A life-course approach to diet, growth and development

In social epidemiology, a comprehensive account of influences across the life course must be undertaken in order to understand why outcomes such as CVD, cancers or diabetes might develop in adulthood⁽ Reference Kawachi and Berkman ^{16 )}. Such an approach takes account of the immediate proximal pathophysiological processes leading to a specific disease outcome, and more recently how genetic, genomic or nutrigenomic factors acting at different time points might be affecting the outcome in question, which we discuss later ^{( 4 ,} Reference Chmurznska ^{17 –} Reference Yaghootkar and Freathy ^{24 )}. However, there are also contextual factors to be considered, such as individual lifestyle choices or health behaviours predisposing to risk factor development, and in turn how so-called meso-level situations such as the work environment, occupation or social class determine those health choices at different life stages, all of these being underpinned by the wider sociopolitical structure⁽ Reference Kawachi and Berkman ^{16 )}. This level of understanding is required if changes are to be wrought by public policy strategies. Study designs to take account of these factors may be longitudinal but also have design challenges to record what is measurable at the level of the individual cohort member and how best to record and examine ecological level factors such as a region or area, or group characteristics such as family membership.

The seminal influence of Barker⁽ Reference Barker, Winter and Osmond ^{1 ,} Reference Barker, Gluckman and Godfrey ^{3 ,} Reference Barker, Larsen and Osmond ^{5 ,} Reference Barker ^{6 )} in reframing our understanding of early life influences on adult chronic disease has created a paradigm shift in our understanding of these processes over the past three decades. In the original Hertfordshire retrospective cohort study, where current health outcomes were linked to original infant anthropometric records, Barker et al. showed the graduated inverse relationship between both birth-weight and weight in infancy and death for CHD at all ages in 10 636 men⁽ Reference Barker, Winter and Osmond ^{1 )}. As they pointed out, there are various reasons for babies to be small at birth and in infancy, including low birth-weight, short stature, placental insufficiencies and failure to thrive⁽ Reference Barker, Gluckman and Godfrey ^{3 )}. Since then many studies have considered cumulative, trajectory and critical period influences on longer term outcomes. It is now well understood that there are two processes at the extremes of birth-weight, whereby larger babies are at greater risk of subsequent diabetes and obesity, in turn related to maternal weight and height⁽ Reference Cnattingus, Villamor and Lagerros ^{8 ,} Reference Poston ^{25 )} and lower birth-weight babies subjected to degrees of intrauterine growth retardation are also susceptible to later adult chronic disease⁽ Reference Barker, Winter and Osmond ^{1 ,} Reference Barker, Larsen and Osmond ^{5 ,} Reference Adair, Fall and Osmond ^{7 ,} Reference Eriksson, Kajantie and Thornburg ^{10 )}. The developmental plasticity hypothesis, whereby there is a critical period during which a system is plastic, followed by loss of plasticity and a fixed functional capacity, is now well accepted as a biological phenomenon, and pregnancy, during which the intrauterine development of all children occurs, is one such critical period shared by all individuals⁽ Reference Barker, Larsen and Osmond ^{5 ,} Reference Eriksson, Kajantie and Thornburg ^{10 ,} Reference Gray, Davey Smith and McConnachie ^{11 )}.

Barker has since extended this concept to define what he calls 100 years of nutritional flow⁽ Reference Barker ^{6 )}. The 1000 d from conception to toddlerhood not only include maternal influences such as her nutritional supply to the fetus during pregnancy and early childhood feeding practices, but also the quality, size and shape of the placenta are increasingly understood to play a role in successful fetal and early childhood growth and development milestones and subsequent long-term health outcomes. The maternal grandmother is especially influential as it is she who makes the grandchild's egg for subsequent fertilisation by the ovulating mother. In this sense, the developmental investment in the index child dates across generations and an ageing adult owes his health status to his own constitutional and environmental exposures but also to that of his parents and grandparents. There is also the genetic influence of both parents and all grandparents, the X-linked effects of the mother, the Y-linked effects of the paternal line, and notably through the maternal line, the mitochondrial DNA transmission⁽ Reference Pembrey, Bygren and Edvinsson ^{26 ,} Reference Fox, Sear and Beise ^{27 )}. Recent studies have shown an association between variants in the ADCY5 gene and birth-weight and up to seven loci identified which account for a similar proportion of the variance in infant birth-weight as smoking⁽ Reference Horikoshi, Yaghootkar and Mook-Kanamori ^{19 )}. This may predispose individuals both to relatively lower birth-weight and a propensity to adult chronic disease.

