Birth weight, components of height and coronary heart disease: evidence from the Whitehall II study

Abstract

Background Previous studies have suggested that shorter leg length, not trunk length, may explain the inverse association between height and coronary heart disease (CHD) risk. However, investigation of the importance of birth weight for these associations has been limited. This study examines associations of measures of stature and birth weight with CHD risk factors (measures of blood pressure and lipids, 2 h glucose, waist–hip ratio and fibrinogen) and incident non-fatal coronary events in middle age.

Methods Data were derived mostly from the Phase 5 (1997–99) clinical screening of the Whitehall II study of British civil servants. The main cross-sectional analyses included 1084 women and 2290 men with complete data.

Results In women total height and leg length were the measures that tended to be most strongly associated with CHD risk factors, while in men leg length demonstrated the closest associations. Although associations between trunk length and CHD risk factors were weaker, trunk length was the component of height that appeared to be most closely associated with coronary events. Associations between birth weight and CHD risk factors and coronary events were generally weaker than for any measure of stature. Adjustment for birth weight had little effect on associations between components of stature and CHD risk factors or events.

Conclusion Findings from this relatively privileged cohort confirmed that shorter leg length underlies the inverse association between height and CHD risk factors in middle-aged women and men. Furthermore, in this study population shorter trunk length was more closely associated with incident, non-fatal coronary events.

An inverse association between adult height and coronary heart disease (CHD) mortality has been demonstrated in the majority,^1–7 but not all,^8,9 of the studies that have reported the association. Similar inverse associations have been demonstrated for many of the major cardiovascular risk factors.^10–13 It has been proposed that the associations between stature and CHD risk can be explained variously by: genetic inheritance; insulin-like growth factor axis; poor lung function; reverse causation; the observed birth weight–CHD association; vessel diameter; and factors affecting childhood growth.^11,12,14

More recent work, which has examined the components of height, has indicated that leg length, rather than trunk length, is the more important predictor of CHD risk in both women and men.^10–13,15 This finding is able further to elucidate the height–CHD association, as leg length is the component of stature most sensitive to environmental influences on childhood growth before puberty.¹⁶ The foetal origins hypothesis suggests that intrauterine exposures are determinants of both height and other risk factors for CHD, such as high blood pressure.^17,18 However, with limited exceptions,^11–13 few studies have examined associations between the components of stature and CHD risk factors or CHD incidence adjusted for birth weight. Clear identification of the part of the life course at which the major cardiovascular risk factors are determined is important both in terms of elucidating the aetiology of CHD and informing public health policy.

The aims of this study are (i) to examine birth weight and components of stature as predictors of CHD risk (CHD risk factors and coronary events) in middle-aged women and men, (ii) to determine whether the associations between measures of stature and CHD risk are independent of birth weight. Data are derived from Whitehall II, a longitudinal study of white-collar British civil servants.

Methods

Population

The target population for Whitehall II was all London-based office staff, aged 35–55, working in 20 Civil Service departments. With a response rate of 73%, the final cohort consisted of 10 308: 6895 men and 3413 women.¹⁹ The true response rate was higher, however, because ∼4% of those invited were not eligible for inclusion. Although mostly white-collar, respondents covered a wide range of grades from office support to permanent secretary (a public officer appointed to head a ministry, the highest grade in the British Civil Service).

Study design

Baseline screening (Phase 1) of the Whitehall II cohort took place between late 1985 and early 1988. This involved a clinical examination and questionnaire. Since baseline, data collection phases including a clinical examination have been carried out in 1992–93, Phase 3, and 1997–99, Phase 5. Intervening phases involved only a self-administered questionnaire.

Measures

Age. Age was calculated from date of birth and the dates of screening at Phases 1 and 5.

Anthropometry. Total height and sitting height (trunk length) were measured to the nearest millimetre using a stadiometer with the participant standing or sitting as tall as possible with the head in the Frankfort plane. Leg length was calculated as the difference between total height and sitting height. Total height was measured at Phases 1 and 5. The components of height were measured only at Phase 5. Birth weight was self-reported and data collected at Phases 3 and 5 were combined.

