk

As a group, twins experience more accelerated postnatal growth than singletons. The fact that CVD risk is not higher in twins suggests that accelerated postnatal growth might not be an important intermediate for ‘fetal origins’ associations.

Some have suggested that the general constraint on growth of twins, which increases with gestational age, leads to symmetrical or proportionate growth restriction, whereas in singletons growth restriction in utero, particularly in late gestation, is disproportionate, such that babies are thin.25 If this is indeed the case, then it is possible that the quality of catch-up growth, in terms of change in body composition, differs between twins and singletons.

Within-cohort vs. within-pair associations between birthweight and measures of CVD risk

Conventionally, twin studies have been used to separate environmental from genetic causes by comparing concordance for a factor in monozygotic and dizygotic twins.26 There is also an opportunity to provide information about causal pathways underlying the widely observed birthweight – CVD risk association. This concept has been suggested previously16,27,28 and a number of the researchers cited in Table 1 have recently used it in their analyses. However, not all ‘fetal origins’ researchers have incorporated this form of analysis into their investigations.29

It is important that we remember that causal pathways may differ according to the outcome of interest.

Shared vs. fetus-specific factors

In cohorts of twins, birthweight can be related to shared maternal (or paternal) factors like maternal nutrition or socio-economic status, as well as factors specific to the individual fetus. However, within-pairs, birthweight discordance cannot relate to shared factors but must relate to differences in factors affecting each individual fetus. This provides opportunities to shed light on the causal pathways underlying associations between birthweight and CVD risk factors seen in cohorts of twins.27 If the association remains in within-pair analyses, factors specific to each individual fetus (like supply of nutrients and oxygen or even postnatal differences like differences in growth) must be involved in the underlying causal pathways (Fig. 2 and hypothetical data in Table 2a). Conversely, if the relationship disappears or substantially diminishes in within-pair analyses, then factors common to the pair (like maternal nutrition or socio-economic status) must be involved (Fig. 2 and Table 2b).

image

Figure 2. Within-cohort vs. within-pair analyses.

Download figure to PowerPoint

Table 2.  Hypothetical results of within-cohort vs. within-pair analyses
  Regression coefficients (β) for difference in systolic blood pressure (mm Hg) per 1 kg increase in birthweight Inference
β within cohortβ within pairs
  • a

    These shared factors may be antecedent (type B) factors like maternal nutrition, confounding (type C) factors like socio-economic status or length of gestation.

  • b

    Another possibility here is that an uncompromised monochorionic MZ twin is programmed for later disease by circulating factors from their compromised co-twin, via interfetal vascular anastomoses.

a. In all twins −2.0−2.0Shared factors are not involved in the association
−2.0−0.5Shared factors are importantly involved in the associationa
b. In MZ and DZ twins separatelyDZ−2.0−1.0Genes not importantly involved in the association
MZ−2.0−1.0
DZ−2.0−1.5Genes may be importantly involved in the associationb
MZ−2.0−0.5

It is, of course, possible that both shared and individual factors are involved, in which case we should see a reduction rather than disappearance of the association in within-pair analyses.

Genetic factors

By examining the association between birthweight and outcomes of interest in DZ and MZ pairs separately, both within the cohort and within pairs, we may be able to shed some light on the role of genetic factors.

If an association is seen between birth and, for example, blood pressure, in both MZ and DZ twins in the population, then we can examine what happens within pairs. If the within-cohort and within-pair associations differ similarly in both MZ and DZ twins, then genes are not importantly involved. Conversely, if the association largely disappears within pairs in MZ twins, but not in DZ twins, then genes may be importantly involved in the association. (See Table 2b for hypothetical examples.)

However, many MZ twins are monochorionic, with vascular communications between the fetal circulations. If one monochorionic twin were experiencing an adverse environment affecting its growth, and if the causal pathway between that adverse environment and programming of later health were mediated through some circulating factor during fetal life, then this circulating factor might also reach the otherwise unaffected co-twin and programme their later health. This could blunt within-pair associations between birthweight and later health in MZ but not DZ twins, and be misinterpreted as a genetic effect.

Data from studies examining within-cohort and within-pair associations

  1. Top of page
  2. Summary
  3. Introduction
  4. How can we use data from twins to shed light on the causal pathways underlying the association between birthweight and later disease risk?
  5. Data from studies examining within-cohort and within-pair associations
  6. Generalisability of findings in cohorts of twins
  7. Acknowledgements
  8. References

Findings from within-cohort vs. within-pair analyses of data from twins, published to date, are shown in Table 1 (all twins combined) and Table 3 (DZ and MZ twins separately). Important points to note are:

Table 3.  Regression coefficients (β) for difference in systolic blood pressure (mm Hg) per 1 kg increase in birthweight in cohorts and within pairs, in DZ and MZ twins separately. Adjusted value (generally for one or more measures of current size, ± age and sex in within-cohort analyses)
  DZ twinsMZ twins
β within cohortβ within pairsN pairsβ within cohortβ within pairsN pairs
  • a

    Two sets of triplets were included.