Cross-generational evidence

In recent years, a number of studies have been published showing the relationship between infant birth-weight and subsequent maternal and paternal morbidity and mortality⁽ Reference Andersen and Osler ^{28 –} Reference Wannamethee, Lawlor and Whincup ^{31 )}. Generally such an association has been shown for lower birth-weight, particularly for mothers. Many studies have attempted to examine how children's growth trajectories relate to their parents’ anthropometric characteristics⁽ Reference Durmus, Arends and Ay ^{9 ,} Reference Gray, Davey Smith and McConnachie ^{11 )}. Prospects for understanding the genetic, epigenetic or nutrigenomic mechanisms have recently been reviewed⁽ Reference Relton and Davey Smith ^{23 )}. These authors define epigenetics as the regulatory processes that control the transcription of information encoded in the DNA sequence into RNA before their transcription into proteins. The tissue specificity of epigenetic patterns is a well-established phenomenon and not all studies report expected associations⁽ Reference Ivanova, Chen and Segonds-Pichon ^{21 )}, even though in principle nutrigenomic effects may be important as for instance DNA methylation which occurs in utero depends on the availability of several nutrients including methionine, vitamins B₆, B₁₂ and folate⁽ Reference Chmurznska ^{17 )}.

A further use of genetic markers is in so-called Mendelian randomisation, whereby an individual receives at random a specific allele from its parent, this process being unlikely to be confounded by any subsequent behavioural characteristic of the individual. It is then possible to adjust for the known influence of this marker on an apparent prospective association between a risk factor and the disease-specific outcome of interest, using genotype as an instrumental variable. This has been shown for the FTO gene in Danish adults in relation to the association of obesity with IHD⁽ Reference Palmer, Nordestgarrard and Benn ^{32 )} and to attenuate greatly the apparently strong association between maternal BMI and offspring adiposity⁽ Reference Lawlor, Timpson and Harbord ^{33 )}. These studies have attempted to address biased or confounded associations in that parents may share genetic or environmental characteristics, but not the direct health behaviours of their offspring⁽ Reference Bonilla, Lawlor and Ben-Shlomo ^{34 )}.

Several animal studies have provided evidence for transgenerational effects of intrauterine exposure⁽ Reference Harrison and Langley-Evans ^{18 ,} Reference Ponzio, Carvalho and Fortes ^{22 ,} Reference Zambrano, Martinez-Samayoa and Bautista ^{35 ,} Reference Torrens, Poston and Hanson ^{36 )}. Female rats, exposed to protein restriction during pregnancy and lactation, were found to have offspring that delivered a progeny with altered growth and metabolism⁽ Reference Zambrano, Martinez-Samayoa and Bautista ^{35 )}. A similar protein restriction trial in pregnant rats reported that second-generation offspring had raised blood pressure and endothelial dysfunction which was passed through the maternal line⁽ Reference Torrens, Poston and Hanson ^{36 )}. In contrast, Harrison and Langley-Evans⁽ Reference Harrison and Langley-Evans ^{18 )} reported that protein restriction during pregnancy results in a phenotype of raised blood pressure and reduced nephron number passed through maternal and paternal lines. More recently, Ponzio et al.⁽ Reference Ponzio, Carvalho and Fortes ^{22 )} have demonstrated that phenotypes resulting from maternal protein restriction are evident in a second and third generation of animals.