CHD risk factors. These were measured at Phase 5. Systolic and diastolic blood pressure were measured in millimetres of mercury (mm Hg). Pulse pressure was calculated as the difference between systolic and diastolic blood pressures.¹³ Total cholesterol, HDL cholesterol, and triglycerides, were measured in a fasting sample of venous serum and 2 h glucose was measured in citrated plasma 2 h after a 75 g anhydrous glucose load. Waist and hip circumferences were measured to the nearest 0.1cm. Fibrinogen was determined in citrated plasma. Fuller details of all measures have been reported previously.²⁰

Socioeconomic position (SEP). Childhood SEP was assessed using a latent variable comprising family conditions before the participants were 16 years of age reported at Phase 5 and self-reported father's social class (Registrar General's Classification) from Phase 1.²¹ Adult SEP was based on employment grade at Phase 5. Twelve non-industrial Civil Service grades were grouped to form three categories: high (senior administrators and managers), intermediate (professional, technical, and executive staff, and middle managers), and low (clerical officers and office support staff).

Health-related behaviours. These were measured at Phase 5. Alcohol consumption was categorized into four levels (three in women) with the highest level (alcohol over the recommended levels) defined as 15 or more units/week for women and 22 or more units/week for men.²² Our current smoking habit measure comprised four categories (never, ex, pipe/cigar, current) with additional adjustment for the number of cigarettes per day in smokers. Weight in kilograms and height in metres were used to calculate body mass index (BMI) as weight/height.²

Total non-fatal coronary (CHD) events. Although most of the anthropometric data were not collected until Phase 5, incident CHD events from Phase 1 to Phase 5 were used as the coronary event outcome, as event data were not available subsequent to Phase 5. Potential cases of definite, non-fatal myocardial infarction (MI) were ascertained by questionnaire items on chest pain,²³ and doctor's diagnosis of heart attack. Details of physician diagnoses and investigation results were sought from clinical records for all potential cases of MI. Twelve-lead resting electrocardiograms were performed at Phase 5 (Siemens Mingorec) and assigned Minnesota codes.²⁴ Based on all available data (from questionnaires, study electrocardiograms, hospital records: acute ECGs and cardiac enzymes), non-fatal MI was defined according to MONICA criteria.²⁵ Classification of MI was carried out blind to other study data independently by two trained coders, with adjudication by a third in the (rare) event of disagreement. Definite angina was recorded for participants who reported symptoms of angina,²⁶ with corroboration in clinical records or abnormalities on a resting ECG, exercise ECG, or coronary angiogram. In addition to definite non-fatal MI and definite angina our total non-fatal CHD events outcome included self-reported cases in the absence of any clinical record evidence of coronary disease. The outcome comprised incident events only and all participants with prevalent CHD (including angina) at Phase 1 were excluded.

Ethical approval

Ethical approval for the Whitehall II study was obtained from the University College London Medical School committee on the ethics of human research.

Study sample and statistical analysis

Of the 10 308 participants at baseline in the Whitehall II study, 8354, 81%, responded at Phase 3 and 7824, 76%, at Phase 5. Of these Phase 5 responders 7269 (93%) attended the screening clinic at Phase 5. From these, participants with missing data for birth weight (2421 missing), or any component of stature (2274 missing), were removed from the dataset leaving a total of 1084 women and 2290 men for analysis. The exclusion of participants with prevalent CHD at baseline (n = 105) or missing values for covariates from the incident event analyses left 2893 participants in the dataset (899 women, 1994 men), who experienced 262 CHD events over the study period: 88 events were in women and 174 were in men, of whom 42 had MI only, 15 had both MI and angina, and 205 had angina only. Participants lost to follow-up, or with missing data were more likely to be older, from the lower grades, and women. Despite these exclusions, comparison of the angina rates with those from the Health Survey for England for 1998 showed rates to be similar.²⁷

