Birthweight in offspring and cardiovascular mortality in their parents, aunts and uncles: a family-based cohort study of 1.35 million births

Abstract Background A link between suboptimal fetal growth and higher risk of cardiovascular disease (CVD) is well documented. It has been difficult to assess the contribution of environmental versus genetic factors to the association, as these factors are closely connected in nuclear families. We investigated the association between offspring birthweight and CVD mortality in parents, aunts and uncles, and examined whether these associations are explained by CVD risk factors. Methods We linked Norwegian data from the Medical Birth Registry, the Cause of Death Registry and cardiovascular surveys. A total of 1 353 956 births (1967–2012) were linked to parents and one maternal and one paternal aunt/uncle. Offspring birthweight and CVD mortality association among all relationships was assessed by hazard ratios (HR) from Cox regressions. The influence of CVD risk factors on the associations was examined in a subgroup. Results Offspring birthweight was inversely associated with CVD mortality among parents and aunts/uncles. HR of CVD mortality for one standard deviation (SD) increase in offspring birthweight was 0.72 (0.69–0.75) in mothers and 0.89 (0.86–0.92) in fathers. In aunts/uncles, the HRs were between 0.90 (0.86–0.95) and 0.93 (0.91–0.95). Adjustment for CVD risk factors in a subgroup attenuated all the associations. Conclusions Birthweight was associated with increased risk of CVD in parents and in aunts/uncles. These associations were largely explained by CVD risk factors. Our findings suggest that associations between offspring birthweight and CVD in adult relatives involve both behavioural variables (especially smoking) and shared genetics relating to established CVD risk factors.


Introduction
A link between suboptimal fetal growth and a higher risk of cardiovascular disease (CVD) has been demonstrated within individuals in several populations. 1-3 Some causal models have been proposed to define a mechanism underlying this association, including intrauterine programming by epigenetic mechanisms 4 and common genetic factors influencing both fetal growth and adult diseases. 5 Alternatively, behavioural/environmental factors may explain the low birthweight(LBW) and CVD risk association. 6 The importance of both genetic and shared environmental factors has been emphasized in previous research. [7][8][9] Some studies report stronger association in mothers than fathers, highlighting the importance of intrauterine factors. 10,11 Moreover, a strong genetic correlation has been found in a genome-wide association study between birthweight (BW) and coronary artery disease, blood pressure and type 2 diabetes, suggesting that the association between BW and adult disease may partly be explained by shared genetic variants. 12 Family studies have reported inverse relationships between offspring BW and CVD mortality in both parents and grandparents, which may implicate common genetic factors. 13,14 As anticipated, maternal smoking during pregnancy was found to be a key confounding factor, 15 suggesting genetic and non-genetic mechanisms in the intergenerational transmission of disease risk. 9,16,17 However, it has been notoriously difficult to separate the contribution of common genetic factors from shared behavioural/socioeconomic circumstances within a nuclear family, because these potential influences are closely linked.
Investigating the offspring BW and CVD mortality association in extended family members such as aunts/uncles provides an alternative approach to studies investigating parental offspring associations. Offspring in principle share on average 50% of their genes with their parents, and they share on average 25% of their genes with their aunts and uncles. We assume that aunts/uncles in most cases belong to households different from their nieces/ nephews, and therefore are less likely to share environmental factors compared with the parents and their offspring.
The objective of this study employing data from the Norwegian Medical Birth Registry and Cause of Death Registry was to investigate if the association observed between offspring BW and parental CVD mortality can also be observed for aunts/uncles, and to explore to what extent these associations are explained by known CVD risk factors such as body mass index (BMI), blood pressure, total cholesterol and smoking. We hypothesized that if shared genes explain the BW and CVD association, we would expect a stronger offspring BW and CVD mortality association in parents than in aunts/uncles, and a similar pattern of association in all four classes of aunts/uncles.

Methods
A cohort was created by linking Norwegian data from cardiovascular health surveys, the Medical Birth Registry, the Cause of Death Registry, the Educational Registry and a multigenerational database containing information on familial relationships for the whole population of Norway. We included offspring (born between 1967 and 2012) with available information on their parents and at least one maternal and one paternal aunt/uncle. Aunts/uncles were defined as full siblings of a parent (sharing both mother and father). Offspring births with gestational age <37 / >44 weeks or BW <1000 g were excluded. The final dataset comprised 1 353 956 births linked to parents and one maternal and one paternal aunt/ uncle ( Figure 1).