Studies of body composition and diet in human subjects have either been limited to one generation or have used proxy markers for diet intake across one or more generations, such as those of the Overkalix studies⁽ Reference Pembrey, Bygren and Edvinsson ^{26 ,} Reference Kaati, Bygren and Pembrey ^{37 )} and the Dutch Winter Hunger study⁽ Reference Heijmans, Tobi and Stein ^{20 )}. By contrast with parental data studies⁽ Reference Andersen and Osler ^{28 –} Reference Wannamethee, Lawlor and Whincup ^{31 ,} Reference Keane, Layte and Harrington ^{38 ,} Reference Davey Smith, Steer and Leary ^{39 )} there are relatively fewer data on how grandparental characteristics affect their grandchildren's growth and development; all such studies have either been retrospective or data linkage in design and none to date had dietary intake data across generations. To date there are just four well-characterised human studies with three generations⁽ Reference Manor and Koupil ^{40 –} Reference Naess, Stoltenberg and Hoff ^{43 )}. Manor and Koupil⁽ Reference Manor and Koupil ^{40 )} showed a U-shaped association between grandchild's birth-weight and circulatory disease mortality in maternal grandmothers only for infants born prior to 1977. There was a U-shaped association for maternal grandfathers’ overall and circulatory disease mortality and for paternal grandfathers there was an inverse association between grandchild's birth-weight and overall mortality. McCarron et al. ⁽ Reference McCarron, Davey Smith and Hattersley ^{41 )} found that the grandchildren of maternal grandparents with reported type 2 diabetes were more likely to be in the top tertile of birth-weight. There was evidence for an inverted U-shaped association between birth-weight of grandchildren and diabetes in paternal grandmothers. Smith et al. ⁽ Reference Smith, Wood and White ^{42 )} have shown an inverse relationship between maternal grandparents’ risk of IHD or cerebrovascular disease and infant birth-weight and did not have data on paternal lineages. In a very large linkage study in Norway, inverse relationships were shown according to both maternal and paternal lineages and cardiovascular outcomes in grandparents; adjusting for maternal smoking accounted for much of the effect on cardiovascular mortality. For grandparental diabetes mortality, U-shaped associations were seen with grandchild birth-weight for the maternal grandmother and inverse associations for all other grandparents⁽ Reference Naess, Stoltenberg and Hoff ^{43 )}.

Prospective studies with information on diet and parental BMI

In an exploration of the literature, we attempted to identify cohort studies that included a measure of nutritional status (BMI or diet assessment) across two or three generations⁽ Reference Alves, Correia, Barros and Azevedo ^{44 –} Reference Wickman, Kull and Pershagen ^{106 )}. Prospective or cohort studies that include specifically information on maternal diet or BMI and that of other family members in relation to index children are summarised in Table 2. As Poston has pointed out, many of these are relatively recently established⁽ Reference Poston ^{25 )}. The studies vary in their objectives; several look at offspring size as the outcome of interest; others include diet and body size as factors associated with child and adult morbidities. Maternal pre or postnatal BMI is the variable most frequently reported (63 %) by these studies followed by paternal postnatal BMI (41 %; Table 2). Diet is less frequently reported with only one-third collecting information on prenatal maternal diet; prenatal paternal diet has only been collected by eight studies.

Table 2. Prospective studies with information on either diet or parental BMI, or both

* Prenatal BMI may be estimated before pregnancy or measured during pregnancy.

† Cohorts of children born to mothers with gestational diabetes mellitus or previous macrosomia.

‡ Cohorts currently recruiting or have recently completed recruitment.

The 1958 birth cohort has described the relationship between birth-weight of the grandchild and the influence of height by the grandparents⁽ Reference Emanuel, Filakti and Alberman ^{57 )}. Only one other study by Davis et al.⁽ Reference Davis, McGonagle and Schoeni ^{51 )}, has described a measure of BMI for grandparents. This study is limited, however, by the fact that grand maternal and grand paternal body composition was aggregated in the reported analysis. To our knowledge, no other studies have reported data available on either the BMI or diet of the proband child and their individual maternal and paternal grandparents. This makes the Lifeways cross-generation cohort study a highly unusual dataset in the literature, in that it contains information on health status, dietary intakes and adult chronic disease outcomes in the grandparental generation related to the health status of their grandchildren⁽ Reference O'Mahony, Fallon and Hannon ^{13 ,} Reference Kelleher, Bury and Murphy ^{14 )}.