Pearson correlation coefficients were calculated between the anthropometric variables. Linear regression was used to assess cross-sectional associations with continuous cardiovascular risk factors and logistic regression with dichotomous risk factors. Values for serum triglyceride and total/HDL cholesterol scores were skewed and required log transformation. The natural logarithm of these variables was used in the regression models. The coefficients derived from these analyses are presented in Tables 3 and 4. Each coefficient represents the age-adjusted absolute change in the risk factor of interest associated with one standard deviation increase in the anthropometric measure of interest. For example, a one standard deviation increase in height in women (6.4 cm; Table 1) was associated with a decrease of 2.21 mm Hg in systolic blood pressure and a decrease of 0.18 g/l in 2 h glucose (Table 3). In the case of log transformed variables, 100 times the coefficient presented represents the percentage change in the risk factor of interest. All analyses were conducted separately in women and men and adjusted initially for current age and then age and birth weight. Cox proportional hazard models were used to determine the hazard ratio (and 95% CI) for CHD events associated with a one standard deviation increase in the anthropometric measures and adjusting for age, birth weight, childhood and adulthood SEP, and health behaviours at baseline (Phase 1). Cox models were initially stratified by sex; however, results were virtually identical in women and men (all sex interaction P-values were >0.7) and pooled estimates were therefore presented.

Results

Table 1 presents descriptive statistics. There are considerable differences between women and men for most measures; the most notable being the proportion in the low employment grades. Mean height at Phase 1 and Phase 5 was identical in both sexes.

Table 2 shows age-adjusted correlations between the measures of stature and birth weight. Correlation coefficients between height and both leg and trunk length are high, reflecting the fact that the latter two are components of the former. Leg length and trunk length are positively and moderately correlated, while correlations between birth weight and the components of stature are somewhat lower, ranging from 0.15 to 0.22.

Cross-sectional associations between birth weight, height, leg length, and trunk length and CHD risk factors are shown in Tables 3 and 4; the associations for height, leg length, and trunk length are presented adjusted for age and adjusted for age and birth weight.

Birth weight

Birth weight was not associated with age—Table 3. In women, inverse associations were observed between birth weight and all the blood pressure measures (the greater the birth weight, the lower the blood pressure) and 2 h glucose. In men, birth weight was inversely associated with systolic blood pressure, 2 h glucose, and fibrinogen.

Height

As expected, shorter height was associated with older age—Table 3. Strong inverse associations were also observed between height and all the blood pressure measures in women but not in men. There were strong associations between height and 2 h glucose, triglyceride level and waist–hip ratio in both sexes, and total cholesterol and total/HDL cholesterol ratio in men. These associations were slightly weakened after adjustment for birth weight, but all remained statistically significant with the exception of triglycerides in women.

Leg length

Leg length was not associated with age at measurement. Strong inverse associations were observed between leg length and pulse pressure and systolic pressure in both sexes and with diastolic pressure in women. There was a strong inverse association between leg length and total cholesterol in men. This association was weaker in women. Longer leg length was associated with a lower total/HDL cholesterol ratio, triglyceride level, and 2 h glucose in both sexes and waist–hip ratio in men. There was no association between leg length and fibrinogen in either sex. These associations were slightly weakened after adjustment for birth weight, but all remained statistically significant with the exception of 2 h glucose in women.

Trunk length

A strong inverse association between age and trunk length was observed in both sexes. There was strong evidence of inverse associations between trunk length and pulse pressure and systolic pressure in women but a positive association between trunk length and diastolic pressure in men. No associations were observed between trunk length and any lipid measure in either sex, with the exception of total cholesterol in men, but strong inverse associations were observed between trunk length and 2 h glucose and waist–hip ratio in both sexes and there was an association with fibrinogen in men. Adjustment for birth weight weakened the associations between trunk length and pulse pressure and systolic pressure in women but strengthened the positive association with diastolic pressure in men. Birth weight adjustment had little effect on the associations between trunk length and 2 h glucose or waist–hip ratio, but the association with fibrinogen in men was weakened.

Coronary events

Evidence of an inverse association between birth weight or leg length and non-fatal CHD events in these data was no more than indicative—Table 5. Associations between height and incident CHD events were almost identical using height measured at Phase 1 and height measured at Phase 5 (n = 2893) and age-adjusted analyses showed evidence of an inverse association between height and CHD events at both Phases. This association remained significant after adjustment for birth weight and childhood SEP but not after further adjustment for adult SEP. There appeared to be strong evidence of an inverse association between trunk length and CHD in age-adjusted analyses. The estimate changed little after adjustment for birth weight and childhood SEP but was considerably reduced after adjustment for adult SEP. As 78% of the coronary events in the present study were angina events we repeated our analyses excluding these. The inverse association between birth weight and MI events was considerably stronger than for our total coronary events outcome and was statistically significant in the fully adjusted model. However, analyses of the measures of stature in this restricted dataset served only to increase the confidence intervals, and none of the associations were statistically significant (data available on request).

