Article Text

Download PDFPDF

The association of childhood height, leg length and other measures of skeletal growth with adult cardiovascular disease: the Boyd–Orr cohort
  1. E Whitley1,
  2. R M Martin1,
  3. G Davey Smith1,
  4. J M P Holly2,
  5. D Gunnell1
  1. 1Department of Social Medicine, Canynge Hall, Bristol, Avon, UK
  2. 2Department of Clinical Sciences at North Bristol, Medical School Unit, Southmead Hospital, Bristol, Avon, UK
  1. Correspondence to Dr Elise Whitley, Department of Social Medicine, Canynge Hall, Whiteladies Road, Bristol BS8 2PR, UK; elise.whitley{at}


Background Taller adults have a reduced risk of cardiovascular disease, and there is some evidence that pre-adolescent exposures, indexed by leg length, underlie this association. Associations with other aspects of skeletal size in childhood have not previously been investigated.

Methods We have examined associations of cardiovascular mortality and morbidity with childhood height, shoulder breadth, leg, trunk and foot length using a cohort of children whose families participated in a 1937–9 survey of diet and health followed up for 59 years.

Results Altogether 2642 traced participants had at least one anthropometric measurement; a subsample (n=1043), completed the Rose angina questionnaire and provided information about doctor-diagnosed ischaemic heart disease (IHD) in 1997–8. Childhood stature was weakly inversely associated with cardiovascular mortality, and leg length was the component with the strongest associations. There was evidence from secondary analyses that childhood anthropometric measurements were inversely related to early (age <65 years) rather than late cardiovascular mortality. Childhood stature was inversely associated with self-reported IHD and associations with leg length were strongest. Associations were somewhat attenuated in models including terms for having been breastfed and socioeconomic position.

Conclusion Pre-adult exposures are more strongly associated with cardiovascular morbidity than mortality, and they affect premature cardiovascular mortality more than later mortality.

  • Cohort
  • coronary heart disease
  • early life origins
  • epidemiology
  • height
  • morbidity
  • mortality SI
  • socioeconomic

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Prospective studies consistently show that taller adults have lower risks of cardiovascular mortality and morbidity.1–6 A 47-year follow-up of a cohort of children who took part in a 1937–9 survey of family diet and health (the Boyd–Orr cohort) indicated that the leg length component of height was the main contributor to this association.7 This stronger association of cardiovascular disease (CVD) with leg length has been replicated in cohorts of adult men and women.8–10 Furthermore, a number of studies show protective associations between long leg length and several cardiovascular risk factors, including insulin resistance,10–14 while associations with trunk length are weaker or absent.

Mechanisms underlying these associations remain unclear. Because prepubertal linear growth occurs more in the leg than the trunk,15 leg length may be a particularly sensitive indicator of prepubertal growth-influencing exposures. Peak growth in other anthropometric measures also occurs at different times; for example, foot length is one of the first components to reach peak growth while shoulder breadth is among the last.15 Specific components of stature may therefore be particularly sensitive indicators of childhood circumstances. The main environmental influences on childhood growth are diet, health and socioeconomic circumstances,16–19 and poor growth in childhood often persists into adulthood with smaller children tending to be smaller as adults.20 21 It is therefore plausible that childhood stature may have long-term associations with adult disease risk.

Our previous study of childhood height and CVD risk in the Boyd–Orr cohort was based on deaths occurring up to 1995.7 Since then the number of deaths has trebled, we have collected self-reported data on non-fatal cardiovascular endpoints, and retrieved additional anthropometric data on childhood foot length and shoulder breadth from original survey records.22 We now investigate two interrelated hypotheses. First, that childhood leg length is more strongly associated with CVD, ischaemic heart disease (IHD) and stroke mortality than trunk, foot or shoulder length and, second, that associations of the anthropometric measures with self-reported prevalent IHD are similar to those with IHD mortality.


The Boyd–Orr cohort

The Boyd–Orr cohort is an historical cohort based on the records of 4999 children from 1343 mainly working class families living in England and Scotland who took part in a detailed survey of family diet and childhood health in 1937–9 ( The original survey examined growth and diet in relation to income, and detailed socioeconomic data are available for all study members. Cohort members also have data on infant feeding, childhood family diet and nutritional status, and 92% have been traced on the National Health Service Central Register. Surviving study members are now aged over 70 years and the research team is notified of all deaths.