  • b

    Zygosity by questionnaire.

  • Birthweight by cmaternal or dself-report.

Dwyer16Childrena,b−6.4−4.939−10.4−6.516
Poulter14Adult femalesd−2.2−3.2237−4.0−4.9167
IJzerman17Adolescents (at rest)c−1.2−5.753−2.8−0.161
Baird18Adultsb,c0.12.6140−0.045.858
Zhang19Children; same sex pairs 2.986 −1.7119
Christensen20Adolescents−0.6−0.4450−2.4−1.1861
Loos21Adult males (at rest)−2.8−2.7440.43.8119
Adult females (at rest)−7.5−4.942−3.0−1.6116
Nowson22Adolescent and adult femalesa,b,c,d−3.3−1.3165−1.8−2.7131
Johansson-Kark23Young adult malesb−0.20.1233−3.8−1.3384
  • Some studies included unpaired twins in within-cohort analyses16,18,21–23 so that within-pair differences were examined in only a subgroup of the twins used for within-cohort analyses.
  • Two studies included triplets.16,22
  • Methodology for measuring blood pressure differed between studies.
  • Type of analysis ‘within cohort’ (whether or not allowance was made for the lack of independence between co-twins) varied between studies. As the type of analysis will certainly affect the estimate of variance, and in any case we do not have 95% confidence interval in every case, we have omitted 95% confidence interval/measure of statistical significance from these tables. It is also important to note that the type of analysis may also affect the regression coefficient (beta).
  • Independent factors included in regression models, other than birthweight, varied between studies.
  • Age of subjects varied between studies.
  • Some studies only included same-sex twins and others also included unlike-sex pairs.
  • The degree of stringency in determination of zygosity varied between studies.
  • The source (and likely accuracy) of birthweight data varied between studies.

Because of these problems we have not attempted meta-analysis of these data. However, it is evident that overall, within the cohorts, there is a small negative association between birthweight and systolic blood pressure, similar to that seen in cohorts of singletons. Of the nine studies here, a positive association was seen in only one. It is also evident that in most (but not all) of these studies there is a reduction in the size of the association when data are analysed within pairs.

Data from DZ and MZ twins separately are more difficult to interpret. The lack of power available because of the relatively small number of total subjects is a critical issue, and further studies are required before general inferences about this question can be developed.

For any summary analysis, we suggest that analysis of pooled data will be preferable to meta-analysis, and will work towards this.

Generalisability of findings in cohorts of twins

  1. Top of page
  2. Summary
  3. Introduction
  4. How can we use data from twins to shed light on the causal pathways underlying the association between birthweight and later disease risk?
  5. Data from studies examining within-cohort and within-pair associations
  6. Generalisability of findings in cohorts of twins
  7. Acknowledgements
  8. References

Whether our findings from analysis of data from twins can be generalised to singletons, comprising the majority of the population, is an important issue. This will be discussed after we have reviewed evidence on the role of factors associated with twinning as determinants of birth size, within-pair birthweight discordance and later health.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. How can we use data from twins to shed light on the causal pathways underlying the association between birthweight and later disease risk?
  5. Data from studies examining within-cohort and within-pair associations
  6. Generalisability of findings in cohorts of twins
  7. Acknowledgements
  8. References