Findings to date in the Lifeways study: pregnancy and early years

The Lifeways cross-generation cohort study was established in 2001–2013 through two maternity hospitals in the Republic of Ireland and was one of many new cohort studies established worldwide in the millennium period⁽ Reference Larsen, Kamper-Jørgensen and Adamson ^{12 ,} Reference O'Mahony, Fallon and Hannon ^{13 ,} Reference Kelleher, Bury and Murphy ^{14 )}. Mothers were recruited at first booking visit, completing a self-administered questionnaire which included a 147 item validated semi-quantitative FFQ. Longitudinal follow-up is ongoing in 2013, with linkage data to hospital and general practice records and examination of children when aged on average 5 and 9 years. The study is one of very few containing data on grandparents of both lineages, with at least one grandparent recruited at baseline. A summary of publications related to diet and cross-generation transmission to date for the Lifeways cohort study is given in Table 3 ⁽ Reference Segonds-Pichon, Hannon and Daly ^{107 –} Reference Viljoen, Shrivastava and Murrin ^{121 )}. These analyses in this cohort have been consistent in showing an influence on health outcomes at different time points of lifestyle factors including diet and of cross-generation and familial associations. This is arguably remarkable given that it is not a large cohort study, although characterised in some detail and with linkage to health records.

Table 3. Cross-generational or longitudinal Influences on health outcomes in the Lifeways cohort study 2001–2013

GMS, General medical services.

We have previously shown a social gradient in relation to baseline dietary intake at recruitment for all adult family members, in that those eligible for GMS had lower intakes of vitamin C and higher total fat intakes, having adjusted for other risk factors⁽ Reference Kelleher, Lotya and O'Hara ^{110 )}. Both age and cohort group were important factors, as might be expected, since the parental and grandparental dietary patterns might be likely to differ and sex differences are also apparent, with higher mean energy intakes in males compared with females. Social gradients at the level of food groups have been reported in the Surveys of Lifestyles, Attitudes and Nutrition at each of the three time points 1998, 2002 and 2007⁽ Reference Kelleher, Friel and Nolan ^{122 ,} Reference Niedhammer, Kerrad and Schütte ^{123 )}, so the Lifeways study corroborates that observation at the level of selected nutrient intakes.

The pregnancy outcomes were also examined in relation to indicators of maternal social position and health status⁽ Reference Segonds-Pichon, Hannon and Daly ^{107 –} Reference Murrin, Fallon and Hannon ^{109 )}. Determinants of mothers’ self-rated health during pregnancy showed a strongly socially graduated pattern and mothers whose own parents were educated to third level had better self-rated health than those whose parents left education early⁽ Reference Segonds-Pichon, Hannon and Daly ^{107 )}. Younger and less socially advantaged mothers tended to have lower birth-weight babies⁽ Reference Murrin, Segonds-Pichon and Fallon ^{108 )}. The dietary patterns of mothers during pregnancy were socially graduated, as was smoking status⁽ Reference Murrin, Fallon and Hannon ^{109 )}. Predictors of pre-term delivery and occupational predictors of pregnancy outcomes were reported in this cohort⁽ Reference Niedhammer, O'Mahony and Daly ^{111 –} Reference Niedhammer, Murrin and O'Mahony ^{112 )}. Mothers required to do shiftwork tended to have a higher risk of pre-term delivery, adjusted for other factors⁽ Reference Niedhammer, O'Mahony and Daly ^{111 )} and this cohort was included in a subsequent meta-analysis, which showed a small effect of shiftwork on risk for this outcome⁽ Reference Bonzini, Palmer and Coggon ^{124 )}.