Measures of height, leg length, foot length and shoulder breadth were made in 1937–9. Height, leg length and shoulder breadth were measured to the nearest millimetre using a stadiometer (height) and steel measuring tape (leg length, shoulder breadth). Leg length was measured as the distance from the summit of the iliac crest to the floor. Biacromial (shoulder) breadth was measured as the distance between the external margins of the acromion processes (bony prominence of the shoulder blades) with the subject seated, with hands on knees and muscles relaxed.22 Foot size was derived by measuring (to the nearest millimetre) chalk imprints of the children's feet made at the survey clinics. A research assistant visited the Rowett Research Institute Archive (where original survey records are held) and measured foot length as the maximum length from the heel to the tip of the toes. The research worker was blind to the subject's vital status. For analysis, the mean of the left and right foot measures was used.

Cause of death was derived from all death certificate fields. Specific causes of interest were (1) all-cause; (2) CVD (ICD9: 390–459; ICD10: I00–I99); (3) IHD (ICD9: 410–414, 429.2; ICD10: I20–25; I51.6); and (4) stroke (ICD9: 430–438; ICD10: I60–I69). An additional outcome was self-reported IHD, derived from questionnaire data collected from a subset of the cohort in 1997–8 (who were alive at the time), and based on subjects' recall of any doctor's diagnosis of myocardial infarction, angina, coronary angioplasty or coronary artery bypass and the Rose angina questionnaire.24

Statistical analysis

Analyses were restricted to study members aged 2–14 years at the time of their measurements. Measurements on younger children are prone to inaccuracy and the number of older subjects was sparse. All anthropometric measures were transformed into sex-specific age standardised z-scores. For some measures, a cubic term for age was required to obtain a good fit and, although this term did not improve the fit for all measures, it was used throughout for consistency. The resultant z-scores measure the number of standard deviations a child's measurement lies above or below the predicted value for his/her age (in months) and sex.

Cox's proportional hazards regression models were used to examine associations of childhood anthropometric measures with mortality, and logistic regression models to examine associations with self-reported IHD. Analyses were restricted to subjects who were alive and resident in Britain on 1 January 1948; deaths occurring before this date were excluded from analyses. In Cox regression models, follow-up started on 1 January 1948 and ended on the date of death, embarkation or end of follow-up (7 March 2007). Robust standard errors were calculated in all models to allow for clustering effects in children from the same household.25

Previously identified factors that may explain leg length associations with adult disease include childhood and adult socioeconomic position. Childhood socioeconomic position was available from baseline household measures: social class of head of household; per capita household income and per capita household food expenditure. Adult socioeconomic position was available for individuals using the Townsend score,26 which is an ecological measure of socioeconomic deprivation comprising census area data on household overcrowding, unemployment, housing tenure and car ownership. Townsend scores were computed using 1991 census data for the area of residence at the time of tracing in the early 1990s or, for those who had died or emigrated before the 1990s, the area of residence at the time of death or emigration. Patterns of infant feeding may also have a long-term impact on growth and CVD risk,27 and so breastfeeding was also included as a possible explanatory variable.

Data on birth weight, adult stature (height and inside leg length) and body mass index were collected by questionnaire from surviving survey members in 1997–8 and are available for between a quarter (birthweight) and a third (adult stature) of the original cohort. The effects of controlling for these factors were examined in separate models. We also investigated whether associations differed in males versus females and in subjects who were under 8 years compared with over 8 years when they were measured. We used formal tests of statistical interaction (effect modification) to investigate differences in associations between subgroups. The cut point of less than 8 or over 8 years is consistent with our previous analyses and ensures that all subjects who were measured aged under 8 years were prepubertal.7 The main analyses were repeated, restricted separately to deaths occurring at ages under 65 years compared with over 65 years as preliminary analyses showed no evidence of an association in the cohort as a whole and our previous analyses,7 carried out when subjects were on average 65 years old, had shown inverse associations between height, leg length and CVD mortality. Finally, the relative strength of associations with childhood and adult height were compared in subjects with both measurements. Children who enter puberty late, perhaps as a consequence of early life exposures, will appear shorter than their peers in childhood but may well achieve the same final height. If these early life exposures also influence CVD risk, then associations with childhood anthropometric measures might be expected to be stronger than those with final adult height.