Measures
BW (in grams) was analysed as a continuous variable and according to categories of offspring BW for gestational age: small for gestational age (SGA), <10th percentile of the BW distribution; large for gestational age (LGA), >90th percentile of the BW distribution; and appropriate for gestational age (AGA), 10th-90th percentiles of BW distribution. 18 Additional data were included for offspring (sex, year of birth and congenital anomalies coded as 'diseases in offspring') 19 and for mothers [age, parity, smoking, diseases before pregnancy (asthma, chronic hypertension, chronic renal disease, urinary tract infection, rheumatoid arthritis, heart disease, diabetes, epilepsy and thyroid diseases), and diseases during pregnancy (vaginal bleeding, glycosuria, hypertension, preeclampsia, eclampsia, gestational diabetes, anaemia, thrombosis and infection]. These maternal and offspring factors could be important confounders for the relationship between BW and CVD mortality in parents. However, to make the analysis comparable between all relationships, we adjusted model 1 for mother's age at offspring birth in every association. Data on age at offspring's birth and the highest level of education (9 years, 10-12 years and 13 years) completed by 2011 were included both for parents and for aunts/uncles. Three large cardiovascular health surveys-the County Study, 20 the Age 40 Program 21 and Cohort Norway (CONOR) 22 -were conducted in Norway during 1974-88, 1985-99 and 1994-2003, respectively. CVD risk factor data-body mass index (BMI; kg/m 2 ), total cholesterol (TC; mmol/L), triglycerides (TG; mmol/L), systolic and diastolic blood pressure (SBP and DBP; mmHg), and smoking-from these health surveys were available in a subgroup ( Figure 1). We used this subgroup to examine the role of traditional CVD risk factors on the association between offspring BW and CVD mortality in parents and in aunts/uncles. In the subgroup, follow-up was started from the date of CVD risk factors measurement in the population surveys.

Statistical analysis
Cox proportional hazard models were used to calculate the hazard ratio (HR) of deaths from CVD, IHD and stroke in parents and in aunts/uncles for a one standard deviation (SD) increase and categories of offspring BW (SGA and LGA with AGA as the reference). Parent's, aunt's/uncle's age was the time axis for the Cox model. Follow-up started at the date of offspring birth and continued up to the parent's/aunt's/uncle's emigration, death or end of the study (30 December 2014). The proportional hazards assumption was examined by plotting the Schoenfeld residuals and was not found to be violated by visual inspection. Total person -years included for the analysis were 30 908 031 (fathers), 31 671 408 (mothers), 29 928 884 (maternal siblings) and 30 020 262 (paternal siblings). Several offspring in our study were nested within the same parents, aunts/uncles. These offspring were clustered on their parents' and aunts'/uncles' identity, using the 'vce cluster' command in Stata. This command effectively adjusts the standard error for within-parents and within- aunts/uncles correlation. Some of the aunts/uncles appeared in the data more than once, as they could be the sibling of several mothers or fathers in the sample.
Modelling was carried out in three stages: Model 1 was adjusted for mother's age at offspring birth (continuous). Model 2 was additionally adjusted for offspring year of birth (continuous), maternal parity (coded as 0, 1 or 2) and maternal diseases before and during pregnancy [coded as 0 (no) or 1 (yes) and disease in offspring at birth (coded as 0 (no) or 1 (yes)]. Model 3 was additionally adjusted for the education of parents, aunts and uncles and marital status of the parents. In the subsample for which CVD risk factor data were accessible, the association between offspring BW and mortality from CVD, IHD and stroke in parents and in aunts/uncles was first adjusted for mother's age, which is comparable with Model 1 in the full dataset. The association was then additionally adjusted for CVD risk factors (BMI, TC, TG, SBP, DBP and smoking) and education of parents, aunts and uncles. To examine specificity of outcomes, whether the paternal association appears to reflect socioeconomic/behavioural confounding, we repeated our analysis with lung cancer mortality as outcome.