Discussion

Of the measures of stature, overall height appeared to demonstrate stronger associations with risk factors for CHD than either leg length or trunk length in women, while leg length demonstrated stronger associations in men. However, in both sexes leg length, rather than trunk length, was the component of height that tended to be most strongly associated with CHD risk factors. In contrast to these findings, the association between leg length and non-fatal coronary events was no more than indicative, but there appeared to be a strong inverse association between trunk length and events, which was weakened on adjustment for adult employment grade. Associations between birth weight and CHD risk factors tended to be weaker than for any measure of stature, and adjustment for birth weight had little effect on associations between height, leg length, or trunk length and CHD risk.

Relationship to previous work

The evidence of an inverse association between birth weight and blood pressure measures in Whitehall II women replicates well-documented associations between these measures.^14,28,29 An inverse association between birth weight and triglyceride levels was observed in women and inverse associations with 2 h glucose were observed for both sexes. A recent review of the relationship between birth weight and blood lipid concentrations in later life concluded that the existing literature showed no consistent association between birth weight and lipid levels, with the possible exception of an inverse association with triglyceride level.³⁰ Similarly, an inverse association between birth weight and fasting glucose has been demonstrated previously,¹⁴ and it has been suggested that insulin resistance may, at least in part, mediate the association between birth weight and CHD.³¹

It is possible, given the large number of correlations between the components of stature and the CHD risk factors, that some of our observed associations were due to chance. At the 5% level of probability 7 of the 132 correlations presented could be thus generated. However, the fact that the number of statistically significant correlations observed was an order of magnitude in excess of that, 72/132, suggests that chance is an unlikely explanation for the majority of our findings. Furthermore, our age-adjusted findings of inverse associations between height and CHD risk factors in both sexes replicate those of a large body of previous work that has documented strong associations between adult height and the major cardiovascular risk factors.^10–12,32 More recent work in the Caerphilly, Midspan, and British Women's Heart and Health (BWHH) cohorts has shown the leg length component of height to be the more important determinant of CHD risk factors in middle-aged and older women and men,^10,11,15 a finding endorsed by the strong associations observed between leg length and CHD risk factors in the present study.

Most previous attempts to separate in utero and post-natal influences on later disease risk have examined correlations between birth weight and the components of stature. This work has shown that there is no consistent evidence that birth weight is more strongly related to leg length than trunk length.³³ However, more recent studies have made some adjustment for birth weight.^11,12,32 Self-reported birth weight data were available for 29% of Midspan participants and measured birth weight was available for all participants in the 1946 birth cohort. Adjustment for birth weight in Midspan did not have any substantial effect on associations between leg length and cholesterol or fibrinogen, but inverse associations with blood pressure in Midspan and the 1946 cohort were slightly weakened.^11,32 Similarly in the present study adjustment for birth weight slightly weakened the inverse associations between measures of stature and CHD risk factors, with the greatest attenuation being observed for measures of blood pressure. However, it should be borne in mind that associations between birth weight and CHD risk factors could be weaker than those for the components of stature because our birth weight measure, which was self-reported rather than measured in clinic, will be subject to greater measurement error. No validation of self-reported birth weights has been carried out in the Whitehall II study. Findings from other studies show the correlation between self-reported and recorded birth weight to be relatively good in younger people but only moderate in elderly respondents.^34,35

A considerable body of work has established birth weight as a strong predictor of CHD in later life,^3,14,31,36 although there is also a body of work, which indicates that exposures over the life course may be more important determinants of later disease risk,³⁷ and not all studies have found an association between birth weight and CHD.³⁸ In the present study the hazard ratio for the association between birth weight and total coronary events was indicative of a protective effect for higher birth weight, an association that reached statistical significance in the fully-adjusted model when the analyses were restricted to MI events. The evidence of an inverse association between height and coronary events observed in our data is well documented in the literature,^1–7 and, in common with our findings, previous work has shown this association to be independent of birth weight in both sexes.^2,3