Of the original 4999 cohort members, 4437 (88.8%) were aged 2–14 years at the time of the survey. Shoulder breadth, height, trunk, leg and foot length were available for 2878 (64.9%), 2960 (66.7%), 2860 (64.5%), 2860 (64.5%) and 2648 (59.7%), respectively, and 2969 (66.9%) had at least one measure. Baseline characteristics of these children are summarised in table 1 using the mean (SD) in children aged 2, 5, 10 and 14 years ± 3 months at measurement. Mean measurements increased with age along with SD, particularly at age 14 years, probably reflecting increasing variation as children reach their adolescent growth spurt. Differences between the sexes were small. Boys were slightly larger at younger ages, while girls, who reach their adolescent growth spurt earlier, were marginally larger at age 14 years. Adult height and inside leg length, available for a subset of those with at least one childhood measure, show clear and expected gender differences.

Table 1

Baseline characteristics of boys and girls aged 2, 5, 10 and 14 years ± 3 months at measurement

Of those subjects with at least one childhood anthropometric measure, 2642 (89.0%) were traced, alive and resident in Britain on 1 January 1948, and are included in the analyses. Untraced subjects were older (8.8 vs 7.8 years); more likely to be female (12.7% vs 9.1%) and of marginally higher childhood socioeconomic position. However, there were no differences in anthropometric measures between traced and untraced subjects. A total of 1043 (39.5%) traced subjects with at least one anthropometric measurement also took part in the 1997–8 survey.

Pairwise correlations between the anthropometric measures (standardised for age and sex) ranged from 0.28 (trunk–leg length) to 0.89 (height–leg length) (p values for all correlations <0.001). The largest correlation coefficients were those with height (height–foot length 0.76; height–shoulder breadth 0.61; height–trunk 0.68) and the smallest were generally those with trunk.

There were 954 (36.1%) deaths during follow-up among traced subjects with at least one anthropometric measurement. Of these, 470 were coded as CVD, 281 as IHD and 92 as stroke. Among those taking part in the 1997–8 survey, there were 198 (19.0%) deaths (96 CVD, 61 IHD and 20 stroke) and 185 (17.7%) cases of self-reported IHD (based on recall of a doctor's diagnosis or the Rose angina questionnaire).

HRs and 95% CI for mortality associated with a 1 SD increase in childhood stature are presented in table 2 for subjects with complete data on all explanatory factors (childhood and adult socioeconomic position and breastfeeding status). Age and sex-adjusted HRs for all-cause mortality were very close to 1 while HRs for CVD mortality were generally consistent with a marginally inverse association (of decreasing hazard with increasing stature). However, the magnitude of these associations was modest and all CI included 1. Additional adjustments for explanatory factors led to a slight attenuation of effects. In the subset of participants with measures of birthweight, adult stature and adult body mass index, controlling for these measures did not substantially alter the associations (not shown).

Table 2

Increase in HR (95% CI) for all-cause, all cardiovascular, IHD and stroke mortality in original cohort associated with 1 SD increase in childhood stature (with robust standard errors)

Just under half the subjects included in the analyses were male. Associations were very similar in men and women, with an isolated, modest suggestion that the inverse association between stature and IHD mortality was restricted to women (HR (95% CI) in men 1.00 (0.86 to 1.16) vs women 0.79 (0.62 to 1.01), p for interaction 0.10). Approximately half the children were aged under 8 years (ie, prepubertal) at the time of measurement. There was no consistent evidence that associations differed by age at measurement.

Table 3 presents ORs for self-reported IHD associated with 1 SD increase in childhood stature in traced study members with anthropometric measurements who also took part in the 1997–8 survey. Increasing childhood stature was associated with decreasing odds of self-reported IHD, with the strongest associations observed for height and leg length. These associations were somewhat attenuated by additional adjustments for factors potentially on the causal pathway but ORs remained below 1.