While many studies report a social gradient to pregnancy outcomes, a few attempts to comprehensively explain the degree to which known health status and lifestyle variables explain that gradient have been made. In the Lifeways study low maternal educational level was associated with an increased risk of this clinical outcome of preterm delivery (hazard ratio 2·14, 95 % CI 1·04, 4·38) and notably the explanatory combination of a material factor such as rented or crowded home, and behavioural factors such as smoking, alcohol consumption and high saturated fat intake during pregnancy reduced the hazard ratio for low education level by 42 %⁽ Reference Niedhammer, Murrin and O'Mahony ^{112 )}. We looked at predictors of low birth-weight of 3 kg or less both in this cohort and in another cohort of disadvantaged Traveller infants, a nomadic disadvantaged group in Ireland and found that maternal smoking and alcohol consumption were important predictors of this outcome, but when these factors were taken account of the birth-weight differential according to social class persisted. We used this cut-off point, as opposed to the clinical cut-off of 2.5 kg, as epidemiological studies show this is associated with later risk of adult chronic disease, providing empirical evidence in an Irish context for findings reported in other cohorts internationally⁽ Reference Hamid, Fitzpatrick and Rowan ^{113 )}.

Follow-up when children averaged age 5 years

A number of analyses have been undertaken across the generations of participating families related to the wave when children now averaged age 5 years⁽ Reference Murrin, Kelly and Tremblay ^{116 –} Reference Murrin, Shrivastava and Kelleher ^{118 )}. Although the responses were received for just 669 of 1126 women originally recruited and not all of these had complete examination data, there was no significant difference in baseline maternal BMI in this sweep's participants compared with that of non-participants, although these mothers were older on average than non-participants. We applied mixed effect models to take account of family relatedness, which enables correlation between family members to be modelled and also to take account of incomplete or variable numbers of family participants. Each family could contain up to seven possible members; in the child the height and weight were measured but for other adult family members the data were self-reported. We found consistent correlations for height between family members in both paternal and maternal lines, but the patterns for BMI were significant only for the maternal line. In these statistical models, education level, and also self-rated health and fruit and vegetable consumption patterns were accounted for because they each showed a social gradient and a relationship with BMI in previous analyses, with a strong significant family effect seen across the generations in the maternal line⁽ Reference Murrin, Kelly and Tremblay ^{116 )}.

The reported evidence in the general literature on the relative influence of paternal and maternal lines on children's BMI is mixed⁽ Reference Durmus, Arends and Ay ^{9 ,} Reference Keane, Layte and Harrington ^{38 ,} Reference Davey Smith, Steer and Leary ^{39 ,} Reference Hui, Nelson and Yu ^{68 ,} Reference Subramanian, Ackerson and Smith ^{125 ,} Reference Kivimaki, Lawlor and Smith ^{126 )}. The Growing Up in Ireland study found that measured parental weight status was correlated with childhood obesity at age 9 years⁽ Reference Keane, Layte and Harrington ^{38 )}. While the Avon Longitudinal Study of Parents and Children study for instance showed no distinction in strength of association between maternal and paternal lines⁽ Reference Davey Smith, Steer and Leary ^{39 )}, the Generation R cohort study also showed a strong maternal pattern, although there was a paternal influence as well⁽ Reference Durmus, Arends and Ay ^{9 )}.

Familial aggregation patterns in dietary intake were also assessed in the Lifeways study. Nuclear family effects were found, with the maternal association being stronger than the paternal association and the maternal grandmothers showed an association both with their own daughters’ diet and with that of their index grandchild's diet at age 5 years. The mother's current diet tended to be more strongly associated with that of her child than her pregnancy consumption, although in the case of non-breastfeeding mothers, dietary fat intake during pregnancy was more strongly associated⁽ Reference Shrivastava, Murrin and Sweeney ^{117 )}. Brion et al. ⁽ Reference Brion, Ness and Rogers ^{46 )}, employing Avon Longitudinal Study of Parents and Children data, found a similar association with regard to prenatal maternal fat intake.