Results
Mean follow-up time (6SD) for the parents and aunts/ uncles was 47 6 5 years. Mean age (years) at the followup was 54 6 9.8 (fathers), 52 6 9.7 (mothers), 55 6 10.4 (maternal siblings), 56 6 10.7 (paternal siblings). During follow-up, 0.29 % of mothers and 1.20 % of fathers died of CVD. The parents, aunts and uncles of the SGA offspring were comparatively younger and less educated than the other two groups. Maternal smoking during pregnancy was associated with lower offspring BW in the subgroup where these data were available. The maximum age of aunts and uncles at follow-up was 74 years. During follow-up, 0.55 % of maternal aunts and 1.68 % of maternal uncles died of CVD. The respective percentages for paternal aunts and uncles were 0.60 % and 1.86 % ( Table 1).

Parental mortality in relation to offspring BW
An inverse association between offspring BW and ageadjusted mortality from CVD, IHD and stroke was observed among mothers and fathers, but was stronger among mothers (Table 2). For all separate causes of death, adding offspring year of birth, maternal parity, maternal 'disease before and during pregnancy' and 'disease in offspring' to the model minimally attenuated the associations in mothers and fathers (Model 2). The effect estimates for 1-SD increase in offspring BW were attenuated marginally in the parents when marital status and educational level were included in Model 3 ( Table 2). The age-adjusted HR (95% CI) for CVD mortality in mothers and fathers of SGA offspring compared with AGA offspring were 2.02 (1.85-2.21) and 1.33 (1.26-1.40), respectively. In LGA offspring a reduced hazard for CVD mortality was observed among mothers and fathers [HR for mothers, 0.74 (0.63-0.86); for fathers, 0.84 (0.78-0.90)]. For IHD and stroke mortality, similar trends in SGA and LGA offspring were observed in both parents (Table 3). We also analysed data according to the sex of the offspring. No difference in association was observed in either parent (Supplementary  Table 1a and b, available as Supplementary data at IJE online).
Aunts' and uncles' mortality in relation to niece/nephew BW LGA (more than 90th percentile of offspring birthweight). d Information on smoking during pregnancy was available in 369 844 mothers. P-value for continuous variables calculated by one-way ANOVA and for categorical variables by chi square test. Continuous variables are given as mean 6 SD and categorical variables are given as percentages.

Discussion
We have shown an inverse association between offspring BW and mortality from CVD and IHD in parents and in their siblings (aunts/uncles). The association was stronger in mothers than in fathers or in aunts/uncles. There were no differences in the estimates among the four classes of aunts/uncles, and the associations among fathers were only slightly stronger than those in aunts/uncles. The associations were to a large extent explained by CVD risk factors.