Table 3

Increase in ORs (95% CI) for self-reported IHD in 1997–8 cohort associated with 1 SD increase in childhood stature (with robust standard errors)

Table 4 presents age and sex-adjusted analyses separately for deaths at ages less than 65 compared with over 65 years. HRs for deaths occurring at over 65 years were consistently close to 1 and there was no evidence of any association between trunk length and mortality at either age. However, increases in the other four anthropometric measurements were consistently associated with decreases in hazard for deaths occurring at less than 65 years and the strongest associations were those with leg length. Formal statistical evidence of a difference was weak (eg, p for difference between early vs late CVD/leg length associations 0.35). Associations were only marginally attenuated by adjustment for breastfeeding, childhood and adult socioeconomic position (SES) (not shown) and the overall pattern of decreasing hazard in earlier deaths versus no association in older deaths remained, particularly for leg length (HR (95% CI) for SD increase in leg length for younger vs older deaths from all CVD 0.85 (0.73 to 1.00) vs 1.01 (0.88 to 1.17); IHD 0.81 (0.66 to 0.99) vs 1.01 (0.84 to 1.23) and stroke 0.84 (0.59 to1.21) vs 0.97 (0.70 to 1.35)).

Table 4

Increase in HR (95% CI) for all-cause, all cardiovascular, IHD and stroke mortality in original cohort associated with 1 SD increase in childhood stature (with robust standard errors) by age at death

Adult height was available for 1026 (38.8%) traced subjects with childhood anthropometric measurements. Associations between mortality and childhood versus adult height were very similar for all causes (HR (95% CI) for childhood 0.97 (0.83 to 1.14) vs adult height 1.01 (0.88 to 1.17)), all CVD (childhood 1.07 (0.85 to 1.35) vs adult 1.08 (0.87 to 1.34)) and IHD (childhood 1.06 (0.79 to 1.42) vs adult 1.09 (0.85 to 1.40)). Stroke associations were somewhat stronger with childhood height (0.78 (0.42 to 1.48)) than with adult height (0.98 (0.60 to 1.59)).


Analyses are based on a large cohort followed from a young age for 59 years with detailed data on socioeconomic position and other childhood circumstances. Anthropometric measurements were made in childhood, well before chronic disease onset, and were standardised for sex and age at measurement. The majority of study participants were traced and followed up and over 900 deaths have been identified, of which approximately half were caused by CVD. The tables summarise a large number of comparisons and we have therefore focused on the magnitude and consistency of HRs and ORs rather than the results of individual significance tests.

Much of the existing evidence on stature and adult disease risk comes from studies of adult anthropometric measurements. However, childhood height, measured here at a mean of 7.9 years, may better reflect prepubertal growth-influencing exposures than final (adult) height, which reflects a combination of childhood growth plus age and duration of pubertal maturation. Moreover, childhood height is more strongly associated with circulating insulin-like growth factor I levels,28 which have been hypothesised as contributing to height–coronary heart disease associations.29

There was no association between childhood stature and all-cause mortality. However, we observed weak associations of decreasing cardiovascular mortality with increasing childhood height and, although CI included 1, HRs were of a similar magnitude to those reported for adult height.2 3 In addition, the leg length component of stature was generally most strongly associated with cardiovascular mortality and, again, HRs were similar to those reported for adult leg length.8–10 We are not aware of any other studies that have considered associations with foot length or shoulder breadth, and we found no consistent evidence of strong associations with these measures.

Analysis of self-reported IHD was based on a subset (approximately 40%) of the original cohort who took part in an additional survey in 1997–8. Increasing childhood stature was associated with decreasing self-reported IHD and associations were stronger than those with mortality, in spite of the reduced number of subjects. However, these differences were compatible with chance, and there are no consistent reports in the literature of differences in associations of stature with cardiovascular morbidity and mortality.