Contrary to the findings of the present study, previous studies that have examined associations between the components of height and CHD have demonstrated a stronger association with leg length than trunk length in both sexes.^10,12 Shorter leg length, but not shorter height or trunk length, was associated with incident CHD events in age-adjusted analyses of middle-aged men in the Caerphilly study.¹⁰ In the BWHH cohort of older women shorter height, leg length, and trunk length were all associated with prevalent CHD in age-adjusted analyses. However, only the associations between height and leg length and CHD remained after adjustment for smoking. Further adjustments, including birth weight, which was available for 33% of the women, did not attenuate these associations.¹² We were concerned that the high proportion of angina events in our CHD events outcome may have influenced our findings. However, repeating the analyses excluding the angina events produced very similar findings to those presented in terms of effect size, although none of the associations was statistically significant owing to small numbers. Both Caerphilly and BWHH are studies of population samples so will include a wider socioeconomic and occupational range than Whitehall II. The largest attenuation of the association between trunk length and CHD events was seen on adjustment for adult social position, measured as Civil Service employment grade. While grade is considered to be a relatively good marker of social position the possibility of residual confounding remains. Although blue-collar employees typically engage in lower rates of leisure-time physical activity than white-collar workers,^39,40 recent data from the EPIC study showed a significantly higher proportion of men in manual social classes in the most physically active groups, although there was no consistent trend for women.⁴¹ A number of other studies similarly have shown physical activity to be higher among blue-collar workers when activity at work is taken into account.^42,43 As Whitehall II participants are almost exclusively white-collar workers we thought that they may do less exercise and thus have greater central obesity than older study populations, which contain a higher proportion of blue-collar workers. Although we had already adjusted for BMI we further examined the effect of adjustment for waist circumference and exercise at Phase 5 on the age and sex-adjusted associations between our anthropometric measures and CHD events. The effect of this was to strengthen, rather than attenuate, the observed protective effects of greater height, leg length, and trunk length. A further possible explanation is that participants in Whitehall II were younger at baseline than either Caerphilly men (aged 45–59 when examined between 1979 and 1983) or BWHH women (aged 60–79 when examined between 1999 and 2001). Associations between the components of height and CHD among younger, less-deprived groups may be different than in these slightly older, more-deprived cohorts.

Methodological considerations

We examined whether our findings were biased because responders to Phase 5 were different from non-responders. To do this we examined associations between height at Phase 1 and CHD risk factors at Phase 1 and tested to see whether the association differed between participants who subsequently attended or did not attend the screening at Phase 5. Despite the fact that non-attendees were more likely to be older, from the lower grades, and women, there was no evidence that the associations between height and CHD risk factors at Phase 1 differed between those who attended the Phase 5 screening and those lost to follow-up. We also examined associations between height at baseline and three CHD outcomes for all the participants in the Whitehall II study. The hazard ratios obtained for these outcomes; all CHD, CHD death or non-fatal MI, and angina were between 0.82 and 0.85—values very similar to that for height at baseline presented in Table 5 (0.86).

Given the discordance between our findings for the CHD risk factors and the coronary events we were concerned that this might be an artefact generated either by the restricted datasets available for our events analyses or by our CHD events outcome.

To investigate the problem of the restricted datasets, we first repeated the analysis of each association between the anthropometric measures and CHD risk using datasets, which included the maximum number of participants. Although findings for the majority of the CHD risk factors and the coronary events were little different from those presented, we did find additional inverse associations between trunk length and diastolic pressure and triglyceride levels in women. Second, we repeated the analysis of each association between the anthropometric measures and CHD risk factors using a dataset restricted to those participants included in the analysis of CHD events. Again the resulting associations were very similar to those presented, with the exception that the associations between birth weight and total cholesterol in women and diastolic pressure in men were strengthened, as were the associations between leg length and diastolic pressure and total cholesterol in women (data available on request).