Maternal dietary intake at both time points was also examined in relation to child's BMI at age 5 years. A social gradient in relation to maternal fruit and vegetable intake was found during pregnancy and also when the child averaged age 5 years. Total energy intake in mothers was greater during pregnancy. Increased odds of overweight or obesity were found in mothers with higher intakes of sugar during pregnancy and higher fat intakes when the child was aged 5 years. Mothers with persistently high intakes of SFA and those who had lowered their sugar consumption since pregnancy were more likely to have overweight or obese children and we discuss the possible biological plausibility of these findings in that paper⁽ Reference Murrin, Shrivastava and Kelleher ^{118 )}. Again, cohort data on this issue are rare. The Southampton women's survey⁽ Reference Moon, Harvey and Robinson ^{127 )} reported that after adjustment for maternal factors including height and duration of breastfeeding maternal n-6 PUFA intake positively predicted fat mass in children aged 4 and 6 years.

Dietary patterns and risk of childhood wheezing or asthma

A specific health outcome of interest that can already be examined in the children themselves is childhood wheeze or asthma, which is strongly suspected to have a dietary basis. The Lifeways study was one of the first to show that general practitioner-reported wheeze or asthma when the child averaged 3 years was inversely associated with maternal fruit and vegetable and fish consumption during pregnancy⁽ Reference Fitz-simon, Fallon and O'Mahony ^{114 )}. In Table 4, we show predictors of reported asthma for children at both time points when they were aged on average 3 and 5 years, replicating the original analysis at the new time point. Various biological, lifestyle and socioeconomic determinants of both child and maternal association that were significant predictors of asthma at the uni-variate level were included as covariates in multivariate models⁽ Reference Viljoen, Shrivastava and O'Brien ^{115 )}. Separate logistic regression models were constructed to examine the effect of food group intake relative to asthma status at each phase of follow-up. Models were adjusted for combinations of birth-weight, sex, region, maternal age, maternal socioeconomic status (measured by education or GMS eligibility), parity, marital status and smoking in pregnancy. Oily fish intake proved to be protective of asthma at both year 3 (odds ratios (OR) = 0·52, 95 % CI 0·30, 0·92) and year 5 (OR = 0·51, 95 % CI 0·28, 0·92). Similarly, vegetable intake in the upper quartile proved protective at year 3 (OR = 0·42, 95 % CI 0·19, 0·93) and year 5 (OR = 0·43, 95 % CI 0·18, 0·99). This shows that mothers of asthmatic children tended to be younger, more disadvantaged and to have a lower oily fish, fruit and vegetable intake during pregnancy. Conversely, high intake of added or spreadable fats were related to asthma at year 3 follow-up (OR = 2·46, 95 % CI 1·34, 4·51). These data are quite consistent over the two time points and notably it is the baseline maternal dietary intake, rather than characteristics at aged 5 years, which show the main effects. The mechanisms through which this may operate are not as yet understood but may be nutrigenomic in origin and several other investigators have examined this question⁽ Reference Devereux and Seaton ^{128 )}. The Lifeways dataset has recently formed part of a pooled analysis as part of the CHICOS (Developing a Research strategy for Child Cohorts in Europe) consortium funded by the FP7 programme to elucidate these pathways further⁽ Reference Larsen, Kamper-Jørgensen and Adamson ^{12 )}; one analysis looked at the influence of maternal fish intake on birth outcomes⁽ Reference Leventakou, Roumeliotaki and Martinez ^{129 )} and another on how birth outcomes might predict risk of childhood asthma⁽ Reference Sonnenschein-van de Voort, Arends and de Jongste ^{130 )}.

Table 4. Predictors of General Practitioner-reported wheeze (10·4 % of n 614) in children aged 3 years and of maternal reported asthma (14·3 % of n 511) in children aged 5 years in the Lifeways Cohort Study

GMS, General medical services; OR, odds ratio; prim, primary; sec, secondary; multip, multipara; nullip, nullipara.