Comparison of results with previous studies and potential mechanisms
The relationship between lower offspring BW and increased risk of CVD among parents and aunts/uncles may support a genetic basis for the association. The relationship observed in parents is consistent with previous studies including both mothers and fathers, 23,24 and with studies indicating a stronger association in mothers than in fathers. 14,25 In contrast, another study reported similar father-offspring and motheroffspring associations for cardiovascular risk factors. 26 To our knowledge, the association between niece/nephew BW and CVD mortality in aunts/uncles has not previously been explored. Therefore, direct comparison of our results with other studies is not possible. However, a number of multigenerational studies, reporting a strong association between grandchild BW and mortality in grandparents, support a genetic influence on the association between BW and CVD. [13][14][15] CVD has a substantial genetic component and several genes, particularly those encoding glucokinase, 5 clotting factors 27 and angiotensinogen, 28 have mutations that are associated with both restricted fetal growth and risk of CVD. A recent study also confirmed genetic influence on the association between LBW and adult hypertension. 29 Additionally, it has been proposed that shared environmental factors, such as smoking, diet and socioeconomic position (SEP), also may contribute to the negative association between BW and CVD risk. 30 Model 1 was adjusted for maternal age at offspring birth. Model 2 was adjusted for Model 1 plus offspring year of birth, parity of mother, mother's diseases before and during pregnancy, diseases in offspring. Model 3 was adjusted for Models 1 and 2 plus parental marital status and education level in parents, aunts and uncles. P-value for difference in effect between mother's and father's mortality from CVD for 1-SD increase in offspring birthweight was <0.001. P-values for difference in effect between maternal aunts' and uncles' and between paternal aunts' and uncles' mortality from CVD for 1-SD increase in offspring birthweight were both >0.37. To investigate the significance of shared familial factors, we extended our analyses and assessed the role of CVD risk factors in the relationship. The attenuation of offspring BW and CVD mortality association in parents and in aunts/uncles after adjustment for CVD risk factors suggests a contribution of familial factors shared not only in a nuclear family, but also in extended families. The impact of CVD risk factors such as blood pressure, lipids and obesity may support a role of shared genes, as these factors are genetically influenced. [31][32][33] However, the contribution of health-related behaviours such as smoking 34 may indicate the importance of shared environmental factors in the association. Smoking behaviour has been linked to genetic variants, 35 but there is little evidence on shared genetic factors linking smoking and LBW. Furthermore, a role of education in BW and CVD mortality association was observed in all familial relationships. Studies have shown a higher obesity and diabetes risk in parents of offspring with higher BWs. [36][37][38] However, we observed an increased CVD mortality among parents of SGA offspring but not with LGA offspring. 10,39,40 These may be two different mechanisms. It might be possible that parental diabetes/ obesity is more relevant to LGA offspring and CVD to SGA offspring.
Multiple potential mechanisms may explain the associations observed between offspring BW and CVD mortality in parents and aunts/uncles. Genetic confounding is one possible explanation, but for a purely genetic model we expect similar strength of associations in parents and half of this strength in aunts/uncles relationships. However, we found a stronger association in mothers than in fathers and aunts/uncles, suggesting that multiple potential  mechanisms are involved in the mother-offspring association. First, intrauterine factors leading to LBW in offspring through malnutrition, poor placental growth and maternal pelvic restriction is one possibility. [41][42][43] Second, a dual action of maternal genes, contributing to fetal growth both by gene inheritance and by affecting the intrauterine environment, could be another mechanism. 44 Third, maternal health-related behaviours such as smoking may have a direct impact on offspring BW and the mother's own risk of CVD.
We expect the genetic association for fathers to be twice that for aunts/uncles, and presumably the environmental/ behavioural association would also be stronger. However, the associations in fathers were only a little stronger than those for aunts/uncles. This reflects that unobserved behavioural confounders, such as alcohol intake and physical activity, may be important in the paternal association. A similar trend of associations with lung cancer mortality also reflects the significance of behavioural confounders in the paternal association. Furthermore, the similar strength of associations with all classes of aunts/uncles is indicative of a genetic link. These associations may also be partly explained by environmental mechanisms, as parents and their siblings share similar home environments, dietary habits and health-related behaviours during early life. However, previous studies investigating offspring BW and parental sibling characteristics have suggested that maternal aunts but not uncles share important links with offspring BW. They propose that genetic effects from mothers are more important than paternal effects. 45,46  A recent large-scale pedigree analysis suggests that assortative mating generates substantial apparent heritability with respect to mortality. 47 Assortative mating might contribute to the mortality associations in our study. Another explanation could be the genetic nurturing phenomenon, suggesting that genetic and environmental mechanisms are interlinked and genetic effects can exert their impact through an environmentally mediated channel. 48 The complete separation of environmental and genetic components that influence CVD mortality is difficult, and an interaction between these factors may further complicate our understanding.

Strengths and weaknesses
Our study is based on data from the nationwide registers, providing a large sample size and comprehensive population coverage. We established a dataset of offspring, parents and their siblings (aunts/uncles), which provides an opportunity to study the association between BW and CVD mortality in family members at different degrees of relatedness. The ability to include data on CVD risk factors adds novelty to the study. We also calculated BW for gestational age, which gives a precise measure of intrauterine fetal growth. Moreover, detailed information on maternal health before and during pregnancy was also included from the registry data. Diet and physical activity, which could be important in the relationship between BW and CVD mortality, were not included in our study. Education level was included as an indicator of SEP. The data on smoking in pregnancy were collected in the Medical Birth Registry from 1998 onwards. Thus, only a few participants with short follow-up have this information, and the effect of smoking during pregnancy cannot be estimated.

Conclusion
We show that offspring BW was associated with increased risk of CVD in parents and in aunts/uncles, and that established CVD risk factors contributed substantially to associations among family members with a known genetic link. This suggests that both behavioural factors, especially smoking, and shared genetic factors in extended family members, involving these established CVD risk factors, play roles in the associations.

Supplementary data
Supplementary data are available at IJE online. cohorts from the CONOR were used in the analyses: The Tromsø Study (IV and V), Troms and Finnmark Health Study, Nord-Trøndelag Health Study (HUNT), Oslo Health Study, Oppland and Hedmark Health Study (OPPHED) and Hordaland Health Study (HUSK).
Conflict of interest: None declared.