To investigate the problem of our CHD events outcome, we first repeated our analyses of CHD events in a dataset that included participants with prevalent CHD at baseline. In these analyses the inverse association between birth weight and CHD events was considerably stronger than in the dataset including incident events only and was statistically significant in the fully adjusted model. However, associations between the measures of stature and CHD events were very similar to those presented (data available on request). We then examined the association between height at baseline and CHD death. The hazard ratio for CHD death was 0.66 (95% CI 0.54–0.82), indicating a much stronger association than for our existing non-fatal CHD events outcome. Unfortunately we were unable to repeat these analyses for the components of height as these were not measured until Phase 5. However, the strength of the association with height at baseline indicates that our CHD events outcome is likely to be an underestimate of effects. If this underestimation was to affect leg length rather than trunk length this could provide an explanation for our contradictory findings.

Shrinkage; which commences around the age of 40 in both sexes,⁴⁴ increases with age, and is associated with some disease processes, has the potential to bias observed associations. Leg length and trunk length are differently affected by shrinkage, which takes place predominantly in the trunk.⁴⁴ Some evidence for this was observed in the present study where trunk, but not leg length, was negatively associated with age, particularly so in women. Disease-related shrinkage may inflate observed associations between stature and CHD (reverse causation), particularly those with trunk length. However, we found little evidence that shrinkage over the study period was responsible for the observed associations: (i) shrinkage was very limited; mean overall height was identical at Phase 1 and Phase 5 in both women and men (Table 1), and change in height among participants event free between baseline and follow-up was 0.42 cm, while among those experiencing a CHD event it was 0.43 cm, P = 0.86; (ii) analyses of coronary events were restricted to participants free of manifest CHD at baseline limiting reverse causation; and (iii) associations between height at Phase 1 and Phase 5 with incident CHD were almost identical (Table 5). However, there remains the strong association between trunk length and age (Table 4). Although participants apparently did not shrink during the study period, our observed association could still be due to age-related shrinkage that occurred before study inception. Differences in height or components of height may also be due to a cohort effect. We were aware that participants born from 1941 onwards, unlike those born before 1941, would have been recipients of war-time food supplements for expectant and nursing mothers, and young children. For this reason we compared height at Phase 1 and Phase 5 for those born before and after 1941. While those born before 1941 were slightly shorter than those born after 1941, the age-adjusted differences were small (<0.14 SD) and not statistically significant. In these data, it was not possible further to differentiate age and cohort effects as the first time these participants were measured was at age 35–55 when older members of the cohort may have already experienced some age-related shrinkage.

Conclusion

In common with previous studies, our findings indicate that in the Whitehall II cohort, leg length is the component of height more closely associated with CHD risk factors in both women and men. Our findings also provide some evidence that associations between measures of stature and CHD risk factors are independent of birth weight. In contrast to our results for CHD risk factors we found trunk length to be the component more closely associated with coronary events. While in part this may be due to our reliance on restricted datasets, it may also be because participants in the Whitehall II study are younger and have experienced less deprivation than previous cohorts in which these associations have been examined. Future studies using data from national birth cohorts will be able to investigate both age and cohort effects on associations between measures of stature and coronary risk.

Contributors

Jane Ferrie wrote the original and successive drafts of the paper and was involved in the data collection for Phase 5. Claudia Langenberg advised on the study design and drafts of the paper. Martin Shipley performed the analyses and advised on drafts of the paper. Michael Marmot designed and directs the Whitehall II study and commented on drafts of the paper.

Conflict of interest

None.

The Whitehall II study has been supported by grants from the Medical Research Council; British Heart Foundation; Health and Safety Executive; Department of Health; National Heart Lung and Blood Institute (HL36310), US, NIH: National Institute on Ageing (AG13196), US, NIH; Agency for Health Care Policy Research (HS06516); and the John D and Catherine T MacArthur Foundation Research Networks on Successful Midlife Development and Socio-economic Status and Health. J.E.F. was supported by the Medical research Council (Grant number G8802774) during the preparation of this paper. C.L. is supported by an MRC Research Training Fellowship. M.J.S. is supported by a grant from the British Heart Foundation. M.G.M. is supported by an MRC Research Professorship. We also thank all participating Civil Service departments and their welfare, personnel, and establishment officers; the Occupational Health and Safety Agency; the Council of Civil Service Unions; all participating civil servants in the Whitehall II study; and all members of the Whitehall II study team. The authors would like to thank George Davey Smith for the ideas on which this paper is based.

References