* P < 0·05

** P = 0·05–0·09

† Adjusted OR are reported for nutritional variables; all other values are from the univariate analysis.

Longitudinal follow-up when children were aged 9 years

Because these analyses to date showed such patterns of association across the generations, it was decided to follow-up adults and children again and between 2011 and 2013 further follow-up has taken place of the families. This included a short self-administered health questionnaire for all adults, with 1587 respondents in 593 families, 53·9 % of the original birth cohort families. A note search was undertaken initially in 2011 for grandparent deaths in the General Registry Office. Grandparental morbidity and mortality patterns were then related also to infant birth-weight. We related also the grandparents' blood pressure and lipoprotein profile at baseline recruitment stage to their grandchildren's birth-weight, the first such report to date and examined indicators of morbidity, stroke and diabetes based on self-report from the baseline questionnaire and general practice records for the grandparents at follow-up when the children averaged age 3 years. Maternal grandmothers' likelihood of both stroke and diabetes were inversely related to infant birth-weight and an inverse or U-shaped association was seen for maternal grandparents' mortality patterns, although not statistically significant. Conversely, the patterns for paternal grandparents and birth-weight were positive⁽ Reference Shrivastava, Murrin and O'Brien ^{120 )} and paternal grandfathers who had now died were more likely to have had higher birth-weight grandchildren, adjusted for grandchildren's gestational age, gender, mothers' age, height, parity, educational status, smoking in pregnancy and grandparents' age, smoking and educational status. Given the novelty of these findings, and the possibility that some form of participant self-selection had influenced the association, we extended the General Registry Office note search in 2012 to all grandparents for whom we had any contact details and not just those who had been examined or completed a questionnaire at baseline. Preliminary analysis of these data reconfirm a positive association between higher birth-weight and paternal grandfathers' mortality⁽ Reference Viljoen, Shrivastava and Murrin ^{121 )}. Our findings in relation to the maternal line were therefore broadly consistent with most other published studies⁽ Reference Smith, Wood and White ^{42 ,} Reference Naess, Stoltenberg and Hoff ^{43 )}. Although the Norwegian linkage study shows a contrasting pattern to ours in relation to the paternal lineage⁽ Reference Naess, Stoltenberg and Hoff ^{43 )}, in that both paternal and maternal grandparental cardiovascular mortality patterns were inversely associated with birth-weight, there was an important effect for maternal smoking on that relationship and secular influences on lifestyle may be explaining the differences in patterns in the cohorts to date. There are very few evidenced studies published on this issue, which merits further research and investigation.

During 2012 we conducted examinations through their general practitioners of height and weight of children and their mothers and took salivary and hair samples for future genotyping, 301 were successfully completed and seventy-six of these mainly Dublin-based children also had blood samples taken for lipoprotein profile. Anthropometric measurements at birth, when children were aged 5 and 9 years are summarised in Table 5. We assessed the representativeness of the children who participated in the 2012 follow-up and had valid height and weight measurements. We compared these children with our original baseline cohort in terms of mothers' age, baseline BMI, education level and medical card holder status⁽ Reference Khalil, Viljoen and O'Brien ^{119 )}. A greater proportion of children at follow-up had mothers who did not hold a medical card (90·3 v. 79·2 %, P < 0·001) and achieved a tertiary level of education (60·3 v. 45·4 %, P < 0·001). Mothers were also slightly older than the cohort at baseline (mean age 32·0 years v. 29·0 years, P < 0·001) but did not differ significantly in BMI. The east coast grandparents are undergoing a cardiovascular risk profile assessment in 2013, including also salivary and hair samples and morbidity follow-up of grandparents through general practitioners is ongoing in 2013. In the next stage of this study, a detailed analysis of these biological outcomes will be undertaken, related to the information gathered in previous sweeps of the study.

Table 5. Anthropometric measurements of Lifeways cross-generation cohort study children at birth, at age 5 and at age 9 years