We thank Janis Baird, Richard IJzerman, Ruth Loos, Caryl Nowson and Neil Poulter for providing additional unpublished data. Ruth Morley is supported by VicHealth, the Victorian Health Promotion Foundation.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. How can we use data from twins to shed light on the causal pathways underlying the association between birthweight and later disease risk?
  5. Data from studies examining within-cohort and within-pair associations
  6. Generalisability of findings in cohorts of twins
  7. Acknowledgements
  8. References
  • 1
    Byrne CD. Phillips DI. Fetal origins of adult disease: epidemiology and mechanisms. Journal of Clinical Pathology 2000; 53: 822828.
  • 2
    Waterland RA, Garza C. Potential mechanisms of metabolic imprinting that lead to chronic disease. American Journal of Clinical Nutrition 1999; 69: 179197.
  • 3
    Grantham-McGregor SM. Small for gestational age, term babies, in the first six years of life. European Journal of Clinical Nutrition 1998; 52: S1: S59–S64.
  • 4
    Jones G, Dwyer T. Birth weight, birth length, and bone density in prepubertal children: evidence for an association that may be mediated by genetic factors. Calcified Tissue International 2000; 67: 304308.
  • 5
    Thompson C, Syddall H, Rodin I, Osmond C, Barker DJ. Birth weight and the risk of depressive disorder in late life. British Journal of Psychiatry 2001; 179: 450455.
  • 6
    Morley R, Dwyer T. Fetal origins of adult disease? Clinical and Experimental Pharmacology and Physiology 2001; 28: 962966.
  • 7
    Kramer MS. Association between restricted fetal growth and adult chronic disease: is it causal? Is it important? American Journal of Epidemiology 2000; 152: 605608.
  • 8
    Gillman MW, Rich-Edwards JW. The fetal origin of adult disease: from sceptic to convert. Paediatric and Perinatal Epidemiology 2000; 14: 192193.
  • 9
    Eriksson JG, Forsen T, Tuomilehto J, Winter PD, Osmond C, Barker DJ. Catch-up growth in childhood and death from coronary heart disease: longitudinal study. British Medical Journal 1999; 318: 427431.
  • 10
    Ben-Shlomo Y, Kuh D. A life course approach to chronic disease epidemiology: conceptual models, empirical challenges and interdisciplinary perspectives. International Journal of Epidemiology 2002; 31: 285293.
  • 11
    Vagero D, Leon D. Ischaemic heart disease and low birth weight: a test of the fetal-origins hypothesis from the Swedish Twin Registry. Lancet 1994; 343: 260263.
  • 12
    Christensen K, Vaupel JW, Holm NV, Yashin AI. Mortality among twins after age 6: fetal origins hypothesis versus twin method. British Medical Journal 1995; 310: 432436.
  • 13
    Christensen K, Wienke A, Skytthe A, Holm NV, Vaupel JW, Yashin AI. Cardiovascular mortality in twins and the fetal origins hypothesis. Twin Research 2001; 4: 344349.
  • 14
    Poulter NR, Chang CL, MacGregor AJ, Snieder H, Spector TD. Association between birth weight and adult blood pressure in twins: historical cohort study. British Medical Journal 1999; 319: 13301333.
  • 15
    De Geus EJ, Posthuma D, Ijzerman RG, Boomsma DI. Comparing blood pressure of twins and their singleton siblings: being a twin does not affect adult blood pressure. Twin Research 2001; 4: 385391.
  • 16
    Dwyer T, Blizzard L, Morley R, Ponsonby AL. Within pair association between birth weight and blood pressure at age 8 in twins from a cohort study. British Medical Journal 1999; 319: 13251329.
  • 17
    IJzerman RG, Stehouwer CD, Boomsma DI. Evidence for genetic factors explaining the birth weight-blood pressure relation. Analysis in twins. Hypertension 2000; 36: 10081012.
  • 18
    Baird J, Osmond C, MacGregor A, Snieder CN, Phillips DIW. Testing the fetal origins hypothesis in twins: the Birmingham twin study. Diabetalogia 2001; 44: 3339.
  • 19
    Zhang J, Brenner RA, Klebanoff MA. Differences in birth weight and blood pressure at age 7 years among twins. American Journal of Epidemiology 2001; 153: 779782.
  • 20
    Christensen K, Stovring H, McGue M. Do genetic factors contribute to the association between birth weight and blood pressure? Journal of Epidemiology and Community Health 2001; 55: 583587.
  • 21
    Loos RJ, Fagard R, Beunen G, Derom C, Vlietinck R. Birth weight and blood pressure in young adults: a prospective twin study. Circulation 2001; 104: 16331638.
  • 22
    Nowson CA, MacInnis RJ, Hopper JL, Alexander JL, Paton LM, Margerison C, et al. Association of birth weight and current body size to blood pressure in female twins. Twin Research 2001; 4: 378384.
  • 23
    Johansson-Kark M, Rasmussen F, De Stavola B, Leon DA. Fetal growth and systolic blood pressure in young adulthood: the Swedish Young Male Twins Study. Paediatric and Perinatal Epidemiology 2002; 16: 200209.
  • 24
    Rolett A, Kiely JL. Maternal sociodemographic characteristics as risk factors for preterm birth in twins versus singletons. Paediatric and Perinatal Epidemiology 2000; 14: 211218.
  • 25
    Phillips DI, Davies MJ, Robinson JS. Fetal growth and the fetal origins hypothesis in twins – problems and perspectives. Twin Research 2001; 4: 327331.
  • 26
    Boyd NF, Dite GS, Stone J, Gunasekara A, English DR, McCredie MR, et al. Heritability of mammographic density, a risk factor for breast cancer. New England Journal of Medicine 2002; 19: 886894.
  • 27
    Dwyer T, Blizzard L, Morley R. Twins and the fetal origins hypothesis: within-pair analyses. Lancet 2002; 359: 22052206.
  • 28
    Leon DA. Twins and fetal programming of blood pressure. Questioning the role of genes and maternal nutrition. British Medical Journal 1999; 319: 13131314.
  • 29
    Huxley R, Neil A, Collins R. Unravelling the fetal origins hypothesis: is there really an inverse association between birthweight and subsequent blood pressure? Lancet 2002; 360: 659665.