Associations were somewhat attenuated after adjustment for childhood and adult socioeconomic position and breastfeeding, particularly those with self-reported IHD. This suggests that childhood stature–IHD associations may index long-term effects of breastfeeding or factors associated with social class, such as income, diet, smoking or obesity. There are plausible mechanisms linking breastfeeding, socioeconomic position, stature and IHD, for example insulin-like growth factor I levels are positively associated with childhood growth28 and inversely associated with adulthood IHD.29

Previous analyses of these data7 reported relatively strong associations of decreasing cardiovascular mortality with increasing childhood height and leg length, which was most marked in children measured at ages less than 8 years. We have repeated this analysis in an updated dataset including three times more cardiovascular deaths and, overall, found only weak evidence of such an association, with no indication of differential effects in children who were measured prepubertally. The previous analysis (followed up to 1995) covered deaths occurring early in follow-up, suggesting that effects may be limited to premature deaths. This idea is supported by analyses subdivided by age at death, in which associations with height, shoulder breadth, foot and leg length were consistently restricted to deaths occurring at ages less than 65 years with no associations with deaths at ages over 65 years. This differential effect in early versus later deaths may be a result of more accurate death certification in younger groups30 leading to more precise estimates of associations, or may indicate a distinct aetiology. However, these results are based on secondary analyses and should be interpreted with caution.

There are limitations to the current analysis. The first is the possibility of measurement error. Anthropometric measurements were made in childhood and, with the exception of foot length, used standard methods. Measurements in children under 2 years were excluded as they were likely to be unreliable, and implausible measurements were also excluded. However, random errors in these measurements may remain, which could lead to an underestimation of stature–morbidity/mortality associations. These effects are likely to be minimal for height associations and greatest for those with foot length. Likewise, measurement error in self-reported IHD may have led to an underestimate of stature–morbidity associations.31

Other unmeasured factors not considered in the analyses may underlie the associations; for example growth-regulating genes may be independently associated with IHD. Alternatively, taller individuals may be more successful in later life and this may protect against IHD rather than some growth-influencing exposure. However, these explanations are purely speculative.

Children with anthropometric measurements were younger, with higher socioeconomic position than those without, whereas untraced subjects were older, more likely to be female and, again, have higher socioeconomic position. In addition, self-reported IHD was collected in 1997–8, and so this analysis is restricted to those who were still alive at the time. Although these restrictions may limit the generalisation of our findings, there is no reason to suppose that they led to any differential bias, and it is reassuring that there were no differences in anthropometric measurements between traced and untraced subjects. In addition, results may not be generalisable to the current era because of differing influences of environmental exposures today compared with the pre-war years. Factors influencing growth and disease in the first half of the 20th century may be less important today, for example bottle-feeding was strongly associated with acute infectious disease morbidity in pre-war infants, but this association is weaker today because of improved sanitary conditions involved in the make-up of formula feeds.

Adult stature, particularly leg length, is associated with CVD. A separate analysis of these data32 indicates that, in addition to genetic influences, anthropometric measurements in childhood are affected by many aspects of diet, housing and socioeconomic position, and that influences are apparent from a young age. The current analysis suggests that these exposures, indexed by childhood stature, are more strongly associated with cardiovascular morbidity and premature mortality than with later mortality.

What is already known on this subject

Taller adults have been shown to have a reduced risk of CVD, and there is some evidence that pre-adolescent exposures, indexed specifically by leg length, may underlie this association. However, childhood stature may better reflect prepubertal growth-influencing exposures than final (adult) stature, which reflects a combination of childhood growth and also age and duration of pubertal maturation. Peak growth in different anthropometric measurements occurs at different times during puberty so that specific components of stature may be particularly sensitive indicators of childhood experiences such as diet, health and socioeconomic circumstances.

What this study adds

Height in childhood was weakly inversely associated with cardiovascular mortality and self-reported IHD. Leg length was the component of stature with the strongest associations. Associations were somewhat attenuated by adjustment for breastfeeding status and childhood socioeconomic position. Secondary analyses suggest that childhood anthropometric measurements were inversely related to early (age <65 years) rather than late cardiovascular mortality.


The authors would like to thank Professor Peter Morgan, director of the Rowett Research Institute for the use of the archive, and in particular Walter Duncan, honorary archivist to the Rowett Institute. They would also like to thank the staff at the NHS Central Register at Southport and Edinburgh, and Professor John Pemberton for information concerning the conduct of the original survey. They wish to acknowledge all the participants and research workers in the original survey in 1937–9. Clare Frobisher, Pauline Emmett and Maria Maynard undertook the re-analyses of the childhood household diet diaries. They authors also thank Mark Taylor, who measured the foot lengths from the archived foot imprints.



  • Funding The World Cancer Research Fund supported the follow-up of the Boyd–Orr cohort and the current analyses.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.